The RoboCup Rescue Simulation domain [1], is a challenging multi agent environment, in witch tasks need to be assigned collaboratively to different types of agents. In this domain, achieving the main goal requires adaptive high performance Cooperation of several agents. The challenging problems in Rescue Simulation are: (1) Lack of information from the environment (Only local information is available for each agent), (2) The dynamic environment makes the available information unreliable after a period of time, (3) Communication between the agents is limited, so its precious to transmit as much data as possible in low size messages. In this paper, we'll show our methods to deal with these bottlenecks in the system.
展开▼