Data-driven model predictive control method and system under time-varying scenario

Abstract

The application discloses a data-driven model predictive control method and system under a time-varying scene, and relates to the field of automatic driving vehicle control. The method comprises the following steps: acquiring measured state data of a target vehicle at a current time, a historical control instruction sequence and a historical measured state data sequence in a current sliding window; determining a current working condition according to the measured state data and a dynamic triggering threshold; generating an optimal predicted control instruction sequence and an optimal predicted state data sequence in a prediction time period based on the current working condition; taking the first control instruction in the sequence as a control instruction of the current time to control the vehicle; and repeating the above steps with the next time as the current time until a preset ending condition is reached. Through self-adaptive identification and differential processing of the working condition, the application solves the problems of low calculation efficiency, poor solution feasibility and redundant steady-state computing power under the time-varying scene, and improves the real-time performance, robustness and computing power efficiency of vehicle trajectory tracking control.