Cooperative control of intelligent robot network

Abstract

The research on robot has been a hot topic nowadays. Within the field, formation control on multi-robot systems has attracted special attention because of its wide usage in various complicated scenarios such as target enclosing and load transport. Among the existing methods for formation control of robot swarms, the leader-follower approach has been proved to be a powerful one. Furthermore, due to the rapid development of deep neural network, deep reinforcement learning has also been a significant technology, which can be applied in robot control field. In this project, we first study one of the latest leader-follower formation control protocol which can deal with time-varying formation tasks and switching communication topologies. In the second stage, we study and discuss the feasibility of applying reinforcement learning methodology in robot control and tracking missions and develop a reinforcement learning approach for a simple robot control task. Finally, we embed reinforcement learning into the leader-follower formation control protocol, which can be a new direction of multi-robot control algorithm design.