PLANNING WITH PREFERENCES

Abstract

Planning with preferences finds solution by considering user preferences. Since obtaining optimal plan is computationalintensive, most of the planers try to find single solution which satisfies the user preferences. In this thesis planner will generate multiple solutions, to provide choice of selecting a plan from the set of solution plans. Those plans will be generated using preferences given by the user. Here two methods of finding solution plans are proposed, one without heuristics and another with heuristics, and those are compared. Modifications to depth first search technique for planner without heuristic and modifications to hill climbing algorithm for planner with heuristic are proposed to search the state space.The heuristic value required to select successive node and for deriving multiple solutions is computed using user preferences, To extract multiple solutions from the relaxed planning graph an algorithm is proposed which is modification to single solution extraction algorithm for relaxed planning graph, in which all possible solutions are extracted, by considering all possible actions to achieve goal predicates, there by computing heuristic value which can guide in obtaining multiple solutions. Even though the state space is limited by the user preferences, it is normally large compared to solution length. Thus heuristic planner is faster.