Font Size: a A A

Multi-Agent Cooperation Mechanism Based On Cooperative Co-Evolution

Posted on:2005-01-25Degree:MasterType:Thesis
Country:ChinaCandidate:Q TangFull Text:PDF
GTID:2168360122981249Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Cooperation is one of the key problems of the autonomous multi-agent robotic system all the time. In the task that multi-agent move multi-object cooperatively, the cooperation effect is not optimal by the protocols existed. So, it is necessary to lucubrate the multi-agent's cooperation mechanism. In this paper, the cooperation mechanism of multi-agent based on cooperative co-evolution is proposed, with which the task can be completed in shorter time.At first, the areas with the same size are assigned to the agents, and each agent searches the objects in its area. Then, with the reference of the agent's ideal cost, the objects are assigned to every agent balanced through the two assignment processes of the task decomposition. And then, the initial populations corresponding to each agent are created in which an individual is made up of the permutation of the objects' number randomly. And then, during the evolutions of the populations, the new individuals are evaluated, and an individual is endowed with higher fitness if it cooperates more effectively with the representatives of the other populations, other wise, lower fitness will be endowed. During the populations' subrogation, the individuals with lower fitness in the father generation are replaced with the new individuals with higher fitness to form a new generation, such that, the populations evolve to the direction of better cooperation among the agents. At last, the objects are moved according the sequence that the agents get after the populations' co-evolution ends.With this mechanism, an experiment that three homogeneous agent robots move eight objects by cooperation has been done, and through which the result that the better cooperation effect by the multi-robot based on cooperative co-evolution is gained compared with the protocols existed. The test process shows that the stability of the mechanism is good, and the algorithm is convergent. Therefore, the cooperation efficiency of the multi-robot system can be improved under this new cooperation mechanism. And the accomplishment of the new cooperation mechanism not only enriches and develops the cooperation theory of the multi-robot, but provides the research of their cooperation behaviors with a new idea.
Keywords/Search Tags:multi-robot system, cooperation, co-evolution, task decomposition
PDF Full Text Request
Related items