Predictive Cruise Control of Connected Vehicle With Online Parameters Learning

The acquisition of multi-dimensional traffic information and constantly increasing computational power enable sophisticated control techniques to be applied in cruise control system. This study proposes a predictive cruise control (PCC) scheme based on model predictive control, which is formulated as a multi-objective nonlinear optimization problem. In order to facilitate the proposed PCC to deal with different driving conditions, a clustering method is used to identify the driving state of the preceding vehicle. Then, Bayesian optimization method is adopted to learn the optimal weighting parameters in the multi-objective optimization function, which can improve the control performance. Simulation results show that 2.83% fuel-saving rate can be obtained by applying Bayesian optimization method compared to fixed weighting parameters while maintaining good tracking ability and driving comfort.