Multi-Agent Reinforcement Learning for Area Coverage in Unknown Environments

The problem of area coverage in unknown environments is crucial in applications such as environmental monitoring, disaster response, and autonomous navigation. Traditional methods struggle to balance exploration, obstacle avoidance, and efficient coverage. In this paper, we propose a deep reinforcement learning-based multi-agent approach that optimizes coverage efficiency while adapting to unknown obstacles. We integrate a two-layer architecture, where the upper layer employs reinforcement learning for global path planning, while the lower layer handles local obstacle avoidance and movement execution. Experimental results demonstrate that our algorithm effectively converges in unknown environments and completes the area coverage task.