Data-driven robust distributed MPC for collision avoidance formation navigation of constrained nonholonomic multi-robot systems

In this work, we consider the robust collision avoidance formation navigation problem for multiple constrained nonholonomic robots with uncertain dynamics. Distributed model predictive control (MPC) based method is proposed in view of its ability to handle the input and state constraints of the robots explicitly. A synchronous non-iterative distributed algorithm is employed which reduces the communication requirement of the system. Furthermore, to enable the state trajectory prediction under uncertain robot dynamics, a data-driven online learning method is proposed to generate an accurate model of the nonholonomic robots adaptively. Based on the proposed control strategy, it is shown that robust collision avoidance formation navigation is successfully achieved while the input and state constraints of the robots are satisfied. Simulation examples are given to demonstrate the performance of the data-driven learning method and the distributed MPC based formation navigation controller.