一种基于时标状态的启发式航天器任务规划算法

深空探测领域对实时性要求较高,在较短时间内找到规划解是深空探测自主任务规划中的一个要求,运用启发式规划算法是达到该要求的方法之一。而深空探测自主任务规划的另外一个特点是需要处理持续动作和数值信息。针对深空探测任务特点,采用规划领域定义语言PDDL,建立深空探测领域中知识模型,描述操作中遇到的时间与资源约束;随后应用以条件数为代价的启发式搜索方法对深空探测规划问题进行求解,并将其与TFD规划器中以动作时间为代价的上下文增强累加启发式搜索方法得到的结果进行对比,得出以条件数为代价的启发式搜索方法在搜索速度方面效果更佳,满足深空探测自主规划任务实时性要求。

Time Stamped States based Heuristic Algorithm for Spacecraft Mission Planning

For real time in deep space exploration, it is a requirement of autonomous mission planning for the explorer to find a plan as soon as possible. A kind of method is to use heuristic algorithm. At the same time, durative actions and numeric information have to be processed. According to these characteristics, this paper adapts planning domain definition language (PDDL) to establish knowledge models and describe time and resource constraints. Then the heuristic algorithm based on condition number is proposed to solve planning problems of deep space exploration. Finally, we compare this heuristic with context-enhanced additive heuristic based on action time in TFD (Temporal Fast Downward) planner. The result of the experiment shows that the heuristic algorithm we proposed is better to solve the planning problems in deep space from the point of view of real time.