Improving Mobility in NDN-Based VANET: A Deep Reinforcement Learning Approach with Deep Prioritization

The utilization of Named Data Network (NDN) in Vehicle Autonomous Network (VANET) has emerged as a prominent research area, aiming to enhance data transmission and distribution. However, effectively addressing the mobility issues in NDN-based VANET poses significant challenges. Existing Q-learning-based geographic routing methods suffer from slow convergence and heavy reliance on dynamic Q-value tables. Moreover, local information-based next-hop selection does not always prioritize optimal global routing forwarding. To overcome these limitations, this study presents a novel routing strategy for in-vehicle named data networks, employing the Deep Prioritized Sarsa algorithm. The proposed approach incorporates fuzzy logic techniques and depth-first search algorithms to obtain and utilize comprehensive global road information. Additionally, it maintains an adaptive fixed-size Q-value table with a customized reward function for vehicle node selection. Simulation results showcase substantial improvements in terms of packet hit rate and end-to-end delay. This research contributes to the advancement of efficient and effective routing strategies for NDN-based VANET, addressing the challenges associated with mobility and data transmission in vehicular networks.