Stigmergic Navigation for Multi-Agent Teams in Complex Environments

Abstract

Summary. Robotic agents in dynamic environments must sometimes navigate using only their local perceptions: for example, in strongly dynamic domains where paths are outdated quickly. Reactive navigation alone has limited power: domain features such as terrain undulation, geometrically complex barriers, and similar obstacles form local maxima and minima that can trap and hinder agents that use it exclusively. Moreover, agents navigating in a purely reactive fashion forget their past discoveries quickly. Preserving this knowledge usually requires that each agent construct a detailed world model as it explores or be forced to rediscover desired goals each time, and share elements of this knowledge explicitly in group situations. The cost of explicit communication can be substantial, however, making it desirable to avoid its use in many domains. In this chapter we present the design and implementation of cooperative methods for reactive navigation: allowing a team of agents to assist one another in their explorations through implicit (stigmergic) communication. These methods range from simple solutions for avoiding specific problems such as individual local maxima, to the construction of sophisticated branching trails. We evaluate these methods individually and in combination using behaviour-based robots in a complex, three-dimensional software environment.