Binary max-sum for multi-team task allocation in RoboCup Rescue

Coordination of agents involved in rescue missions is an important open research problem. We focus on the RoboCup Rescue Simulation (RCS) challenge, where different teams of agents perform urban rescue operations. Previous approaches typically cast such coordination problem as separate single-team allocation problems, and solve them separately. Our first key contribution is to focus on the max-sum approach, which has been successfully applied in this setting. We show that it is possible to reduce the computational complexity associated to maxsum from exponential to polynomial time. Our empirical evaluation shows that, by using our approach, the fire brigades team obtains significantly better results when compared to state-ofthe-art approaches. Our second key contribution is a methodology that allows teams in RCS to make joint allocations. Specifically, our approach supports a modular design, where teams are independently modeled and subsequently connected via well-defined coordination points. To the best of our knowledge, this is the first task-assignment approach in the literature that enables teams in RCS to make simultaneous joint allocations. Experiments with fire brigades and police agents show that teams employing inter-team coordination are significantly more effective than uncoordinated teams.