Proximal Policy Optimization With Adaptive Weight Adjustment for Airborne Integrated Sensing and Communication

Abstract

Integrated sensing and communication (ISAC) technology has emerged as a potential enabler for efficient data transmission and sensing in autonomous aerial vehicle (UAV) networks. In this letter, we investigate a UAV-ISAC system, where a UAV furnishes downlink communication and simultaneous sensing for mobile ground nodes (GNs). To address this non-convex optimization problem and prevent retraining required due to dynamic environments, we design a new proximal policy optimization (PPO)-based approach that jointly optimizes the UAV trajectory and resource assignment while guaranteeing the required throughput, sensing fairness (SF) and age of information (AoI) among GNs. We integrate a new adaptive reward weight adjustment (ARWA) mechanism into the PPO framework to guide the UAV operations dynamically. Simulations corroborate that the proposed ARWA-PPO scheme can ameliorate the system throughput by 12% and increase the minimum GN throughput by 19%, compared to the existing baselines, including PPO.