Robustness in the Presence of Task Differentiation in Robot Ensembles

Abstract

In the last fifteen years, much interest has been focused on the deployment of large teams of autonomous robots for applications such as environmental monitoring, surveillance and reconnaissance, and automated parts inspection for manufacturing. The objective is to leverage the team’s inherent redundancy to simultaneously cover wide regions and achieve massive parallelization in task execution while remaining robust to individual failures. Despite recent successes, significant challenges remain, in part, due to the difficulties associated with managing and coordinating the various redundancies that exist in a large team of homogeneous agents. In this chapter, we present an ensemble approach towards the design of distributed control and communication strategies for the dynamic allocation of a team of robots to a set of tasks. This approach uses a class of stochastic hybrid systems to model the robot team dynamics as a continuous-time Markov jump process. The main advantage is a lower-dimensional representation of the team dynamics that is amenable to system-level analysis of the team’s performance in the presence of task differentiation. We show how such analysis can be further used to design and optimize individual robot control policies through simulations and experimental validation.