Collaborative Area Coverage Algorithm Based on Iterative Teacher-student Training

To address the problem of cooperative area coverage path planning for unmanned aerial vehicles (UAVs) without prior information, this paper proposes an iterative teacher-student training framework that integrates curriculum learning with the heterogeneous-agent proximal policy optimization (HAPPO) algorithm. A complex coverage task is decomposed into a sequence of sub-tasks from simple to complex by designing a multi-stage curriculum training method, which significantly improves the training efficiency and convergence speed of the HAPPO algorithm. To solve the poor generalization capability of deep reinforcement learning algorithms in multi-scenario applications, an iterative teacher-student learning mechanism is introduced based on curriculum training. Specifically, a model converged through primary curriculum training serves as the teacher network to guide the student network via knowledge distillation during secondary curriculum training, thereby enhancing the model's adaptability to different environments. Simulated results demonstrate that the proposed algorithm has a faster convergence speed during the training process by introducing the curriculum training and the iterative teacher-student learning mechanism, and achieves shorter area coverage time and lower cell revisit rate during the testing phase, providing an efficient and robust solution for UAV cooperative area coverage tasks.