嫦娥三号“玉兔号”巡视器行为规划方法

为支持"玉兔号"巡视器完成对月面较大空间范围的科学探测,对动态任务进行快速规划,提出了一种基于人工智能领域智能规划技术的自动化任务规划方法。提出了行为持续时间和行为效果动态确定的时态规划模型(TP~(DDDE)),设计了描述TP~(DDDE)问题的规划领域定义语言PDDL——PDDL~(DDDE),以及针对TP~(DDDE)问题的启发式规划算法,运用Landmark知识分析规划问题结构,从而设计了合理反映动作前提评估顺序的启发函数。在嫦娥三号任务中,本文方法支撑"玉兔号"巡视器圆满完成了预期的科学探测任务。

Activity planning method for Chang' E-3 "Jade Rabbit" rover

To enable the "Jade Rabbit" rover to plan fast in a dynamic work environment on the lunar surface, an activity planning model called temporal planning with dynamic duration and dynamic effect (TP^DD&DE) was developed based on the automated planning technology of artificial intelligence. This model extends temporal planning to incorporate external computing procedures and to support durations of actions to be determined dynamically. To describe tasks of TP^DD&DE, an extension of the planning domain definition language(PDDL), called PDDL^DD&DE, was designed. A planning system that is based on the heuristic state space search framework was developed to solve the planning problem. By utilizing the "proposition Landmarks analysis" technique, we designed a heuristic function that accounts for reasonable orderings of actions' preconditions and is capable of getting better goal distance estimates for states. With the successful completion of Chang'E-3 mission, the proposed method played a key role in supporting the "Jade Rabbit" rover to fulfill its exploration tasks on the moon.