Research And Implementation Of The Sensory Graphplan Based On Heuristic Search

Recently, due to intelligent planning's broad application in many fields, some scholars proposed uncertain planning problem. However, classical Graphplan cannot solve the uncertain planning problem. It is difficult for uncertain planning problems in intelligent planning domains, but they more approach the real world problems. Some planners for uncertain planning problems are developed, which are based on Graphplan, such as SGP, CFF and PFF. Sensory Graphplan-SGP is the most famous one that Daniel S. Weld,Corin R. Anderson and David E. Smith proposed in 1998.The uncertain planning algorithms have some restrictions, e.g., its algorithm is not simple and time complexity is very high. Also, it doesn't use heuristic research, and the research from the initial situation is slow. So on the whole the performance of planner SGP is highly poor. we propose a new algorithm based on SGP for uncertain planning problems to solve this problem.This paper makes research on a kind of complex planning problems—uncertain planning problems. The paper applies a novel intelligent planning algorithm which is Sensory Graphplan with heuristic research. In contrast to the existing methods, the algorithm adopts the heuristic methods used in FF. At the same time, it introduces the relax action also used in FF. Hence the algorithm needn't deal with mutex relations, and advances the approach of forward graph expansion. The proposed method improves the planner solving ability, both theoretical and practical application of great value.