Multi-robot task allocation based on utility and distributed computing and centralized determination

This paper proposes a Distributed Computing and Centralized Determination (DCCD) method to solve the multi-robot task allocation problem, in order to maximize the utility of the whole robot system. First, a utility model is presented which takes the cost for executing tasks and the quality of task completing time into consideration. DCCD employs each robot to compute and provide sub-plans for executing one or multiple tasks. Then, the task manager forms allocations for accomplishing tasks using all the sub-plans and determine the optimal one according to the utility model. Compared with fully-centralized allocation, this method can reduce the computation largely for task manager. Theoretical analysis and simulation verify the effectiveness of DCCD, and shows that DCCD can obtain global optimal allocation comparing with the fact that the widely-used single-item and combinational auction methods can only obtain local optimal solution.