Density-Adaptive UAV ISAC: A DRL Framework for Joint Beam Control and Trajectory Design

High-density vehicular networks present significant challenges for simultaneous sensing and communication due to dynamic channel conditions and varying traffic patterns. This paper addresses the joint optimization problem of beam control, trajectory planning, and power allocation in UAV-enabled Integrated Sensing and Communication (ISAC) systems. We propose a deep reinforcement learning approach based on Adaptive Proximal Policy Optimization (Adaptive PPO) that jointly optimizes discrete actions for beam direction, UAV movement, and power allocation in a unified action space, dynamically adapting resource allocation between sensing and communication functions based on real-time vehicle density, channel quality, and detection accuracy. Numerical analysis using a bidirectional highway scenario demonstrates that our Adaptive PPO-based approach outperforms conventional baseline methods across all evaluated metrics, providing a robust solution for next-generation vehicular communication networks.