Vehicle predictive control based on the recursive subspace identification method

Vehicle active control is an important technology for vehicle safety and performance. Most vehicle control systems are based on models, and the effects of model-based controllers are more dependent on the model parameters. In order to improve the adaptability and the veracity of a control system, a new vehicle predictive control method is proposed. Based on the recursive optimized version of the predictor-based subspace identification method, the vehicle model can be identified from input–output data. According to the predicted outputs and the optimal criterion, two predictive controllers are derived. One is called the recursive subspace predictive controller, and the other is called the recursive subspace predictive controller with integrator, which is an improved form to handle integrated white noise. Compared with the outputs from traditional predictive control, the predicted outputs in this paper are obtained directly from the subspace identification process at each step, which avoids the difficult process of solving Diophantine equations. In this paper, a simulation vehicle model, which is made up of a seven-degree-of-freedom vehicle model and a magic formula tyre model, is used to verify the effectiveness of the controllers. By numerical examples, it can be shown that the proposed methods not only are effective for vehicle control in the linear domain but also can track the desired values well and can improve the effect of vehicle control even when the lateral tyre forces reach saturation.