Multi-Robot Task Allocation Based on Combinatorial Auction

Abstract

In recent years, the rapid development of robots has brought many conveniences to the production and life of human beings. The research on robots is of great significance. Task allocation is one of the most important issues in the field of multi-robot research. As the working environment of robots is inherently highly dynamic and uncertain, multi-robots also have the problem of communication and cooperation, how to quickly and accurately allocate tasks in these complex environments is very important. In this paper, aiming at the real-time task assignment problem of agreed time coupling, the multi-robot task allocation topology of mixed committee decision-making is adopted. The two intelligent agents of manipular and mobile service robot are heterogeneous, and the task allocation of multi-service robot based on combined auction is realized. Based on the traditional method of simulating the market mechanism, this paper discarded the single commodity auction method that caused low allocation efficiency, followed the mechanism of using combined auction to improve efficiency, and proposed a “social welfare maximization” mechanism which considers the budget and improved from lock-in auction method, each agent takes the future cost into consideration in the bidding process, so as to maximize the overall benefits of the multi-robot system. In addition, solving the high-dimensional NP hard problem of largescale robot task allocation usually leads to an explosion of dimensions, this paper proposed a dimension-reduced combined commodities auction method. The subtasks decomposed from the parent task are divided into types and numbers of service locations and service objects, and a combination of lower dimensionality is selected to form each group of combined commodities that can be classified as traveling salesman problems. Finally, three combined commodities auctions at different time points were conducted in a simulated restaurant scene. The experimental results show that the multi-service robot task allocation method based on combined auctions proposed in this paper is fast, efficient and accurate.