The Description And Analysis For Reasoning About Action Based On Colored Petri Net

Reasoning about action of intelligent agent is surely one of the most exciting and challenging areas in Artificial Intelligence.Intelligent agents are embedded in an incomplete,dynamic environment,and this requires the agents have higher cognitive capabilities such as reasoning about action,devising a plan,and drawing logical conclusions from sensor information acquired,to find a plan from the initial state to the final state.Petri Nets,as a formal tool special suitable to specify currency and synchronization and having graphical representations and rich analysis approaches, are very suitable to represent and simulate the systems characterized as synchronization,concurrency,and dynamics.This paper describes reasoning about action based on the advanced net of Petri Nets—Colored Petri Nets and the main works are as follows:(1)Based on Colored Petri Nets a new formalized representation method—CPNSRAA net is introduced.The states,actions and the compound relations among actions(such as sequence,iteration,condition,synchronization concurrency,conflict, true concurrency and interruption) are constructed by using CPNSRAA net.(2)A method has provided for constructing CPNSRAA net system according the initial conditions,goal of the system and the pre-conditions and post-conditions of action,which can make the agent reason about actions automatically and goal-oriented in the dynamic environment.An algorithm used to produce action sequence in CPNSRAA net system to reach goal has also provided.(3)A CPNMRAA net model has introduced to model the multi-agent systems, which can be divided as:agent module and environment module.The agent module contains action module and communication module,while the environment module contains synchronization module,communication processing module and action processing module.The case of multi-agent reasoning about action in robot office environment are modeled and analyzed by using CPNMRAA model.(4)The researching contents are simulated.The case of simple agent reasoning about action in robot office environment and the case with time have simulated,which proves CPNSRAA net as a feasible and effective representation for reasoning about action can reason the action sequence to achieve the goal perfectly,and the boundedness and liveness of the CPNRAA net we modeled also are verified by the simulating reports.Our paper takes the advantage of the asynchronism and concurrency of the formal method of Petri net,and does some tries on using Petri net to represent reasoning about action of agents,which lays the foundation for the future research and the spread for the technology of reasoning about action of agents.