Abstract:

To address the coupling problem and decision-making problem of multiple tasks such as task allocation and path planning of multi-agents in complex application scenarios， a design method of hybrid intelligent decision-making framework for multi-agent and multi-coupling tasks is proposed. Firstly， the advantages of single agent multi-task hybrid framework and multi-agent distributed collaborative control， a hybrid intelligent decision-making framework for multi-agent and multi-coupling tasks is designed. Secondly， the strategy network of the framework and the training controller for the strategy network are designed and a coupling relationship matrix based on coupling relationships is proposed to achieve efficient training of multi-agents and multi-tasks in face of collaborative decision-making problems. Finally， this paper modeled， trained algorithm and simulated in simulation environment，and compared with the tradition method to verifies the effectiveness and advantages of the proposed method.

Key words: intelligent decision, deep reinforcement learning, multi-agent, path planning, task allocation