Advances in formation control of multi-robot systems using reinforcement learning and neural networks
Published in International Journal of Dynamics and Control, 2021
The problem of formation in multi-robot systems has broad applications and gained considerable recognition in the past two decades among researchers. The purpose of this paper is to provide an overview of a new concept of learning-based control using adaptive and bio-inspired techniques in multi-robot systems. We survey recent advances in this research area to identify the strengths, limitations, open challenges, and research frontiers. We review the critical issues in formation control with a focus on the primary control strategies of different tasks (i.e., trajectory tracking, collision avoidance, etc.) In addition, we provide a comparison of the state-of-the-art in multi-robot formation and classify different approaches of learning-based control using artificial neural networks and bio-inspired techniques. For each approach, we analyze how formation can be done in different scenarios, which are considered as instances of the formation control problem, i.e., path planning, and in various application domains, including exploration, object transportation, target tracking, etc. Finally, we discuss open challenges and offer a range of possible research directions. The goal of this study is to provide guidance for future research on employing learning-based control in multi-robot systems.
Recommended citation: Salimi, M., Pasquier, P. (2021). "Advances in formation control of multi-robot systems using reinforcement learning and neural networks" Submitted to the 2021 Proceedings of the International Journal of Dynamics and Control, 2021.