Online data-driven event-triggered model predictive control for path following of autonomous vehicles

In this paper, a data-driven linear time-varying event-triggered model predictive control (ET–LTV–MPC) framework based on the Koopman operator is proposed to address the problems of complex modeling, limited communication resources and control difficulties in path following control of autonomous vehicles. Firstly, Koopman-operator-based modeling approach is exploited to characterize the nonlinear dynamics of autonomous vehicles. Thus, the linear time-varying state-space model are in real time identified by recursive least squares algorithm in order to achieve online updating of path following model of autonomous vehicles. Secondly, a novel event-triggered scheme is designed to effectively balance the communication assumption and path following control performance. Finally, MPC solution is presented under the constructed linear time varying models and the developed event-triggered scheme. Simulation experiments demonstrate the effectiveness of the proposed data-driven ET–LTV–MPC scheme.