Abstract:

Aimed at the preceding vehicular acceleration prediction problem in automated cruise systems, a learning predictive control strategy is proposed based on Gaussian process regression to meet people’s requirements for safety, comfort, and economy of vehicles. First, the method of Gaussian process regression is used to build the learning modeling of preceding vehicular acceleration. Then, the learning model is combined with the inter-vehicle kinematics models to define the predictive model of the vehicular automated cruise system. After that, the learning predictive controller is estabilished for the vehicular automated cruise system through optimizing the safety, driving comfort, and economy indexes online. Finally, under accelerating-decelerating classical driving scenarios, the effectiveness of the method proposed is compared with that of the traditional predictive cruise control on the CarSim/Simulink co-simulation platform. The results show that the method proposed is more effective and superior to traditional control strategies.

Key words: model predictive control, Gaussian process regression, automated cruise systems, automated vehicles