Reinforcement Learning-Based AI-Powered Adaptive Routing Protocols for Urban VANETs

Due to the static and reactive nature of the traditional routing protocols like Ad-hoc On-Demand Distance Vector (AODV) and Dynamic Source Routing (DSR), theses protocols are not very efficient and face crucial challenges such as dynamic traffic patterns, frequent topology changes and high mobility, which are highly anticipated in the Urban Vehicular Ad Hoc Networks (VANETs).To overcome these issues, this paper proposes a reinforcement learning (RL) based adaptive routing protocol which is AI powered and customized according to the needs of modern VANETs.The model incorporates Q-learning, wherein each vehicle is treated as an autonomous learning agent.This agent selects optimal path at real time based on local feedback about vehicle mobility, neighbouring spots and traffic density.The model exhibits high simulation accuracy by integrating SUMO for real time mobility simulation and NS3 for network behaviour.To optimize the pivotal performance parameters such as routing overhead, packet delivery ratio, end to end delay, throughput, hop count a specialized reward function is designed.The proposed model is evaluated for the performance with previous popular models such as AODV and DSR across different modern traffic scenarios.Undoubtedly it has given improved responsiveness, scalability and routing optimality and reliability in complex modern traffic environments.