A Norm Improvement Method In An Exception Scenario

Agent system and Multi-Agent System(MAS)is a valuable research direction in the field of artificial intelligence and computer.MAS is an efficient system with internal cooperation and independent division of labor.MAS is large and complex,it can separate multiple independent single Agent systems.The core problem of MAS is to control the coordinated interaction between agents and the normal operation of single Agent system.In order to maintain the correct operation of the system,there is a norm set inside the MAS and the single Agent system,reasonable interaction and mutual restriction between agents are controlled by norms.Within a single Agent system,norms are used to effectively restrict and maintain agents from performing illegal behaviors.In the face of such a complex and large MAS,there will be exceptions in the system due to various reasons.Restricting the behavior of each Agent system can avoid the exceptions in the MAS.The efficient method to solve the exceptions and maintain correct operation of the system in the Agent system can be traced back to the norms.By improving the norms and making up for the rigorous of the norms,and the exceptions can be well avoided.This paper mainly studies the norm improvement based on an exception scenario in the single Agent system,solving the exception in the Agent system by norm improvement.In the Agent system,when an executed norm was violated,it is called there is an exception in the system,and the norm is in an exception state.The exception is the research scenario to improve the rigor of preconditions of norms.When the precondition of a norm is not rigorous enough,the expected postcondition cannot be achieved after execution,and the norm was violated.In this paper,based on the recombination in evolutionary computing theory,we propose an exception handling method of an Agent system based on norm improvement.Each executed norm is defined as a node,the triggering relationship between two norms is expressed as a directed edges,and the execution of all norms in a scenario is graphed as a directed acyclic normative graph.By mining the effective information in the normative graph and analyzing the association between the environment states before and after the node execution,the key path for norm improvement and solving the exception can be accurately located,and the exception processing and norm improvement process can be completed synchronously.Postcondition is the expected target after the execution of norms.Exception handling method gives the norm in the exception state an ability to learn from the postcondition of other rules.The way of learning is recombination,and the precondition of the norm in the exception state can be improved.When improving the precondition of a norm,it is adaptive to be consistent with the authority of the norm itself.Finally,the simulation experiment simulated the trigger situation between norms and the execution of exception scenarios,realized the exception handling method of an Agent system based on norm improvement,the norms in the exception states were improved,and verified the feasibility of the exception handling method to improve the norms.Norms are improved in the exception scenarios,and the improvement of norms avoids the exceptions in the Agent system,which is a complementary process.From the perspective of the Agent system,the exception handling method assists the Agent system to improve the ability of exception handling,and the stability of the Agent system is improved.From the perspective of evolutionary computation,a norm is regarded as a gene.The mutual learning ability between norms makes norms evolve continuously.An evolutionary gene will be stronger and better to adapt to the changes of dynamic environment,and the probability of gene mutation will be reduced at the same time.