Spatio-Temporal Clustering of Tasks for Swap-Based Negotiation Protocols in Multi-Agent Systems

Path planning in robotic agents consists in determining, on a network of places and routes modeling the environment, a sequence of resources to be visited in order to carry out a set of tasks. In multi-agent systems, agents may cooperate to decrease the task execution costs; the Contract Net Protocol is an approach to negotiation of tasks based on the announcement-bid-award mechanism. If the agents do not own money, they can decrease their costs only by swapping tasks with other agents; unfortunately, swapping only single tasks may trap negotiation in local minima of the cost. In this paper we show how the utility of negotiations can be increased by allowing tasks to be swapped in non-disjoint clusters. Clustering of tasks is carried out, in a fuzzy fashion, according to two orthogonal dimensions which consider, respectively, the spatial disposition of the resources within the environment and the temporal progression of their time windows; the approach is scalable to support other dimensions as well.