Developing a pervasive edge computing environment for Vehicular Communication using modified Reinforcement Learning in Routing and Dynamic Traffic Flow Prediction

AbstractVehicular networking in smart autonomous connected vehicle communications evolved with high mobility and due to high dynamics in an urban environment, new challenges are addressed by academicians and researchers for providing better support. Dynamic changes of vehicular nodes position, routing in Vehicular Adhoc Network (VANET) using existing traditional networking routing algorithms may not provide optimal solution for efficient communication. Also predicting or forecasting traffic flow in VANET can be improved through sharing traffic information in real-time using intelligent transportation systems. In this paper we proposed modified reinforcement learning algorithm that supports for optimal route identification for dynamically disconnected vehicles in urban environment by considering its previous state and predicts flow of traffic generated by vehicles in various time interval. Experimental result shows better performance in routing parameters like packet delivery ratio, routing overhead, latency and predicting traffic flow by proposed algorithms achieves significant accomplishments comparing to existing algorithms.