A Deep Reinforcement Learning Based UAV Trajectory Planning Method For Integrated Sensing And Communications Networks

With the development of information technology, the next-generation information technologies such as artificial intelligence, digital twin and reconfigurable intelligent surface have become key research areas for current 6G networks. In addition, to improve the end-to-end information processing capability in the next-generation networks and better meet the demand for high-speed communication and high-precision sensing for digital twin, Virtual Reality and Augmented Reality immersive services in the 6G networks, integrated sensing and communications (ISAC) has emerged. To deal with the conflict between high-quality communication services and low-latency sensing targets in an ISAC architecture, this paper investigates a UAV-assisted ISAC system in which the UAV adopts a flight-hover-communication protocol. In particular, the UAV communicates with Internet of Things (IoT) devices during the hovering period, while it senses the location of targets during the flying period. To maximize the number of connected IoT devices and minimize the energy consumption of the UAV, a deep reinforcement learning (DRL) based trajectory planning algorithm is designed. The numerical results demonstrate that the proposed algorithm can effectively detect sensor devices as well as collect sensor data.