| Intelligent planning is one of the most important research fields in AI. Classical planningproblem supposes the information about the world is totally complete and when an action isperformed, the effects of the action are definite, which limits the classical planning algorithmsonly to solve small-scaled model planning problems. To solve real problems, nowadays,people have attached importance to the research on incomplete information and uncertaineffects. Probabilistic planning has extremely aroused attention because it can quantitativelydescribe uncertain information accurately. Blum & Furst developed Probabilistic Planning inthe Graphplan Framework.In an environment where multiple agents exist, they would cooperate and collaboratewith each other to solve the problem of the user. This cooperation exists because the domainof the problem may be too big to be handled by a single agent or the problem require agents,having different knowledge or skills, to contribute to the final solution of the problem. Agentscommunicate and cooperate to solve a problem, to resolve conflicts, and to achievecoordination in the system. An agent may also interact with other agents to satisfy a personalagenda in its plan, where they may interact by negotiating or bidding to perform or delegate atask. Agent-based architectures offer modularity, robustness, separation of concerns, andother advantages of a distributed system. These characteristics have led agent technology togain applications in wide areas.By analysis of PGP, we introduce a novel approach of Probabilistic Planning modelbased on Multi-Agent technique, which is called PGP Based on Multi-Agent (MAPGP) in thispaper. Depending on Agents' parallel execution, MAPGP will appear at higher speedcompared with PGP. Extending planning graphs, information propagation and searching for aplan, included in MAPGP, are expounded. Finally, We summarize the value of Multi-Agenttechnique in intelligent planning in the future. |
PDF Full Text Request