Multi-robot task allocation clustering based on game theory

A cooperative game theory framework is proposed to solve multi-robot task allocation (MRTA) problems. In particular, a cooperative game is built to assess the performance of sets of robots and tasks so that the Shapley value of the game can be used to compute their average marginal contribution. This fact allows us to partition the initial MRTA problem into a set of smaller and simpler MRTA subproblems, which are formed by ranking and clustering robots and tasks according to their Shapley value. A large-scale simulation case study illustrates the benefits of the proposed scheme, which is assessed using a genetic algorithm (GA) as a baseline method. The results show that the game theoretical approach outperforms GA both in performance and computation time for a range of problem instances. •Cooperative game theory tools are considered to deal with MRTA problems.•Robots and tasks are defined and ranked in a game according to their Shapley value.•An algorithm is proposed to group the players into balanced clusters.•Randomized methods are applied to large problems to relieve the computational load.•The feasibility is assessed in a large scenario and contrasted with a genetic approach.