A decentralized approach to multi-robot formation initialization

摘要

This paper presents a novel solution to the formation initialization problem, in which autonomous mobile robots must arrange themselves into a specific geometric configuration without centralized control, without explicit inter-agent communication, and using only information from local sensors which is necessarily incomplete. Our solution to this fundamental multi-robot coordination problem is based on satisficing theory, a paradigm for decision making in which individual options are evaluated by comparing potential gains with projected costs. An option can be justified as adequate if its benefits outweigh its costs. In multi-agent systems inclined to cooperation, satisficing offers advantages relative to traditional utility-maximization approaches. We present experimental results that demonstrate cooperative, emergent multi-agent behaviour. We compare the performance of various enhancements within the satisficing framework across a collection of scenarios that differ in starting positions, target formations, number of agents involved, and the presence of static obstacles. The results suggest that satisficing is an attractive alternative for the synthesis of cooperative multi-robot systems.