Distributed Iterative Learning Formation control for Nonholonomic Multiple Wheeled Mobile Robots with Channel Noise

Abstract

In this paper, we studied the robust formation control issue of multiple non-holonomic wheel mobile robots (WMRs) with nonlinear characteristics and considered the channel noise and switching communication topology, a distributed iterative learning formation control (DILFC) scheme using information interaction between robots is proposed. Firstly, the formation tracking error with consensus information is constructed, and the relationship between formation error and channel noise is obtained from the nonlinear system model of mobile robot. Next, the controller is designed based on the prediction and the current learning term between robots, and the switching topology is introduced into the formation algorithm in the form of piecewise function. The sufficient condition and norm upper bound for the formation tracking stability of the system are obtained by theoretical analysis. The results show that although the channel noise accumulates in both the time domain and iteration domain, the validity of formation tracking can be guaranteed by adjusting the sampling time of the system. To illustrate the effectiveness of the proposed scheme, numerical simulation results of a group of WMRs are presented.

Simulated Autonomous Driving Using Reinforcement Learning: A Comparative Study on Unity’s ML-Agents Framework

“…One key advantage of using simulation environments for the training of RL agents is the ability to easily generate large amounts of data for training and testing. This is particularly important for tasks such as autonomous driving [17] and autonomous path planning [18,19] and tracking control [20,21], where real-world data can be difficult and expensive to collect. Simulation environments also allow for greater control over the training environment, enabling researchers to easily vary factors such as the layout of the track or the weather conditions [22].…”