A combinatorial auction framework for decentralised task allocation

This paper proposes a framework for decentralised task allocation in which agents find a global solution sharing local performance metrics. Leveraging the PAUSE procedure for Combinatorial Auctions we outline a decentralised version of in order to solve a task allocation problem. The algorithm has polynomial communications overheads and it can also accommodate a variety of objective functions and task dependencies. To investigate its performance, some numerical experiments were carried out and their results were compared to those of an optimal solver and a representative alternative algorithm, namely CBBA, solving the problem with three representative objective functions: MinSum, MinMax and MinAve. The results show that, overall, the proposed algorithm outperforms the current state of the art alternative, moreover the results for MinSum show that it always finds optimal solutions while it remains close to the optimal for MinMax and MinAve.