In the large-scale disaster (LSD) rescue operation, an effective cooperative strategy for the rescue team can considerably increase the rescue speed, effectively use the limited resource and decrease the casualty and the damage of property during the disaster period. However, the dynamical changed environment and un-predicated multi-task arriving make the rescue teams cooperation different from normal cooperation scenario. The narrow long-range communication bandwidth between rescue teams makes it worse. This proposal suggests a multi-agent cooperative strategy that helps rescue team cooperatively explore the dynamical changed environment, compose global environment information map and allocate different tasks to different multi-agent group. In the present project, every rescue individual is considered as an intelligent agent. By applying the “dynamical sub-task roaming” algorithm proposed in this project and utilizing the high-speed bandwidth of the short-range communication between rescue agents, the system can implement most of cooperative work. To verify the efficiency of the strategy, an earthquake simulation system will be developed and a physically implemented heterogeneous robots system will be build. At the same time, the performance of the proposed cooperative strategy will be tested in the simulation system


      Since an effective cooperative strategy for the rescue team can considerably increase the rescue speed, effectively use the limited resource and decrease the damage of property and casualty during the disaster period, we will concentrate on the fundamental theoretical issues that impinge on cooperation. The Multi-agent cooperative (MAC) project proposed in this project will focus on the researching cooperative strategy for heterogeneous rescue teams in a LSD environment. To simplify the complexity of the system, we consider each individual in the rescue team as an intelligent agent, whether this individual is a person with PDA or an intelligent robot. That also suits our imagination about future rescue operation. We image in the future, the rescue teams are composed of high intelligent robots specially design for different purpose. 

      As the initial research of the LSD rescue strategy, most of the work of the project will be based on a LSD computer simulation system. As the most often happened LSD and caused a great amount of damage, earthquake has been studied from different perspective. In MAC project, an earthquake simulation environment will be developed and used as the test LSD simulation platform for testing the rescue agents’ cooperative strategy. In this earthquake simulation system, a simple LSD rescue scenario is adopted. The rescue works in the simulation platform are simplified to three tasks: rescuing human victims, mitigating spread fires and clearing the blocked roads. The mobile rescue agents used in the rescue operation are all highly intelligent robots that are specially designed as: ambulance teams, fire brigades and police. The rescue teams in the simulation system do not include any human rescue personnel. Since the robots are specially designed, they cannot switch their rescue job with other kind of robots. 

      To find an effective cooperative strategy for those rescue robots and implementing this strategy in this earthquake simulation environment, the work list below will be completed in the MAC project: 

· Implementing the computer simulation platform of an earthquake environment.
· Developing a suitable communication infrastructure for the thousands of heterogeneous rescue robots.
· Developing virtual robot teams act as the rescue agents in the disaster simulation system. The robot teams should be composed of different kinds of robots that can handle different disaster situations. And the robot teams can stand loss of a portion of the robot team by developing robust task-allocation methods.
· Devising and achieving cooperative strategy of heterogeneous agents for searching, rescuing the victims and mitigating the damage caused by the disaster.
· Building a real heterogeneous robots system in the MRL lab and implementing the multi-agent cooperative algorithm on this physically implemented multi-robot system and measure its effect.