DRL-Based User Selection and Bandwidth Allocation in ISAC-Assisted IoV Networks

With the rapid development of the Internet of Vehicles (IoV), there is an increasing demand for efficient, reliable, and low-latency communication technologies. Integrated sensing and communication (ISAC) technology offers improved resource utilization, enhanced performance, and better support for dynamic and complex traffic environments, increasing the reliability and robustness of vehicular networks. In this paper, we focus on ISAC-assisted IoV networks and formulate a joint user selection and bandwidth allocation problem to maximise the performance of sensing with guaranteeing the performance of communication. We consider the miss detection ratio as the sensing metric, which is defined by the ratio of the number of undetected vehicles to the number of total vehicles in a time duration in the IoV scenario. Given this complex and dynamic temporal coupling problem, we study a clustering-based algorithm for user selection and propose a deep reinforcement learning (DRL) approach based on proximal policy optimization (PPO) algorithm for optimizing bandwidth allocation between sensing and communication. Meanwhile, we generate real-time vehicles and environmental data using Simulation of Urban Mobility (SUMO). Simulation results validate better performance of the proposed algorithm on miss detection ratio compared to other benchmarks.