Deep reinforcement learning path planning and task allocation for multi-robot collaboration
Zhixian Li, Nianfeng Shi, Liguo Zhao, Mengxia Zhang
Abstract
In the current technological landscape, Multi-Robot Systems (MRS) have become crucial for complex tasks, with applications in industrial automation , search and rescue , and intelligent transportation . However, existing techniques face challenges in path planning and task allocation, particularly regarding adaptability, real-time decision-making, and efficiency. Deep Reinforcement Learning (DRL) has emerged as a promising solution due to its robust learning capabilities. To address these challenges, we propose an innovative DRL-MPC-GNNs model that integrates Deep Reinforcement Learning, Model Predictive Control (MPC), and Graph Neural Networks (GNNs). Our model aims to optimize path planning and task allocation in multi-robot systems. Through rigorous experiments in simulated environments, we validated our model’s effectiveness, demonstrating significant improvements in path planning precision, task allocation efficiency, and inter-robot collaboration performance. These results highlight our model’s practicality and offer new insights for future research and applications in multi-robot systems. Overall, our integrated model addresses key issues in multi-robot collaboration, contributing an innovative solution to the field’s development. This research provides a novel approach for path planning and task allocation in multi-robot systems, laying a solid foundation for deploying intelligent robotic systems in complex and dynamic environments.