Enchères pour le Maintien des Communications lors de l'Allocation de Tâches

Abstract

— In the last ten years, various industrial applications of multi-robot systems have emerged. Among these are systems executing pick and delivery missions, such as the well known autonomous terminal in Rotterdam, as well as many systems tasked with surveillance missions. A key step in the execution of any multi-robot mission is the resolution of the multi-robot task allocation problem. It consists in assigning the tasks of the mission to the robots. If one needs to optimize an objective representing the performances of the system while assigning the tasks, it becomes an integer programming problem. To solve multi-robot task allocation, researchers have proposed many approaches. Among these proposals, a class of approximate methods, based on auction mechanisms, has drawn the attention of researchers for their ability to quickly reallocate tasks if this improves the execution of the mission. In this paper, we introduce a new term in the evaluation of bids for an auction based task allocation protocol. This new term enables us account for the connectivity of the communication network, which represents the communication links between the robots. The connectivity of the communication network is key to the efficiency of auction-based methods, as they need to share auction messages as broadly as possible to produce efficient allocations. We evaluated our method in a surveillance scenario. We derived theoretical bounds of the time complexity of the evaluation of the bids and of the size of the data shared during the mission. We demonstrated through simulation experiments that improved communications increase the robustness of the multi-robot system to dynamic events.