Multi-Agent Coalition Cooperative Path Finding for Tasks with Release Time
Wang Le, Ming ChenZhi, Tan Ming, Ye GuoDong, Yu Shan Sun, Jun Wang · 2025
Current research on multi-agent path planning primarily focuses on scenarios without considering task release times. However, in real-world applications, tasks often have release time constraints, and agents are heterogeneous, making it difficult for a single agent to independently complete complex tasks. To address these challenges, this paper proposes a coalition algorithm for Path Finding for Tasks with Release Time (PF-TRT). The proposed system leverages a prediction model to estimate the potential release times of tasks and dynamically adjusts the task queues of each agent in real time. To further validate the effectiveness of task release time handling, deep reinforcement learning (DQN) is incorporated in the final experiments to derive optimal strategies for path planning in dynamic environments. This integration ultimately enhances the overall performance of the multi-agent system.