An Extension to Least Committed Graphplan for Handling Conditional Effects

Since Graphplan was proposed, intelligent planning has become a research focus in AI. This paper provides a new algorithm that extends Least Committed Graphplan to handling conditional effects. Different from the previous work on it, we employ the sub-mutex relations between action components, which makes the construction of planning graph much faster. In addition, we distinguish brotherhood components and non-brotherhood components, and treat them respectively during solution extraction, in which efficiency of plan synthesis can be improved greatly. The algorithm can find the valid plan just as Graphplan can do, and the problems handled by the algorithm are more than those of Graphplan. Moreover, because of wide application of intelligent planning, our research is very helpful to the development of robotology, natural language understanding, intelligent agent and so on.