A Plan Recognition Algorithm Based on the Probabilistic Goal Graph

Based on the Goal Graph, we construct a new structure named Probabilistic Goal Graph (PGG), which uses the goal nodes instead of state nodes, and adds the observation nodes. According to the background knowledge, we calculate the probability distribution of the observed actions, and then calculate the probability of each goal node, ultimately get an optimal plan which has maximum probability to achieve the goal. By putting the probabilistic values into the goal graph, it can be effectively recognized the partial observed actions, and overcomes the defects that the goal recognition algorithm can not find the optimal plan.