Hybrid DRL-Based Sensing and Age of Information Optimization for UAV-Enabled ISCC

Abstract

In low-altitude economy (LAE), the deployment of unmanned aerial vehicles (UAVs) provides substantial convenience and enhances operational efficiency. This letter investigates a joint resource and trajectory optimization problem in a UAV-enabled integrated sensing, computation, and communication (ISCC) network where the UAV senses and processes the status information from sensing targets (STs), and then sends the computed results to the data collection center (DC). Aiming to maximize the sensing data volume while minimizing the age of information (AoI), we jointly optimize the UAV’s sensing schedule, number of sensing trials, time allocation, transmit power, CPU frequency, and trajectory. The problem is formulated as a Markov decision process (MDP), and a hybrid deep reinforcement learning (DRL) framework is proposed to derive optimal policies. Specifically, we adopt the twin delayed deep deterministic policy gradient (TD3) framework and enhance it with a hybrid-baseline prioritized experience replay (PER) mechanism, denoted as HPTD3. Simulation results demonstrate that the proposed approach significantly outperforms benchmark schemes.