Multi-Task Coalition Generation of Multi-Agent System with Limited Resource
Ning Qiang, Feng Ju Kang · Advanced materials research · 2014
A new fitness function is introduced in order to maximize the number of task served by the multi-agent system (MAS) with limited resource, while the tasks information remains unknown until the system found them one by one. The new fitness function not only considers to maximize the profit of the system which can be seen as to maximize the remaining resource of the system in the case of the MAS with limited resource, but also takes the balance of remaining resource in to account and it can makes a compromise between them. This paper uses an improved discrete particle swarm optimization to optimize the coalition of MAS. In order to improve the performance of the algorithm we redefine the particle velocity and position update formula. The simulation results show the effectiveness and superiority of the proposed fitness function and optimization algorithm.