Optimizing roundabout safety: using UAVs and computer vision for driver behavior analysis

Optimizing roundabout safety: using UAVs and computer vision for driver behavior analysis | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Optimizing roundabout safety: using UAVs and computer vision for driver behavior analysis Chang Saar Chai, Jit Boon Bong, Kennedy Kwong Shin Tiong, Lam Tatt Soon This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5680878/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Roundabouts are safer than standard intersections, but traffic accidents remain a major public health issue. Driver behaviour plays a critical role in accident occurrence at roundabouts. To enhance safety, this study examines the relationship between driver behaviours and roundabout traffic accidents. The study utilised Unmanned Aerial Vehicles (UAVs) to collect field data and process it using a computer vision tool. Driver behaviours were studied regarding average speed, pattern, gap acceptance, and idle time. The maximum speed exceeds 40 km/h, with drivers often travelling above safe limits. They also tend to accelerate inside the roundabout and decelerate at the entry and exit. Also, high circulating speeds reduce gap size, limiting roundabout entry and reducing capacity. Light vehicles have shorter idle times of less than six seconds, while heavy vehicles, due to size and inertia, require longer reaction times and larger gaps, leading to long idle times. The research contributes significantly to roundabout safety, efficiency and urban planning. It aids in designing future roundabouts suited for local traffic capacity and behaviour patterns while promoting safe and efficient intersections. This study also highlights the potential of using advanced technologies, like UAVs and computer vision, for further traffic analysis. Ultimately, understanding driver behaviour is key to improving traffic safety and efficiency. Roundabout Traffic Management Driver Behaviors Unmanned Aerial Vehicles (UAVs) Computer Vision Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Traffic accidents remain a significant public health concern, with an estimated 1.35 million fatalities annually, equivalent to 3,700 deaths per day, between 2010 and 2019, as reported by the Malaysia Road Fatalities Index [ 1 ]. In Selangor alone, 10,675 road accident fatalities were recorded during the same period [ 2 ]. While roundabouts have been widely recognised for their effectiveness in reducing accidents, occasional road incidences highlight the challenges of ensuring complete road safety [ 3 ]. Roundabouts enhance traffic flow by minimising crossing conflict points and providing a self-regulating traffic system that eliminates the need for traffic signals. Their continuous traffic movement reduces delays and congestion, improving urban mobility [ 4 ]. However, driver behaviour plays a crucial role in roundabout safety and efficiency. Driver decisions and actions, such as gap acceptance, speed regulation, and manoeuvring techniques, are key to determining traffic safety. Despite being an intangible cognitive process, driver intention can be inferred through observable behaviours, including pedal manoeuvres and mirror checks, which indicate how drivers interact with their surroundings [ 5 ]. Traffic accidents are predominantly attributed to human factors, with driver behaviour identified as a leading cause of crashes [ 6 , 7 ]. In Selangor, many drivers have admitted to engaging in risky driving behaviours, including tailgating, improper overtaking, and disregarding traffic signals [ 8 ]. Although various safety initiatives have been implemented, shaping positive driving attitudes and behaviours remains an overlooked aspect of road safety management. Driving behaviour is influenced by external traffic conditions and internal factors such as decision-making, mental health, and substance use, which affect driver reactions and risk perception [ 9 ]. There is a strong interrelationship between driver behaviour and traffic congestion, as aggressive driving tendencies increase crash risk and traffic bottlenecks [ 10 ]. Xing, et al. [ 5 ] found that positive drivers exhibit higher driving efficiency and fewer errors, whereas those in a negative mood are more prone to mistakes. Despite these insights, the relationship between driver behaviour and crash risks at roundabouts remains insufficiently explored. Given that roundabouts differ from conventional intersections regarding lane interactions, entry-exit manoeuvres, and merging conflicts, a more detailed examination of driver behaviour and its influence on roundabout safety is necessary. This study explores driver behaviours and their correlations with accident occurrences at roundabouts. This research builds on previous studies by explicitly examining driving patterns at roundabouts, which are often generalised across various intersection types. In this research, a methodology was developed to analyse the complex relationship between driver behaviours and roundabout safety outcomes. Ultimately, this study aims to improve roundabout traffic management by examining the link between driver behaviours, lane performance, and accident risks, offering insights that can guide more effective traffic planning and policy development. 2. Literature Review 2.1 Research Gap The bibliography includes 1,000 carefully selected references, grouped into four clusters that reflect the wide range of topics covered. To enhance the analysis, this study uses the Visualisation of Similarities (VOS) Viewer, a well-known tool for bibliometric analysis. Through this process, 6,526 relevant publications were identified, mainly focusing on traffic management, driver behaviour, safety, roundabouts, unmanned aerial vehicles (UAVs), and computer vision. Expanding the dataset this way helps capture broader research trends in transportation studies. The VOSviewer analysis revealed an essential gap in the literature. Limited research was done on how driver behaviour in Malaysia affects road safety at roundabouts. This finding highlights the need for further study to understand this issue better. Also, the analysis identifies key research clusters, particularly in UAVs and computer vision, which have a strong potential to improve traffic safety and control (Fig 1). While UAVs and computer vision benefit traffic analysis, research on their application remains limited, presenting an opportunity for further exploration in future studies. 2.2 Driver Behaviors Understanding and addressing driver behaviour is critical in managing road traffic injuries and crashes, which remain a leading global cause of mortality and morbidity. Atombo, et al. [11] showed that speeding was the most common violation in Ghana, followed by overtaking, which drivers engaged in when they believed would enhance their performance. Additionally, drivers with strong control beliefs were more likely to commit these violations, highlighting the need for targeted interventions to improve road safety. Also, in Iran, some mental disorders can increase the risk of road accidents [12]. A study in Malaysia has identified that factors such as driver attitudes and sociodemographic traits significantly influence highway speeding behaviours [13]. Shandhana Rashmi and Marisamynathan [14] summarised aberrant driving behaviours and assessed their impacts on crash risk and driving performance. Their findings show that approximately 90 per cent (%) of crash causes are attributed to road users, with drivers being the primary contributing factor. Adedeji, et al. [15] revealed that less-educated drivers are more likely to misinterpret communication cues, increasing the risk of accidents. Drivers in South Africa depend on formal and informal signals. Misinterpreting these signals can lead to moderate to high-risk accidents, emphasising the need for proper driver education on communication. The education level and gender of drivers also significantly impact their comprehension of communication, leading to traffic accidents. Therefore, these factors are significant to road safety. A similar finding was also obtained by Papantoniou, et al. [16] in Greece: (1) female, (2) older, (3) lower education levels, and (4) more driving experience drivers are more likely to commit driving errors. They also pointed out that drivers in rural areas are more prone to risky driving situations and errors. 2.3 Unmanned Aerial Vehicles (UAVs) Given that driver behaviours are crucial in traffic accidents, there is a growing need for innovative technologies to monitor and manage them to ensure road safety. One such technology is the UAV. UAVs offer a unique advantage in real-time traffic monitoring, enabling authorities to observe driver behaviour and flow from an aerial perspective [17]. UAVs can be classified based on their aerodynamic design, landing mechanism, and level of autonomy. Aerodynamically, they fall into four main categories: fixed-wing, flapping-wing, ducted-fan, and multi-rotor, which include tricopters, quadcopters, hexacopters, and octocopters [18, 19]. Their landing system categorises UAVs as vertical takeoff and landing (VTOL) or horizontal takeoff and landing (HTOL) [19]. Additionally, they can be grouped by autonomy level, ranging from manually piloted UAVs to fully autonomous systems capable of independent operation [20]. Table 1 outlines the advantages and disadvantages of different UAV types. Table 1 Advantages and disadvantages based on UAV type [21, 22] Types of UAVs Advantages Disadvantages Single Rotor VTOL and hover flight heavy payload long flight time hard control system high safety risk Multi Rotor VTOL and hover flight high manoeuvrability limited payload short flight time Fixed Wing long flight time large coverage fast flight speed requires space for launching or landing hard control system no VTOL and hover flight Hybrid VTOL flight long flight time undesirable for hovering or forward flight still in development The use of UAVs in Road Traffic Monitoring (RTM) has gained momentum in recent years, particularly in urban areas where traffic congestion and accidents are prevalent. Many researchers utilised UAV for RTM, such as the work done by Elloumi, et al. [23], Byun, et al. [24], Dronova, et al. [25], Gupta and Verma [26] and Liu and Bai [27] and. With their ability to capture high-resolution footage from various angles, UAVs can provide real-time insights into traffic patterns, driver behaviour, and hazardous road conditions. This data can be integrated with advanced computer vision software, such as GoodVision, to enhance decision-making in traffic management. Through such technologies, traffic authorities can better identify risky behaviours and intervene more effectively, potentially reducing the number of accidents caused by driver error. 2.4 Computer Vision Computer vision (CV) is a field of artificial intelligence (AI) that enables computers to interpret and analyse visual data, allowing them to extract meaningful information from images and videos. In road traffic monitoring, CV plays a crucial role in processing UAV footage to detect traffic flow patterns, identify risky behaviours, and assess road conditions. It can automate traffic analysis, reducing the need for manual data collection and improving the accuracy of traffic management strategies. This work selected GoodVision [28] to process drone recording. GoodVision is a cloud-based traffic analysis software using AI and CV technology to analyse traffic video footage. The GoodVision Insight application offers two customisable video processing types tailored to project requirements and camera types. Table 2 shows the video processing types and the required camera height. Table 2 The video processing types and their camera height for data processing [28]. Types Data Processing Camera Height (meters) Traffic Camera (fixed) 5 to 30 Drone Camera (hover) High Drone Low Drone 30 to 250 Up to 30 2.5 Traffic safety study from cognate disciplines Integrating UAV and CV tools exemplifies how technological advancements transform traffic safety research. By offering real-time traffic analysis, GoodVision [28] provides valuable data that contributes to understanding driver behaviours and traffic conditions. Adopting these tools in traffic safety research fosters interdisciplinary collaborations that combine psychology, AI, and computer vision, leading to more effective and informed traffic safety strategies at roundabouts and other critical areas. Traffic psychology and behaviour incorporate driver perception, cognition and decision-making, which are crucial to developing accident countermeasures [29]. Numerous research has been done on factors affecting roundabout safety, such as Sheykhfard, et al. [30], Distefano, et al. [31] and Distefano, et al. [32]. In general, driver behaviour plays a crucial role in the safety of roundabouts, affecting the traffic accident rate and regional driving experience. Advances in AI and computer vision capture and analyse real-time driver behaviours, allowing for precise assessment of traffic conflicts and safety at roundabouts. Many researchers utilised AI and computer vision in fieldwork to obtain safety data, such as Bhavsar, et al. [33], Scholl, et al. [34], St-Aubin, et al. [35], St-Aubin, et al. [36] and Zaki, et al. [37]. These works show that AI and CV are promising tools for obtaining traffic safety data at roundabouts, providing more actionable data for roundabout safety studies. However, despite these promising advancements, a significant research gap exists regarding the relationship between driver behaviours and road safety in Malaysian roundabouts. While research from various disciplines has extensively covered general traffic safety and AI applications, limited studies address the unique behaviours and contributing factors specific to Malaysian drivers at roundabouts. Furthermore, while CV tools like GoodVision are effective in capturing and analysing traffic data, their application to Malaysian traffic scenarios, particularly concerning driver behaviour at roundabouts, remains underexplored. This gap presents a promising opportunity for further research, especially in integrating UAVs, CV, and AI technologies to understand and improve traffic safety at roundabouts in Malaysia. Addressing this gap would significantly advance traffic safety strategies tailored to local contexts and help bridge the knowledge gap. Therefore, this research aims to investigate the impacts of driver behaviour on traffic safety, efficiency, and overall road management. The research outcomes will contribute valuable data to enhance urban road safety and efficiency. While this study does not incorporate advanced modelling techniques such as agent-based simulation, structural equation modelling, or machine learning, its methodology remains valid. It is consistent with several peer-reviewed works that have relied on observational and descriptive approaches to traffic behaviour analysis. For example, St-Aubin et al. (2013) used computer vision techniques to analyse trajectory and behaviour patterns at Canadian roundabouts without employing complex statistical models. Though exploratory, their research is recognised for its methodological rigour and contribution to video-based safety diagnostics. Similarly, Bhavsar et al. (2023) employed UAVs to observe traffic violations at Indian roundabouts and relied primarily on frequency analysis and speed metrics to draw meaningful conclusions for urban traffic management. Zaki, Sayed, and Cheung (2013) also demonstrated that valuable behavioural insights can be extracted from computer vision tools without applying inferential statistics, particularly in cyclist trajectory analysis. These precedents collectively validate the methodological stance of the present study, which emphasises the importance of empirical data extraction, pattern recognition, and visual analytics in establishing a foundational understanding of driver behaviour in real-world environments. Beyond methodological alignment, this study advances the literature in several important ways. First, it offers a novel geographic contribution by focusing on a Malaysian multiple-lane, flyover-covered urban roundabout, an infrastructure context that is underrepresented in global research. While prior studies concentrated on North American or Indian intersections, the current research responds to the urgent need for localised data in Southeast Asia, where cultural driving norms, enforcement rigour, and infrastructure design differ. Second, this study integrates multiple behavioural dimensions into a unified analysis framework, such as average speed, gap acceptance, and idle time. Unlike earlier works that isolated specific behaviours, the present research triangulates these metrics to form a more holistic picture of driver dynamics. Third, GoodVision's acceleration heatmap represents a visual innovation by identifying high-risk zones for abrupt acceleration and deceleration, facilitating spatial diagnostics for planners and engineers. This visual layer supplements traditional numeric data with intuitive cues for traffic intervention. Furthermore, while earlier studies primarily described behavioural trends, this study links those trends to concrete policy and planning recommendations. These include suggestions for roundabout geometric redesign, improved speed management, targeted law enforcement, and strategies to reduce vehicular idle time, which directly affect road safety, environmental impact, and traffic efficiency. Lastly, the study demonstrates field-level innovation by adapting commercially available UAV technology (Da-Jiang Innovations (DJI) Air 2S) to Malaysian urban traffic contexts, ensuring data quality while maintaining operational feasibility in resource-constrained settings. Taken together, these contributions reinforce both the validity and the applied relevance of this study. Although it does not employ advanced modelling, its empirical depth, contextual specificity, and policy orientation make it a substantive advancement in the growing UAV- and computer vision-based traffic behaviour analysis fields. 3. Methodology This research proposed a methodology framework to collect higher-quality data for a more in-depth analysis (Fig 2). A drone was used to collect traffic flow data in a roundabout. The specifications of the drone used can be found in Appendix A. The footage will be uploaded to GoodVision [28] for processing before being turned into driver behaviour parameters for analysis. Initially, insufficient data was collected for GoodVision [28] to process. Therefore, the data collection process was repeated, as shown by the dotted line. In the second collection, the video was recorded from an alternative viewpoint at 60 meters (m), compared to the previous 28 m, to enhance coverage and accuracy. Fieldwork was conducted at the New Pantai Expressway Roundabout Kewajipan (Lebuhraya New Pantai Bulatan Kewajipan), with a coordinate of 3.0732°N, 101.5930°E. This roundabout has an inscribed circle diameter (ICD) of 90 m, measured from the outer edge of the kerb surrounding the circulating lanes (Fig 3). It has two circulating lanes surrounding the central island, allowing vehicles to circulate clockwise in the roundabout. This roundabout has four legs, with each leg having two-lane entries. This research introduces leg numbering to ease the description. Additionally, the roundabout is integrated with three layers of flyovers, one linked to the Kelana-Subang Link and two towards Sunway City. This roundabout was selected as it carries about 3,000 vehicles per hour daily, providing sufficient data samples for analysis. The fieldwork was conducted on 22 May 2024, a sunny Friday, to ensure traffic data normality. During fieldwork, a drone was deployed from 1400 to 1600 to capture traffic flows and patterns. Each battery can last for 20 minutes. Hence, six batteries were used to cover a two-hour collection. The drone was flown 60 meters from the roundabout to capture roundabout traffic flow. Vehicles on the three layers of flyovers were excluded as this study focuses only on driver behaviour in the roundabout. Observations were conducted covertly. Since the flight height exceeded the line of sight of the drivers, the natural driver behaviours (without drone impact) were recorded. A drone recording was uploaded to GoodVision [28] to extract traffic flow data. Previous work, such as Humoody and Younis [38] and Bong, et al. [17], used GoodVision [28] in analysing their roundabouts data. The first group of researchers validated data using simulation software, while the second validated by comparing GoodVision [28] data with local field data. Their findings show that data extracted from GoodVision [28] matches their secondary sources, indicating that GoodVision [28] provides reliable data sources for traffic analysis. The GoodVision Insight application lets users draw virtual lines to exact vehicle time gaps. Figure 4 shows an example of extracting time gap at Leg 1 entry using three virtual lines. The time gaps are crucial to study driver behaviours in average vehicle speed, gap acceptance and idle time, which will be discussed in the following. This research investigates three forms of driver behaviour: average speed and its pattern, gap acceptance, and idle time. Average speed, in kilometres per hour (km/h), is the speed vehicles travel through the roundabout. It differs with the location in a roundabout, such as high-speed (70 to 80 km/h) approaching the roundabout, low speed (10 to 20 km/h) at the yield line and back to high speed when leaving the roundabout [39]. Considering the speed values change with locations, the average speed and its patterns at various locations in roundabouts were included in the studies. Gap acceptance usually happens in unsignalised intersections controlled by priority, such as roundabouts [40, 41]. As vehicles approach a roundabout entry, they will decelerate to seek a gap from the circulating flow, enter the intersection if the gap is larger than or equal to the critical gap, or stop waiting for the next one. Gap acceptances are often measured in headway distribution of circulating vehicles, critical gap, and following gap [42]. Idle time refers to how long vehicles stop or move at very low speeds, often due to yielding or congestion. 4. Results and Discussion 4.1 Average Speed Analysis In this work, average speed was analysed based on turning movement. For example, for Leg 1, the average speed of the left turn (to Leg 2), through (to Leg 3), right turn (to Leg 4) and U-turn (to Leg 1) were recorded. Out of 16 movements, twelve could not be captured due to the blockage of three layers of the flyover. For instance, the left turn from Leg 3 was blocked by the flyover of Exit to Lebuhraya Pantai Baru, as illustrated in Fig 5. The Subang-Kelana Jaya Link blocked the entry and exit from Leg 2. Therefore, only four movements could be captured. The maximum and average speeds were recorded and compared (Table 3). Table 3 Average and maximum speeds when vehicles perform turning movements at Roundabout Kewajipan. Turning movements Average Speed (km/h) Maximum Speed (km/h) Leg 1 to Leg 2 Left-turn 39 47 Leg 1 to Leg 3 Through - - Leg 1 to Leg 4 Right-turn - - Leg 1 to Leg 1 U-turn - - Leg 2 to Leg 1 Left-turn - - Leg 2 to Leg 3 Through 32 44 Leg 2 to Leg 4 Right-turn - - Leg 2 to Leg 2 U-turn - - Leg 3 to Leg 1 Left-turn - - Leg 3 to Leg 2 Through - - Leg 3 to Leg 4 Right-turn - - Leg 3 to Leg 3 U-turn - - Leg 4 to Leg 1 Left-turn 26 37 Leg 4 to Leg 2 Through 35 44 Leg 4 to Leg 3 Right-turn - - Leg 4 to Leg 4 U-turn - - Remarks - indicates data is unavailable due to obstruction by three layers of flyovers. From the table, the highest average speed was detected at Leg 1, followed by Legs 4, 2 and 3. Three legs hit more than 40 km/h for maximum speed. According to Ahmad and Rastogi [43], the roundabout is constructed with a 30 km/h speed restriction. If any vehicle exceeds or goes below this limit, it might be applied as a safety indicator. An average maximum speed of 43 km/h is recorded from the field, indicating that the drivers at this roundabout may tend to travel at unsafe speeds, which could compromise overall traffic safety. The risks associated with high vehicle speeds are well-documented in road safety literature. A study by Doecke, et al. [44] discussed the impact of high speed correlated with the risk of serious injury. The likelihood of severe harm sharply increases as speed rises. For instance, a head-on collision at 28 km/h carries only a 1% risk of serious injury. However, this risk escalates to approximately 50% at speeds exceeding 76 km/h. Other collisions also become significantly more dangerous at higher speeds: side impacts at 51 km/h, frontal impacts at 64 km/h, and rear-end crashes at 67 km/h are all associated with a substantially increased risk of serious injury. Average speed data are crucial for assessing the efficiency and safety of a roundabout. High average speed improves efficient traffic flow but might raise safety concerns. High average speed indicates smooth traffic flow, and vehicles experience minimal delay. However, vehicles entering the roundabout at high speed are unlikely to slow down and yield circulating vehicles, promoting the clashing between entry and circulating vehicles (which indicates less safe driving conditions) [45-47]. On the contrary, low average speed suggests vehicles moving slowly due to congestion, indicating safer driving conditions. The trade-off between efficiency and safety mostly depends on the roundabout geometry that manages vehicle speeds (safety) and accommodates traffic volume (efficiency). In this study, the trade-off of a single-lane roundabout with high traffic volume includes reducing the lane deflection or narrowing lane width to slow down the vehicle speed while maintaining the roundabout capacity. The acceleration heatmap of GoodVision (2024) was used to display the average speed pattern in different locations in the roundabout. GoodVision (2024) shows areas with decelerating traffic as hot (red) while areas with accelerating traffic are cold (blue) instead of specifying speed metric in defining the areas (to what extent acceleration is indicated as cold or deceleration as hot). Hence, the results visually portray acceleration areas without a specific acceleration range. Figure 5 identifies the acceleration and deceleration in the studied roundabout. The slow-speed area appeared at all leg entries, represented by hot red. At the entry, drivers tend to slow down to seek a gap from the circulating flow to enter the roundabout. Thus, a deceleration was detected at this zone. Conversely, the circulating lane was prioritised based on the give-way rule [48, 49]. They proceed with high speed as they do not need to yield to entry flow. Hence, this area is cold blue, showing acceleration in the roundabout. The circulating lanes in front of Legs 2, 3 and 4 tend to have lower acceleration (yellowish) than Leg 1. This observation, again, can be linked back to the priority rule. AlKheder, et al. [50] mentioned that almost 80% of their sample always followed the priority rule when utilising a roundabout, with only 20% refusing to prioritise circulating vehicles. A similar condition could happen in this roundabout. Some entry vehicles adopted reverse priority by forcing themselves to enter the roundabout. In this condition, circulating vehicles decelerate to yield to the entry to avoid a collision. Hence, deceleration was observed in these areas. In short, the heatmap identifies the hot zone for accelerating within the roundabout and decelerating at entry and exit. The high-speed circulating area is often at high risk for collisions or near-misses when entry drivers seek gaps to enter the roundabout. Safe precautions should be taken in this area, including implementing police intervention to regulate traffic flows on each leg, installing a traffic signal to reduce the speed in circulating lanes or speed management by introducing speed limits within the roundabout. The practitioners could update current guidelines or standards limiting the maximum speed within roundabouts or encourage single-lane designs in areas with large vehicle traffic, reducing the need for sudden braking and lane changes. Policymakers could also enforce stricter yielding laws at roundabout entries to help maintain safer interactions between entry and circulating vehicles. 4.2 Gap Acceptance Analysis In this work, the gap acceptance by the entry vehicles was studied using circulating speed inside the roundabout. This roundabout operates under the give-way rule, meaning that vehicles seek a gap from the circulating flow to enter the roundabout. Hence, parameters that could represent entry and circulating flows were investigated. GoodVision provides circulating speed and time gaps between entry vehicles, which best present entry and circulation in this condition. The circulating speed reflects the gap availability in the roundabout. A slow circulating speed creates smaller gaps than fast ones, making more gaps available and increasing the traffic volume that can enter the roundabout [51, 52]. The time gaps were collected when two entry vehicles passed through the same virtual line. It reflects the drivers' ability to accept the gap to enter the roundabout. A short time gap illustrates a quick flow into the roundabout, representing a high gap acceptance behaviour. Therefore, the time gaps between entry vehicles were plotted against the average speed of the circulating flow (Fig 7). The relationship should convey how the speed of the circulating flow influences the time gaps between entry vehicles, thereby reflecting driver gap acceptance behaviour. Across all legs, a positive correlation is observed between the time gap of entry vehicles and the speed of vehicles on the circulating lane. The time gap between entry vehicles increased as the circulating speed increased. As the circulating speed increases, the gap size and availability in the stream shrink [51, 52]. With fewer chances (gaps) in the circulating flow, drivers must seek acceptable gaps more [53, 54]. This scenario leads to fewer vehicles entering the roundabout, which is reflected by the extended time gap between the two entry vehicles. A long time gap shows an entry vehicle stuck at the yield line for too long to seek an acceptable gap (rejecting more gaps in the circulating flow). This slow gap acceptance behaviour caused by small or limited gaps in the circulating flow matches the findings earlier. The outliers indicate that drivers exhibit slow gap acceptance behaviours at low circulating speeds. For example, Leg 4 has about nine vehicles that take a long time (time gaps of more than 25 seconds (s)) to enter the roundabout when the circulating speed is less than 40 km/h. This scenario can relate to the driver's decision to merge into the roundabout. Matured drivers tend to have better awareness of the rules of the road, more experience in varied driving conditions and scenarios, a stronger ability to react calmly under stress, and a larger sense of responsibility to other road users. Observation shows that during low traffic flows, the likelihood of accidents significantly increases due to overspeeding. This scenario occurs as the roundabout geometry discourages high-speed entry [43]. Despite outliers, the time gap between entry vehicles and the circulating speed shows strong positive correlations ranging from 0.74 to 0.83, implying a reliable and predictable relationship between the two variables. Besides the coefficient of determination, this work also conducted the Pearson correlation test to validate the relationship between the time gap of entry vehicles and the speed of vehicles on the circulating lane. The null hypothesis is "there is no relationship between the two variables", and the alternative hypothesis is "there is a relationship between the two variables". Table 4 shows the correlation results for all legs. The test also shows a strong positive relationship between the two variables, with a positive coefficient of more than 0.8. The p-value also validates the relationship. The small p-values from all legs reject the null hypothesis, showing a statistically significant positive relationship b between the time gap of entry vehicles and the speed of vehicles on the circulating lane. Table 4 Correlation between the time gap of entry vehicles and the speed of vehicles on the circulating lane using the coefficient of determination and Pearson coefficient. Sample size Coefficient of determination, r-squared Pearsons coefficient, r (p-value) Leg 1 412 0.778 0.857 (6.599E-120) Leg 2 132 0.742 0.829 (1.5118E-34) Leg 3 406 0.826 0.855 (5.3019E-117) Leg 4 264 0.743 0.835 (4.7222E-70) The findings indicate that higher circulating speeds reduce the time gaps between vehicles in the stream, decreasing overall roundabout capacity. The reduced gap could also cause safety issues. A short time gap suggests a higher likelihood of collisions or near-misses at the roundabout entry. As drivers accept the reduced gap, they are more likely to quickly merge into the circulating flow, potentially unaware of the upcoming traffic. [55]. Hence, the collision rate between entry and circulating vehicles increases and the safety drops. Several improvements can be made through roundabout geometry adjustment. The number of circulating lanes can be increased to distribute the traffic flow equally, creating more gaps between two successive circulating vehicles. Also, the circulating path curvature could be adjusted to create a circulating lane with a higher turning degree. This design requires additional steering manoeuvres, reducing the speed and improving the safety. This modification can be applied to other roundabouts, given that there is allowance in the central island or areas close to circulating lanes. Improvements can also be made through traffic flow management, such as implementing traffic signals to slow down the circulating speed or implementing police intervention to regulate traffic flow, ensuring vehicles on the saturated leg have more chance to enter the roundabout and clear the congestion. 4.3 Idle Time Analysis Idle time was determined at entry and exit for each leg based on five vehicle classifications (car, bus, van, truck and heavy truck). This research did not define time range, i.e., how many seconds is considered long or short, due to limited works available to benchmark the defined time range. Therefore, the idle time between legs and vehicle classifications was compared. Table 5 shows the average idle time at entries and exits of all four legs. Table 5 Average idle time of Minor Stream Before Entering/Exiting Major Stream Average Idle Time (seconds) Car Bus Van Truck Heavy Truck Total idle time by leg Leg 1 (Entry) 4.00 - 4.53 7.17 10.59 26.29 Leg 1 (Exit) 2.34 - 1.15 3.19 2.55 9.23 Leg 2 (Entry) 4.47 - 4.28 4.05 4.78 17.58 Leg 2 (Exit) 8.75 - 7.57 - - 16.32 Leg 3 (Entry) 3.74 3.13 1.90 2.99 - 12.06 Leg 3 (Exit) 2.50 6.99 3.56 3.14 - 16.19 Leg 4 (Entry) 6.27 4.44 7.82 3.45 - 21.98 Leg 4 (Exit) 11.06 23.91 - 6.18 - 41.15 Average. idle time based on vehicle type 5.39 9.62 4.40 4.31 5.97 Remarks - indicates no observation at this leg. For exit, Leg 4 has the highest average idle time (41.15 s) among the legs. The result was caused by the large idle time from the bus (23.91 s) and some from the car (11.06 s). The long idle time is often caused by driver hesitation due to a lack of familiarity with roundabouts or pedestrian volume increment [56, 57]. Conversely, Leg 1 has the lowest idle time of 9.23 s among the legs. All vehicles exit the roundabout smoothly, with the highest idle time being only 3.19 s (by truck). Regarding entry, Leg 1 has the longest time compared to the other. This leg has high inflow demand (668 pcu/h), likely to form queues and congestion. Many drivers attempt to enter the roundabout but must wait for sufficient gaps in the circulating flow [58]. They must wait for traffic ahead to clear, leading to a long idle time. Another possible reason could be that drivers are less aggressive in accepting the gap to enter the roundabout. Hence, they spend more time seeking new gaps, leading to long idle time. Leg 3 is the shortest, indicating that drivers can quickly accept the gap in circulating flow and enter the roundabout. A short average gap allows more vehicles to enter but indicates higher collision risks. Entry vehicles might have insufficient time to merge into the circulating flow safely as circulating vehicles are close (as represented by a small gap). This statement is supported by Singh, et al. [59], highlighting that the relatively short time between the end of a right-turning vehicle and the arrival of a moving vehicle at the conflict zone increases the likelihood of critical crossing conflicts. For vehicle classification, cars, vans and trucks have an average idle time of about four to five seconds. They are smaller in size compared to buses and heavy trucks. The size advantage allows them to move actively and fill in the gap between the large ones instead of staying static, explaining their short idle time. Yang, et al. [60] also stated that light vehicles tend to overtake heavy vehicles to fill the spaces between them, further supporting this behaviour. The bus and heavy truck have a high idle time of 9.62 s and 5.97 s, respectively. Both vehicles are huge and have large inertia, leading to a longer reaction time to move than light vehicles [61, 62]. Heavy vehicles must seek a large enough gap for the whole vehicle to merge into the circulating flow. The longer the time required to accept a gap, the longer heavy vehicles stay static. Therefore, long idle time was observed. High idle time impacts congestion, as vehicles are stuck in traffic for long periods. This scenario later leads to environmental effects such as air pollution due to increments in emissions and fuel consumption [63, 64]. High idle time can often be solved by promoting carpools in busy traffic areas and public transport. Intelligence transportation systems that provide real-time traffic data could help to solve this. They can help drivers to avoid areas with high idle time by providing alternative routes. 4.4 Practical and Policy Implications The findings of this study carry significant practical and policy implications, particularly in the context of urban traffic management and road safety in Malaysia. Identifying behavioural patterns such as excessive speeds within roundabouts, aggressive merging by heavy vehicles, and extended idle times among smaller vehicles provides data-driven evidence to inform infrastructure design, enforcement strategies, and driver education programs. For example, acceleration heatmaps revealed zones of abrupt speed changes, suggesting the need for geometric redesigns at entry and exit points to encourage smoother transitions and reduce conflict risks. From a policy standpoint, these insights can support revisions to Malaysia's existing Public Work Department (PWD) road design standards, especially Arahan Teknik (Jalan) 11/87, which governs roundabout design. Currently, the guidelines emphasise geometric dimensions and signage but lack behavioural calibration based on empirical UAV and computer vision data. This research addresses that gap by offering actionable recommendations grounded in local driver behaviour, including the introduction of lane-specific speed regulations, stricter yield enforcement at entries, and context-sensitive adjustments to lane widths and curvature. Moreover, the observed disparity in behaviour between vehicle classes underscores the potential benefit of adaptive traffic control systems and real-time monitoring solutions, such as integrating AI-enabled UAV surveillance for traffic violation detection and congestion prediction. Beyond infrastructure and enforcement, this study also offers insights that align with national initiatives under Malaysia's Road Safety Plan 2022–2030, which aims to reduce road traffic fatalities by 50%. The behavioural findings, particularly regarding driver aggressiveness and decision-making under congestion, support the plan's strategic pillars on safe road users and speed. Public awareness campaigns and driver education modules can be tailored to address the specific behavioural deficiencies identified, such as the tendency to disregard yielding rules or the hesitation of smaller vehicles to merge, which collectively impact roundabout performance and safety. Furthermore, the environmental implications of high idle times, such as increased emissions and fuel consumption, are significant, particularly in Malaysia's broader sustainability goals outlined in the Low Carbon Cities Framework (LCCF) and Green Technology Master Plan. Reducing congestion through behaviour-informed design and policy can significantly lower urban transportation emissions. Overall, this study bridges a crucial gap between observed traffic behaviour and actionable policy response, enabling Malaysian authorities to move beyond traditional, geometry-centric approaches and adopt a more behaviour-sensitive, technology-integrated framework for roundabout safety and efficiency. 5. Conclusions and recommendations Driver behaviour is crucial for traffic safety and efficiency at roundabouts. UAVs and CV technologies offer valuable tools for monitoring and analysing traffic patterns. The data collected from these technologies can provide insights for designing safer roundabouts and developing targeted traffic management strategies. By using these advanced tools, roundabout safety and overall traffic management could be improved significantly. This research utilised these technologies to investigate three driver behaviours, including average speed, gap acceptance, and idle time, to understand their impact on roundabout performance and safety. The key findings include: The maximum speed hit more than 40 km/h, showing that drivers tend to travel at unsafe speeds. Acceleration often happens inside the roundabout, and deceleration occurs at the entry and exit of the roundabout. A high circulating speed decreases the gap size and availability in the flow. Drivers tend to reject a small gap and seek a large one, leading to fewer vehicles entering the roundabouts and reducing capacity. Light vehicles (cars, vans, and trucks) have short idle times due to flexibility in occupying space in the traffic flow. Heavy vehicles have a longer idle time due to large size and inertia, requiring longer reaction time to move and a large gap to enter the roundabout. This study enhances roundabout traffic management by addressing gaps in theoretical models, which often generalise driver behaviour rather than capturing real-world complexities. While roundabouts are theoretically designed to improve safety and traffic flow, actual driver behaviours, such as average speed, gap acceptance and idle time, are not always fully considered. By focusing on the New Pantai Expressway Roundabout Kewajipan and using empirical data, this research provides insights on practical and policy standspoint tailored to Malaysian urban roundabout conditions. These findings contribute to more effective roundabout design, management, and assessment, offering valuable guidance for policymakers, traffic planners, and organisations like the Malaysian Public Works Department to enhance traffic efficiency and safety. This study was conducted on a two-lane roundabout, with traffic flow recorded over two hours on a sunny working day when high traffic volume was expected. The findings primarily reflect driver behaviour at this specific location and timeframe. Future research can expand by analysing data from multi-lane roundabouts in rural areas, exploring different intersection types, extending data collection to peak hours, or conducting fieldwork under varied weather and lighting conditions. Considering various factors could expand the database, providing a more comprehensive understanding of driver behaviour and improving the relevance and reliability of this study. Additionally, case studies on other multi-lane roundabouts in Malaysia or comparisons with international roundabouts could further validate and generalise the findings. *Disclaimer: This article uses a language model for language editing purposes. Declarations *Disclaimer: This article uses a language model for language editing purposes. Ethical approval Taylor's University Human Ethics has confirmed that no ethical approval is required. Informed Consent Informed consent was obtained from all individual participants included in the study. Data availability The datasets generated during and/or analysed during the current study are available from the corresponding author upon reasonable request. Funding Declaration The authors declare that this study was conducted without any external funding. Clinical trial number A clinical trial number is not applicable. Competing Interest The authors declare no competing interests. Author Contribution Jit Boon Bong and ChangSaar Chai reviewed and revised the manuscript. Kennedy Kwong Shin Tiong collected data and prepared the first draft. Lam Tatt Soon collected data and screened the preliminary data. References Ministry of Transport Malaysia. "Malaysia Road Fatalities Index." https://www.mot.gov.my/en/land/safety/malaysia-road-fatalities-index. (accessed 14 November 2023. Royal Malaysia Police. "Jumlah kematian disebabkan kemalangan jalan raya mengikut tahun dan negeri." https://archive.data.gov.my/data/ms_MY/dataset/jumlah-kematian-disebabkan-kemalangan-jalan-raya-mengikut-tahun-dan-negeri/resource/68eaabde-60d1-40a2-bd2c-a43be44de21c?view_id=ea241f92-624a-4013-8ddc-58a6d0aff7a8 (accessed 14 November 2023. H. A. Zubaidi, J. C. Anderson, and S. Hernandez, "Understanding roundabout safety through the application of advanced econometric techniques," International Journal of Transportation Science and Technology, vol. 9, no. 4, pp. 309-321, 2020, doi: 10.1016/j.ijtst.2020.03.001. M. Singh and R. Kaur, "Review study on roundabout," International Journal of Research Publication and Reviews, vol. 1, no. 6, pp. 64-69, 2020. Y. Xing, C. Lv, and D. Cao, "Chapter 5 - Driver behavior recognition in driver intention inference systems," in Advanced Driver Intention Inference : Elsevier, 2020, pp. 99-134. D. McCarty and H. W. Kim, "Risky behaviors and road safety: An exploration of age and gender influences on road accident rates," PLoS One, vol. 19, no. 1, 2024, doi: 10.1371/journal.pone.0296663. L. Hu, X. Bao, H. Wu, and W. Wu, "A study on correlation of traffic accident tendency with driver characters using in-depth traffic accident data," Journal of Advanced Transportation, vol. 2020, no. 1, 2020, doi: 10.1155/2020/9084245. H. W. You, A. Abdul Rahman, and L. H. Hendri Dwisatrya, "Dataset of driving behaviours in Selangor, Malaysia," Data in Brief, vol. 31, p. 105783, 2020, doi: 10.1016/j.dib.2020.105783. N. S. Azman, R. Ilyas, N. Isah, and M. K. A. Ibrahim, "Riding behaviour of motorcyclists in the Klang Valley ", 2022. G. Li et al. , "Influence of traffic congestion on driver behavior in post-congestion driving," Accident Analysis & Prevention, vol. 141, p. 105508, 2020, doi: 10.1016/j.aap.2020.105508. C. Atombo, C. Wu, M. Zhong, and H. Zhang, "Investigating the motivational factors influencing drivers intentions to unsafe driving behaviours: Speeding and overtaking violations," Transportation Research Part F: Traffic Psychology and Behaviour, vol. 43, pp. 104-121, 2016, doi: 10.1016/j.trf.2016.09.029. S. S. Alavi, M. R. Mohammadi, H. Souri, S. Mohammadi Kalhori, F. Jannatifard, and G. Sepahbodi, "Personality, driving behavior and mental disorders factors as predictors of road traffic accidents based on logistic regression," (in eng), Iranian Journal of Medical Sciences, vol. 42, no. 1, pp. 24-31, 2017. F. F. Mohamad, A. S. Saifizul, and J. Mohamad, "Are sociodemographic characteristics and attitude good predictors of speeding behavior among drivers on Malaysia federal roads?," Traffic Injury Prevention, vol. 20, no. 5, pp. 478-483, 2019, doi: 10.1080/15389588.2019.1612057. B. Shandhana Rashmi and S. Marisamynathan, "Factors affecting truck driver behavior on a road safety context: A critical systematic review of the evidence," Journal of Traffic and Transportation Engineering (English Edition), vol. 10, no. 5, pp. 835-865, 2023, doi: 10.1016/j.jtte.2023.04.006. J. A. Adedeji, X. E. Feikie, and D. Dijanic, "Driver-to-driver communication skills: Contributing factors to traffic accidents on South African roads?," Sustainability, vol. 16, no. 11, 2024, doi: 10.3390/su16114833. P. Papantoniou, G. Yannis, and E. Christofa, "Which factors lead to driving errors? A structural equation model analysis through a driving simulator experiment," IATSS Research, vol. 43, no. 1, pp. 44-50, 2019, doi: 10.1016/j.iatssr.2018.09.003. J. B. Bong, E. E. M. Chong, and C. S. Choo, "UAV-based lane-changing reduction using traffic video analyser," in Australian Engineering Congress (AAEC2022) , Kuching, Sarawak, Malaysia, 13-15 July 2022 2023, pp. 285-297. M. Farajijalal, H. Eslamiat, V. Avineni, E. Hettel, and C. Lindsay, "Safety systems for emergency landing of civilian Unmanned Aerial Vehicles (UAVs)—A comprehensive review," Drones, vol. 9, no. 2, p. 141, 2025, doi: 10.3390/drones9020141. M. El Adawy et al. , "Design and fabrication of a fixed-wing Unmanned Aerial Vehicle (UAV)," Ain Shams Engineering Journal, vol. 14, no. 9, 2023, doi: 10.1016/j.asej.2022.102094. S. Bemposta Rosende, S. Ghisler, J. Fernández-Andrés, and J. Sánchez-Soriano, "Implementation of an edge-computing vision system on reduced-board computers embedded in UAVs for intelligent traffic management," Drones, vol. 7, no. 11, 2023, doi: doi:10.3390/drones7110682. A. Erceg, B. Činčurak Erceg, and A. Vasilj, "Unmanned aircraft systems in logistics – legal regulation and worldwide examples toward use in Croatia," in 17th International Scientific Conference Business Logistics in Modern Management (2017) , Osijek, Croatia, 12-13 October 2017 2017. W. Greenwood William, P. Lynch Jerome, and D. Zekkos, "Applications of UAVs in civil infrastructure," Journal of Infrastructure Systems, vol. 25, no. 2, 2019, doi: 10.1061/(ASCE)IS.1943-555X.0000464. M. Elloumi, R. Dhaou, B. Escrig, H. Idoudi, and L. A. Saidane, "Monitoring road traffic with a UAV-based system," in 2018 IEEE Wireless Communications and Networking Conference (WCNC) , Barcelona, Spain, 15-18 April 2018 2018, pp. 1-6, doi: 10.1109/WCNC.2018.8377077. S. Byun, I.-K. Shin, J. Moon, J. Kang, and S.-I. Choi, "Road traffic monitoring from UAV images using deep learning networks," Remote Sensing, vol. 13, no. 20, 2021, doi: doi:10.3390/rs13204027. O. Dronova et al. , "Unmanned aerial vehicles as element of road traffic safety monitoring," Transportation Research Procedia, vol. 63, pp. 2308-2314, 2022, doi: 10.1016/j.trpro.2022.06.263. H. Gupta and O. P. Verma, "Monitoring and surveillance of urban road traffic using low altitude drone images: A deep learning approach," Multimedia Tools and Applications, vol. 81, no. 14, pp. 19683–19703, 2022, doi: 10.1007/s11042-021-11146-x. S. Liu and Y. Bai, "Multiple UAVs collaborative traffic monitoring with intention-based communication," Computer Communications, vol. 210, pp. 116-129, 2023, doi: 10.1016/j.comcom.2023.08.005. GoodVision . (2025). J. A. Groeger and J. A. Rothengatter, "Traffic psychology and behaviour," Transportation Research Part F: Traffic Psychology and Behaviour, vol. 1, no. 1, pp. 1-9, 1998, doi: 10.1016/S1369-8478(98)00007-2. A. Sheykhfard, F. Haghighi, S. Saeidi, M. SafariTaherkhani, and S. Das, "Understanding the influence of environmental factors on driver speed: A structural equation modeling analysis," IATSS Research, vol. 48, no. 3, pp. 427-439, 2024, doi: 10.1016/j.iatssr.2024.08.001. N. Distefano et al. , "Physiological and driving behaviour changes associated to different road intersections," European Transport, vol. 2020, no. 77, 2020, doi: 10.48295/ET.2020.77.4. N. Distefano, S. Leonardi, and G. Pulvirenti, "Factors with the greatest influence on drivers' judgment of roundabouts safety. An analysis based on web survey in Italy," IATSS Research, vol. 42, no. 4, pp. 265-273, 2018, doi: 10.1016/j.iatssr.2018.04.002. Y. M. Bhavsar, M. S. Zaveri, M. S. Raval, and S. B. Zaveri, "Vision-based investigation of road traffic and violations at urban roundabout in India using UAV video: A case study," Transportation Engineering, vol. 14, 2023, doi: 10.1016/j.treng.2023.100207. L. Scholl, M. Elagaty, B. Ledezma-Navarro, E. Zamora, and L. Miranda-Moreno, "A surrogate video-based safety methodology for diagnosis and evaluation of low-cost pedestrian-safety countermeasures: The case of Cochabamba, Bolivia," Sustainability, vol. 11, no. 17, 2019, doi: doi:10.3390/su11174737. P. St-Aubin, N. Saunier, and L. Miranda-Moreno, "Large-scale automated proactive road safety analysis using video data," Transportation Research Part C: Emerging Technologies, vol. 58, pp. 363-379, 2015, doi: 10.1016/j.trc.2015.04.007. P. St-Aubin, N. Saunier, L. F. Miranda-Moreno, and K. Ismail, "Use of computer vision data for detailed driver behavior analysis and trajectory interpretation at roundabouts," Transportation Research Record, vol. 2389, no. 1, pp. 65-77, 2013, doi: 10.3141/2389-07. M. H. Zaki, T. Sayed, and A. Cheung, "Computer vision techniques for the automated collection of cyclist data," Transportation Research Record, vol. 2387, no. 1, pp. 10-19, 2013, doi: 10.3141/2387-02. M. A. Humoody and N. A. Younis, "Congestion identification for multilane highway by vehicle lane change as driver behavior," Tikrit Journal of Engineering Sciences, vol. 29, no. 1, pp. 46-59, 2022, doi: 10.25130/tjes.29.1.5. B. W. Robinson et al. , "Roundabouts: An informational guide," United States Federal Highway Administration, 2000. R. Guo, L. Liu, and W. Wang, "Review of roundabout capacity based on gap acceptance," Journal of Advanced Transportation, vol. 2019, no. 1, 2019, doi: 10.1155/2019/4971479. K. Shaaban and H. Hamad, "Group gap acceptance: A new method to analyze driver behavior and estimate the critical gap at multilane roundabouts," Journal of Advanced Transportation, vol. 2018, no. 1, 2018, doi: 10.1155/2018/1350679. J. Dahl and C. Lee, "Empirical estimation of capacity for roundabouts using adjusted gap-acceptance parameters for trucks," Transportation Research Record, vol. 2312, no. 1, pp. 34-45, 2012, doi: 10.3141/2312-04. A. Ahmad and R. Rastogi, "Influence of roundabout size and circulating traffic flow on critical gaps, follow-up times and speeds," Indian Highways, vol. 45, pp. 37-48, 2017. S. D. Doecke, M. R. J. Baldock, C. N. Kloeden, and J. K. Dutschke, "Impact speed and the risk of serious injury in vehicle crashes," Accident Analysis & Prevention, vol. 144, 2020, doi: 10.1016/j.aap.2020.105629. CD116 Geometric design of roundabouts , Highways Agency, London, United Kingdom, 2020. Ishanya, S. Marathe, and Y. Suresh, "A study on rotary intersection at Mangaluru – A case study of Nanthur Junction," International Journal of Current Engineering and Scientific Research (IJCESR), vol. 4, no. 5, 2017. I. Mahdalova, V. Krivda, and V. Skvain, "Influence of roundabout inscribed circle diameter to the traffic safety," Applied Mechanics and Materials, vol. 409-410, pp. 1122-1125, 2013, doi: 10.4028/www.scientific.net/AMM.409-410.1122. X. Chen and M. S. Lee, "A case study on multi-lane roundabouts under congestion: Comparing software capacity and delay estimates with field data," Journal of Traffic and Transportation Engineering (English Edition), vol. 3, no. 2, pp. 154-165, 2016, doi: 10.1016/j.jtte.2016.03.005. N. Iscan, "Evaluation the level of service at a roundabout: A case study on Al-Ibrahimeya roundabout in Alexandria, Egypt," School of Architecture and the Built Environment (ABE), Civil and Architectural Engineering, Transport planning., KTH Royal Institute of Technology, TRITA-ABE-MBT-19617, 2019. S. AlKheder, F. Al-Rukaibi, and A. Al-Faresi, "Driver behavior at Kuwait roundabouts and its performance evaluation," IATSS Research, vol. 44, no. 4, pp. 272-284, 2020, doi: 10.1016/j.iatssr.2020.03.004. H. G. Demir and Y. K. Demir, "A comparison of traffic flow performance of roundabouts and signalized intersections: A case study in Nigde," The Open Transportation Journal, vol. 14, pp. 120-132, 2020, doi: 10.2174/1874447802014010120. R. Azimi, G. Bhatia, R. Rajkumar, and P. Mudalige, "V2V-Intersection Management at Roundabouts," SAE International Journal of Passenger Cars - Mechanical Systems, vol. 6, no. 2, pp. 681-690, 2013, doi: 10.4271/2013-01-0722. O. Boualam, A. Borsos, C. Koren, and V. Nagy, "Impact of autonomous vehicles on roundabout capacity," Sustainability, vol. 14, no. 4, p. 2203, 2022, doi: 10.3390/su14042203. A.-K. Ekman, "Calibration of traffic models in SIDRA," Department of Science and Technology, Linköping University, LIU-ITN-TEK-A-13/006-SE, 2013. S. Kim and J. Choi, "Safety analysis of roundabout designs based on geometric and speed characteristics," KSCE Journal of Civil Engineering, vol. 17, no. 6, pp. 1446-1454, 2013, doi: 10.1007/s12205-013-0177-4. R. Akçelik, "The relationship between capacity and driver behaviour," presented at the National Roundabout Conference, Transportation Research Board, Kansas City, Missouri, United States, 18-21 May 2008, 2008. A. R. Alozi and M. Hussein, "Multi-criteria comparative assessment of unconventional roundabout designs," International Journal of Transportation Science and Technology, vol. 11, no. 1, pp. 158-173, 2022, doi: 10.1016/j.ijtst.2021.03.006. K. Gkyrtis and A. Kokkalis, "An overview of the efficiency of roundabouts: Design aspects and contribution toward safer vehicle movement," Vehicles, vol. 6, no. 1, pp. 433-449, 2024, doi: 10.3390/vehicles6010019. D. Singh, P. Das, and I. Ghosh, "Conflict-based safety evaluations at unsignalized intersections using surrogate safety measures," Heliyon, vol. 10, no. 5, 2024, doi: 10.1016/j.heliyon.2024.e27665. X. F. Yang, J. R. Wang, and X. Z. Wang, "Multi-lane lane-changing model under the influence of heavy vehicles," Advanced Materials Research, vol. 838-841, pp. 2117-2120, 2014, doi: 10.4028/www.scientific.net/AMR.838-841.2117. E. Macioszek, "The passenger car equivalent factors for heavy vehicles on turbo roundabouts," (in English), Frontiers in Built Environment, vol. 5, 2019, doi: 10.3389/fbuil.2019.00068. N. Kang and H. Nakamura, "An analysis of heavy vehicle impact on roundabout entry capacity in Japan," Transportation Research Procedia, vol. 15, pp. 308-318, 2016, doi: 10.1016/j.trpro.2016.06.026. K. Alkhaledi, "Evaluating the operational and the environmental benefits of a smart roundabout," The South African Journal of Industrial Engineering, vol. 26, p. 191, 2015, doi: 10.7166/26-2-1025. S. Mandavilli, E. R. Russell, and M. J. Rys, "Impact of modern roundabouts on vehicular emissions," in Proceedings of the 2003 Mid-Continent Transportation Research Symposium , Ames, Iowa, United States, 21-22 August 2003 2003. Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5680878","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":444827791,"identity":"705d6409-7279-41f2-a9a2-7b43eab0fd1f","order_by":0,"name":"Chang Saar Chai","email":"","orcid":"","institution":"Taylor's University","correspondingAuthor":false,"prefix":"","firstName":"Chang","middleName":"Saar","lastName":"Chai","suffix":""},{"id":444827792,"identity":"1a4fe3e9-f29e-4527-b329-3f55449166ff","order_by":1,"name":"Jit Boon Bong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6klEQVRIiWNgGAWjYBACxhk8IMqGgYGZgQ0maECMljQStDBIgLUcBhFEamGe3XvwccGv8/YGx3mPPebdc0eOgb15mwRjzmHcDptzLtl4Zt/txA2H+dKNeZ49M2bgOVYmwbgNj5YZOWbSvD23E8wO85hJ8xw4nNggkWNGjJZz9jAt9Q3yb4jQwvPjAFANREsCMEAIaAH5hbchOXH/YR5zwzkHDhu28aQVWyRuS8epxRAUYjx/7Owl+8+YPXhz4LA8P/vhjTc+brPGraUBZFUbkgg4dhIYmnFqkQeTfzAl6nBqGQWjYBSMghEHAK2MUyslDY5TAAAAAElFTkSuQmCC","orcid":"","institution":"Swinburne University of Technology Sarawak Campus","correspondingAuthor":true,"prefix":"","firstName":"Jit","middleName":"Boon","lastName":"Bong","suffix":""},{"id":444827793,"identity":"838e2f35-94d0-44ad-9673-4bcf1428446c","order_by":2,"name":"Kennedy Kwong Shin Tiong","email":"","orcid":"","institution":"Taylor's University","correspondingAuthor":false,"prefix":"","firstName":"Kennedy","middleName":"Kwong Shin","lastName":"Tiong","suffix":""},{"id":444827794,"identity":"8281ce98-053f-4862-afae-89b87c8a7e37","order_by":3,"name":"Lam Tatt Soon","email":"","orcid":"","institution":"Taylor's University","correspondingAuthor":false,"prefix":"","firstName":"Lam","middleName":"Tatt","lastName":"Soon","suffix":""}],"badges":[],"createdAt":"2024-12-20 05:08:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5680878/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5680878/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":81029283,"identity":"766e10ab-1384-4dc5-a512-60386670e5fc","added_by":"auto","created_at":"2025-04-21 11:08:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":256153,"visible":true,"origin":"","legend":"\u003cp\u003eThe VOSViewer co-occurrence keyword map shows that roads and streets, driver behaviour, computer vision and unmanned aerial vehicles are the four main focuses from the 1,000 references.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/76aed2c4b48730f91bf56407.png"},{"id":81029251,"identity":"385b739f-0a1e-4485-8ec6-6ed7fceca956","added_by":"auto","created_at":"2025-04-21 11:08:27","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":58221,"visible":true,"origin":"","legend":"\u003cp\u003eThe proposed methodology framework in this research.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/89a571609dab5c4132a68f42.png"},{"id":81030546,"identity":"24898341-a350-4525-ab74-fb808fde3a4a","added_by":"auto","created_at":"2025-04-21 11:16:27","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1963736,"visible":true,"origin":"","legend":"\u003cp\u003eNew Pantai Expressway Roundabout Kewajipan at Subang Jaya, Selangor, Malaysia.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/3acbd8427a0e1eefcf6d3212.png"},{"id":81029246,"identity":"6c191cd3-9c75-4e11-a17a-65bcae297c60","added_by":"auto","created_at":"2025-04-21 11:08:26","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":606979,"visible":true,"origin":"","legend":"\u003cp\u003eVirtual line setup at GoodVision to obtain traffic flow data at Roundabout Kewajipan.\u003c/p\u003e\n\u003cp\u003eRemarks\u003c/p\u003e\n\u003cp\u003eTG(1)W, TG(2)W, \u0026nbsp;\u0026nbsp;and MAJOR W serve as reference indicators for defining virtual lines in this research. \u0026nbsp;\u0026nbsp;Users may customise their indicators as needed.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/a6cf664f17c333b7522fa42f.png"},{"id":81029293,"identity":"dee17f45-b878-44f6-ab20-5286731564f1","added_by":"auto","created_at":"2025-04-21 11:08:29","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1950910,"visible":true,"origin":"","legend":"\u003cp\u003eVehicle speeds from entry to exit of each leg at Roundabout Kewajipan.\u003c/p\u003e\n\u003cp\u003eRemarks\u003c/p\u003e\n\u003cp\u003eThe arrows \u0026nbsp;\u0026nbsp;indicate the direction of vehicle flow in the roundabout.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/d750746565c9e93218b6203f.png"},{"id":81029304,"identity":"db6423f9-8323-405c-bf78-4046795753f4","added_by":"auto","created_at":"2025-04-21 11:08:30","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":606736,"visible":true,"origin":"","legend":"\u003cp\u003eAcceleration heatmap of Roundabout Kewajipan shows that high speed (cold blue) is found at the circulating lane and low speed at the entry and exit (yellow to red hot).\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/4b708e4830b3ab20981ea03e.png"},{"id":81029257,"identity":"298877b2-f8f9-42d5-848f-2fa5f2cda59e","added_by":"auto","created_at":"2025-04-21 11:08:27","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":184836,"visible":true,"origin":"","legend":"\u003cp\u003eThe relationship between the time gaps between entry vehicles and the average speed of the circulating flow at (a) Leg 1, (b) Leg 2, (c) Leg 3 and (d) Leg 4.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/687b0bb529859fc67ea72936.png"},{"id":85147856,"identity":"eeaefcdd-291e-46a1-8ea5-a7eb9f353247","added_by":"auto","created_at":"2025-06-22 14:16:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8134823,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/282407ff-5a0b-43f3-b40d-a628a8432917.pdf"},{"id":81029254,"identity":"c2709e7c-1da7-450f-b4bd-db92ff6f8115","added_by":"auto","created_at":"2025-04-21 11:08:27","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":14383,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-5680878/v1/4bac2ebbefca0b8b451cb91f.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Optimizing roundabout safety: using UAVs and computer vision for driver behavior analysis","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eTraffic accidents remain a significant public health concern, with an estimated 1.35\u0026nbsp;million fatalities annually, equivalent to 3,700 deaths per day, between 2010 and 2019, as reported by the Malaysia Road Fatalities Index [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. In Selangor alone, 10,675 road accident fatalities were recorded during the same period [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. While roundabouts have been widely recognised for their effectiveness in reducing accidents, occasional road incidences highlight the challenges of ensuring complete road safety [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRoundabouts enhance traffic flow by minimising crossing conflict points and providing a self-regulating traffic system that eliminates the need for traffic signals. Their continuous traffic movement reduces delays and congestion, improving urban mobility [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. However, driver behaviour plays a crucial role in roundabout safety and efficiency. Driver decisions and actions, such as gap acceptance, speed regulation, and manoeuvring techniques, are key to determining traffic safety. Despite being an intangible cognitive process, driver intention can be inferred through observable behaviours, including pedal manoeuvres and mirror checks, which indicate how drivers interact with their surroundings [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTraffic accidents are predominantly attributed to human factors, with driver behaviour identified as a leading cause of crashes [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. In Selangor, many drivers have admitted to engaging in risky driving behaviours, including tailgating, improper overtaking, and disregarding traffic signals [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Although various safety initiatives have been implemented, shaping positive driving attitudes and behaviours remains an overlooked aspect of road safety management. Driving behaviour is influenced by external traffic conditions and internal factors such as decision-making, mental health, and substance use, which affect driver reactions and risk perception [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThere is a strong interrelationship between driver behaviour and traffic congestion, as aggressive driving tendencies increase crash risk and traffic bottlenecks [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Xing, et al. [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] found that positive drivers exhibit higher driving efficiency and fewer errors, whereas those in a negative mood are more prone to mistakes. Despite these insights, the relationship between driver behaviour and crash risks at roundabouts remains insufficiently explored. Given that roundabouts differ from conventional intersections regarding lane interactions, entry-exit manoeuvres, and merging conflicts, a more detailed examination of driver behaviour and its influence on roundabout safety is necessary.\u003c/p\u003e \u003cp\u003eThis study explores driver behaviours and their correlations with accident occurrences at roundabouts. This research builds on previous studies by explicitly examining driving patterns at roundabouts, which are often generalised across various intersection types. In this research, a methodology was developed to analyse the complex relationship between driver behaviours and roundabout safety outcomes. Ultimately, this study aims to improve roundabout traffic management by examining the link between driver behaviours, lane performance, and accident risks, offering insights that can guide more effective traffic planning and policy development.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003ch2\u003e2.1 Research Gap\u003c/h2\u003e\n\u003cp\u003eThe bibliography includes 1,000 carefully selected references, grouped into four clusters that reflect the wide range of topics covered. To enhance the analysis, this study uses the Visualisation of Similarities (VOS) Viewer, a well-known tool for bibliometric analysis. Through this process, 6,526 relevant publications were identified, mainly focusing on traffic management, driver behaviour, safety, roundabouts, unmanned aerial vehicles (UAVs), and computer vision. Expanding the dataset this way helps capture broader research trends in transportation studies.\u003c/p\u003e\n\u003cp\u003eThe VOSviewer analysis revealed an essential gap in the literature. Limited research was done on how driver behaviour in Malaysia affects road safety at roundabouts. This finding highlights the need for further study to understand this issue better. Also, the analysis identifies key research clusters, particularly in UAVs and computer vision, which have a strong potential to improve traffic safety and control (Fig 1). While UAVs and computer vision benefit traffic analysis, research on their application remains limited, presenting an opportunity for further exploration in future studies.\u003c/p\u003e\n\u003ch2\u003e2.2 Driver Behaviors\u003c/h2\u003e\n\u003cp\u003eUnderstanding and addressing driver behaviour is critical in managing road traffic injuries and crashes, which remain a leading global cause of mortality and morbidity. Atombo, et al. [11] showed that speeding was the most common violation in Ghana, followed by overtaking, which drivers engaged in when they believed would enhance their performance. Additionally, drivers with strong control beliefs were more likely to commit these violations, highlighting the need for targeted interventions to improve road safety. Also, in Iran, some mental disorders can increase the risk of road accidents [12]. A study in Malaysia has identified that factors such as driver attitudes and sociodemographic traits significantly influence highway speeding behaviours [13]. Shandhana Rashmi and Marisamynathan [14] summarised aberrant driving behaviours and assessed their impacts on crash risk and driving performance.\u0026nbsp;Their findings show that approximately 90\u0026nbsp;per cent (%)\u0026nbsp;of crash causes are attributed to road users, with drivers being the primary contributing factor.\u003c/p\u003e\n\u003cp\u003eAdedeji, et al. [15] revealed that less-educated drivers are more likely to misinterpret communication cues, increasing the risk of accidents. Drivers in South Africa depend on formal and informal signals. Misinterpreting these signals can lead to moderate to high-risk accidents, emphasising the need for proper driver education on communication. The education level and gender of drivers also significantly impact their comprehension of communication, leading to traffic accidents. Therefore, \u0026nbsp;these factors are significant to road safety. A similar finding was also obtained by Papantoniou, et al. [16] in Greece: (1) female, (2) older, (3) lower education levels, and (4) more driving experience drivers are more likely to commit driving errors. They also pointed out that drivers in rural areas are more prone to risky driving situations and errors.\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e2.3 Unmanned Aerial Vehicles (UAVs)\u003c/h2\u003e\n\u003cp\u003eGiven that driver behaviours are crucial in traffic accidents, there is a growing need for innovative technologies to monitor and manage them to ensure road safety. One such technology is the UAV. UAVs offer a unique advantage in real-time traffic monitoring, enabling authorities to observe driver behaviour and flow from an aerial perspective [17]. UAVs can be classified based on their aerodynamic design, landing mechanism, and level of autonomy. Aerodynamically, they fall into four main categories: fixed-wing, flapping-wing, ducted-fan, and multi-rotor, which include tricopters, quadcopters, hexacopters, and octocopters [18, 19]. Their landing system categorises UAVs as vertical takeoff and landing (VTOL) or horizontal takeoff and landing (HTOL) [19]. Additionally, they can be grouped by autonomy level, ranging from manually piloted UAVs to fully autonomous systems capable of independent operation [20]. Table 1 outlines the advantages and disadvantages of different UAV types.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;1 Advantages and disadvantages based on UAV type [21, 22]\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eTypes of UAVs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eAdvantages\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eDisadvantages\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eSingle Rotor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003eVTOL and hover flight\u003c/li\u003e\n \u003cli\u003eheavy payload\u003c/li\u003e\n \u003cli\u003elong flight time\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003ehard control system\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ehigh safety risk\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eMulti Rotor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003eVTOL and hover flight\u0026nbsp;\u003c/li\u003e\n \u003cli\u003ehigh manoeuvrability\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003elimited payload\u003c/li\u003e\n \u003cli\u003eshort flight time\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eFixed Wing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003elong flight time\u003c/li\u003e\n \u003cli\u003elarge coverage\u003c/li\u003e\n \u003cli\u003efast flight speed\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003erequires space for launching or landing\u003c/li\u003e\n \u003cli\u003ehard control system\u003c/li\u003e\n \u003cli\u003eno VTOL and hover flight\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eHybrid\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003eVTOL flight\u003c/li\u003e\n \u003cli\u003elong flight time\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cul\u003e\n \u003cli\u003eundesirable for hovering or forward flight\u003c/li\u003e\n \u003cli\u003estill in development\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eThe use of UAVs in Road Traffic Monitoring (RTM) has gained momentum in recent years, particularly in urban areas where traffic congestion and accidents are prevalent. Many researchers utilised UAV for RTM, such as the work done by Elloumi, et al.\u0026nbsp;[23], Byun, et al. [24], Dronova, et al. [25], Gupta and Verma [26] and Liu and Bai [27] and. With their ability to capture high-resolution footage from various angles, UAVs can provide real-time insights into traffic patterns, driver behaviour, and hazardous road conditions. This data can be integrated with advanced computer vision software, such as GoodVision, to enhance decision-making in traffic management. Through such technologies, traffic authorities can better identify risky behaviours and intervene more effectively, potentially reducing the number of accidents caused by driver error.\u003c/p\u003e\n\u003ch2\u003e2.4 Computer Vision\u003c/h2\u003e\n\u003cp\u003eComputer vision (CV) is a field of artificial intelligence (AI) that enables computers to interpret and analyse visual data, allowing them to extract meaningful information from images and videos. In road traffic monitoring, CV plays a crucial role in processing UAV footage to detect traffic flow patterns, identify risky behaviours, and assess road conditions. It can automate traffic analysis, reducing the need for manual data collection and improving the accuracy of traffic management strategies. This work selected GoodVision [28] to process drone recording. GoodVision is a cloud-based traffic analysis software using AI and CV technology to analyse traffic video footage.\u0026nbsp;The GoodVision Insight application offers two customisable video processing types tailored to project requirements and camera types. Table 2 shows the video processing types and the required camera height.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;2 The video processing types and their camera height for data processing [28].\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 36px;\"\u003e\n \u003cp\u003eTypes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 63px;\"\u003e\n \u003cp\u003eData Processing Camera Height (meters)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 36px;\"\u003e\n \u003cp\u003eTraffic Camera (fixed)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63px;\"\u003e\n \u003cul\u003e\n \u003cli\u003e5 to 30\u0026nbsp;\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 36px;\"\u003e\n \u003cp\u003eDrone Camera (hover)\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003eHigh Drone\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eLow Drone\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 63px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e30 to 250\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eUp to 30\u0026nbsp;\u003c/li\u003e\n \u003c/ul\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e2.5 Traffic safety study from cognate disciplines\u003c/h2\u003e\n\u003cp\u003eIntegrating UAV and CV tools exemplifies how technological advancements transform traffic safety research. By offering real-time traffic analysis, GoodVision [28] provides valuable data that contributes to understanding driver behaviours and traffic conditions. Adopting these tools in traffic safety research fosters interdisciplinary collaborations that combine psychology, AI, and computer vision, leading to more effective and informed traffic safety strategies at roundabouts and other critical areas. Traffic psychology and behaviour incorporate driver perception, cognition and decision-making, which are crucial to developing accident countermeasures [29]. Numerous research has been done on factors affecting roundabout safety, such as Sheykhfard, et al. [30], Distefano, et al. [31] and Distefano, et al. [32]. In general, driver behaviour plays a crucial role in the safety of roundabouts, affecting the traffic accident rate and regional driving experience.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAdvances in AI and computer vision capture and analyse real-time driver behaviours, allowing for precise assessment of traffic conflicts and safety at roundabouts. Many researchers utilised AI and computer vision in fieldwork to obtain safety data, such as Bhavsar, et al. [33], Scholl, et al. [34], St-Aubin, et al. [35], St-Aubin, et al. [36] and Zaki, et al. [37]. These works show that AI and CV are promising tools for obtaining traffic safety data at roundabouts, providing more actionable data for roundabout safety studies. However, despite these promising advancements, a significant research gap exists regarding the relationship between driver behaviours and road safety in Malaysian roundabouts. While research from various disciplines has extensively covered general traffic safety and AI applications, limited studies address the unique behaviours and contributing factors specific to Malaysian drivers at roundabouts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFurthermore, while CV tools like GoodVision are effective in capturing and analysing traffic data, their application to Malaysian traffic scenarios, particularly concerning driver behaviour at roundabouts, remains underexplored. This gap presents a promising opportunity for further research, especially in integrating UAVs, CV, and AI technologies to understand and improve traffic safety at roundabouts in Malaysia. Addressing this gap would significantly advance traffic safety strategies tailored to local contexts and help bridge the knowledge gap. Therefore, this research aims to investigate the impacts of driver behaviour on traffic safety, efficiency, and overall road management. The research outcomes will contribute valuable data to enhance urban road safety and efficiency.\u003c/p\u003e\n\u003cp\u003eWhile this study does not incorporate advanced modelling techniques such as agent-based simulation, structural equation modelling, or machine learning, its methodology remains valid. It is consistent with several peer-reviewed works that have relied on observational and descriptive approaches to traffic behaviour analysis. For example, St-Aubin et al. (2013) used computer vision techniques to analyse trajectory and behaviour patterns at Canadian roundabouts without employing complex statistical models. Though exploratory, their research is recognised for its methodological rigour and contribution to video-based safety diagnostics. Similarly, Bhavsar et al. (2023) employed UAVs to observe traffic violations at Indian roundabouts and relied primarily on frequency analysis and speed metrics to draw meaningful conclusions for urban traffic management. Zaki, Sayed, and Cheung (2013) also demonstrated that valuable behavioural insights can be extracted from computer vision tools without applying inferential statistics, particularly in cyclist trajectory analysis. These precedents collectively validate the methodological stance of the present study, which emphasises the importance of empirical data extraction, pattern recognition, and visual analytics in establishing a foundational understanding of driver behaviour in real-world environments.\u003c/p\u003e\n\u003cp\u003eBeyond methodological alignment, this study advances the literature in several important ways. First, it offers a novel geographic contribution by focusing on a Malaysian multiple-lane, flyover-covered urban roundabout, an infrastructure context that is underrepresented in global research. While prior studies concentrated on North American or Indian intersections, the current research responds to the urgent need for localised data in Southeast Asia, where cultural driving norms, enforcement rigour, and infrastructure design differ. Second, this study integrates multiple behavioural dimensions into a unified analysis framework, such as average speed, gap acceptance, and idle time. Unlike earlier works that isolated specific behaviours, the present research triangulates these metrics to form a more holistic picture of driver dynamics.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThird, GoodVision\u0026apos;s acceleration heatmap represents a visual innovation by identifying high-risk zones for abrupt acceleration and deceleration, facilitating spatial diagnostics for planners and engineers. This visual layer supplements traditional numeric data with intuitive cues for traffic intervention. Furthermore, while earlier studies primarily described behavioural trends, this study links those trends to concrete policy and planning recommendations. These include suggestions for roundabout geometric redesign, improved speed management, targeted law enforcement, and strategies to reduce vehicular idle time, which directly affect road safety, environmental impact, and traffic efficiency. Lastly, the study demonstrates field-level innovation by adapting commercially available UAV technology (Da-Jiang Innovations (DJI) Air 2S) to Malaysian urban traffic contexts, ensuring data quality while maintaining operational feasibility in resource-constrained settings.\u003c/p\u003e\n\u003cp\u003eTaken together, these contributions reinforce both the validity and the applied relevance of this study. Although it does not employ advanced modelling, its empirical depth, contextual specificity, and policy orientation make it a substantive advancement in the growing UAV- and computer vision-based traffic behaviour analysis fields.\u003c/p\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThis research proposed a methodology framework to collect higher-quality data for a more in-depth analysis (Fig 2). A drone was used to collect traffic flow data in a roundabout. The specifications of the drone used can be found in Appendix A. The footage will be uploaded to GoodVision [28] for processing before being turned into driver behaviour parameters for analysis. Initially, insufficient data was collected for GoodVision [28] to process. Therefore, the data collection process was repeated, as shown by the dotted line. In the second collection, the video was recorded from an alternative viewpoint at 60 meters (m), compared to the previous 28 m, to enhance coverage and accuracy.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFieldwork was conducted at the New Pantai Expressway Roundabout Kewajipan (Lebuhraya New Pantai Bulatan Kewajipan), with a coordinate of 3.0732\u0026deg;N, 101.5930\u0026deg;E. This roundabout has an inscribed circle diameter (ICD) of 90 m, measured from the outer edge of the kerb surrounding the circulating lanes (Fig 3). It has two circulating lanes surrounding the central island, allowing vehicles to circulate clockwise in the roundabout. This roundabout has four legs, with each leg having two-lane entries. This research introduces leg numbering to ease the description. Additionally, the roundabout is integrated with three layers of flyovers, one linked to the Kelana-Subang Link and two towards Sunway City. This roundabout was selected as it carries about 3,000 vehicles per hour daily, providing sufficient data samples for analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe fieldwork was conducted on 22 May 2024, a sunny Friday, to ensure traffic data normality. During fieldwork, a drone was deployed from 1400 to 1600 to capture traffic flows and patterns. Each battery can last for 20 minutes. Hence, six batteries were used to cover a two-hour collection. The drone was flown 60 meters from the roundabout to capture roundabout traffic flow. Vehicles on the three layers of flyovers were excluded as this study focuses only on driver behaviour in the roundabout. Observations were conducted covertly. Since the flight height exceeded the line of sight of the drivers, the natural driver behaviours (without drone impact) were recorded.\u003c/p\u003e\n\u003cp\u003eA drone recording was uploaded to GoodVision [28] to extract traffic flow data. Previous work, such as Humoody and Younis [38] and Bong, et al. [17], used GoodVision [28] in analysing their roundabouts data. The first group of researchers validated data using simulation software, while the second validated by comparing GoodVision [28] data with local field data. Their findings show that data extracted from GoodVision [28] matches their secondary sources, indicating that GoodVision [28] provides reliable data sources for traffic analysis. The GoodVision Insight application lets users draw virtual lines to exact vehicle time gaps. Figure 4 shows an example of extracting time gap at Leg 1 entry using three virtual lines. The time gaps are crucial to study driver behaviours in average vehicle speed, gap acceptance and idle time, which will be discussed in the following.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis research investigates three forms of driver behaviour: average speed and its pattern, gap acceptance, and idle time. Average speed, in kilometres per hour (km/h), is the speed vehicles travel through the roundabout. It differs with the location in a roundabout, such as high-speed (70 to 80 km/h) approaching the roundabout, low speed (10 to 20 km/h) at the yield line and back to high speed when leaving the roundabout [39]. Considering the speed values change with locations, the average speed and its patterns at various locations in roundabouts were included in the studies. Gap acceptance usually happens in unsignalised intersections controlled by priority, such as roundabouts [40, 41]. As vehicles approach a roundabout entry, they will decelerate to seek a gap from the circulating flow, enter the intersection if the gap is larger than or equal to the critical gap, or stop waiting for the next one. Gap acceptances are often measured in headway distribution of circulating vehicles, critical gap, and following gap [42]. Idle time refers to how long vehicles stop or move at very low speeds, often due to yielding or congestion.\u003c/p\u003e"},{"header":"4. Results and Discussion","content":"\u003ch2\u003e4.1 Average Speed Analysis\u003c/h2\u003e\n\u003cp\u003eIn this work, average speed was analysed based on turning movement. For example, for Leg 1, the average speed of the left turn (to Leg 2), through (to Leg 3), right turn (to Leg 4) and U-turn (to Leg 1) were recorded. Out of 16 movements, twelve could not be captured due to the blockage of three layers of the flyover. For instance, the left turn from Leg 3 was blocked by the flyover of Exit to Lebuhraya Pantai Baru, as illustrated in Fig 5. The Subang-Kelana Jaya Link blocked the entry and exit from Leg 2. Therefore, only four movements could be captured. The maximum and average speeds were recorded and compared (Table 3).\u003c/p\u003e\n\u003cp\u003eTable 3 Average and maximum speeds when vehicles perform turning movements at Roundabout Kewajipan.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"102%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eTurning movements\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003eAverage Speed (km/h)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003eMaximum Speed\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e(km/h)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 1 to Leg 2\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eLeft-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e47\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 1 to Leg 3\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eThrough\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 1 to Leg 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eRight-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 1 to Leg 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eU-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 2 to Leg 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eLeft-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 2 to Leg 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eThrough\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 2 to Leg 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eRight-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 2 to Leg 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eU-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 3 to Leg 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eLeft-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 3 to Leg 2\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eThrough\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 3 to Leg 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eRight-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 3 to Leg 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eU-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 4 to Leg 1\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eLeft-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 4 to Leg 2\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eThrough\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 4 to Leg 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eRight-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 18px;\"\u003e\n \u003cp\u003eLeg 4 to Leg 4\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 24px;\"\u003e\n \u003cp\u003eU-turn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 25px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 31px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eRemarks\u003cbr\u003e\u0026nbsp;- indicates data is unavailable due to obstruction by three layers of flyovers.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eFrom the table, the highest average speed was detected at Leg 1, followed by Legs 4, 2 and 3. Three legs hit more than 40 km/h for maximum speed. According to Ahmad and Rastogi [43], the roundabout is constructed with a 30 km/h speed restriction. If any vehicle exceeds or goes below this limit, it might be applied as a safety indicator. An average maximum speed of 43 km/h is recorded from the field, indicating that the drivers at this roundabout may tend to travel at unsafe speeds, which could compromise overall traffic safety.\u003c/p\u003e\n\u003cp\u003eThe risks associated with high vehicle speeds are well-documented in road safety literature. A study by Doecke, et al. [44] discussed the impact of high speed correlated with the risk of serious injury. The likelihood of severe harm sharply increases as speed rises. For instance, a head-on collision at 28 km/h carries only a 1% risk of serious injury. However, this risk escalates to approximately 50% at speeds exceeding 76 km/h. Other collisions also become significantly more dangerous at higher speeds: side impacts at 51 km/h, frontal impacts at 64 km/h, and rear-end crashes at 67 km/h are all associated with a substantially increased risk of serious injury.\u003c/p\u003e\n\u003cp\u003eAverage speed data are crucial for assessing the efficiency and safety of a roundabout. High average speed improves efficient traffic flow but might raise safety concerns. High average speed indicates smooth traffic flow, and vehicles experience minimal delay. However, vehicles entering the roundabout at high speed are unlikely to slow down and yield circulating vehicles, promoting the clashing between entry and circulating vehicles (which indicates less safe driving conditions) [45-47]. On the contrary, low average speed suggests vehicles moving slowly due to congestion, indicating safer driving conditions. The trade-off between efficiency and safety mostly depends on the roundabout geometry that manages vehicle speeds (safety) and accommodates traffic volume (efficiency). In this study, the trade-off of a single-lane roundabout with high traffic volume includes reducing the lane deflection or narrowing lane width to slow down the vehicle speed while maintaining the roundabout capacity.\u003c/p\u003e\n\u003cp\u003eThe acceleration heatmap of GoodVision (2024) was used to display the average speed pattern in different locations in the roundabout. GoodVision (2024) shows areas with decelerating traffic as hot (red) while areas with accelerating traffic are cold (blue) instead of specifying speed metric in defining the areas (to what extent acceleration is indicated as cold or deceleration as hot). Hence, the results visually portray acceleration areas without a specific acceleration range.\u003c/p\u003e\n\u003cp\u003eFigure 5 identifies the acceleration and deceleration in the studied roundabout. The slow-speed area appeared at all leg entries, represented by hot red. At the entry, drivers tend to slow down to seek a gap from the circulating flow to enter the roundabout. Thus, a deceleration was detected at this zone. Conversely, the circulating lane was prioritised based on the give-way rule [48, 49]. They proceed with high speed as they do not need to yield to entry flow. Hence, this area is cold blue, showing acceleration in the roundabout.\u003c/p\u003e\n\u003cp\u003eThe circulating lanes in front of Legs 2, 3 and 4 tend to have lower acceleration (yellowish) than Leg 1. This observation, again, can be linked back to the priority rule. AlKheder, et al. [50] mentioned that almost 80% of their sample always followed the priority rule when utilising a roundabout, with only 20% refusing to prioritise circulating vehicles. A similar condition could happen in this roundabout. Some entry vehicles adopted reverse priority by forcing themselves to enter the roundabout. In this condition, circulating vehicles decelerate to yield to the entry to avoid a collision. Hence, deceleration was observed in these areas.\u003c/p\u003e\n\u003cp\u003eIn short, the heatmap identifies the hot zone for accelerating within the roundabout and decelerating at entry and exit. The high-speed circulating area is often at high risk for collisions or near-misses when entry drivers seek gaps to enter the roundabout. Safe precautions should be taken in this area, including implementing police intervention to regulate traffic flows on each leg, installing a traffic signal to reduce the speed in circulating lanes or speed management by introducing speed limits within the roundabout. The practitioners could update current guidelines or standards limiting the maximum speed within roundabouts or encourage single-lane designs in areas with large vehicle traffic, reducing the need for sudden braking and lane changes. Policymakers could also enforce stricter yielding laws at roundabout entries to help maintain safer interactions between entry and circulating vehicles. \u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e4.2 Gap Acceptance Analysis\u003c/h2\u003e\n\u003cp\u003eIn this work, the gap acceptance by the entry vehicles was studied using circulating speed inside the roundabout. This roundabout operates under the give-way rule, meaning that vehicles seek a gap from the circulating flow to enter the roundabout. Hence, parameters that could represent entry and circulating flows were investigated. GoodVision provides circulating speed and time gaps between entry vehicles, which best present entry and circulation in this condition. The circulating speed reflects the gap availability in the roundabout. A slow circulating speed creates smaller gaps than fast ones, making more gaps available and increasing the traffic volume that can enter the roundabout [51, 52]. The time gaps were collected when two entry vehicles passed through the same virtual line. It reflects the drivers\u0026apos; ability to accept the gap to enter the roundabout. A short time gap illustrates a quick flow into the roundabout, representing a high gap acceptance behaviour. Therefore, the time gaps between entry vehicles were plotted against the average speed of the circulating flow (Fig 7). The relationship should convey how the speed of the circulating flow influences the time gaps between entry vehicles, thereby reflecting driver gap acceptance behaviour.\u003c/p\u003e\n\u003cp\u003eAcross all legs, a positive correlation is observed between the time gap of entry vehicles and the speed of vehicles on the circulating lane.\u0026nbsp;The time gap between entry vehicles increased as the circulating speed increased. As the circulating speed increases, the gap size and availability in the stream shrink [51, 52]. With fewer chances (gaps) in the circulating flow, drivers must seek acceptable gaps more [53, 54]. This scenario leads to fewer vehicles entering the roundabout, which is reflected by the extended time gap between the two entry vehicles. A long time gap shows an entry vehicle stuck at the yield line for too long to seek an acceptable gap (rejecting more gaps in the circulating flow). This slow gap acceptance behaviour caused by small or limited gaps in the circulating flow matches the findings earlier.\u003c/p\u003e\n\u003cp\u003eThe outliers indicate that drivers exhibit slow gap acceptance behaviours at low circulating speeds. For example, Leg 4 has about nine vehicles that take a long time (time gaps of more than 25 seconds (s)) to enter the roundabout when the circulating speed is less than 40 km/h. This scenario can relate to the driver\u0026apos;s decision to merge into the roundabout. Matured drivers tend to have better awareness of the rules of the road, more experience in varied driving conditions and scenarios, a stronger ability to react calmly under stress, and a larger sense of responsibility to other road users. Observation shows that during low traffic flows, the likelihood of accidents significantly increases due to overspeeding. This scenario occurs as the roundabout geometry discourages high-speed entry [43]. Despite outliers, the time gap between entry vehicles and the circulating speed shows strong positive correlations ranging from 0.74 to 0.83, implying a reliable and predictable relationship between the two variables.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBesides the coefficient of determination, this work also conducted the Pearson correlation test to validate the relationship between the time gap of entry vehicles and the speed of vehicles on the circulating lane. The null hypothesis is \u0026quot;there is no relationship between the two variables\u0026quot;, and the alternative hypothesis is \u0026quot;there is a relationship between the two variables\u0026quot;. Table 4 shows the correlation results for all legs. The test also shows a strong positive relationship between the two variables, with a positive coefficient of more than 0.8. The p-value also validates the relationship. The small p-values from all legs reject the null hypothesis, showing a statistically significant positive relationship b between the time gap of entry vehicles and the speed of vehicles on the circulating lane.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 4 Correlation between the time gap of entry vehicles and the speed of vehicles on the circulating lane using the coefficient of determination and Pearson coefficient.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 137px;\"\u003e\n \u003cp\u003eSample size\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 201px;\"\u003e\n \u003cp\u003eCoefficient of determination, r-squared\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003ePearsons coefficient, r (p-value)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eLeg 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 137px;\"\u003e\n \u003cp\u003e412\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 201px;\"\u003e\n \u003cp\u003e0.778\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.857 (6.599E-120)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eLeg 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 137px;\"\u003e\n \u003cp\u003e132\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 201px;\"\u003e\n \u003cp\u003e0.742\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.829 (1.5118E-34)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eLeg 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 137px;\"\u003e\n \u003cp\u003e406\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 201px;\"\u003e\n \u003cp\u003e0.826\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.855 (5.3019E-117)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 76px;\"\u003e\n \u003cp\u003eLeg 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 137px;\"\u003e\n \u003cp\u003e264\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 201px;\"\u003e\n \u003cp\u003e0.743\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 188px;\"\u003e\n \u003cp\u003e0.835 (4.7222E-70)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eThe findings indicate that higher circulating speeds reduce the time gaps between vehicles in the stream, decreasing overall roundabout capacity. The reduced gap could also cause safety issues. A short time gap suggests a higher likelihood of collisions or near-misses at the roundabout entry. As drivers accept the reduced gap, they are more likely to quickly merge into the circulating flow, potentially unaware of the upcoming traffic. [55]. Hence, the collision rate between entry and circulating vehicles increases and the safety drops.\u003c/p\u003e\n\u003cp\u003eSeveral improvements can be made through roundabout geometry adjustment. The number of circulating lanes can be increased to distribute the traffic flow equally, creating more gaps between two successive circulating vehicles. Also, the circulating path curvature could be adjusted to create a circulating lane with a higher turning degree. This design requires additional steering manoeuvres, reducing the speed and improving the safety. This modification can be applied to other roundabouts, given that there is allowance in the central island or areas close to circulating lanes. Improvements can also be made through traffic flow management, such as implementing traffic signals to slow down the circulating speed or implementing police intervention to regulate traffic flow, ensuring vehicles on the saturated leg have more chance to enter the roundabout and clear the congestion.\u003c/p\u003e\n\u003ch2\u003e4.3 Idle Time Analysis\u003c/h2\u003e\n\u003cp\u003eIdle time was determined at entry and exit for each leg based on five vehicle classifications (car, bus, van, truck and heavy truck). This research did not define time range, i.e., how many seconds is considered long or short, due to limited works available to benchmark the defined time range. Therefore, the idle time between legs and vehicle classifications was compared. Table 5 shows the average idle time at entries and exits of all four legs.\u003c/p\u003e\n\u003cp\u003eTable\u0026nbsp;5 Average idle time of Minor Stream Before Entering/Exiting Major Stream\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"5\"\u003e\n \u003cp\u003eAverage Idle Time (seconds)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eCar\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eBus\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eVan\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTruck\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eHeavy Truck\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTotal idle time by leg\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 1 (Entry)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e7.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e10.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e26.29\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 1 (Exit)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9.23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 2 (Entry)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e17.58\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 2 (Exit)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e8.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e7.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e16.32\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 3 (Entry)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e12.06\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 3 (Exit)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e16.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 4 (Entry)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e7.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e21.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eLeg 4 (Exit)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e11.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e23.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e41.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAverage. idle time based on vehicle type\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e5.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e5.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003e\n \u003cp\u003eRemarks\u003c/p\u003e\n \u003cp\u003e- indicates no observation at this leg.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eFor exit, Leg 4 has the highest average idle time (41.15 s) among the legs. The result was caused by the large idle time from the bus (23.91 s) and some from the car (11.06 s). The long idle time is often caused by driver hesitation due to a lack of familiarity with roundabouts or pedestrian volume increment [56, 57]. Conversely, Leg 1 has the lowest idle time of 9.23 s among the legs. All vehicles exit the roundabout smoothly, with the highest idle time being only 3.19 s (by truck).\u003c/p\u003e\n\u003cp\u003eRegarding entry, Leg 1 has the longest time compared to the other. This leg has high inflow demand (668 pcu/h), likely to form queues and congestion. Many drivers attempt to enter the roundabout but must wait for sufficient gaps in the circulating flow [58]. They must wait for traffic ahead to clear, leading to a long idle time. Another possible reason could be that drivers are less aggressive in accepting the gap to enter the roundabout. Hence, they spend more time seeking new gaps, leading to long idle time.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eLeg 3 is the shortest, indicating that drivers can quickly accept the gap in circulating flow and enter the roundabout. A short average gap allows more vehicles to enter but indicates higher collision risks. Entry vehicles might have insufficient time to merge into the circulating flow safely as circulating vehicles are close (as represented by a small gap). This statement is supported by Singh, et al. [59], highlighting that the relatively short time between the end of a right-turning vehicle and the arrival of a moving vehicle at the conflict zone increases the likelihood of critical crossing conflicts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFor vehicle classification, cars, vans and trucks have an average idle time of about four to five seconds. They are smaller in size compared to buses and heavy trucks. The size advantage allows them to move actively and fill in the gap between the large ones instead of staying static, explaining their short idle time. Yang, et al. [60] also stated that light vehicles tend to overtake heavy vehicles to fill the spaces between them, further supporting this behaviour.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe bus and heavy truck have a high idle time of 9.62 s and 5.97 s, respectively. Both vehicles are huge and have large inertia, leading to a longer reaction time to move than light vehicles [61, 62]. Heavy vehicles must seek a large enough gap for the whole vehicle to merge into the circulating flow. The longer the time required to accept a gap, the longer heavy vehicles stay static. Therefore, long idle time was observed.\u003c/p\u003e\n\u003cp\u003eHigh idle time impacts congestion, as vehicles are stuck in traffic for long periods. This scenario later leads to environmental effects such as air pollution due to increments in emissions and fuel consumption [63, 64]. High idle time can often be solved by promoting carpools in busy traffic areas and public transport. Intelligence transportation systems that provide real-time traffic data could help to solve this. They can help drivers to avoid areas with high idle time by providing alternative routes.\u003c/p\u003e\n\u003ch2\u003e4.4 Practical and Policy Implications\u003c/h2\u003e\n\u003cp\u003eThe findings of this study carry significant practical and policy implications, particularly in the context of urban traffic management and road safety in Malaysia. Identifying behavioural patterns such as excessive speeds within roundabouts, aggressive merging by heavy vehicles, and extended idle times among smaller vehicles provides data-driven evidence to inform infrastructure design, enforcement strategies, and driver education programs. For example, acceleration heatmaps revealed zones of abrupt speed changes, suggesting the need for geometric redesigns at entry and exit points to encourage smoother transitions and reduce conflict risks.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFrom a policy standpoint, these insights can support revisions to Malaysia\u0026apos;s existing Public Work Department (PWD) road design standards, especially Arahan Teknik (Jalan) 11/87, which governs roundabout design. Currently, the guidelines emphasise geometric dimensions and signage but lack behavioural calibration based on empirical UAV and computer vision data. This research addresses that gap by offering actionable recommendations grounded in local driver behaviour, including the introduction of lane-specific speed regulations, stricter yield enforcement at entries, and context-sensitive adjustments to lane widths and curvature. Moreover, the observed disparity in behaviour between vehicle classes underscores the potential benefit of adaptive traffic control systems and real-time monitoring solutions, such as integrating AI-enabled UAV surveillance for traffic violation detection and congestion prediction.\u003c/p\u003e\n\u003cp\u003eBeyond infrastructure and enforcement, this study also offers insights that align with national initiatives under Malaysia\u0026apos;s Road Safety Plan 2022\u0026ndash;2030, which aims to reduce road traffic fatalities by 50%. The behavioural findings, particularly regarding driver aggressiveness and decision-making under congestion, support the plan\u0026apos;s strategic pillars on safe road users and speed. Public awareness campaigns and driver education modules can be tailored to address the specific behavioural deficiencies identified, such as the tendency to disregard yielding rules or the hesitation of smaller vehicles to merge, which collectively impact roundabout performance and safety.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFurthermore, the environmental implications of high idle times, such as increased emissions and fuel consumption, are significant, particularly in Malaysia\u0026apos;s broader sustainability goals outlined in the Low Carbon Cities Framework (LCCF) and Green Technology Master Plan. Reducing congestion through behaviour-informed design and policy can significantly lower urban transportation emissions. Overall, this study bridges a crucial gap between observed traffic behaviour and actionable policy response, enabling Malaysian authorities to move beyond traditional, geometry-centric approaches and adopt a more behaviour-sensitive, technology-integrated framework for roundabout safety and efficiency.\u003c/p\u003e"},{"header":"5. Conclusions and recommendations","content":"\u003cp\u003eDriver behaviour is crucial for traffic safety and efficiency at roundabouts. UAVs and CV technologies offer valuable tools for monitoring and analysing traffic patterns. The data collected from these technologies can provide insights for designing safer roundabouts and developing targeted traffic management strategies. By using these advanced tools, roundabout safety and overall traffic management could be improved significantly. This research utilised these technologies to investigate three driver behaviours, including average speed, gap acceptance, and idle time, to understand their impact on roundabout performance and safety. The key findings include:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eThe maximum speed hit more than 40 km/h, showing that drivers tend to travel at unsafe speeds. Acceleration often happens inside the roundabout, and deceleration occurs at the entry and exit of the roundabout.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eA high circulating speed decreases the gap size and availability in the flow. Drivers tend to reject a small gap and seek a large one, leading to fewer vehicles entering the roundabouts and reducing capacity.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eLight vehicles (cars, vans, and trucks) have short idle times due to flexibility in occupying space in the traffic flow. Heavy vehicles have a longer idle time due to large size and inertia, requiring longer reaction time to move and a large gap to enter the roundabout.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThis study enhances roundabout traffic management by addressing gaps in theoretical models, which often generalise driver behaviour rather than capturing real-world complexities. While roundabouts are theoretically designed to improve safety and traffic flow, actual driver behaviours, such as average speed, gap acceptance and idle time, are not always fully considered. By focusing on the New Pantai Expressway Roundabout Kewajipan and using empirical data, this research provides insights on practical and policy standspoint tailored to Malaysian urban roundabout conditions. These findings contribute to more effective roundabout design, management, and assessment, offering valuable guidance for policymakers, traffic planners, and organisations like the Malaysian Public Works Department to enhance traffic efficiency and safety.\u003c/p\u003e \u003cp\u003eThis study was conducted on a two-lane roundabout, with traffic flow recorded over two hours on a sunny working day when high traffic volume was expected. The findings primarily reflect driver behaviour at this specific location and timeframe. Future research can expand by analysing data from multi-lane roundabouts in rural areas, exploring different intersection types, extending data collection to peak hours, or conducting fieldwork under varied weather and lighting conditions. Considering various factors could expand the database, providing a more comprehensive understanding of driver behaviour and improving the relevance and reliability of this study. Additionally, case studies on other multi-lane roundabouts in Malaysia or comparisons with international roundabouts could further validate and generalise the findings.\u003c/p\u003e \u003cp\u003e*Disclaimer: This article uses a language model for language editing purposes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e*Disclaimer: This article uses a language model for language editing purposes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTaylor\u0026apos;s University Human Ethics has confirmed that no ethical approval is required.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInformed Consent\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInformed consent was obtained from all individual participants included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated during and/or analysed during the current study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that this study was conducted without any external funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA clinical trial number is not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contribution\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJit Boon Bong and ChangSaar Chai reviewed and revised the manuscript. Kennedy Kwong Shin Tiong collected data and prepared the first draft. Lam Tatt Soon collected data and screened the preliminary data.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eMinistry of Transport Malaysia. \u0026quot;Malaysia Road Fatalities Index.\u0026quot; https://www.mot.gov.my/en/land/safety/malaysia-road-fatalities-index. (accessed 14 November 2023.\u003c/li\u003e\n \u003cli\u003eRoyal Malaysia Police. \u0026quot;Jumlah kematian disebabkan kemalangan jalan raya mengikut tahun dan negeri.\u0026quot; https://archive.data.gov.my/data/ms_MY/dataset/jumlah-kematian-disebabkan-kemalangan-jalan-raya-mengikut-tahun-dan-negeri/resource/68eaabde-60d1-40a2-bd2c-a43be44de21c?view_id=ea241f92-624a-4013-8ddc-58a6d0aff7a8 (accessed 14 November 2023.\u003c/li\u003e\n \u003cli\u003eH. A. Zubaidi, J. C. Anderson, and S. Hernandez, \u0026quot;Understanding roundabout safety through the application of advanced econometric techniques,\u0026quot; \u003cem\u003eInternational Journal of Transportation Science and Technology,\u0026nbsp;\u003c/em\u003evol. 9, no. 4, pp. 309-321, 2020, doi: 10.1016/j.ijtst.2020.03.001.\u003c/li\u003e\n \u003cli\u003eM. Singh and R. Kaur, \u0026quot;Review study on roundabout,\u0026quot; \u003cem\u003eInternational Journal of Research Publication and Reviews,\u0026nbsp;\u003c/em\u003evol. 1, no. 6, pp. 64-69, 2020.\u003c/li\u003e\n \u003cli\u003eY. Xing, C. Lv, and D. Cao, \u0026quot;Chapter 5 - Driver behavior recognition in driver intention inference systems,\u0026quot; in \u003cem\u003eAdvanced Driver Intention Inference\u003c/em\u003e: Elsevier, 2020, pp. 99-134.\u003c/li\u003e\n \u003cli\u003eD. McCarty and H. W. Kim, \u0026quot;Risky behaviors and road safety: An exploration of age and gender influences on road accident rates,\u0026quot; \u003cem\u003ePLoS One,\u0026nbsp;\u003c/em\u003evol. 19, no. 1, 2024, doi: 10.1371/journal.pone.0296663.\u003c/li\u003e\n \u003cli\u003eL. Hu, X. Bao, H. Wu, and W. Wu, \u0026quot;A study on correlation of traffic accident tendency with driver characters using in-depth traffic accident data,\u0026quot; \u003cem\u003eJournal of Advanced Transportation,\u0026nbsp;\u003c/em\u003evol. 2020, no. 1, 2020, doi: 10.1155/2020/9084245.\u003c/li\u003e\n \u003cli\u003eH. W. You, A. Abdul Rahman, and L. H. Hendri Dwisatrya, \u0026quot;Dataset of driving behaviours in Selangor, Malaysia,\u0026quot; \u003cem\u003eData in Brief,\u0026nbsp;\u003c/em\u003evol. 31, p. 105783, 2020, doi: 10.1016/j.dib.2020.105783.\u003c/li\u003e\n \u003cli\u003eN. S. Azman, R. Ilyas, N. Isah, and M. K. A. Ibrahim, \u0026quot;Riding behaviour of motorcyclists in the Klang Valley \u0026quot;, 2022.\u003c/li\u003e\n \u003cli\u003eG. Li\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e, \u0026quot;Influence of traffic congestion on driver behavior in post-congestion driving,\u0026quot; \u003cem\u003eAccident Analysis \u0026amp; Prevention,\u0026nbsp;\u003c/em\u003evol. 141, p. 105508, 2020, doi: 10.1016/j.aap.2020.105508.\u003c/li\u003e\n \u003cli\u003eC. Atombo, C. Wu, M. Zhong, and H. Zhang, \u0026quot;Investigating the motivational factors influencing drivers intentions to unsafe driving behaviours: Speeding and overtaking violations,\u0026quot; \u003cem\u003eTransportation Research Part F: Traffic Psychology and Behaviour,\u0026nbsp;\u003c/em\u003evol. 43, pp. 104-121, 2016, doi: 10.1016/j.trf.2016.09.029.\u003c/li\u003e\n \u003cli\u003eS. S. Alavi, M. R. Mohammadi, H. Souri, S. Mohammadi Kalhori, F. Jannatifard, and G. Sepahbodi, \u0026quot;Personality, driving behavior and mental disorders factors as predictors of road traffic accidents based on logistic regression,\u0026quot; (in eng), \u003cem\u003eIranian Journal of Medical Sciences,\u0026nbsp;\u003c/em\u003evol. 42, no. 1, pp. 24-31, 2017.\u003c/li\u003e\n \u003cli\u003eF. F. Mohamad, A. S. Saifizul, and J. Mohamad, \u0026quot;Are sociodemographic characteristics and attitude good predictors of speeding behavior among drivers on Malaysia federal roads?,\u0026quot; \u003cem\u003eTraffic Injury Prevention,\u0026nbsp;\u003c/em\u003evol. 20, no. 5, pp. 478-483, 2019, doi: 10.1080/15389588.2019.1612057.\u003c/li\u003e\n \u003cli\u003eB. Shandhana Rashmi and S. Marisamynathan, \u0026quot;Factors affecting truck driver behavior on a road safety context: A critical systematic review of the evidence,\u0026quot; \u003cem\u003eJournal of Traffic and Transportation Engineering (English Edition),\u0026nbsp;\u003c/em\u003evol. 10, no. 5, pp. 835-865, 2023, doi: 10.1016/j.jtte.2023.04.006.\u003c/li\u003e\n \u003cli\u003eJ. A. Adedeji, X. E. Feikie, and D. Dijanic, \u0026quot;Driver-to-driver communication skills: Contributing factors to traffic accidents on South African roads?,\u0026quot; \u003cem\u003eSustainability,\u0026nbsp;\u003c/em\u003evol. 16, no. 11, 2024, doi: 10.3390/su16114833.\u003c/li\u003e\n \u003cli\u003eP. Papantoniou, G. Yannis, and E. Christofa, \u0026quot;Which factors lead to driving errors? A structural equation model analysis through a driving simulator experiment,\u0026quot; \u003cem\u003eIATSS Research,\u0026nbsp;\u003c/em\u003evol. 43, no. 1, pp. 44-50, 2019, doi: 10.1016/j.iatssr.2018.09.003.\u003c/li\u003e\n \u003cli\u003eJ. B. Bong, E. E. M. Chong, and C. S. Choo, \u0026quot;UAV-based lane-changing reduction using traffic video analyser,\u0026quot; in \u003cem\u003eAustralian Engineering Congress (AAEC2022)\u003c/em\u003e, Kuching, Sarawak, Malaysia, 13-15 July 2022 2023, pp. 285-297.\u003c/li\u003e\n \u003cli\u003eM. Farajijalal, H. Eslamiat, V. Avineni, E. Hettel, and C. Lindsay, \u0026quot;Safety systems for emergency landing of civilian Unmanned Aerial Vehicles (UAVs)\u0026mdash;A comprehensive review,\u0026quot; \u003cem\u003eDrones,\u0026nbsp;\u003c/em\u003evol. 9, no. 2, p. 141, 2025, doi: 10.3390/drones9020141.\u003c/li\u003e\n \u003cli\u003eM. El Adawy\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e, \u0026quot;Design and fabrication of a fixed-wing Unmanned Aerial Vehicle (UAV),\u0026quot; \u003cem\u003eAin Shams Engineering Journal,\u0026nbsp;\u003c/em\u003evol. 14, no. 9, 2023, doi: 10.1016/j.asej.2022.102094.\u003c/li\u003e\n \u003cli\u003eS. Bemposta Rosende, S. Ghisler, J. Fern\u0026aacute;ndez-Andr\u0026eacute;s, and J. S\u0026aacute;nchez-Soriano, \u0026quot;Implementation of an edge-computing vision system on reduced-board computers embedded in UAVs for intelligent traffic management,\u0026quot; \u003cem\u003eDrones,\u0026nbsp;\u003c/em\u003evol. 7, no. 11, 2023, doi: doi:10.3390/drones7110682.\u003c/li\u003e\n \u003cli\u003eA. Erceg, B. Činčurak Erceg, and A. Vasilj, \u0026quot;Unmanned aircraft systems in logistics \u0026ndash; legal regulation and worldwide examples toward use in Croatia,\u0026quot; in \u003cem\u003e17th International Scientific Conference Business Logistics in Modern Management (2017)\u003c/em\u003e, Osijek, Croatia, 12-13 October 2017 2017.\u003c/li\u003e\n \u003cli\u003eW. Greenwood William, P. Lynch Jerome, and D. Zekkos, \u0026quot;Applications of UAVs in civil infrastructure,\u0026quot; \u003cem\u003eJournal of Infrastructure Systems,\u0026nbsp;\u003c/em\u003evol. 25, no. 2, 2019, doi: 10.1061/(ASCE)IS.1943-555X.0000464.\u003c/li\u003e\n \u003cli\u003eM. Elloumi, R. Dhaou, B. Escrig, H. Idoudi, and L. A. Saidane, \u0026quot;Monitoring road traffic with a UAV-based system,\u0026quot; in \u003cem\u003e2018 IEEE Wireless Communications and Networking Conference (WCNC)\u003c/em\u003e, Barcelona, Spain, 15-18 April 2018 2018, pp. 1-6, doi: 10.1109/WCNC.2018.8377077.\u003c/li\u003e\n \u003cli\u003eS. Byun, I.-K. Shin, J. Moon, J. Kang, and S.-I. Choi, \u0026quot;Road traffic monitoring from UAV images using deep learning networks,\u0026quot; \u003cem\u003eRemote Sensing,\u0026nbsp;\u003c/em\u003evol. 13, no. 20, 2021, doi: doi:10.3390/rs13204027.\u003c/li\u003e\n \u003cli\u003eO. Dronova\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e, \u0026quot;Unmanned aerial vehicles as element of road traffic safety monitoring,\u0026quot; \u003cem\u003eTransportation Research Procedia,\u0026nbsp;\u003c/em\u003evol. 63, pp. 2308-2314, 2022, doi: 10.1016/j.trpro.2022.06.263.\u003c/li\u003e\n \u003cli\u003eH. Gupta and O. P. Verma, \u0026quot;Monitoring and surveillance of urban road traffic using low altitude drone images: A deep learning approach,\u0026quot; \u003cem\u003eMultimedia Tools and Applications,\u0026nbsp;\u003c/em\u003evol. 81, no. 14, pp. 19683\u0026ndash;19703, 2022, doi: 10.1007/s11042-021-11146-x.\u003c/li\u003e\n \u003cli\u003eS. Liu and Y. Bai, \u0026quot;Multiple UAVs collaborative traffic monitoring with intention-based communication,\u0026quot; \u003cem\u003eComputer Communications,\u0026nbsp;\u003c/em\u003evol. 210, pp. 116-129, 2023, doi: 10.1016/j.comcom.2023.08.005.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eGoodVision\u003c/em\u003e. (2025).\u003c/li\u003e\n \u003cli\u003eJ. A. Groeger and J. A. Rothengatter, \u0026quot;Traffic psychology and behaviour,\u0026quot; \u003cem\u003eTransportation Research Part F: Traffic Psychology and Behaviour,\u0026nbsp;\u003c/em\u003evol. 1, no. 1, pp. 1-9, 1998, doi: 10.1016/S1369-8478(98)00007-2.\u003c/li\u003e\n \u003cli\u003eA. Sheykhfard, F. Haghighi, S. Saeidi, M. SafariTaherkhani, and S. Das, \u0026quot;Understanding the influence of environmental factors on driver speed: A structural equation modeling analysis,\u0026quot; \u003cem\u003eIATSS Research,\u0026nbsp;\u003c/em\u003evol. 48, no. 3, pp. 427-439, 2024, doi: 10.1016/j.iatssr.2024.08.001.\u003c/li\u003e\n \u003cli\u003eN. Distefano\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e, \u0026quot;Physiological and driving behaviour changes associated to different road intersections,\u0026quot; \u003cem\u003eEuropean Transport,\u0026nbsp;\u003c/em\u003evol. 2020, no. 77, 2020, doi: 10.48295/ET.2020.77.4.\u003c/li\u003e\n \u003cli\u003eN. Distefano, S. Leonardi, and G. Pulvirenti, \u0026quot;Factors with the greatest influence on drivers\u0026apos; judgment of roundabouts safety. An analysis based on web survey in Italy,\u0026quot; \u003cem\u003eIATSS Research,\u0026nbsp;\u003c/em\u003evol. 42, no. 4, pp. 265-273, 2018, doi: 10.1016/j.iatssr.2018.04.002.\u003c/li\u003e\n \u003cli\u003eY. M. Bhavsar, M. S. Zaveri, M. S. Raval, and S. B. Zaveri, \u0026quot;Vision-based investigation of road traffic and violations at urban roundabout in India using UAV video: A case study,\u0026quot; \u003cem\u003eTransportation Engineering,\u0026nbsp;\u003c/em\u003evol. 14, 2023, doi: 10.1016/j.treng.2023.100207.\u003c/li\u003e\n \u003cli\u003eL. Scholl, M. Elagaty, B. Ledezma-Navarro, E. Zamora, and L. Miranda-Moreno, \u0026quot;A surrogate video-based safety methodology for diagnosis and evaluation of low-cost pedestrian-safety countermeasures: The case of Cochabamba, Bolivia,\u0026quot; \u003cem\u003eSustainability,\u0026nbsp;\u003c/em\u003evol. 11, no. 17, 2019, doi: doi:10.3390/su11174737.\u003c/li\u003e\n \u003cli\u003eP. St-Aubin, N. Saunier, and L. Miranda-Moreno, \u0026quot;Large-scale automated proactive road safety analysis using video data,\u0026quot; \u003cem\u003eTransportation Research Part C: Emerging Technologies,\u0026nbsp;\u003c/em\u003evol. 58, pp. 363-379, 2015, doi: 10.1016/j.trc.2015.04.007.\u003c/li\u003e\n \u003cli\u003eP. St-Aubin, N. Saunier, L. F. Miranda-Moreno, and K. Ismail, \u0026quot;Use of computer vision data for detailed driver behavior analysis and trajectory interpretation at roundabouts,\u0026quot; \u003cem\u003eTransportation Research Record,\u0026nbsp;\u003c/em\u003evol. 2389, no. 1, pp. 65-77, 2013, doi: 10.3141/2389-07.\u003c/li\u003e\n \u003cli\u003eM. H. Zaki, T. Sayed, and A. Cheung, \u0026quot;Computer vision techniques for the automated collection of cyclist data,\u0026quot; \u003cem\u003eTransportation Research Record,\u0026nbsp;\u003c/em\u003evol. 2387, no. 1, pp. 10-19, 2013, doi: 10.3141/2387-02.\u003c/li\u003e\n \u003cli\u003eM. A. Humoody and N. A. Younis, \u0026quot;Congestion identification for multilane highway by vehicle lane change as driver behavior,\u0026quot; \u003cem\u003eTikrit Journal of Engineering Sciences,\u0026nbsp;\u003c/em\u003evol. 29, no. 1, pp. 46-59, 2022, doi: 10.25130/tjes.29.1.5.\u003c/li\u003e\n \u003cli\u003eB. W. Robinson\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e, \u0026quot;Roundabouts: An informational guide,\u0026quot; United States Federal Highway Administration, 2000.\u003c/li\u003e\n \u003cli\u003eR. Guo, L. Liu, and W. Wang, \u0026quot;Review of roundabout capacity based on gap acceptance,\u0026quot; \u003cem\u003eJournal of Advanced Transportation,\u0026nbsp;\u003c/em\u003evol. 2019, no. 1, 2019, doi: 10.1155/2019/4971479.\u003c/li\u003e\n \u003cli\u003eK. Shaaban and H. Hamad, \u0026quot;Group gap acceptance: A new method to analyze driver behavior and estimate the critical gap at multilane roundabouts,\u0026quot; \u003cem\u003eJournal of Advanced Transportation,\u0026nbsp;\u003c/em\u003evol. 2018, no. 1, 2018, doi: 10.1155/2018/1350679.\u003c/li\u003e\n \u003cli\u003eJ. Dahl and C. Lee, \u0026quot;Empirical estimation of capacity for roundabouts using adjusted gap-acceptance parameters for trucks,\u0026quot; \u003cem\u003eTransportation Research Record,\u0026nbsp;\u003c/em\u003evol. 2312, no. 1, pp. 34-45, 2012, doi: 10.3141/2312-04.\u003c/li\u003e\n \u003cli\u003eA. Ahmad and R. Rastogi, \u0026quot;Influence of roundabout size and circulating traffic flow on critical gaps, follow-up times and speeds,\u0026quot; \u003cem\u003eIndian Highways,\u0026nbsp;\u003c/em\u003evol. 45, pp. 37-48, 2017.\u003c/li\u003e\n \u003cli\u003eS. D. Doecke, M. R. J. Baldock, C. N. Kloeden, and J. K. Dutschke, \u0026quot;Impact speed and the risk of serious injury in vehicle crashes,\u0026quot; \u003cem\u003eAccident Analysis \u0026amp; Prevention,\u0026nbsp;\u003c/em\u003evol. 144, 2020, doi: 10.1016/j.aap.2020.105629.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eCD116 Geometric design of roundabouts\u003c/em\u003e, Highways Agency, London, United Kingdom, 2020.\u003c/li\u003e\n \u003cli\u003eIshanya, S. Marathe, and Y. Suresh, \u0026quot;A study on rotary intersection at Mangaluru \u0026ndash; A case study of Nanthur Junction,\u0026quot; \u003cem\u003eInternational Journal of Current Engineering and Scientific Research (IJCESR),\u0026nbsp;\u003c/em\u003evol. 4, no. 5, 2017.\u003c/li\u003e\n \u003cli\u003eI. Mahdalova, V. Krivda, and V. Skvain, \u0026quot;Influence of roundabout inscribed circle diameter to the traffic safety,\u0026quot; \u003cem\u003eApplied Mechanics and Materials,\u0026nbsp;\u003c/em\u003evol. 409-410, pp. 1122-1125, 2013, doi: 10.4028/www.scientific.net/AMM.409-410.1122.\u003c/li\u003e\n \u003cli\u003eX. Chen and M. S. Lee, \u0026quot;A case study on multi-lane roundabouts under congestion: Comparing software capacity and delay estimates with field data,\u0026quot; \u003cem\u003eJournal of Traffic and Transportation Engineering (English Edition),\u0026nbsp;\u003c/em\u003evol. 3, no. 2, pp. 154-165, 2016, doi: 10.1016/j.jtte.2016.03.005.\u003c/li\u003e\n \u003cli\u003eN. Iscan, \u0026quot;Evaluation the level of service at a roundabout: A case study on Al-Ibrahimeya roundabout in Alexandria, Egypt,\u0026quot; School of Architecture and the Built Environment (ABE), Civil and Architectural Engineering, Transport planning., KTH Royal Institute of Technology, TRITA-ABE-MBT-19617, 2019.\u003c/li\u003e\n \u003cli\u003eS. AlKheder, F. Al-Rukaibi, and A. Al-Faresi, \u0026quot;Driver behavior at Kuwait roundabouts and its performance evaluation,\u0026quot; \u003cem\u003eIATSS Research,\u0026nbsp;\u003c/em\u003evol. 44, no. 4, pp. 272-284, 2020, doi: 10.1016/j.iatssr.2020.03.004.\u003c/li\u003e\n \u003cli\u003eH. G. Demir and Y. K. Demir, \u0026quot;A comparison of traffic flow performance of roundabouts and signalized intersections: A case study in Nigde,\u0026quot; \u003cem\u003eThe Open Transportation Journal,\u0026nbsp;\u003c/em\u003evol. 14, pp. 120-132, 2020, doi: 10.2174/1874447802014010120.\u003c/li\u003e\n \u003cli\u003eR. Azimi, G. Bhatia, R. Rajkumar, and P. Mudalige, \u0026quot;V2V-Intersection Management at Roundabouts,\u0026quot; \u003cem\u003eSAE International Journal of Passenger Cars - Mechanical Systems,\u0026nbsp;\u003c/em\u003evol. 6, no. 2, pp. 681-690, 2013, doi: 10.4271/2013-01-0722.\u003c/li\u003e\n \u003cli\u003eO. Boualam, A. Borsos, C. Koren, and V. Nagy, \u0026quot;Impact of autonomous vehicles on roundabout capacity,\u0026quot; \u003cem\u003eSustainability,\u0026nbsp;\u003c/em\u003evol. 14, no. 4, p. 2203, 2022, doi: 10.3390/su14042203.\u003c/li\u003e\n \u003cli\u003eA.-K. Ekman, \u0026quot;Calibration of traffic models in SIDRA,\u0026quot; Department of Science and Technology, Link\u0026ouml;ping University, LIU-ITN-TEK-A-13/006-SE, 2013.\u003c/li\u003e\n \u003cli\u003eS. Kim and J. Choi, \u0026quot;Safety analysis of roundabout designs based on geometric and speed characteristics,\u0026quot; \u003cem\u003eKSCE Journal of Civil Engineering,\u0026nbsp;\u003c/em\u003evol. 17, no. 6, pp. 1446-1454, 2013, doi: 10.1007/s12205-013-0177-4.\u003c/li\u003e\n \u003cli\u003eR. Ak\u0026ccedil;elik, \u0026quot;The relationship between capacity and driver behaviour,\u0026quot; presented at the National Roundabout Conference, Transportation Research Board, Kansas City, Missouri, United States, 18-21 May 2008, 2008.\u003c/li\u003e\n \u003cli\u003eA. R. Alozi and M. Hussein, \u0026quot;Multi-criteria comparative assessment of unconventional roundabout designs,\u0026quot; \u003cem\u003eInternational Journal of Transportation Science and Technology,\u0026nbsp;\u003c/em\u003evol. 11, no. 1, pp. 158-173, 2022, doi: 10.1016/j.ijtst.2021.03.006.\u003c/li\u003e\n \u003cli\u003eK. Gkyrtis and A. Kokkalis, \u0026quot;An overview of the efficiency of roundabouts: Design aspects and contribution toward safer vehicle movement,\u0026quot; \u003cem\u003eVehicles,\u0026nbsp;\u003c/em\u003evol. 6, no. 1, pp. 433-449, 2024, doi: 10.3390/vehicles6010019.\u003c/li\u003e\n \u003cli\u003eD. Singh, P. Das, and I. Ghosh, \u0026quot;Conflict-based safety evaluations at unsignalized intersections using surrogate safety measures,\u0026quot; \u003cem\u003eHeliyon,\u0026nbsp;\u003c/em\u003evol. 10, no. 5, 2024, doi: 10.1016/j.heliyon.2024.e27665.\u003c/li\u003e\n \u003cli\u003eX. F. Yang, J. R. Wang, and X. Z. Wang, \u0026quot;Multi-lane lane-changing model under the influence of heavy vehicles,\u0026quot; \u003cem\u003eAdvanced Materials Research,\u0026nbsp;\u003c/em\u003evol. 838-841, pp. 2117-2120, 2014, doi: 10.4028/www.scientific.net/AMR.838-841.2117.\u003c/li\u003e\n \u003cli\u003eE. Macioszek, \u0026quot;The passenger car equivalent factors for heavy vehicles on turbo roundabouts,\u0026quot; (in English), \u003cem\u003eFrontiers in Built Environment,\u0026nbsp;\u003c/em\u003evol. 5, 2019, doi: 10.3389/fbuil.2019.00068.\u003c/li\u003e\n \u003cli\u003eN. Kang and H. Nakamura, \u0026quot;An analysis of heavy vehicle impact on roundabout entry capacity in Japan,\u0026quot; \u003cem\u003eTransportation Research Procedia,\u0026nbsp;\u003c/em\u003evol. 15, pp. 308-318, 2016, doi: 10.1016/j.trpro.2016.06.026.\u003c/li\u003e\n \u003cli\u003eK. Alkhaledi, \u0026quot;Evaluating the operational and the environmental benefits of a smart roundabout,\u0026quot; \u003cem\u003eThe South African Journal of Industrial Engineering,\u0026nbsp;\u003c/em\u003evol. 26, p. 191, 2015, doi: 10.7166/26-2-1025.\u003c/li\u003e\n \u003cli\u003eS. Mandavilli, E. R. Russell, and M. J. Rys, \u0026quot;Impact of modern roundabouts on vehicular emissions,\u0026quot; in \u003cem\u003eProceedings of the 2003 Mid-Continent Transportation Research Symposium\u003c/em\u003e, Ames, Iowa, United States, 21-22 August 2003 2003.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Roundabout, Traffic Management, Driver Behaviors, Unmanned Aerial Vehicles (UAVs), Computer Vision","lastPublishedDoi":"10.21203/rs.3.rs-5680878/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5680878/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRoundabouts are safer than standard intersections, but traffic accidents remain a major public health issue. Driver behaviour plays a critical role in accident occurrence at roundabouts. To enhance safety, this study examines the relationship between driver behaviours and roundabout traffic accidents. The study utilised Unmanned Aerial Vehicles (UAVs) to collect field data and process it using a computer vision tool. Driver behaviours were studied regarding average speed, pattern, gap acceptance, and idle time. The maximum speed exceeds 40 km/h, with drivers often travelling above safe limits. They also tend to accelerate inside the roundabout and decelerate at the entry and exit. Also, high circulating speeds reduce gap size, limiting roundabout entry and reducing capacity. Light vehicles have shorter idle times of less than six seconds, while heavy vehicles, due to size and inertia, require longer reaction times and larger gaps, leading to long idle times. The research contributes significantly to roundabout safety, efficiency and urban planning. It aids in designing future roundabouts suited for local traffic capacity and behaviour patterns while promoting safe and efficient intersections. This study also highlights the potential of using advanced technologies, like UAVs and computer vision, for further traffic analysis. Ultimately, understanding driver behaviour is key to improving traffic safety and efficiency.\u003c/p\u003e","manuscriptTitle":"Optimizing roundabout safety: using UAVs and computer vision for driver behavior analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-21 11:08:12","doi":"10.21203/rs.3.rs-5680878/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cd842046-bc7a-4f9d-927d-ec56c9093d72","owner":[],"postedDate":"April 21st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-06-22T14:08:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-21 11:08:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5680878","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5680878","identity":"rs-5680878","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}