A Multi-Agent Path Planning Method based on Improved MADDPG

To effectively address the multi-agent path planning problem in complicated environments, this study proposes an improved Multi-Agent Deep Deterministic Policy Gradient (MADDPG) method for multi-agent path planning, with multi-unmanned aerial vehicles (MUAVs) as the research subject. The proposed method incorporates a directional reward function that considers the dynamic changes in the UAVs' actions over two consecutive steps. This ensures stable flight direction and smooth trajectory adjustments based on the surrounding environment. Additionally, the method includes reward functions related to coverage range and movement cost, aiming to overcome challenges such as path instability and long planning times in multi-agent path planning. Experimental results show that compared to traditional methods, the proposed approach significantly reduces the time required for MUAVs to complete tasks, increases path smoothness by nearly 96%, and improves the cumulative reward value.