This work considers the problem of using multiple small, low-cost robots, with a limited range of local communication ability, to collaboratively search and engage an indeterminate number of tasks in an unknown large-scale hostile area. In this project, we provide an approach, which we refer to this schemas the gradual expansion-based exploring approach (GEBEA), to explore environment by the cooperating robots.  We also propose a task allocation approach for solving the multi-robot multi-task allocation problem. The key idea in these approaches is to make the robots automatically build up and maintain a dynamically stable ad hoc network when they are exploring or engaging tasks. That makes the robots get the advantage of behaving both collaboratively and decentralized, which normally is only available in the system that has global communication. In this case, although each robot only has limited range of communication, by using the ad hoc communication network, all of the information obtained by any robot can be shared by the others. This allows robots to make decisions based on global information and reduce the possibility of overlap exploration. At the same time, the dynamic central control architecture generated by the robots, which finds the target task during the exploration period, can generate an efficient task solution.

      A simulator has been implemented to test our approaches. According to the result of the simulation, the approach has a better performance than the parallel alignment approach  when they are tested in an unknown large-scale static environment that has indeterminate number of tasks. To test our collaborative strategy in a dynamical environment, our next step is to implement this collaborative search and engage strategy in a large-scale disaster rescue simulator. The implementation of the simulator is based on the RoboCup Rescue Simulator (RCRSS), a computer simulation system providing a virtual environment where large-scale disaster such as earthquakes can be simulated and heterogeneous rescue agents, which simulate the real world rescue robots, can collaborate in the task of disaster mitigation.