Multi-agent-based Auctions for Multi-robot Exploration

Abstract

Aiming at improving the efficiency of the multi-robot exploration, this paper presents a novel auction algorithm based on the contact net protocol of the multi-agent theory. On the basis of the analysis of the limitations of the usual auction methods to the multi-robot exploration, the contract net protocol in the multi-agent theory is presented to design a completely distributive structure for the multi-robot exploration in order to increase the tolerance and reliability of the whole system. Then an improved multi-round single-item auction algorithm is presented to assign the frontiers to the multiple robots effectively. The theoretical analysis and the simulation experiments prove that the algorithm can lead to the effective coordination of multi-robot exploration with the decreasing computation compared to the combinatorial auctions