Behavior Tree Generation Study for Multi-agent

Deep reinforcement learning algorithms perform well in the field of control and artificial intelligence, but still face problems such as poor interpretability. Combined with the structural advantages of behavior tree, research on behavior tree generation and optimization based on reinforcement learning has become one of the ways to solve the problem. However, due to the high flexibility of behavior tree, the generation mechanism based on data-driven methods often has difficulties. Therefore, the paper mainly proposes a normalization method of behavior tree, which is applicable to the research of behavior tree generation based on multi-agent reinforcement learning. On this basis, combined with the classification tree model, the framework and algorithm of behavior tree generation for multi-agent systems are designed to provide a basis for the subsequent research on the interpretability of multi-agent reinforcement learning.