机器人问题求解中的评价函数及矛盾和冗余的消除

本文提出了一种求解复杂的机器人问题的算法,该算法采用子问题求解与分层策略相结合,在动作规划过程中,建立了初始状态下的前指针和目标状态下的后指针,并形成一个决定动作排序优先级的评价函数,按不同的空间和时间段的差异,将全局的动作序列划分为若干子区域。本算法证明只需在子区域范围内消除冗余和矛盾,即可大大地提高系统的效率,在子区域范围内出现矛盾时,用优先级高的动作取代优先级低的动作,变矛盾为冗余,最后保留排在最前面的动作,而删除后面相同的动作,即消除了冗余。

Formulation of the Evaluation Function and Elimination of the Redundance and Contradiction in the Robot Problem Solving

This paper proposes an algorithm which combines sub-problem solving with hierarchical strategies to solve complicated robot problems successfully. It is well-known that,in hierarchical planning,the greatest difficulty is the elimination of redundance and contradiction. To solve this problem,the algorithm formulates an evaluation function of action priority in terms of the front pointer in initial state and the back pointer in goal state in the process of action planning. It then divides the long global action sequence into several sub-sequences according to the difference between the spaces and time slices. When contradictions occur in these sub-sequences, they are turned into redundances by the replacement of the actions of lower priority with those of higher priority. Finally, the redundances are also eliminated by the retention of the foremost actions and the elimination of all the i-dentical actions behind them. It shows that the efficiency of the system can be considerably improved when the redundances and contradictions are eliminated in the sub-sequences. Furthermore, the problem of the goals interaction is solved by a smart combination of the evaluation function and non-linear technique.