New Results on Cooperative Multi-Vehicle Deterministic Learning Control: Design and Validation in Gazebo Simulation

In this paper, new results on the cooperative deterministic learning (CDL) control method originally proposed in [1] for a group of unicycle-type ground vehicles are presented by considering a generalized nonholonomic uncertain vehicle dynamics. The new controller is capable of (i) controlling the vehicles to their respective desired reference trajectories; (ii) locally accurately learning/identifying, during the real-time control process, the vehicle's uncertain dynamics using radial basis function neural networks; and (iii) re-utilizing the learned knowledge to control the multi-vehicle system with guaranteed control performance and significantly reduced computational complexity. In addition, a Gazebo-based simulator is developed, based on which simulation validations have been conducted for the proposed algorithm.