Task Allocation And Cooperation For MAS In The Dynamic Environment

The multi-agent system (MAS), has become an important research areas of artificial intelligence. How to handle the cooperation among agents is the key to play the advantage of the MAS. By advanced structure and cooperation mechanism, MAS can be capable of solving all kinds of complex tasks.Based on the research of agent architecture and multi-agent architecture, this thesis constructs a MAS structure for task-level cooperation, also, the cooperation and task allocation problem of MAS is studied.Firstly, this article detailed the coordination mechanism of multi-agent system. For the collision avoidance and arrest problem of multi-robot, a strategy is presented which based on the consultation and the will. An objective function is optimized that obtain the optimal capture point surrounding"invader". The robots first converge on the capture point, after the"invader"is captured, push it to the goal point. Combine the move to goal behavior and collision avoidance behavior by consultation to make sure no collision.Secondly, in order to take into account many uncertainties during the process of robot ability evaluating, Set Pair Analysis is introduced to establish a multi-index synthetic evaluation model for MAS. This dissertation depicts the robots ability to complete task by identical degree, difference degree and closeness degree. Set-pairs proximity, which is analyzed by the method of Euclidean space weighted distance, is used as fitness between robot and task. This evaluation model provides a reliable basis for task allocation of MAS.At last, it briefly introduces the basic method of task allocation, proposes a multi-task allocation mechanism based on auction contract net, and a measure based on matrix reduction. In the past bidding mechanism, agents give the competing values according to their own knowledge, with subjective judgment. But in the modified bidding mechanism, agents only want to issue their abilities to manage agent within the prescribed time. After obtaining the information, manage agent comprehensive evaluates the fitness between robot and task according to the index. Then it realizes the rapid and effective optimal allocation based on fitness matrix.It is of innovation and pioneering on the research contents and research approach. Task allocation model proposed in this article has some reference value on further research of multi-agent system. The thesis also completes some fundamental work to the further research and popular application of MAS in the future.