Design and Implementation of an AI-Driven Smart Waste Sorting System: A Java-Based Simulation for Enhancing Recycling Efficiency

Design and Implementation of an AI-Driven Smart Waste Sorting System: A Java-Based Simulation for Enhancing Recycling Efficiency | 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 Design and Implementation of an AI-Driven Smart Waste Sorting System: A Java-Based Simulation for Enhancing Recycling Efficiency Taiyeba Tasnim, Boishik Bonik, Attick Bonik, Tasrif Ahamed Sabid, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8135228/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 This study presents the design and implementation of an AI-driven smart waste sorting system supported by a Java-based simulation framework to enhance recycling efficiency. The research addresses persistent challenges in manual waste sorting—such as misclassification, high labor dependency, and contamination—by integrating machine learning models for waste identification and a real-time decision workflow simulated through Java. A multi-class classification model was trained using CNN, SVM, and Random Forest algorithms, with CNN achieving the highest accuracy (94.8%). The simulation replicates waste flow dynamics, sorting decisions, and throughput variations, enabling systematic evaluation of model performance under different scenarios. Experimental results show significant improvements in sorting accuracy, purity rate, and operational efficiency compared to traditional manual sorting. The proposed system demonstrates that AI-powered classification, combined with a modular Java simulation environment, can serve as an effective tool for advancing intelligent waste management technologies. This work offers practical insights for future smart waste infrastructures and provides a reusable platform for researchers to test and optimize sorting algorithms before real-world deployment. Smart waste management machine learning-based waste classification Java simulation environment convolutional neural networks automated waste sorting Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Global waste generation has increased at an unprecedented rate over the past decades, driven by rapid urbanization, population growth, and industrial expansion. According to the World Bank, global municipal solid waste (MSW) production is projected to reach 3.40 billion tons annually by 2050, up from 2.01 billion tons in 2018 (Kaza et al., 2018). This surge in waste generation poses severe environmental, social, and economic challenges for both developed and developing nations. Traditional waste management methods—especially manual sorting—are becoming increasingly unsustainable due to their high labor costs, inaccuracy, and exposure to hazardous materials (Wilson et al., 2015). Manual waste sorting often results in contamination, inconsistent classification, and reduced recycling efficiency, limiting the effectiveness of municipal recycling programs (Mwanza & Mbohwa, 2017). Consequently, researchers and governments worldwide are turning toward automated and intelligent waste sorting systems as an innovative and scalable solution. The growing interest in smart waste management technologies is supported by advancements in artificial intelligence (AI), machine learning (ML), computer vision, robotics, and Internet of Things (IoT) devices. These technologies enable machines to identify, categorize, and sort waste with high accuracy and reliability (Zeng et al., 2019). AI-driven systems can recognize distinct waste types—such as plastic, paper, glass, metal, and organic waste—through image-based classification, sensor data, or neural network models (Bai et al., 2021). IoT-enabled waste bins equipped with sensors can detect fill levels, monitor waste composition, and communicate with centralized systems to optimize waste collection routes, thereby reducing operational costs and environmental impact (Abolfazli et al., 2014). As global nations increasingly adopt smart city initiatives, AI-integrated waste management solutions have become a focal point for innovation and environmental sustainability (Sharholy et al., 2018). Despite these technological advancements, traditional waste sorting workflows remain slow, highly error-prone, and expensive, especially in regions lacking advanced infrastructure. Most developing nations struggle with inefficient collection systems, outdated sorting technologies, and minimal access to automated facilities (Suthar & Singh, 2015). The absence of reliable automation increases the burden on municipal authorities and limits the effectiveness of recycling initiatives (Hoornweg & Bhada-Tata, 2012). Manual sorting also exposes workers to harmful substances, increasing occupational health hazards and reducing system efficiency (Mavropoulos & Newman, 2015). Furthermore, in many regions, waste is mixed at the source, making manual identification extremely challenging and directly impacting the purity of recovered materials (Alavi, 2016). These inefficiencies highlight the urgent need for AI-driven automated waste sorting systems that can operate with precision and handle diverse waste streams. An additional challenge is the lack of simulation tools for designing, testing, and validating waste sorting algorithms. While many studies implement machine learning models for waste classification, most developments focus solely on algorithmic accuracy rather than system-level validation (Yang & Thung, 2016). Without simulation platforms, researchers cannot easily test how machine learning models behave under different waste-flow conditions, varying object speeds, lighting variations, or multiple waste types arriving simultaneously. Also, simulations provide a valuable environment to iterate on system designs before deploying physical prototypes, reducing costs and risks (Reynaud et al., 2020). Yet, very few studies provide interactive or real-time simulation tools for evaluating smart waste sorting mechanisms. A major limitation identified in the current literature is the overreliance on Python-based frameworks, particularly TensorFlow, PyTorch, and OpenCV, for the implementation of waste classification models (Menon et al., 2020). Although Python offers strong ML libraries, very few studies adopt Java-based simulation frameworks, despite Java’s advantages in cross-platform deployment, GUI development, scalability, and industrial application environments (Sestoft, 2020). Java’s ability to integrate ML models using libraries like DeepLearning4J or WEKA makes it a suitable choice for building robust simulations with real-time visualization. However, this potential is underexplored in existing research, creating a clear knowledge gap. Another prominent gap is that existing smart bin systems—which integrate sensors or simple detection mechanisms—rarely evaluate how algorithms perform under changing or complex waste-flow scenarios. Most studies only test recognition accuracy using static datasets but do not simulate dynamic sorting workflows involving mechanical movement, classifier decision delays, or continuous waste streams (Rad et al., 2017). Additionally, very few investigations combine AI waste classification, Java simulation, and performance analysis into a unified research framework. As a result, system-level impacts such as throughput, misclassification rates, decision latency, and overall recycling efficiency remain largely unexplored. The current research aims to address these limitations by designing and implementing a complete AI-driven smart waste sorting system integrated into a Java-based simulation environment. The primary objective is to develop a robust and interactive simulation platform capable of testing machine learning–based waste classification under real-time conditions. This study trains machine learning models—such as Convolutional Neural Networks (CNNs), Support Vector Machines (SVMs), and Random Forests (RF)—to classify waste into multiple categories. The trained models are then integrated into a Java-based simulation where waste objects move dynamically along a virtual conveyor belt, and real-time classification decisions determine the sorting actions. The simulation evaluates multiple performance parameters, including sorting accuracy, classification speed, throughput, and system robustness. First, it introduces a Java-based simulation environment tailored for testing and evaluating AI-driven waste sorting algorithms, addressing the current absence of such tools in the literature. Second, it develops a machine learning model for multi-class waste identification, focusing on image-based classification using a standard dataset. Third, it provides a comparative evaluation of multiple machine learning algorithms, including CNN, SVM, and RF, to identify the most effective model for real-time waste classification. Fourth, it integrates the selected ML model into a real-time waste sorting workflow simulation, enabling system-level performance analysis and identifying efficiency improvements over manual sorting methods. Overall, this research advances the field of smart waste management by providing a unified framework combining AI, simulation, and performance assessment. The proposed system can serve as a foundational model for future physical implementations that involve robotic arms, conveyor belts, and IoT-enabled smart bins. By leveraging AI-driven classification and Java-based simulation, this study demonstrates an innovative pathway toward enhancing global recycling efficiency and supporting sustainable urban waste management systems. 2. Literature Review Effective waste management has become an increasingly critical global concern due to rapid urbanization, heightened consumption patterns, and the limited capacity of existing disposal technologies. The literature surrounding smart waste management systems, automated sorting technologies, machine learning–based classification, and simulation environments has expanded significantly over the past decade. This section reviews previous studies organized into four main areas: ( 1 ) global waste management and recycling challenges, ( 2 ) smart waste systems and technological advancements, ( 3 ) machine learning and computer vision for waste classification, and ( 4 ) the role of simulation tools in designing and evaluating automated waste sorting systems, followed by the gap analysis. 2.1 Global Waste Management and Recycling Challenges Global municipal solid waste (MSW) production continues to rise dramatically, posing significant economic, environmental, and infrastructural challenges. According to the World Bank, global waste generation is projected to reach 3.40 billion tons by 2050 (Kaza et al., 2018). This increase is driven primarily by population growth, industrialization, and urban expansion (Hoornweg & Bhada-Tata, 2012). Many municipalities continue to rely on traditional waste management practices such as manual sorting, landfilling, and unregulated dumping, which are inadequate for modern waste volumes (Wilson et al., 2015). Manual waste sorting is labor-intensive, hazardous, and prone to errors (Mwanza & Mbohwa, 2017). Workers are frequently exposed to infectious, chemical, and sharp wastes, leading to serious health risks (Couth & Trois, 2012). Additionally, manual sorting decreases recycling efficiency due to inconsistent classification and contamination of materials (Pires et al., 2011). These limitations have driven the adoption of automated waste sorting systems and intelligent classification technologies. In developing regions, waste management faces structural weaknesses such as lack of funding, poor infrastructure, inefficient collection systems, and low public awareness (Suthar & Singh, 2015; Guerrero et al., 2013). The absence of advanced facilities and automated sorting technologies further compounds the challenge (Alavi, 2016). Thus, improving waste sorting through intelligent systems has become essential for enhancing recycling efficiency and achieving sustainable waste management outcomes. 2.2 Smart Waste Systems and Technological Advancements The evolution of smart waste management technologies aligns closely with the rise of smart cities and Industry 4.0. IoT-enabled smart bins equipped with ultrasonic sensors, RFID tags, and fill-level monitoring systems help optimize waste collection routes and reduce operational costs (Abolfazli et al., 2014; Idris et al., 2018). These systems enable real-time monitoring and provide data-driven insights for municipal waste authorities (Longhi et al., 2012). Computer vision and sensor-based technologies have also been increasingly adopted to improve automated waste sorting accuracy. For instance, hyperspectral imaging systems have been used to identify specific material types, including plastics and metals, based on spectral signatures (Aldayarov et al., 2020). Robotics-based sorting systems use mechanical arms guided by visual detection algorithms to sort waste in real time (Rocha et al., 2019). Earlier smart bin designs primarily relied on basic sensing mechanisms and rule-based algorithms (Singh et al., 2017). However, modern systems leverage AI-based decision-making to handle dynamic waste streams, classify materials, and provide automated responses to varying conditions (Sharholy et al., 2018). As smart waste management technologies evolve, integration with machine learning has become increasingly prominent. 2.3 Machine Learning and Computer Vision in Waste Classification Machine learning (ML) and deep learning (DL) techniques have significantly advanced the field of automated waste classification. Convolutional Neural Networks (CNNs), in particular, have demonstrated high accuracy in image-based waste classification tasks due to their ability to extract high-level visual features (Yang & Thung, 2016; Bai et al., 2021). Transfer learning models such as VGG16, ResNet50, MobileNet, and InceptionV3 have been widely used for recognizing recyclable materials (Saadat et al., 2021). In addition to CNNs, classical ML algorithms such as Support Vector Machines (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN) have been applied in waste classification based on image features or sensor-derived attributes (Rad et al., 2017; Alqahtani & Kavakli, 2020). SVM is known for its strong generalization performance, while RF provides robust multi-class classification with reduced overfitting (Breiman, 2001). Researchers have developed various datasets to support ML-based waste classification. The TrashNet dataset introduced by Yang and Thung (2016) paved the way for numerous studies on recyclable materials identification. Later studies expanded the dataset or introduced new image repositories to improve classification robustness under different lighting, orientation, and environmental conditions (Garcia-Garin et al., 2020). Deep learning–based detection frameworks such as YOLO, Faster R-CNN, and SSD have also been employed for multi-class waste detection and real-time sorting (Jain et al., 2022). These frameworks support object localization in addition to classification, improving overall sorting accuracy. However, existing studies mostly focus on developing accurate classification algorithms, while system-level simulation and integration of these models into automated sorting workflows remain largely unexplored. 2.4 Simulation Tools for Automated Waste Sorting Systems Simulation plays a crucial role in evaluating the performance of automated systems before physical implementation. Simulation tools allow researchers to test the behavior of algorithms under different operational conditions and optimize system design at minimal cost (Reynaud et al., 2020). In waste management, simulations have been used to model landfill operations, collection routes, and recycling systems (Mavropoulos & Newman, 2015). Despite the significance of simulations, only a limited number of studies integrate ML models into dynamic waste sorting simulations. Most existing works rely on Python-based prototypes without real-time simulation components (Menon et al., 2020). Java, with its robust GUI frameworks (JavaFX, Swing) and cross-platform capabilities, has been underutilized in waste sorting simulations despite its adoption in other industrial simulations (Sestoft, 2020). Simulation frameworks allow for testing conveyor belt behavior, object movement speed, classifier decision delays, and mechanical sorting actions before creating costly physical prototypes (Abdelrahman et al., 2019). In robotics research, simulations such as ROS-Gazebo provide testing environments, but these tools are rarely applied specifically to waste sorting research (Lenz et al., 2015). Therefore, the lack of Java-based simulation environments integrating machine learning waste classifiers constitutes a significant research gap. 2.5 Gaps Identified in the Literature Based on the reviewed literature, several gaps emerge: Lack of Java-Based Simulation Frameworks: Most waste classification systems rely on Python, while Java-based real-time simulation environments integrating ML models are scarce. Limited Studies Combining AI + Simulation: Most studies focus on algorithmic accuracy rather than full workflow simulation involving conveyor movements, decision delays, or actuator logic. Minimal System-Level Performance Evaluation: Few works evaluate throughput, misclassification rate, decision speed, or sorting efficiency using integrated ML models. Inadequate Testing Under Dynamic Conditions: Existing smart bin systems do not simulate real-time conditions such as varied waste flow speeds or multi-object detection. This research addresses these gaps by integrating machine learning classification models with a Java-based simulation platform that replicates real-time waste sorting operations, providing a holistic performance evaluation framework. 3. Methodology The methodology adopted for this research integrates data-driven machine learning, simulation engineering, and performance evaluation to develop an AI-driven smart waste sorting system. The overall workflow follows a structured pipeline consisting of dataset acquisition, image preprocessing, machine learning model development, Java-based simulation implementation, and system evaluation. The methodological design ensures that both computational and system-level aspects are rigorously examined to validate the effectiveness of the proposed waste sorting framework. 3.1 Research Framework The research methodology is grounded in a sequential yet iterative design framework. Initially, a comprehensive dataset was constructed using publicly available waste image repositories. Following this, advanced preprocessing techniques were applied to prepare the data for machine learning models. Subsequently, multiple algorithms including Convolutional Neural Network (CNN), Support Vector Machine (SVM), and Random Forest (RF) were trained and evaluated to determine the most reliable waste classification model. The best-performing model was exported and integrated into a custom Java-based simulation engine, which emulates the real-world workflow of an automated waste sorting facility. Finally, a performance evaluation was conducted under several simulation scenarios to assess system accuracy, throughput, and efficiency. This framework enables a holistic experiment-to-simulation connection where the AI model directly influences virtual sorting behavior. 3.2 Dataset Acquisition and Preprocessing To ensure comprehensive coverage of waste categories, the study utilized two publicly available datasets: the TrashNet dataset and the Waste Classification dataset from Kaggle. These datasets collectively provided images from six primary categories—plastic, paper, metal, glass, organic, and mixed waste. A total of 3,617 images were curated after removing duplicates and low-quality samples. Each category contained a variable number of samples, ensuring a diverse and representative dataset. Table 1 Dataset Characteristics Category No. of Images Source Description Plastic 482 TrashNet Bottles, wrappers, containers Paper 594 TrashNet Newspapers, sheets, cartons Metal 410 TrashNet Aluminum cans, metal scrap Glass 501 TrashNet Bottles, jars, glassware Organic 980 Kaggle Food waste, leaves, biodegradable matter Mixed 650 Kaggle Non-categorized waste objects To prepare the dataset for machine learning, several preprocessing steps were performed. First, all images were resized to 224 × 224 pixels to maintain uniformity. Pixel values were normalized to the 0–1 range to ensure optimal CNN learning behavior. Augmentation techniques—such as random rotation, zooming, and horizontal flipping—were used to enhance dataset variability and minimize overfitting. Labels were encoded numerically and the dataset was divided into an 80/10/10 train–validation–test split. This distribution allowed the models to learn effectively while maintaining evaluation integrity. Table 2 Data Split Distribution Category Train Validation Test Plastic 385 48 49 Paper 475 60 59 Metal 326 42 42 Glass 401 50 50 Organic 784 98 98 Mixed 520 65 65 3.3 Machine Learning Model Development Three core machine learning algorithms were developed and compared in this study: a Convolutional Neural Network (CNN), a Support Vector Machine (SVM), and a Random Forest (RF) classifier. The CNN served as the primary deep learning architecture due to its ability to learn spatial and hierarchical features. The model consisted of sequential convolutional layers, pooling layers, a fully connected dense layer, and a softmax output layer for multi-class classification. The model was trained using the Adam optimizer, categorical cross-entropy loss, a batch size of 32, and 30 epochs.In contrast, the SVM model used handcrafted features extracted through Histogram of Oriented Gradients (HOG). Images were converted to grayscale before feature extraction. The SVM classifier used a radial basis function (RBF) kernel to capture non-linear relationships. The Random Forest model operated on dimensionality-reduced pixel data using Principal Component Analysis (PCA), retaining 300 principal components for optimal performance. The RF model utilized 200 decision trees, enabling robust classification through ensemble learning. Performance for all models was assessed using precision, recall, accuracy, and F1-score. Among the three, the CNN demonstrated the highest classification accuracy, making it the final model integrated into the Java simulation. Table 3 Model Performance Comparison Model Accuracy Precision Recall F1-Score CNN 94.2% 93.7% 94.2% 94.0% SVM 87.5% 86.4% 87.2% 86.7% RF 82.1% 81.0% 80.5% 80.7% 3.4 Java-Based Simulation Engine A custom Java-based simulation was designed to mimic the real-world functioning of an automated waste sorting facility. The simulation includes modules for waste item creation, conveyor belt movement, AI classification integration, decision-making logic, robotic arm control, and bin management. The system architecture is multi-layered and modular, enabling easy adjustments of simulation parameters such as waste flow rates and bin capacity. The simulation begins by generating virtual waste items, each linked with an image selected from the test dataset. The image is passed into the embedded TensorFlow Lite model, which outputs the predicted category. Subsequently, the decision engine instructs the robotic arm simulation to direct the item toward the appropriate waste bin. A graphical dashboard developed using JavaFX provides real-time visualization of bin usage, sorting actions, throughput, and time-based performance metrics. This Java-based environment provides a controlled platform to analyse the behavior of AI-powered waste sorting systems under varying operational conditions. 3.5 Simulation Workflow and Algorithms The simulation workflow consists of sequential steps that connect image classification with mechanical movement. First, the system loads the waste image, performs classification through the CNN model, and retrieves the predicted category. The classification result triggers the sorting logic, which controls the movement of a simulated robotic arm to deposit the waste item into the corresponding bin. Detailed logs are captured for each processed item to record accuracy and throughput. Two main algorithms govern the system: the AI-based classification algorithm and the sorting decision logic algorithm. The classification algorithm preprocesses the images, applies the CNN model, extracts the probability distribution, and selects the class with the highest probability. The decision logic algorithm maps each predicted class to a specific bin and updates simulation statistics. These steps collectively ensure realistic and accurate sorting behavior. 3.6 Evaluation Metrics and Experimental Scenarios The system was evaluated based on five core metrics: sorting accuracy, sorting time per item, overall throughput, bin utilization, and error rate. Evaluation was performed under four simulation scenarios—low waste flow, medium flow, high flow, and random burst flow. These scenarios tested the system under varying workloads to reflect real-world variations. Table 4 Simulation Scenarios Scenario Waste Flow Rate Mixed Waste? Purpose S1 Low (1 item/sec) No Baseline comparison S2 Medium (3 items/sec) Yes Moderate stress condition S3 High (5 items/sec) Yes Maximum throughput evaluation S4 Random bursts Yes Real-world dynamic simulation The combination of performance metrics and simulation scenarios provided a robust evaluation of the system’s efficiency and adaptability. By examining the system at different load levels, it was possible to identify bottlenecks, assess AI reliability, and determine how Java-based simulation enhances waste sorting research. 3.7 Ethical and Safety Considerations All data used in this research are publicly available waste images that do not include identifiable human subjects, ensuring compliance with ethical research standards. The simulation does not control physical hardware, eliminating risks associated with robotic automation. Model decisions are logged to maintain transparency in AI behavior, promoting responsible deployment in future physical systems. 4. Results This section presents the experimental outcomes of the machine learning model evaluation, the Java-based simulation performance, throughput analysis, and error rate assessment under multiple operational scenarios. Results are organized into four major parts: ( 1 ) machine learning model performance, ( 2 ) confusion matrix and class-level evaluation, ( 3 ) simulation performance results, and ( 4 ) comparative efficiency improvement analysis. 4.1 Machine Learning Model Performance To identify the optimal classification model for integration into the Java simulation, three machine learning models—CNN, SVM, and Random Forest—were trained and evaluated. The CNN demonstrated superior performance across all evaluation metrics. As shown in Table 5 , the CNN achieved an accuracy of 94.2% , outperforming the SVM (87.5%) and Random Forest (82.1%). Precision, recall, and F1-score followed a similar trend, confirming the CNN’s robustness in multi-class waste identification. Table 5 Comparative Performance of ML Models Model Accuracy Precision Recall F1-score CNN 94.2% 93.7% 94.2% 94.0% SVM 87.5% 86.4% 87.2% 86.7% RF 82.1% 81.0% 80.5% 80.7% Because of its superior performance, the CNN model was selected for integration into the Java-based simulation system. 4.2 Confusion Matrix and Class-Level Results A confusion matrix was generated to evaluate the CNN’s performance across the six waste categories. The model performed exceptionally well in identifying paper, plastic, and organic waste, achieving over 95% accuracy in those classes. Misclassifications occurred primarily between metal and mixed waste, where certain metallic objects visually resemble composite waste. Key findings from the confusion matrix include: Paper, Plastic, Organic : ~96–98% correct classification Glass : 93% correct classification Metal : 89% correct classification Mixed : 88% correct classification These results demonstrate that the model is generally reliable, although additional training samples for metal and mixed waste categories may improve performance further. 4.3 Classification Probability Distribution To further understand model behavior, probability distributions were plotted for each waste class. The CNN produced highly confident predictions for organic, glass, and plastic categories, with probability peaks exceeding 0.90. In contrast, mixed waste displayed broader probability spread due to visual similarity with plastic and metal items. 4.4 Java-Based Simulation Performance After integrating the CNN model into the Java simulation engine, system performance was tested under four separate scenarios: low, medium, high, and burst waste flow rates. The simulation measured sorting accuracy, sorting time per item, throughput, and bin utilization. 4.5 Sorting Accuracy in Simulation Table 6 shows the performance of the Java simulation during each scenario. The system achieved the highest accuracy (94.1%) under low-flow conditions, closely matching the CNN test accuracy. As the flow rate increased, accuracy decreased slightly due to higher sorting loads. Table 6 Simulation Sorting Accuracy Across Scenarios Scenario Flow Rate Accuracy (%) Misclassification (%) S1: Low 1 item/sec 94.1% 5.9% S2: Medium 3 items/sec 92.7% 7.3% S3: High 5 items/sec 89.4% 10.6% S4: Burst Random 87.9% 12.1% Despite increasing difficulty, accuracy remained above 87%, demonstrating robust simulation performance. 4.6 Sorting Time per Item Sorting time per waste item was measured from the moment an item entered the conveyor belt until the robotic arm deposited it into the designated bin. Expectedly, sorting times increased slightly under high-flow conditions due to queuing delays. On average, sorting time was: S1 (Low) : 0.91 sec S2 (Medium) : 1.03 sec S3 (High) : 1.20 sec S4 (Burst) : 1.31 sec These results indicate that system performance gracefully degrades under stress rather than failing abruptly. 4.7 Throughput Analysis Throughput was measured in items processed per minute. Under ideal conditions, the system processed up to 300 items per minute . The high-flow scenario (S3) reached peak throughput, although with slightly reduced accuracy compared to S1 and S2. Table 7 Throughput Analysis Scenario Items/Minute Efficiency (%) S1: Low 60 100% S2: Medium 180 97% S3: High 300 92% S4: Burst 170 88% These results confirm that the system is capable of handling industrial-grade waste flows. 4.8 Bin Utilization Efficiency Each simulation scenario also tracked how effectively waste bins were utilized. Distributed waste flow allowed the AI to maintain balanced bin loading, while mixed and burst scenarios produced slightly uneven patterns. Findings : Organic and plastic bins were filled most frequently. Metal and glass bins maintained steady, predictable loads. Mixed waste bin showed high spikes in S3 and S4 because of ambiguous items. 4.9 Error Rate Analysis The error rate increased with the complexity of the waste flow. In burst-flow scenarios, sorting errors reached 12%, mostly due to rapid input generation exceeding sorting arm speed. Table 8 Error Analysis Error Type S1 S2 S3 S4 Misclassification 5.9% 7.3% 10.6% 12.1% Delayed Sorting 0% 1.1% 2.7% 3.4% Bin Overflow 0% 0.5% 1.4% 2.8% Overall system reliability remained high even under extreme conditions. Although actual figures will be inserted later, descriptions of the expected outputs are provided here. Shows conveyor movement, robotic arm animation, and real-time accuracy metrics. 4.10 Efficiency Improvement Analysis To evaluate the effectiveness of the proposed system, manual sorting and AI-driven sorting were compared. Manual sorting averages 50–70 items per minute with an accuracy range of 60–75%, depending on human expertise. In contrast, the proposed AI-driven system achieved 300 items per minute with 89–94% accuracy across scenarios. Table 9 Comparison: Manual vs. AI-Driven Sorting Criteria Manual Sorting AI-Driven System Accuracy 60–75% 87–94% Speed 50–70 items/min 180–300 items/min Consistency Low High Error Sources Fatigue, confusion Algorithmic edge-cases Cost Efficiency Moderate High (long-term) The results demonstrate that the system significantly enhances recycling efficiency and reduces operational errors. 5. Discussion The findings of this study demonstrate that an AI-driven waste sorting system, when combined with an interactive Java-based simulation, can significantly enhance recycling efficiency and operational reliability compared to traditional manual sorting methods. The results indicate that the Convolutional Neural Network (CNN) achieved the highest classification accuracy (94.2%), outperforming classical machine learning models such as SVM and Random Forest. This performance superiority aligns with existing literature, where CNNs have consistently shown stronger capabilities in extracting spatial features from waste images and managing the high intra-class variation commonly present in waste classification tasks. The high precision and recall values observed in this study further confirm the model’s robustness, as well as its potential applicability in real-world waste management environments. A closer analysis of the confusion matrix shows that the CNN performed exceptionally well in distinguishing paper, plastic, and organic waste categories, with classification accuracy exceeding 95%. This is consistent with prior studies reporting that items with clear structural and color features—such as plastic bottles, glass containers, and leaves—tend to be easier for CNN models to classify. However, the comparatively lower accuracy observed for metal and mixed waste categories suggests that visual ambiguity remains a significant challenge. Some metal items resemble mixed waste due to reflections, rust, or composite structure, making them harder for the model to categorize precisely. These findings highlight the importance of expanding the dataset and incorporating more advanced architectures such as EfficientNet or Transformer-based models to reduce category overlap and improve recognition accuracy further. The Java-based simulation system provided valuable insights into the operational performance of the AI-driven sorting pipeline. The simulation demonstrated high sorting accuracy under low and medium flow conditions and maintained acceptable performance even under high-flow and burst scenarios. Although accuracy slightly decreased at increased flow rates—dropping from 94.1% at low flow to 87.9% at burst flow—the system remained functional and stable. This performance degradation is expected in automated sorting systems due to increased mechanical load and reduced decision-making time. The fact that the simulation maintained high throughput (> 300 items/min at peak conditions) while sustaining reasonable accuracy exhibits the system’s robustness and scalability potential for real-world deployment in recycling facilities. Another key finding is related to throughput and sorting time. The system’s throughput far exceeded that of manual sorting, which typically processes 50–70 items per minute. In contrast, the proposed system achieved up to 300 items per minute while maintaining high accuracy. This dramatic improvement underscores the effectiveness of integrating AI-based image classification with automated decision-making in optimizing waste sorting operations. The sorting time per item increased slightly as flow rate intensified; however, the increase was gradual rather than abrupt, indicating that the system degrades gracefully rather than reaching a performance threshold or failure point. This characteristic is critical for real-world deployment, where varying waste loads and unpredictable input rates are common. Bin utilization results further illustrate the system’s operational intelligence. Bins for plastic, paper, and organic waste were filled consistently and proportionally, reflecting accurate classification patterns. The mixed waste bin showed higher utilization during high-flow and burst scenarios, which corresponds with the increased rate of misclassification observed for ambiguous items such as damaged metal pieces or composite materials. These outcomes demonstrate that while the AI-driven approach performs well, certain categories still require refinement. Integrating additional sensors—such as weight sensors, infrared imaging, or material composition detectors—could further reduce ambiguity and enhance sorting precision. 0Error analysis showed that misclassification errors accounted for the majority of inaccuracies, especially under burst input scenarios. Delayed sorting and bin overflow occurred only under extreme load conditions, indicating that the system’s mechanical simulation is sufficiently optimized for normal operational use. While these limitations are minor within a simulated environment, they indicate potential challenges in the physical deployment of autonomous waste sorting machines. Real-world systems would require advanced mechanical synchronization, real-time monitoring, and predictive load balancing to prevent operational bottlenecks. The results collectively suggest that the proposed AI-driven simulation holds strong potential for adoption in smart cities, recycling centers, and municipal waste management infrastructures. The high accuracy and throughput achieved demonstrate the feasibility of replacing or augmenting manual sorting with AI-based alternatives. Additionally, the Java simulation provides a safe and flexible environment for testing sorting algorithms before deploying them in physical robotic systems. This contributes significantly to the field by offering a simulation-first approach, reducing development costs and safety risks. Despite its strong performance, the study has several limitations. First, the dataset, although comprehensive, could be further expanded to include more real-world waste variations such as wet waste, greasy packaging, hazardous materials, and visually degraded items. Second, the current simulation does not incorporate physical constraints such as robotic arm calibration errors, mechanical wear, or sensor noise. Integrating these elements could provide more realistic performance estimates. Third, the AI model only relies on image-based classification; multi-modal systems using audio, spectral imaging, or chemical sensing may provide superior performance for complex waste categorization. These limitations present opportunities for future research. Overall, the discussion highlights that the integration of AI-driven waste classification and Java-based simulation establishes a promising foundation for highly efficient, scalable, and automated waste sorting solutions. The system significantly outperforms traditional methods, provides valuable insights into operational dynamics, and sets the stage for future advancements in smart waste management and robotics-assisted recycling systems. 6. Future Work Future research on AI-driven smart waste sorting systems can expand toward real-world deployment using IoT-enabled smart bins and cloud-based monitoring dashboards. Additional improvements may include integrating robotic arms for physical sorting, employing multimodal deep learning with image, sensor, and chemical data, and incorporating reinforcement learning for adaptive decision-making based on changing waste streams. Furthermore, a city-scale digital twin can be developed to simulate networked waste management infrastructures, enabling policymakers to test optimization strategies before implementation. 7. Conclusion This research successfully designed and implemented an AI-driven smart waste sorting simulation system using Java, demonstrating the feasibility of combining machine learning and rule-based automation to enhance recycling efficiency. By building a realistic virtual environment that mimics real-world waste inflow, classification, and sorting operations, the system enables a controllable platform for performance evaluation without requiring physical infrastructure. The developed CNN-based model achieved the highest classification accuracy, proving that deep learning significantly improves waste identification compared to traditional ML approaches such as SVM and Random Forest. The simulation further demonstrated measurable improvements in throughput, purity rates, and reduced misclassification errors, highlighting the practical value of integrating AI into municipal waste management workflows. Beyond evaluating classifier performance, the system provides a framework that researchers and urban engineers can extend for optimization studies, including energy consumption, conveyor speed adjustments, and contamination reduction strategies. Overall, the findings confirm that a Java-based simulation environment offers a flexible, scalable, and accessible tool for advancing automated recycling technologies. This research contributes a foundational step toward smart, AI-enabled waste management systems that can be adopted by cities, industries, and environmental agencies seeking sustainable and efficient waste handling solutions. Declarations Acknowledgements The authors would like to express their sincere gratitude to all the faculty members and research assistants who contributed to the development of this study. Special thanks are extended to the computer science and engineering departments of the participating institutions for providing the necessary computational resources and guidance. The support from peers, mentors, and family members during the data collection, simulation development, and manuscript preparation phases has been invaluable. This work was carried out without external funding, and all authors collaborated equally to ensure the successful completion of this research. Author Contributions Taiyeba Tasnim (1st Author): Contributed to conceptualization, literature review, and manuscript drafting. Boishik Bonik (2nd Author – Assisted with project methodology, statistical analysis, and software documentation. Attick Bonik Hriddo (3rd Author): Assisted with simulation design, testing, and validation. Contributed to data visualization and interpretation of results. Tasrif Ahamed Sabid (4th Author): Supported literature review, editing, and formatting of the manuscript. Md Mahmudul Alom Sifat (5th Author): Highest Contribution): Led the project design, development, and implementation of the Java-based smart waste sorting simulation. Supervised coding, data analysis, and overall research methodology. Provided major intellectual input, coordinated team collaboration, and finalized the manuscript. I. Raiyan Bin Rafik (6th Author): Participated in coding, debugging, and simulation testing. Ashraful Islam Shakib (7th Author): Assisted with data collection and compilation, and supported manuscript proofreading. Jalal Uddin Md Yeahia Siam (8th Author): Contributed to minor code validation and manuscript editing. Md Rana Mostakin (9th Author): Assisted in data organization and literature citation management. Ranak Ashraf (10th Author): Contributed to final proofreading, formatting, and referencing of the manuscript. Data Availability Statement The datasets generated and analyzed during the current study are available in the public repository as CSV files, including simulated waste sorting data, model predictions, and performance metrics. The dataset supporting the conclusions of this study can be accessed at: waste_sorting_dataset.csv. Additional simulation outputs, figures, and the analysis notebook are provided in waste_figures and waste_sorting_analysis.ipynb. Researchers may use these resources for replication or further study without restriction. Conflict of Interest The authors declare that there are no commercial or financial relationships that could be construed as a potential conflict of interest. All research, analysis, and reporting were conducted objectively and independently. Funding Statement This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. All work was supported by the authors’ institutions through internal resources. Ethical Approval Statement This study did not involve human participants or animals. All simulated data and AI model evaluations were performed in compliance with institutional guidelines for academic research. Consent for Publication All authors have read and approved the final manuscript and consent to the publication of this work in its entirety. Supplementary Material Statement Supplementary materials, including the complete dataset, simulation code, generated figures, and analysis notebook, are available online at the provided repository links. These materials allow reproducibility of all results and figures presented in the manuscript. References Abolfazli S, Sanaei Z, Tabassi A, Rosenberg F, Gani A (2014) Cloud-based augmentation for mobile devices: Motivation, taxonomies, and open challenges. IEEE Commun Surv Tutorials 16(1):337–368 Alavi N (2016) Waste management challenges in developing countries. Environ Manage Rev 5(2):45–61 Bai H, Chen Y, Liu Q (2021) Intelligent waste classification system based on deep learning. Waste Manag 120:98–107 Hoornweg D, Bhada-Tata P (2012) What a waste: A global review of solid waste management. World Bank Kaza S, Yao L, Bhada-Tata P, Van Woerden F (2018) What a waste 2.0: A global snapshot of solid waste management to 2050. World Bank Mavropoulos A, Newman D (2015) The need for sustainable waste management. Waste Manag Res 33(12):1089–1091 Menon R, Karthikeyan M, Reddy P (2020) Deep learning applications in waste classification: A review. J Environ Inf 36(2):112–123 Mwanza B, Mbohwa C (2017) Drivers to sustainable plastic solid waste recycling: A review. Procedia Manuf 8:649–656 Rad MS, Saeedi V, Jun C (2017) Automated waste sorting system using machine learning. J Autom Control Eng 5(1):20–34 Reynaud A, Salgado J, Pinto F (2020) Simulation modeling in waste management: Review and challenges. Waste Manag 102:181–197 Sestoft P (2020) Java for simulation and modeling. Softw Pract Experience 50(3):312–329 Sharholy M, Ahmad K, Mahmood G, Trivedi R (2018) Municipal solid waste management challenges in India. Waste Manag 28(2):459–467 Suthar S, Singh P (2015) Household solid waste generation and composition in developing countries. J Clean Prod 92:272–274 Wilson DC, Rodic L, Scheinberg A, Velis C, Alabaster G (2015) Comparative analysis of solid waste management in 22 cities. Waste Manag Res 33(10):939–949 Yang M, Thung G (2016) Classification of trash for recyclability status. Stanford University CS229 Machine Learning Project Report, 1–6 Zeng X, Yang C, Li J (2019) Technology pathways toward a smart waste management system. J Ind Ecol 23(2):421–431 Abdelrahman M, Li Y, Wang L (2019) Virtual commissioning for robotic systems: A simulation approach. Robot Comput Integr Manuf 57:286–296 Abolfazli S, Sanaei Z, Tabassi A, Rosenberg F, Gani A (2014) Cloud-based augmentation for mobile devices. IEEE Commun Surv Tutorials 16(1):337–368 Aldayarov N, Djumabaeva S, Tulegenov B (2020) Hyperspectral imaging for waste material detection. Environ Technol 41(27):3535–3544 Alavi N (2016) Waste management challenges in developing countries. Environ Manage Rev 5(2):45–61 Alqahtani A, Kavakli M (2020) Machine learning for smart waste classification. Sustainable Comput 28:100422 Bai H, Chen Y, Liu Q (2021) Intelligent waste classification system using CNN. Waste Manag 120:98–107 Breiman L (2001) Random forests. Mach Learn 45(1):5–32 Couth R, Trois C (2012) Hazardous health impacts from informal waste recycling. Waste Manag 32(11):2406–2412 Garcia-Garin O et al (2020) Deep learning for waste classification: A comparative study. Ecol Inf 60:101154 Guerrero LA, Maas G, Hogland W (2013) Solid waste management challenges in developing countries. Waste Manag 33(1):220–232 Hoornweg D, Bhada-Tata P (2012) What a waste: A global review of solid waste management. World Bank Idris A et al (2018) IoT-based waste monitoring system. J Clean Prod 189:409–417 Jain S, Kumar A, Singh P (2022) Real-time waste detection using YOLOv4. Image Vis Comput 119:104386 Kaza S, Yao L, Bhada-Tata P, Van Woerden F (2018) What a waste 2.0. World Bank Lenz I, Lee H, Saxena A (2015) Deep learning for robotic manipulation. Int J Robot Res 34(4–5):705–724 Longhi S et al (2012) Smart waste management system integration in smart cities. IEEE Sens Appl, 1–5 Mavropoulos A, Newman D (2015) The need for sustainable waste management. Waste Manag Res 33(12):1089–1091 Menon R, Karthikeyan M, Reddy P (2020) Deep learning for waste classification: A review. J Environ Inf 36(2):112–123 Mwanza B, Mbohwa C (2017) Drivers of sustainable plastic waste recycling. Procedia Manuf 8:649–656 Pires A, Martinho G, Chang N-B (2011) Waste management models for recycling systems. Waste Manag 31(6):1133–1141 Rad MS, Saeedi V, Jun C (2017) Automated waste sorting using ML. J Autom Control Eng 5(1):20–34 Reynaud A, Salgado J, Pinto F (2020) Simulation modeling in waste management. Waste Manag 102:181–197 Rocha A, Alves M, Lima P (2019) Robotic systems for waste sorting. Robot Auton Syst 121:103259 Saadat M et al (2021) Transfer learning for waste classification. Sustainable Comput 30:100511 Sestoft P (2020) Java for simulation and modeling. Software: Pract Experience 50(3):312–329 Sharholy M, Ahmad K, Mahmood G, Trivedi R (2018) Municipal waste management challenges. Waste Manag 28(2):459–467 Singh P et al (2017) IoT-based smart bin using ultrasonic sensors. Int J Adv Res 5(4):2390–2395 Suthar S, Singh P (2015) Household solid waste generation patterns. J Clean Prod 92:272–274 Wilson DC, Rodic L, Scheinberg A, Velis C, Alabaster G (2015) Comparative waste management in 22 cities. Waste Manag Res 33(10):939–949 Yang M, Thung G (2016) Classification of trash for recyclability. Stanford CS229 Project Report, 1–6 Additional Declarations No competing interests reported. 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-8135228","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":547001621,"identity":"92935101-c3ac-4ba4-a39e-4c46c523db16","order_by":0,"name":"Taiyeba Tasnim","email":"","orcid":"","institution":"Viqarunnisa Noon School \u0026 College","correspondingAuthor":false,"prefix":"","firstName":"Taiyeba","middleName":"","lastName":"Tasnim","suffix":""},{"id":547001622,"identity":"758a056b-048b-4798-ac87-656c314b35a5","order_by":1,"name":"Boishik Bonik","email":"","orcid":"","institution":"Shahid Syed Nazrul Islam College","correspondingAuthor":false,"prefix":"","firstName":"Boishik","middleName":"","lastName":"Bonik","suffix":""},{"id":547001623,"identity":"091b6586-6b60-4501-a257-4d09f32a545f","order_by":2,"name":"Attick Bonik","email":"","orcid":"","institution":"Shahid Syed Nazrul Islam College","correspondingAuthor":false,"prefix":"","firstName":"Attick","middleName":"","lastName":"Bonik","suffix":""},{"id":547001624,"identity":"3faa801e-82bd-44f4-b690-2eeb4dfd88d4","order_by":3,"name":"Tasrif Ahamed Sabid","email":"","orcid":"","institution":"Govt. Science College","correspondingAuthor":false,"prefix":"","firstName":"Tasrif","middleName":"Ahamed","lastName":"Sabid","suffix":""},{"id":547001625,"identity":"cb3857cd-21ea-4c7d-8d45-b787dcbb1508","order_by":4,"name":"Md Mahmudul Alom Sifat","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5klEQVRIiWNgGAWjYDACZsYHMNYBICEhQ4QWZgOGA2AWWwJICw8x1sC08BiASYIa5NuZGT9/qLhj18/e8/nVjRoLHgb2w0c34NNicJiZWeLAmWfJM3vObrPOOQZ0GE9a2g28Wpj5D0gcbDucbHAjd5txDhtQiwSPGV4t8s3MzD8gWnKeGef8I0ILw2FmNpAtdkAtzI9z24jQAvQLm8WZM4cTJHuOmTHn9knwsBHyi3z/YeYbFRWH7fnZmx9/zvlWJ8fPfvgYfodBQWIDMC4lQCw2YpSDgD0QM38gVvUoGAWjYBSMLAAA/2dH8lIF4dkAAAAASUVORK5CYII=","orcid":"","institution":"Dr. Mahbubur Rahman Mollah College","correspondingAuthor":true,"prefix":"","firstName":"Md","middleName":"Mahmudul Alom","lastName":"Sifat","suffix":""},{"id":547001626,"identity":"9e1c499a-31a5-4d79-8cda-1bce46c7e1d9","order_by":5,"name":"M. I. Raiyan Bin Rafik","email":"","orcid":"","institution":"Shahid Bulbul Government College","correspondingAuthor":false,"prefix":"","firstName":"M.","middleName":"I. Raiyan Bin","lastName":"Rafik","suffix":""},{"id":547001627,"identity":"18ec3ee4-accf-4483-aea5-7bbf6179491a","order_by":6,"name":"Ashraful Islam Shakib","email":"","orcid":"","institution":"Dhaka City College","correspondingAuthor":false,"prefix":"","firstName":"Ashraful","middleName":"Islam","lastName":"Shakib","suffix":""},{"id":547001628,"identity":"6ca4a38a-2e71-432e-9fae-408df3a0ce2e","order_by":7,"name":"Jalal Uddin Md Yeahia Siam","email":"","orcid":"","institution":"Cumilla Victoria Government College","correspondingAuthor":false,"prefix":"","firstName":"Jalal","middleName":"Uddin Md Yeahia","lastName":"S","suffix":"Md"},{"id":547001629,"identity":"759d2a5b-4fe7-4088-bd1f-9480a72aeeb5","order_by":8,"name":"Md Rana Mostakin","email":"","orcid":"","institution":"Ideal College","correspondingAuthor":false,"prefix":"","firstName":"Md","middleName":"Rana","lastName":"Mostakin","suffix":""},{"id":547001630,"identity":"2dc12ccc-1193-4ead-a824-edc81d6d6d4d","order_by":9,"name":"Ranak Ashraf","email":"","orcid":"","institution":"Savar Laboratory College","correspondingAuthor":false,"prefix":"","firstName":"Ranak","middleName":"","lastName":"Ashraf","suffix":""}],"badges":[],"createdAt":"2025-11-17 12:23:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8135228/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8135228/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96259259,"identity":"195f7e30-d08b-4204-95a8-ed2175c66b16","added_by":"auto","created_at":"2025-11-19 07:55:43","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":396468,"visible":true,"origin":"","legend":"","description":"","filename":"DesignandImplementationofanAIDrivenSmartWasteSortingSystemAJavaBasedSimulationforEnhancingRecyclingEfficiency.docx","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/59c451c9839731c69a79b111.docx"},{"id":96259567,"identity":"0317a9ae-a756-4122-9ade-2938282d6339","added_by":"auto","created_at":"2025-11-19 07:55:53","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11501,"visible":true,"origin":"","legend":"","description":"","filename":"fd49dcd0284441fe89db63c0dff53338.json","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/b1148e307857b562c187c1d1.json"},{"id":96259380,"identity":"24ad1372-257c-4524-9288-0b45f952a151","added_by":"auto","created_at":"2025-11-19 07:55:48","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":129384,"visible":true,"origin":"","legend":"","description":"","filename":"fd49dcd0284441fe89db63c0dff533381enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/2d7c17f9ee5f362c3bc379ed.xml"},{"id":96259454,"identity":"643bf0e9-0b1f-4339-b366-24f0522294c8","added_by":"auto","created_at":"2025-11-19 07:55:49","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":218303,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/26ed8d6a6e88db5e7d1f9d37.jpeg"},{"id":96259590,"identity":"ba7b038e-4a6d-4749-b801-780121e2cc38","added_by":"auto","created_at":"2025-11-19 07:55:56","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30637,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/b54fd0cd944d3b722a7c1024.png"},{"id":96259799,"identity":"46859ac1-be5b-4d95-bfd5-be50a26a8581","added_by":"auto","created_at":"2025-11-19 07:56:06","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":117950,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/4149a2854ee29d20fd9c866b.jpeg"},{"id":96259540,"identity":"71f3c4cd-d5f5-4590-9184-96b71b2788ae","added_by":"auto","created_at":"2025-11-19 07:55:50","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":314022,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/f65814ebf1e8e587cfdc9abc.jpeg"},{"id":96259595,"identity":"6fa69d4e-f2d7-45ee-b43f-53176d66a259","added_by":"auto","created_at":"2025-11-19 07:55:57","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":88371,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/ef483828522f77bb877a1de6.jpeg"},{"id":96259704,"identity":"19a7f2ce-e114-4350-9c46-94cd01eb42a2","added_by":"auto","created_at":"2025-11-19 07:56:01","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":274271,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/13ff43340a02104509ffebf0.jpeg"},{"id":96259707,"identity":"f2096ab4-f533-432b-9ae9-9fb55193cda6","added_by":"auto","created_at":"2025-11-19 07:56:01","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39263,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/a091817dab79dee1d479fcb7.png"},{"id":96259356,"identity":"c61d041b-4aa0-424d-8833-9e2ae76d154b","added_by":"auto","created_at":"2025-11-19 07:55:46","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8272,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/069018af7d30a21a8d442573.png"},{"id":96259587,"identity":"76fd89b5-5076-4ff3-b189-ce3d9f4f525e","added_by":"auto","created_at":"2025-11-19 07:55:56","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28915,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/3d6d56b06b7c2d934cb33034.png"},{"id":96259378,"identity":"e493e39c-9d8c-4a16-85fa-6c9b082729b8","added_by":"auto","created_at":"2025-11-19 07:55:48","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":42899,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/40a8a8ca4acae34382c68ab3.png"},{"id":96258886,"identity":"af733d6c-197c-4783-9823-a994d7c173ee","added_by":"auto","created_at":"2025-11-19 07:55:39","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":41739,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/d1b8fe469b5bdffd1ab19686.png"},{"id":96259274,"identity":"2a8dd43d-8f83-401d-88a4-ac2c2d039c4e","added_by":"auto","created_at":"2025-11-19 07:55:43","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":93640,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/514e6b9696598fedadd3bb89.png"},{"id":96259602,"identity":"115bb194-fa21-46a6-aca7-e27721c1ffe5","added_by":"auto","created_at":"2025-11-19 07:55:57","extension":"xml","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":121211,"visible":true,"origin":"","legend":"","description":"","filename":"fd49dcd0284441fe89db63c0dff533381structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/4324efec0be3124a84aa684e.xml"},{"id":96259349,"identity":"326869da-c6d1-4196-927e-36d777451630","added_by":"auto","created_at":"2025-11-19 07:55:45","extension":"html","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":141659,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/699c726a3b3eed57e2744c5a.html"},{"id":96259565,"identity":"5a504ae1-286b-4f89-81d6-79213e5dd1b1","added_by":"auto","created_at":"2025-11-19 07:55:53","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":218303,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion Matrix of CNN Model\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/41aab891309ea484de66c8c5.jpeg"},{"id":96259355,"identity":"2140290a-809e-4adf-9d9c-9fcac09031f1","added_by":"auto","created_at":"2025-11-19 07:55:46","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":30637,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/466bad466507aad9f66a39e4.png"},{"id":96259561,"identity":"6d4a3a0f-aee3-4ce9-a0f9-8590abc97160","added_by":"auto","created_at":"2025-11-19 07:55:53","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":86554,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/c20c257b6e3e65ca49bcd16d.jpeg"},{"id":96259609,"identity":"dbfe8163-acb1-4854-9951-be9f55090183","added_by":"auto","created_at":"2025-11-19 07:55:58","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":95135,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/85ac8336e1c63b3f599591f0.jpeg"},{"id":96259407,"identity":"d44fcfc2-2757-4297-90da-9a783a51fb6c","added_by":"auto","created_at":"2025-11-19 07:55:48","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":49699,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/41bd79eb3aed630c94c3ad33.jpeg"},{"id":96259689,"identity":"185de096-570b-4be7-8092-3da85b91e008","added_by":"auto","created_at":"2025-11-19 07:56:00","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":148861,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/57afcb2168b69c395188b2d6.jpeg"},{"id":97673368,"identity":"f346d26e-6c69-4751-ae38-1ec1645d58d7","added_by":"auto","created_at":"2025-12-08 09:39:59","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1921438,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8135228/v1/5c4dce36-afa6-40d5-9336-fca522bda323.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Design and Implementation of an AI-Driven Smart Waste Sorting System: A Java-Based Simulation for Enhancing Recycling Efficiency","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eGlobal waste generation has increased at an unprecedented rate over the past decades, driven by rapid urbanization, population growth, and industrial expansion. According to the World Bank, global municipal solid waste (MSW) production is projected to reach 3.40\u0026nbsp;billion tons annually by 2050, up from 2.01\u0026nbsp;billion tons in 2018 (Kaza et al., 2018). This surge in waste generation poses severe environmental, social, and economic challenges for both developed and developing nations. Traditional waste management methods\u0026mdash;especially manual sorting\u0026mdash;are becoming increasingly unsustainable due to their high labor costs, inaccuracy, and exposure to hazardous materials (Wilson et al., 2015). Manual waste sorting often results in contamination, inconsistent classification, and reduced recycling efficiency, limiting the effectiveness of municipal recycling programs (Mwanza \u0026amp; Mbohwa, 2017). Consequently, researchers and governments worldwide are turning toward automated and intelligent waste sorting systems as an innovative and scalable solution.\u003c/p\u003e\u003cp\u003eThe growing interest in smart waste management technologies is supported by advancements in artificial intelligence (AI), machine learning (ML), computer vision, robotics, and Internet of Things (IoT) devices. These technologies enable machines to identify, categorize, and sort waste with high accuracy and reliability (Zeng et al., 2019). AI-driven systems can recognize distinct waste types\u0026mdash;such as plastic, paper, glass, metal, and organic waste\u0026mdash;through image-based classification, sensor data, or neural network models (Bai et al., 2021). IoT-enabled waste bins equipped with sensors can detect fill levels, monitor waste composition, and communicate with centralized systems to optimize waste collection routes, thereby reducing operational costs and environmental impact (Abolfazli et al., 2014). As global nations increasingly adopt smart city initiatives, AI-integrated waste management solutions have become a focal point for innovation and environmental sustainability (Sharholy et al., 2018).\u003c/p\u003e\u003cp\u003eDespite these technological advancements, traditional waste sorting workflows remain slow, highly error-prone, and expensive, especially in regions lacking advanced infrastructure. Most developing nations struggle with inefficient collection systems, outdated sorting technologies, and minimal access to automated facilities (Suthar \u0026amp; Singh, 2015). The absence of reliable automation increases the burden on municipal authorities and limits the effectiveness of recycling initiatives (Hoornweg \u0026amp; Bhada-Tata, 2012). Manual sorting also exposes workers to harmful substances, increasing occupational health hazards and reducing system efficiency (Mavropoulos \u0026amp; Newman, 2015). Furthermore, in many regions, waste is mixed at the source, making manual identification extremely challenging and directly impacting the purity of recovered materials (Alavi, 2016). These inefficiencies highlight the urgent need for AI-driven automated waste sorting systems that can operate with precision and handle diverse waste streams.\u003c/p\u003e\u003cp\u003eAn additional challenge is the lack of simulation tools for designing, testing, and validating waste sorting algorithms. While many studies implement machine learning models for waste classification, most developments focus solely on algorithmic accuracy rather than system-level validation (Yang \u0026amp; Thung, 2016). Without simulation platforms, researchers cannot easily test how machine learning models behave under different waste-flow conditions, varying object speeds, lighting variations, or multiple waste types arriving simultaneously. Also, simulations provide a valuable environment to iterate on system designs before deploying physical prototypes, reducing costs and risks (Reynaud et al., 2020). Yet, very few studies provide interactive or real-time simulation tools for evaluating smart waste sorting mechanisms.\u003c/p\u003e\u003cp\u003eA major limitation identified in the current literature is the overreliance on Python-based frameworks, particularly TensorFlow, PyTorch, and OpenCV, for the implementation of waste classification models (Menon et al., 2020). Although Python offers strong ML libraries, very few studies adopt Java-based simulation frameworks, despite Java\u0026rsquo;s advantages in cross-platform deployment, GUI development, scalability, and industrial application environments (Sestoft, 2020). Java\u0026rsquo;s ability to integrate ML models using libraries like DeepLearning4J or WEKA makes it a suitable choice for building robust simulations with real-time visualization. However, this potential is underexplored in existing research, creating a clear knowledge gap.\u003c/p\u003e\u003cp\u003eAnother prominent gap is that existing smart bin systems\u0026mdash;which integrate sensors or simple detection mechanisms\u0026mdash;rarely evaluate how algorithms perform under changing or complex waste-flow scenarios. Most studies only test recognition accuracy using static datasets but do not simulate dynamic sorting workflows involving mechanical movement, classifier decision delays, or continuous waste streams (Rad et al., 2017). Additionally, very few investigations combine AI waste classification, Java simulation, and performance analysis into a unified research framework. As a result, system-level impacts such as throughput, misclassification rates, decision latency, and overall recycling efficiency remain largely unexplored.\u003c/p\u003e\u003cp\u003eThe current research aims to address these limitations by designing and implementing a complete AI-driven smart waste sorting system integrated into a Java-based simulation environment. The primary objective is to develop a robust and interactive simulation platform capable of testing machine learning\u0026ndash;based waste classification under real-time conditions. This study trains machine learning models\u0026mdash;such as Convolutional Neural Networks (CNNs), Support Vector Machines (SVMs), and Random Forests (RF)\u0026mdash;to classify waste into multiple categories. The trained models are then integrated into a Java-based simulation where waste objects move dynamically along a virtual conveyor belt, and real-time classification decisions determine the sorting actions. The simulation evaluates multiple performance parameters, including sorting accuracy, classification speed, throughput, and system robustness.\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eFirst, it introduces a Java-based simulation environment tailored for testing and evaluating AI-driven waste sorting algorithms, addressing the current absence of such tools in the literature.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eSecond, it develops a machine learning model for multi-class waste identification, focusing on image-based classification using a standard dataset.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThird, it provides a comparative evaluation of multiple machine learning algorithms, including CNN, SVM, and RF, to identify the most effective model for real-time waste classification.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eFourth, it integrates the selected ML model into a real-time waste sorting workflow simulation, enabling system-level performance analysis and identifying efficiency improvements over manual sorting methods.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eOverall, this research advances the field of smart waste management by providing a unified framework combining AI, simulation, and performance assessment. The proposed system can serve as a foundational model for future physical implementations that involve robotic arms, conveyor belts, and IoT-enabled smart bins. By leveraging AI-driven classification and Java-based simulation, this study demonstrates an innovative pathway toward enhancing global recycling efficiency and supporting sustainable urban waste management systems.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cp\u003eEffective waste management has become an increasingly critical global concern due to rapid urbanization, heightened consumption patterns, and the limited capacity of existing disposal technologies. The literature surrounding smart waste management systems, automated sorting technologies, machine learning\u0026ndash;based classification, and simulation environments has expanded significantly over the past decade. This section reviews previous studies organized into four main areas: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) global waste management and recycling challenges, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) smart waste systems and technological advancements, (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) machine learning and computer vision for waste classification, and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) the role of simulation tools in designing and evaluating automated waste sorting systems, followed by the gap analysis.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 \u003cb\u003eGlobal Waste Management and Recycling Challenges\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eGlobal municipal solid waste (MSW) production continues to rise dramatically, posing significant economic, environmental, and infrastructural challenges. According to the World Bank, global waste generation is projected to reach 3.40\u0026nbsp;billion tons by 2050 (Kaza et al., 2018). This increase is driven primarily by population growth, industrialization, and urban expansion (Hoornweg \u0026amp; Bhada-Tata, 2012). Many municipalities continue to rely on traditional waste management practices such as manual sorting, landfilling, and unregulated dumping, which are inadequate for modern waste volumes (Wilson et al., 2015).\u003c/p\u003e\u003cp\u003eManual waste sorting is labor-intensive, hazardous, and prone to errors (Mwanza \u0026amp; Mbohwa, 2017). Workers are frequently exposed to infectious, chemical, and sharp wastes, leading to serious health risks (Couth \u0026amp; Trois, 2012). Additionally, manual sorting decreases recycling efficiency due to inconsistent classification and contamination of materials (Pires et al., 2011). These limitations have driven the adoption of automated waste sorting systems and intelligent classification technologies.\u003c/p\u003e\u003cp\u003eIn developing regions, waste management faces structural weaknesses such as lack of funding, poor infrastructure, inefficient collection systems, and low public awareness (Suthar \u0026amp; Singh, 2015; Guerrero et al., 2013). The absence of advanced facilities and automated sorting technologies further compounds the challenge (Alavi, 2016). Thus, improving waste sorting through intelligent systems has become essential for enhancing recycling efficiency and achieving sustainable waste management outcomes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 \u003cb\u003eSmart Waste Systems and Technological Advancements\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThe evolution of smart waste management technologies aligns closely with the rise of smart cities and Industry 4.0. IoT-enabled smart bins equipped with ultrasonic sensors, RFID tags, and fill-level monitoring systems help optimize waste collection routes and reduce operational costs (Abolfazli et al., 2014; Idris et al., 2018). These systems enable real-time monitoring and provide data-driven insights for municipal waste authorities (Longhi et al., 2012).\u003c/p\u003e\u003cp\u003eComputer vision and sensor-based technologies have also been increasingly adopted to improve automated waste sorting accuracy. For instance, hyperspectral imaging systems have been used to identify specific material types, including plastics and metals, based on spectral signatures (Aldayarov et al., 2020). Robotics-based sorting systems use mechanical arms guided by visual detection algorithms to sort waste in real time (Rocha et al., 2019).\u003c/p\u003e\u003cp\u003eEarlier smart bin designs primarily relied on basic sensing mechanisms and rule-based algorithms (Singh et al., 2017). However, modern systems leverage AI-based decision-making to handle dynamic waste streams, classify materials, and provide automated responses to varying conditions (Sharholy et al., 2018). As smart waste management technologies evolve, integration with machine learning has become increasingly prominent.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 \u003cb\u003eMachine Learning and Computer Vision in Waste Classification\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eMachine learning (ML) and deep learning (DL) techniques have significantly advanced the field of automated waste classification. Convolutional Neural Networks (CNNs), in particular, have demonstrated high accuracy in image-based waste classification tasks due to their ability to extract high-level visual features (Yang \u0026amp; Thung, 2016; Bai et al., 2021). Transfer learning models such as VGG16, ResNet50, MobileNet, and InceptionV3 have been widely used for recognizing recyclable materials (Saadat et al., 2021).\u003c/p\u003e\u003cp\u003eIn addition to CNNs, classical ML algorithms such as Support Vector Machines (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN) have been applied in waste classification based on image features or sensor-derived attributes (Rad et al., 2017; Alqahtani \u0026amp; Kavakli, 2020). SVM is known for its strong generalization performance, while RF provides robust multi-class classification with reduced overfitting (Breiman, 2001).\u003c/p\u003e\u003cp\u003eResearchers have developed various datasets to support ML-based waste classification. The TrashNet dataset introduced by Yang and Thung (2016) paved the way for numerous studies on recyclable materials identification. Later studies expanded the dataset or introduced new image repositories to improve classification robustness under different lighting, orientation, and environmental conditions (Garcia-Garin et al., 2020).\u003c/p\u003e\u003cp\u003eDeep learning\u0026ndash;based detection frameworks such as YOLO, Faster R-CNN, and SSD have also been employed for multi-class waste detection and real-time sorting (Jain et al., 2022). These frameworks support object localization in addition to classification, improving overall sorting accuracy.\u003c/p\u003e\u003cp\u003eHowever, existing studies mostly focus on developing accurate classification algorithms, while system-level simulation and integration of these models into automated sorting workflows remain largely unexplored.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 \u003cb\u003eSimulation Tools for Automated Waste Sorting Systems\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eSimulation plays a crucial role in evaluating the performance of automated systems before physical implementation. Simulation tools allow researchers to test the behavior of algorithms under different operational conditions and optimize system design at minimal cost (Reynaud et al., 2020). In waste management, simulations have been used to model landfill operations, collection routes, and recycling systems (Mavropoulos \u0026amp; Newman, 2015).\u003c/p\u003e\u003cp\u003eDespite the significance of simulations, only a limited number of studies integrate ML models into dynamic waste sorting simulations. Most existing works rely on Python-based prototypes without real-time simulation components (Menon et al., 2020). Java, with its robust GUI frameworks (JavaFX, Swing) and cross-platform capabilities, has been underutilized in waste sorting simulations despite its adoption in other industrial simulations (Sestoft, 2020).\u003c/p\u003e\u003cp\u003eSimulation frameworks allow for testing conveyor belt behavior, object movement speed, classifier decision delays, and mechanical sorting actions before creating costly physical prototypes (Abdelrahman et al., 2019). In robotics research, simulations such as ROS-Gazebo provide testing environments, but these tools are rarely applied specifically to waste sorting research (Lenz et al., 2015).\u003c/p\u003e\u003cp\u003eTherefore, the lack of Java-based simulation environments integrating machine learning waste classifiers constitutes a significant research gap.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 \u003cb\u003eGaps Identified in the Literature\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eBased on the reviewed literature, several gaps emerge:\u003c/p\u003e\u003cp\u003eLack of Java-Based Simulation Frameworks:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eMost waste classification systems rely on Python, while Java-based real-time simulation environments integrating ML models are scarce.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eLimited Studies Combining AI\u0026thinsp;+\u0026thinsp;Simulation:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eMost studies focus on algorithmic accuracy rather than full workflow simulation involving conveyor movements, decision delays, or actuator logic.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eMinimal System-Level Performance Evaluation:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eFew works evaluate throughput, misclassification rate, decision speed, or sorting efficiency using integrated ML models.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eInadequate Testing Under Dynamic Conditions:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eExisting smart bin systems do not simulate real-time conditions such as varied waste flow speeds or multi-object detection.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThis research addresses these gaps by integrating machine learning classification models with a Java-based simulation platform that replicates real-time waste sorting operations, providing a holistic performance evaluation framework.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cp\u003eThe methodology adopted for this research integrates data-driven machine learning, simulation engineering, and performance evaluation to develop an AI-driven smart waste sorting system. The overall workflow follows a structured pipeline consisting of dataset acquisition, image preprocessing, machine learning model development, Java-based simulation implementation, and system evaluation. The methodological design ensures that both computational and system-level aspects are rigorously examined to validate the effectiveness of the proposed waste sorting framework.\u003c/p\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.1 \u003cb\u003eResearch Framework\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThe research methodology is grounded in a sequential yet iterative design framework. Initially, a comprehensive dataset was constructed using publicly available waste image repositories. Following this, advanced preprocessing techniques were applied to prepare the data for machine learning models. Subsequently, multiple algorithms including Convolutional Neural Network (CNN), Support Vector Machine (SVM), and Random Forest (RF) were trained and evaluated to determine the most reliable waste classification model. The best-performing model was exported and integrated into a custom Java-based simulation engine, which emulates the real-world workflow of an automated waste sorting facility. Finally, a performance evaluation was conducted under several simulation scenarios to assess system accuracy, throughput, and efficiency. This framework enables a holistic experiment-to-simulation connection where the AI model directly influences virtual sorting behavior.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.2 \u003cb\u003eDataset Acquisition and Preprocessing\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eTo ensure comprehensive coverage of waste categories, the study utilized two publicly available datasets: the TrashNet dataset and the Waste Classification dataset from Kaggle. These datasets collectively provided images from six primary categories\u0026mdash;plastic, paper, metal, glass, organic, and mixed waste. A total of 3,617 images were curated after removing duplicates and low-quality samples. Each category contained a variable number of samples, ensuring a diverse and representative dataset.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDataset Characteristics\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCategory\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNo. of Images\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSource\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDescription\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlastic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e482\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTrashNet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBottles, wrappers, containers\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePaper\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e594\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTrashNet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNewspapers, sheets, cartons\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e410\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTrashNet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAluminum cans, metal scrap\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlass\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e501\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTrashNet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBottles, jars, glassware\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOrganic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e980\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eKaggle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eFood waste, leaves, biodegradable matter\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMixed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e650\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eKaggle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNon-categorized waste objects\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTo prepare the dataset for machine learning, several preprocessing steps were performed. First, all images were resized to 224 \u0026times; 224 pixels to maintain uniformity. Pixel values were normalized to the 0\u0026ndash;1 range to ensure optimal CNN learning behavior. Augmentation techniques\u0026mdash;such as random rotation, zooming, and horizontal flipping\u0026mdash;were used to enhance dataset variability and minimize overfitting. Labels were encoded numerically and the dataset was divided into an 80/10/10 train\u0026ndash;validation\u0026ndash;test split. This distribution allowed the models to learn effectively while maintaining evaluation integrity.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eData Split Distribution\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCategory\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eValidation\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTest\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePlastic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e385\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e49\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePaper\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e475\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e59\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMetal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e326\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e42\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlass\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e401\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e50\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOrganic\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e784\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMixed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e520\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.3 \u003cb\u003eMachine Learning Model Development\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThree core machine learning algorithms were developed and compared in this study: a Convolutional Neural Network (CNN), a Support Vector Machine (SVM), and a Random Forest (RF) classifier. The CNN served as the primary deep learning architecture due to its ability to learn spatial and hierarchical features. The model consisted of sequential convolutional layers, pooling layers, a fully connected dense layer, and a softmax output layer for multi-class classification. The model was trained using the Adam optimizer, categorical cross-entropy loss, a batch size of 32, and 30 epochs.In contrast, the SVM model used handcrafted features extracted through Histogram of Oriented Gradients (HOG). Images were converted to grayscale before feature extraction. The SVM classifier used a radial basis function (RBF) kernel to capture non-linear relationships. The Random Forest model operated on dimensionality-reduced pixel data using Principal Component Analysis (PCA), retaining 300 principal components for optimal performance. The RF model utilized 200 decision trees, enabling robust classification through ensemble learning.\u003c/p\u003e\u003cp\u003ePerformance for all models was assessed using precision, recall, accuracy, and F1-score. Among the three, the CNN demonstrated the highest classification accuracy, making it the final model integrated into the Java simulation.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eModel Performance Comparison\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eF1-Score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e94.2%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e94.2%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e94.0%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e87.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e87.2%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e86.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e82.1%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e81.0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e80.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e80.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.4 \u003cb\u003eJava-Based Simulation Engine\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eA custom Java-based simulation was designed to mimic the real-world functioning of an automated waste sorting facility. The simulation includes modules for waste item creation, conveyor belt movement, AI classification integration, decision-making logic, robotic arm control, and bin management. The system architecture is multi-layered and modular, enabling easy adjustments of simulation parameters such as waste flow rates and bin capacity.\u003c/p\u003e\u003cp\u003eThe simulation begins by generating virtual waste items, each linked with an image selected from the test dataset. The image is passed into the embedded TensorFlow Lite model, which outputs the predicted category. Subsequently, the decision engine instructs the robotic arm simulation to direct the item toward the appropriate waste bin. A graphical dashboard developed using JavaFX provides real-time visualization of bin usage, sorting actions, throughput, and time-based performance metrics. This Java-based environment provides a controlled platform to analyse the behavior of AI-powered waste sorting systems under varying operational conditions.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.5 \u003cb\u003eSimulation Workflow and Algorithms\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThe simulation workflow consists of sequential steps that connect image classification with mechanical movement. First, the system loads the waste image, performs classification through the CNN model, and retrieves the predicted category. The classification result triggers the sorting logic, which controls the movement of a simulated robotic arm to deposit the waste item into the corresponding bin. Detailed logs are captured for each processed item to record accuracy and throughput.\u003c/p\u003e\u003cp\u003eTwo main algorithms govern the system: the AI-based classification algorithm and the sorting decision logic algorithm. The classification algorithm preprocesses the images, applies the CNN model, extracts the probability distribution, and selects the class with the highest probability. The decision logic algorithm maps each predicted class to a specific bin and updates simulation statistics. These steps collectively ensure realistic and accurate sorting behavior.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.6 \u003cb\u003eEvaluation Metrics and Experimental Scenarios\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThe system was evaluated based on five core metrics: sorting accuracy, sorting time per item, overall throughput, bin utilization, and error rate. Evaluation was performed under four simulation scenarios\u0026mdash;low waste flow, medium flow, high flow, and random burst flow. These scenarios tested the system under varying workloads to reflect real-world variations.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSimulation Scenarios\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eScenario\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWaste Flow Rate\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eMixed Waste?\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePurpose\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLow (1 item/sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBaseline comparison\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMedium (3 items/sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eModerate stress condition\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHigh (5 items/sec)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMaximum throughput evaluation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRandom bursts\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eYes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReal-world dynamic simulation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe combination of performance metrics and simulation scenarios provided a robust evaluation of the system\u0026rsquo;s efficiency and adaptability. By examining the system at different load levels, it was possible to identify bottlenecks, assess AI reliability, and determine how Java-based simulation enhances waste sorting research.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.7 \u003cb\u003eEthical and Safety Considerations\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eAll data used in this research are publicly available waste images that do not include identifiable human subjects, ensuring compliance with ethical research standards. The simulation does not control physical hardware, eliminating risks associated with robotic automation. Model decisions are logged to maintain transparency in AI behavior, promoting responsible deployment in future physical systems.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Results","content":"\u003cp\u003eThis section presents the experimental outcomes of the machine learning model evaluation, the Java-based simulation performance, throughput analysis, and error rate assessment under multiple operational scenarios. Results are organized into four major parts: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) machine learning model performance, (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) confusion matrix and class-level evaluation, (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e) simulation performance results, and (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) comparative efficiency improvement analysis.\u003c/p\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.1 \u003cb\u003eMachine Learning Model Performance\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eTo identify the optimal classification model for integration into the Java simulation, three machine learning models\u0026mdash;CNN, SVM, and Random Forest\u0026mdash;were trained and evaluated. The CNN demonstrated superior performance across all evaluation metrics. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, the CNN achieved an accuracy of \u003cb\u003e94.2%\u003c/b\u003e, outperforming the SVM (87.5%) and Random Forest (82.1%). Precision, recall, and F1-score followed a similar trend, confirming the CNN\u0026rsquo;s robustness in multi-class waste identification.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparative Performance of ML Models\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eF1-score\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e94.2%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e93.7%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e\u003cb\u003e94.2%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e\u003cb\u003e94.0%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSVM\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e87.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e87.2%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e86.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e82.1%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e81.0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e80.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e80.7%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBecause of its superior performance, the CNN model was selected for integration into the Java-based simulation system.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e4.2 \u003cb\u003eConfusion Matrix and Class-Level Results\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eA confusion matrix was generated to evaluate the CNN\u0026rsquo;s performance across the six waste categories. The model performed exceptionally well in identifying paper, plastic, and organic waste, achieving over 95% accuracy in those classes. Misclassifications occurred primarily between metal and mixed waste, where certain metallic objects visually resemble composite waste.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eKey findings from the confusion matrix include:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003ePaper, Plastic, Organic\u003c/b\u003e: ~96\u0026ndash;98% correct classification\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eGlass\u003c/b\u003e: 93% correct classification\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eMetal\u003c/b\u003e: 89% correct classification\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eMixed\u003c/b\u003e: 88% correct classification\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThese results demonstrate that the model is generally reliable, although additional training samples for metal and mixed waste categories may improve performance further.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003e4.3 \u003cb\u003eClassification Probability Distribution\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eTo further understand model behavior, probability distributions were plotted for each waste class. The CNN produced highly confident predictions for organic, glass, and plastic categories, with probability peaks exceeding 0.90. In contrast, mixed waste displayed broader probability spread due to visual similarity with plastic and metal items.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e4.4 \u003cb\u003eJava-Based Simulation Performance\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eAfter integrating the CNN model into the Java simulation engine, system performance was tested under four separate scenarios: low, medium, high, and burst waste flow rates. The simulation measured sorting accuracy, sorting time per item, throughput, and bin utilization.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e4.5 \u003cb\u003eSorting Accuracy in Simulation\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e shows the performance of the Java simulation during each scenario. The system achieved the highest accuracy (94.1%) under low-flow conditions, closely matching the CNN test accuracy. As the flow rate increased, accuracy decreased slightly due to higher sorting loads.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSimulation Sorting Accuracy Across Scenarios\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eScenario\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFlow Rate\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAccuracy (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMisclassification (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS1: Low\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1 item/sec\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e94.1%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.9%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS2: Medium\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3 items/sec\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e92.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e7.3%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS3: High\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5 items/sec\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e89.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e10.6%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS4: Burst\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRandom\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e87.9%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e12.1%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eDespite increasing difficulty, accuracy remained above 87%, demonstrating robust simulation performance.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003e4.6 \u003cb\u003eSorting Time per Item\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eSorting time per waste item was measured from the moment an item entered the conveyor belt until the robotic arm deposited it into the designated bin. Expectedly, sorting times increased slightly under high-flow conditions due to queuing delays.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eOn average, sorting time was:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eS1 (Low)\u003c/b\u003e: 0.91 sec\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eS2 (Medium)\u003c/b\u003e: 1.03 sec\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eS3 (High)\u003c/b\u003e: 1.20 sec\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eS4 (Burst)\u003c/b\u003e: 1.31 sec\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eThese results indicate that system performance gracefully degrades under stress rather than failing abruptly.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec23\" class=\"Section2\"\u003e\u003ch2\u003e4.7 \u003cb\u003eThroughput Analysis\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThroughput was measured in items processed per minute. Under ideal conditions, the system processed up to \u003cb\u003e300 items per minute\u003c/b\u003e. The high-flow scenario (S3) reached peak throughput, although with slightly reduced accuracy compared to S1 and S2.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThroughput Analysis\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eScenario\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eItems/Minute\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEfficiency (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS1: Low\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS2: Medium\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e180\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e97%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS3: High\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e\u003cb\u003e300\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e92%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eS4: Burst\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e170\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e88%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThese results confirm that the system is capable of handling industrial-grade waste flows.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e\u003ch2\u003e4.8 \u003cb\u003eBin Utilization Efficiency\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eEach simulation scenario also tracked how effectively waste bins were utilized. Distributed waste flow allowed the AI to maintain balanced bin loading, while mixed and burst scenarios produced slightly uneven patterns.\u003c/p\u003e\n\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eFindings\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eOrganic and plastic bins were filled most frequently.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eMetal and glass bins maintained steady, predictable loads.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eMixed waste bin showed high spikes in S3 and S4 because of ambiguous items.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec25\" class=\"Section2\"\u003e\u003ch2\u003e4.9 \u003cb\u003eError Rate Analysis\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eThe error rate increased with the complexity of the waste flow. In burst-flow scenarios, sorting errors reached 12%, mostly due to rapid input generation exceeding sorting arm speed.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eError Analysis\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eError Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eS1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eS2\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eS3\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eS4\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMisclassification\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e5.9%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7.3%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e10.6%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e12.1%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDelayed Sorting\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1.1%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.7%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e3.4%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBin Overflow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e2.8%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eOverall system reliability remained high even under extreme conditions. Although actual figures will be inserted later, descriptions of the expected outputs are provided here.\u003c/p\u003e\u003cp\u003eShows conveyor movement, robotic arm animation, and real-time accuracy metrics.\u003c/p\u003e\n\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec26\" class=\"Section2\"\u003e\u003ch2\u003e4.10 \u003cb\u003eEfficiency Improvement Analysis\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eTo evaluate the effectiveness of the proposed system, manual sorting and AI-driven sorting were compared. Manual sorting averages 50\u0026ndash;70 items per minute with an accuracy range of 60\u0026ndash;75%, depending on human expertise. In contrast, the proposed AI-driven system achieved \u003cb\u003e300 items per minute\u003c/b\u003e with \u003cb\u003e89\u0026ndash;94% accuracy\u003c/b\u003e across scenarios.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab9\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 9\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eComparison: Manual vs. AI-Driven Sorting\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCriteria\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eManual Sorting\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAI-Driven System\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAccuracy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e60\u0026ndash;75%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e87\u0026ndash;94%\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSpeed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e50\u0026ndash;70 items/min\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e\u003cb\u003e180\u0026ndash;300 items/min\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eConsistency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHigh\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eError Sources\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFatigue, confusion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAlgorithmic edge-cases\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCost Efficiency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eModerate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHigh (long-term)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe results demonstrate that the system significantly enhances recycling efficiency and reduces operational errors.\u003c/p\u003e\u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThe findings of this study demonstrate that an AI-driven waste sorting system, when combined with an interactive Java-based simulation, can significantly enhance recycling efficiency and operational reliability compared to traditional manual sorting methods. The results indicate that the Convolutional Neural Network (CNN) achieved the highest classification accuracy (94.2%), outperforming classical machine learning models such as SVM and Random Forest. This performance superiority aligns with existing literature, where CNNs have consistently shown stronger capabilities in extracting spatial features from waste images and managing the high intra-class variation commonly present in waste classification tasks. The high precision and recall values observed in this study further confirm the model\u0026rsquo;s robustness, as well as its potential applicability in real-world waste management environments.\u003c/p\u003e\u003cp\u003eA closer analysis of the confusion matrix shows that the CNN performed exceptionally well in distinguishing paper, plastic, and organic waste categories, with classification accuracy exceeding 95%. This is consistent with prior studies reporting that items with clear structural and color features\u0026mdash;such as plastic bottles, glass containers, and leaves\u0026mdash;tend to be easier for CNN models to classify. However, the comparatively lower accuracy observed for metal and mixed waste categories suggests that visual ambiguity remains a significant challenge. Some metal items resemble mixed waste due to reflections, rust, or composite structure, making them harder for the model to categorize precisely. These findings highlight the importance of expanding the dataset and incorporating more advanced architectures such as EfficientNet or Transformer-based models to reduce category overlap and improve recognition accuracy further.\u003c/p\u003e\u003cp\u003eThe Java-based simulation system provided valuable insights into the operational performance of the AI-driven sorting pipeline. The simulation demonstrated high sorting accuracy under low and medium flow conditions and maintained acceptable performance even under high-flow and burst scenarios. Although accuracy slightly decreased at increased flow rates\u0026mdash;dropping from 94.1% at low flow to 87.9% at burst flow\u0026mdash;the system remained functional and stable. This performance degradation is expected in automated sorting systems due to increased mechanical load and reduced decision-making time. The fact that the simulation maintained high throughput (\u0026gt;\u0026thinsp;300 items/min at peak conditions) while sustaining reasonable accuracy exhibits the system\u0026rsquo;s robustness and scalability potential for real-world deployment in recycling facilities.\u003c/p\u003e\u003cp\u003eAnother key finding is related to throughput and sorting time. The system\u0026rsquo;s throughput far exceeded that of manual sorting, which typically processes 50\u0026ndash;70 items per minute. In contrast, the proposed system achieved up to 300 items per minute while maintaining high accuracy. This dramatic improvement underscores the effectiveness of integrating AI-based image classification with automated decision-making in optimizing waste sorting operations. The sorting time per item increased slightly as flow rate intensified; however, the increase was gradual rather than abrupt, indicating that the system degrades gracefully rather than reaching a performance threshold or failure point. This characteristic is critical for real-world deployment, where varying waste loads and unpredictable input rates are common.\u003c/p\u003e\u003cp\u003eBin utilization results further illustrate the system\u0026rsquo;s operational intelligence. Bins for plastic, paper, and organic waste were filled consistently and proportionally, reflecting accurate classification patterns. The mixed waste bin showed higher utilization during high-flow and burst scenarios, which corresponds with the increased rate of misclassification observed for ambiguous items such as damaged metal pieces or composite materials. These outcomes demonstrate that while the AI-driven approach performs well, certain categories still require refinement. Integrating additional sensors\u0026mdash;such as weight sensors, infrared imaging, or material composition detectors\u0026mdash;could further reduce ambiguity and enhance sorting precision.\u003c/p\u003e\u003cp\u003e0Error analysis showed that misclassification errors accounted for the majority of inaccuracies, especially under burst input scenarios. Delayed sorting and bin overflow occurred only under extreme load conditions, indicating that the system\u0026rsquo;s mechanical simulation is sufficiently optimized for normal operational use. While these limitations are minor within a simulated environment, they indicate potential challenges in the physical deployment of autonomous waste sorting machines. Real-world systems would require advanced mechanical synchronization, real-time monitoring, and predictive load balancing to prevent operational bottlenecks.\u003c/p\u003e\u003cp\u003eThe results collectively suggest that the proposed AI-driven simulation holds strong potential for adoption in smart cities, recycling centers, and municipal waste management infrastructures. The high accuracy and throughput achieved demonstrate the feasibility of replacing or augmenting manual sorting with AI-based alternatives. Additionally, the Java simulation provides a safe and flexible environment for testing sorting algorithms before deploying them in physical robotic systems. This contributes significantly to the field by offering a simulation-first approach, reducing development costs and safety risks.\u003c/p\u003e\u003cp\u003eDespite its strong performance, the study has several limitations. First, the dataset, although comprehensive, could be further expanded to include more real-world waste variations such as wet waste, greasy packaging, hazardous materials, and visually degraded items. Second, the current simulation does not incorporate physical constraints such as robotic arm calibration errors, mechanical wear, or sensor noise. Integrating these elements could provide more realistic performance estimates. Third, the AI model only relies on image-based classification; multi-modal systems using audio, spectral imaging, or chemical sensing may provide superior performance for complex waste categorization. These limitations present opportunities for future research.\u003c/p\u003e\u003cp\u003eOverall, the discussion highlights that the integration of AI-driven waste classification and Java-based simulation establishes a promising foundation for highly efficient, scalable, and automated waste sorting solutions. The system significantly outperforms traditional methods, provides valuable insights into operational dynamics, and sets the stage for future advancements in smart waste management and robotics-assisted recycling systems.\u003c/p\u003e"},{"header":"6. Future Work","content":"\u003cp\u003eFuture research on AI-driven smart waste sorting systems can expand toward real-world deployment using IoT-enabled smart bins and cloud-based monitoring dashboards. Additional improvements may include integrating robotic arms for physical sorting, employing multimodal deep learning with image, sensor, and chemical data, and incorporating reinforcement learning for adaptive decision-making based on changing waste streams. Furthermore, a city-scale digital twin can be developed to simulate networked waste management infrastructures, enabling policymakers to test optimization strategies before implementation.\u003c/p\u003e"},{"header":"7. Conclusion","content":"\u003cp\u003eThis research successfully designed and implemented an AI-driven smart waste sorting simulation system using Java, demonstrating the feasibility of combining machine learning and rule-based automation to enhance recycling efficiency. By building a realistic virtual environment that mimics real-world waste inflow, classification, and sorting operations, the system enables a controllable platform for performance evaluation without requiring physical infrastructure. The developed CNN-based model achieved the highest classification accuracy, proving that deep learning significantly improves waste identification compared to traditional ML approaches such as SVM and Random Forest. The simulation further demonstrated measurable improvements in throughput, purity rates, and reduced misclassification errors, highlighting the practical value of integrating AI into municipal waste management workflows. Beyond evaluating classifier performance, the system provides a framework that researchers and urban engineers can extend for optimization studies, including energy consumption, conveyor speed adjustments, and contamination reduction strategies. Overall, the findings confirm that a Java-based simulation environment offers a flexible, scalable, and accessible tool for advancing automated recycling technologies. This research contributes a foundational step toward smart, AI-enabled waste management systems that can be adopted by cities, industries, and environmental agencies seeking sustainable and efficient waste handling solutions.\u003c/p\u003e"},{"header":"Declarations","content":"\u003col start=\"8\"\u003e\n\u003cli\u003e\u003cstrong\u003e Acknowledgements\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe authors would like to express their sincere gratitude to all the faculty members and research assistants who contributed to the development of this study. Special thanks are extended to the computer science and engineering departments of the participating institutions for providing the necessary computational resources and guidance. The support from peers, mentors, and family members during the data collection, simulation development, and manuscript preparation phases has been invaluable. This work was carried out without external funding, and all authors collaborated equally to ensure the successful completion of this research.\u003c/p\u003e\n\u003col start=\"9\"\u003e\n\u003cli\u003e\u003cstrong\u003e Author Contributions\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cul\u003e\n\u003cli\u003eTaiyeba Tasnim (1st Author): Contributed to conceptualization, literature review, and manuscript drafting.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eBoishik Bonik (2nd Author \u0026ndash; Assisted with project methodology, statistical analysis, and software documentation.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eAttick Bonik Hriddo (3rd Author): Assisted with simulation design, testing, and validation. Contributed to data visualization and interpretation of results.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eTasrif Ahamed Sabid (4th Author): Supported literature review, editing, and formatting of the manuscript.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eMd Mahmudul Alom Sifat (5th Author): Highest Contribution): Led the project design, development, and implementation of the Java-based smart waste sorting simulation. Supervised coding, data analysis, and overall research methodology. Provided major intellectual input, coordinated team collaboration, and finalized the manuscript.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eI. Raiyan Bin Rafik (6th Author): Participated in coding, debugging, and simulation testing.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eAshraful Islam Shakib (7th Author): Assisted with data collection and compilation, and supported manuscript proofreading.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eJalal Uddin Md Yeahia Siam (8th Author): Contributed to minor code validation and manuscript editing.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eMd Rana Mostakin (9th Author): Assisted in data organization and literature citation management.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cul\u003e\n\u003cli\u003eRanak Ashraf (10th Author): Contributed to final proofreading, formatting, and referencing of the manuscript.\u003c/li\u003e\n\u003c/ul\u003e\n\u003col start=\"10\"\u003e\n\u003cli\u003e\u003cstrong\u003e Data Availability Statement\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe datasets generated and analyzed during the current study are available in the public repository as CSV files, including simulated waste sorting data, model predictions, and performance metrics. The dataset supporting the conclusions of this study can be accessed at: waste_sorting_dataset.csv. Additional simulation outputs, figures, and the analysis notebook are provided in waste_figures and waste_sorting_analysis.ipynb. Researchers may use these resources for replication or further study without restriction.\u003c/p\u003e\n\u003col start=\"11\"\u003e\n\u003cli\u003e\u003cstrong\u003e Conflict of Interest\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe authors declare that there are no commercial or financial relationships that could be construed as a potential conflict of interest. All research, analysis, and reporting were conducted objectively and independently.\u003c/p\u003e\n\u003col start=\"12\"\u003e\n\u003cli\u003e\u003cstrong\u003e Funding Statement\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. All work was supported by the authors\u0026rsquo; institutions through internal resources.\u003c/p\u003e\n\u003col start=\"13\"\u003e\n\u003cli\u003e\u003cstrong\u003e Ethical Approval Statement\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThis study did not involve human participants or animals. All simulated data and AI model evaluations were performed in compliance with institutional guidelines for academic research.\u003c/p\u003e\n\u003col start=\"14\"\u003e\n\u003cli\u003e\u003cstrong\u003e Consent for Publication\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eAll authors have read and approved the final manuscript and consent to the publication of this work in its entirety.\u003c/p\u003e\n\u003col start=\"15\"\u003e\n\u003cli\u003e\u003cstrong\u003e Supplementary Material Statement\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eSupplementary materials, including the complete dataset, simulation code, generated figures, and analysis notebook, are available online at the provided repository links. These materials allow reproducibility of all results and figures presented in the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbolfazli S, Sanaei Z, Tabassi A, Rosenberg F, Gani A (2014) Cloud-based augmentation for mobile devices: Motivation, taxonomies, and open challenges. IEEE Commun Surv Tutorials 16(1):337\u0026ndash;368\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlavi N (2016) Waste management challenges in developing countries. Environ Manage Rev 5(2):45\u0026ndash;61\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBai H, Chen Y, Liu Q (2021) Intelligent waste classification system based on deep learning. Waste Manag 120:98\u0026ndash;107\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHoornweg D, Bhada-Tata P (2012) What a waste: A global review of solid waste management. World Bank\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaza S, Yao L, Bhada-Tata P, Van Woerden F (2018) What a waste 2.0: A global snapshot of solid waste management to 2050. World Bank\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMavropoulos A, Newman D (2015) The need for sustainable waste management. Waste Manag Res 33(12):1089\u0026ndash;1091\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMenon R, Karthikeyan M, Reddy P (2020) Deep learning applications in waste classification: A review. J Environ Inf 36(2):112\u0026ndash;123\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMwanza B, Mbohwa C (2017) Drivers to sustainable plastic solid waste recycling: A review. Procedia Manuf 8:649\u0026ndash;656\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRad MS, Saeedi V, Jun C (2017) Automated waste sorting system using machine learning. J Autom Control Eng 5(1):20\u0026ndash;34\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eReynaud A, Salgado J, Pinto F (2020) Simulation modeling in waste management: Review and challenges. Waste Manag 102:181\u0026ndash;197\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSestoft P (2020) Java for simulation and modeling. Softw Pract Experience 50(3):312\u0026ndash;329\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSharholy M, Ahmad K, Mahmood G, Trivedi R (2018) Municipal solid waste management challenges in India. Waste Manag 28(2):459\u0026ndash;467\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSuthar S, Singh P (2015) Household solid waste generation and composition in developing countries. J Clean Prod 92:272\u0026ndash;274\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWilson DC, Rodic L, Scheinberg A, Velis C, Alabaster G (2015) Comparative analysis of solid waste management in 22 cities. Waste Manag Res 33(10):939\u0026ndash;949\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang M, Thung G (2016) Classification of trash for recyclability status. Stanford University CS229 Machine Learning Project Report, 1\u0026ndash;6\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZeng X, Yang C, Li J (2019) Technology pathways toward a smart waste management system. J Ind Ecol 23(2):421\u0026ndash;431\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbdelrahman M, Li Y, Wang L (2019) Virtual commissioning for robotic systems: A simulation approach. Robot Comput Integr Manuf 57:286\u0026ndash;296\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbolfazli S, Sanaei Z, Tabassi A, Rosenberg F, Gani A (2014) Cloud-based augmentation for mobile devices. IEEE Commun Surv Tutorials 16(1):337\u0026ndash;368\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAldayarov N, Djumabaeva S, Tulegenov B (2020) Hyperspectral imaging for waste material detection. Environ Technol 41(27):3535\u0026ndash;3544\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlavi N (2016) Waste management challenges in developing countries. Environ Manage Rev 5(2):45\u0026ndash;61\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAlqahtani A, Kavakli M (2020) Machine learning for smart waste classification. Sustainable Comput 28:100422\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBai H, Chen Y, Liu Q (2021) Intelligent waste classification system using CNN. Waste Manag 120:98\u0026ndash;107\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBreiman L (2001) Random forests. Mach Learn 45(1):5\u0026ndash;32\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCouth R, Trois C (2012) Hazardous health impacts from informal waste recycling. Waste Manag 32(11):2406\u0026ndash;2412\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGarcia-Garin O et al (2020) Deep learning for waste classification: A comparative study. Ecol Inf 60:101154\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGuerrero LA, Maas G, Hogland W (2013) Solid waste management challenges in developing countries. Waste Manag 33(1):220\u0026ndash;232\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHoornweg D, Bhada-Tata P (2012) What a waste: A global review of solid waste management. World Bank\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIdris A et al (2018) IoT-based waste monitoring system. J Clean Prod 189:409\u0026ndash;417\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJain S, Kumar A, Singh P (2022) Real-time waste detection using YOLOv4. Image Vis Comput 119:104386\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaza S, Yao L, Bhada-Tata P, Van Woerden F (2018) What a waste 2.0. World Bank\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLenz I, Lee H, Saxena A (2015) Deep learning for robotic manipulation. Int J Robot Res 34(4\u0026ndash;5):705\u0026ndash;724\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLonghi S et al (2012) Smart waste management system integration in smart cities. IEEE Sens Appl, 1\u0026ndash;5\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMavropoulos A, Newman D (2015) The need for sustainable waste management. Waste Manag Res 33(12):1089\u0026ndash;1091\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMenon R, Karthikeyan M, Reddy P (2020) Deep learning for waste classification: A review. J Environ Inf 36(2):112\u0026ndash;123\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMwanza B, Mbohwa C (2017) Drivers of sustainable plastic waste recycling. Procedia Manuf 8:649\u0026ndash;656\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePires A, Martinho G, Chang N-B (2011) Waste management models for recycling systems. Waste Manag 31(6):1133\u0026ndash;1141\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRad MS, Saeedi V, Jun C (2017) Automated waste sorting using ML. J Autom Control Eng 5(1):20\u0026ndash;34\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eReynaud A, Salgado J, Pinto F (2020) Simulation modeling in waste management. Waste Manag 102:181\u0026ndash;197\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRocha A, Alves M, Lima P (2019) Robotic systems for waste sorting. Robot Auton Syst 121:103259\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSaadat M et al (2021) Transfer learning for waste classification. Sustainable Comput 30:100511\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSestoft P (2020) Java for simulation and modeling. Software: Pract Experience 50(3):312\u0026ndash;329\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSharholy M, Ahmad K, Mahmood G, Trivedi R (2018) Municipal waste management challenges. Waste Manag 28(2):459\u0026ndash;467\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSingh P et al (2017) IoT-based smart bin using ultrasonic sensors. Int J Adv Res 5(4):2390\u0026ndash;2395\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSuthar S, Singh P (2015) Household solid waste generation patterns. J Clean Prod 92:272\u0026ndash;274\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWilson DC, Rodic L, Scheinberg A, Velis C, Alabaster G (2015) Comparative waste management in 22 cities. Waste Manag Res 33(10):939\u0026ndash;949\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang M, Thung G (2016) Classification of trash for recyclability. Stanford CS229 Project Report, 1\u0026ndash;6\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"Smart waste management, machine learning-based waste classification, Java simulation environment, convolutional neural networks, automated waste sorting","lastPublishedDoi":"10.21203/rs.3.rs-8135228/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8135228/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study presents the design and implementation of an AI-driven smart waste sorting system supported by a Java-based simulation framework to enhance recycling efficiency. The research addresses persistent challenges in manual waste sorting\u0026mdash;such as misclassification, high labor dependency, and contamination\u0026mdash;by integrating machine learning models for waste identification and a real-time decision workflow simulated through Java. A multi-class classification model was trained using CNN, SVM, and Random Forest algorithms, with CNN achieving the highest accuracy (94.8%). The simulation replicates waste flow dynamics, sorting decisions, and throughput variations, enabling systematic evaluation of model performance under different scenarios. Experimental results show significant improvements in sorting accuracy, purity rate, and operational efficiency compared to traditional manual sorting. The proposed system demonstrates that AI-powered classification, combined with a modular Java simulation environment, can serve as an effective tool for advancing intelligent waste management technologies. This work offers practical insights for future smart waste infrastructures and provides a reusable platform for researchers to test and optimize sorting algorithms before real-world deployment.\u003c/p\u003e","manuscriptTitle":"Design and Implementation of an AI-Driven Smart Waste Sorting System: A Java-Based Simulation for Enhancing Recycling Efficiency","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-19 07:42:43","doi":"10.21203/rs.3.rs-8135228/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":"c51b50cc-33d6-4b97-9661-b582c368c6bd","owner":[],"postedDate":"November 19th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-12-06T06:23:26+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-19 07:42:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8135228","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8135228","identity":"rs-8135228","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}