Low-altitude Secure Communication Driven by Deep Reinforcement Learning: An Integrated Sensing and Communication Design

To address the physical layer security challenges in low-altitude Unmanned Aerial Vehicle (UAV) communications, this paper proposes an Integrated Sensing And Communication (ISAC) scheme. For the proposed ISAC scheme, an online optimization framework for UAV trajectory and communication resource allocation is developed using Deep Reinforcement Learning (DRL). In the proposed scheme, artificial noise transmitted by a communication UAV is reused to simultaneously sense and jam a potential eavesdropping UAV, thereby enhancing secure communications for ground users. By estimating and predicting the state of the eavesdropping UAV, the trajectory and resource allocation design problem is reformulated as a Markov decision process. Using the Deep Deterministic Policy Gradient (DDPG) algorithm, the optimal framework is learned over time, dynamically optimizing the communication UAV’s trajectory and resource allocation to maximize long-term sensing and secure communication performance. Simulation results demonstrate that the proposed scheme achieves a superior trade-off between sensing and security without degrading sensing performance and outperforms baseline methods in terms of secure communication performance. This validates the performance gains achieved through sensing and online trajectory design, as well as the potential and superior performance of applying DRL to the integrated design of sensing, communication, and trajectory.