Multi-agent System Dynamic Task Allocation Research

In today’s society, the complex systems are related to the task allocation issues in each domain Dynamic task allocation is to allocate tasks to each performer in the system real-timely and reasonably, so that the system resources can be fully used and the production efficiency can be improved in this way.The Multi-Agent system can simulate all kinds of complicated system in the practical applications and it allows resolving a variety of system task allocation problems by applying the Multi-Agent system, therefore, the study has important research significance.Taking Multi-Agent system as a study model, with referencing to the research achievements and technology methods on task allocation theory, this report has made a discussion on task allocation of the two structures of the Multi-Agent system.Firstly, regarding the centralized system structure, the dynamic task allocation model which is based on the Markov Decision Processes, MDP can be taken to discuss the optimization of the task allocation scheme, secondly, in terms of the distributed system structure, build task allocation model based on extended contract net protocol, standardize the dynamic task allocation process which is based on extended contract net protocol (ECNP) by describing the operation framework and process of the task allocation system.For centralized dynamic task allocation issues, According to the target of maximizing the system operation income, Environmental uncertainty and the dynamic task allocation sequence decision requirements. The article constructs the corresponding markov decision process model, proves optimum solution exists when the state space and action space is limited, and discusses the iteration algorithm.The simulation experiment shows that since the dynamic task allocation based on the MDP has considered the long-term gains of the system, Although some decisions in some time point are not instant and optimal, the expected total revenue of long-term operation system is better than the instant optimal allocation plan. Therefore, Dynamic task allocation modelRegarding distributed dynamic task allocation issues, on the basis of building the theory of extended contract net, It has proposed basic idea of solving dynamic task allocation through the description of two level framework, which based on dynamic task allocation of MAS. Standardize the operation procedure from the preparation, invitation of tender, bidding and release. The dynamic task allocation based on ECNP can effectively direct task allocation problem of complex system under the distributed environment.In general, the dynamic task allocation study on the basis of MAS is not only the need of real application, but also the inevitable tendency of development of theory...