A Explainable Method for Task Planning Scheme Based on Hierarchical Reinforcement Learning

In view of the low transparency and insufficient explainability of task planning schemes based on intelligent algorithms, this paper proposes a explainable method of task planning schemes based on hierarchical intensive learning. Firstly, the technical architecture of the results of the hierarchical intensive learning task planning scheme is proposed, the results of the task planning scheme are divided into macro-strategy layer and micro-strategy layer, then the specific steps of the hierarchical intensive learning method are given, and finally the backtracking process of macro and micro strategies is explained through the task decomposition diagram. The simulation proves that this Hierarchical Reinforcement learning structure based on state conditions can clearly show the action selection logic in different states, and then help the commander to better understand and optimize the action strategy of the agent.