Hierarchical Extraction, Planning and Behavior Tree Process Control Based on Historical Trajectory

In recent years, the research of task planning has made great progress. The use of hierarchical information to reduce the planning scope and time is one of the main ways of planning implementation. However, at present, the layered information of planners such as HTN generally needs to be set manually, so how to obtain the layered information automatically and efficiently is a big challenge. In order to deal with these challenges, this paper designs a hierarchical extraction algorithm based on PDDL description and historical planning trajectory, and innovatively deals with the fixed order problem between the operators at the same level through causal inference, so as to avoid the planning failure caused by the contradictions of them. In addition, a hierarchical programming algorithm for PDDL is designed based on hierarchical forward programming, which guarantees the completeness of solutions compared with reverse programming. Finally, the algorithm uses the behavior tree to control the process of the solution of the planner, and uses another distributed multi -agent scheduling system based on the behavior tree to allocate the task execution relationship, so that the algorithm can adapt to the constraints of the start and end of the collaborative task.