Adaptive Navigation Control for Swarms of Autonomous Mobile Robots

Abstract

Deploying a large number of resource-constrained mobile robots performing a common group task may offer many advantages in efficiency, costs per system, and fault-tolerance (Sahin, 2005). Therefore, robot swarms are expected to perform missions in a wide variety of applications such as environment and habitat monitoring, exploration, odor localization, medical service, and search-and-rescue. In order to perform the above-mentioned tasks successfully, one of the most important concerns is how to enable swarms of simple robots to autonomously navigate toward a specified destination in the presence of obstacles and dead-end passageways as seen in Fig. 1. From the standpoint of the decentralized coordination, the motions of individual robots need to be controlled to support coordinated collective behavior. We address the coordinated navigation of a swarm of mobile robots through a cluttered environment without hitting obstacles and being trapped in dead-end passageways. Our study is motivated by the observation that schools of fish exhibit emergent group behavior. For instance, when schools of fish are faced with obstacles, they can split themselves into a plurality of smaller groups to avoid collision and then merge into a single group after passing around the obstacles (Wilson, 1976). It is also worth noting that a group of fish facing a dead end can get out of the area. Based on the observation of schooling behavior in fish, this work aims to present a novel adaptive group behavior, enabling large-scale robot swarms with limited sensing capabilities to navigate toward a goal that is visible only to a limited number of robots. In particular, the coordinated navigation is achieved without using any leader, identifiers, common coordinate system, and explicit communication. Under such a minimal robot model, the adaptive navigation scheme exploits the geometric local interaction which allows three neighboring robots to form an equilateral triangle. Specifically, the proposed algorithm allows robot swarms to 1) navigate while maintaining equilateral triangular lattices, 2) split themselves into multiple groups while maintaining a uniform distance of each other, 3) merge into a single group while maintaining a uniform distance of each other,