A new hybrid task sharing method for cooperative multi agent systems

In this paper, we discuss a novel method for task sharing and cooperation between Firefighter Agents in Roborescue simulation environment. This method consists of three main sections, which has an effect on the agents´ decision-making process, based on various parameters. The first section is the selection of agent´s individual goal. The second section is the supervision of the fire hall center. The aim of this section is the supervision of the center on selecting goals by firefighters, arranging the firefighter groups and producing appropriate advice for guiding the groups´ activities into optimal solution. The third section is selecting the group´s goal by cooperation with other group´s members. The results have been assessed from different viewpoints such as primary reaction speed, number of passed messages and final score based on T-Test statistical distribution. Overall, this method has better results in comparison with the usual reactive, distributed, and centralized task allocation methods. There are some important advantages to this method such as appropriate primary reaction speed, robustness to destruction of some agents or center, robustness to limited communication and higher final score