Multilevel adaptation for autonomous mobile robot formations

Abstract

This paper studies adaptive behavior within multiagent autonomous robot formation-keeping. The goal of the research is to develop control laws for each robot in a formation such that the formation as a whole maintains a certain geometric configuration in the face of changing terrain conditions and actuator constraints. It is a main result of this paper that adaptation can occur at several different levels in a formation control architecture; just as a single robot may adapt to a new environment, a robot-formatio n as a whole may also adapt to a new environment. Terrain uncertainty and related actuator saturation are identified as important limitations to formation-kee ping in uncertain environments. In order to account for terrain uncertainty, individual robots are equipped with parameter estimators to allow implementation of direct adaptive control. Using these adaptive controllers, it is shown that formation error reduces to zero under non-saturated conditions. Actuator saturation brings new challenges to formation-kee ping. It is shown that rough terrain can easily lead to actuator saturation. For this case, a new ethological control scheme is introduced. It is shown that in this case the entire formation adapts to the weakest robot.