Robust UAV Spectrum Scheduling Under Incomplete Information: A Fuzzy Framework Based on PPO-Assisted Evolutionary Reinforcement Learning
Lu Sun, Shengen Zhao, Liangtian Wan, Kaihui Liu, Jie Wang, Yuan Yuan, Yun Lin, Bo Ai
Abstract
In unmanned aerial vehicle (UAV) search and rescue missions, UAV swarms equipped with joint radar-communication (JRC) equipment play a pivotal role by collaboratively detecting terrain in rescue zones, locating buried personnel, and transmitting critical information in real time to command centers, serving as a core technology to enhance mission efficiency. However, complex rescue environments often involve incomplete information and jamming interference, increasing scheduling uncertainty and reducing system robustness. To address this challenge, this paper presents a realistic spectrum scheduling model tailored for UAV search and rescue scenarios. The model integrates a fuzzy rule-based system (FRBS) to quantify and manage interference uncertainties arising from incomplete data, enabling robust decision-making for dynamic spectrum resource allocation. Further, an evolutionary reinforcement learning (ERL)-based framework is proposed to develop a proximal policy optimization (PPO)-enhanced clustered pied kingfisher optimizer (PCPKO). By deeply fusing adaptive clustering strategies with PPO’s dynamic hyperparameter tuning mechanism, PCPKO effectively balances global exploration and local exploitation capabilities in spectrum allocation. Simulation results show that PCPKO significantly outperforms traditional algorithms in spectrum utilization, system stability, and global optimization. Compared with the benchmark algorithms, PCPKO improves spectrum utilization by 20% to 34% and reduces the frequency conflict rate by 5% to 20%. This validates its excellent anti-interference capability and robustness in complex environments, providing an efficient and reliable solution for UAV search and rescue spectrum scheduling.