Hierarchical model predictive control strategy based on Q-Learning algorithm for hybrid electric vehicle platoon

Abstract

In view of the problem that hybrid electric vehicle (HEV) platoon energy management cannot adapt to working condition and online implementation, this paper proposes hierarchical model predictive control strategy based on Q-Learning algorithm. Firstly, the upper-level controller obtains the speed and position information of the preceding vehicle by vehicle-to-vehicle (V2V) communication. Model predictive control (MPC) is implemented to achieve platoon longitudinal control. The target speed of following vehicle is calculated and transmit to the driver model. Then, the driver’s power demand is figured out on the basis of the difference between target and actual speed. The lower-level controller uses the Q-learning algorithm to allocate the energy of HEV platoon based on driver’s power demand and state of charge (SOC) at the current moment. Finally, the simulation model of Chongqing Yubei actual working condition is established by using MATLAB/Simulink software. The simulation results show that in the upper-level controller, the average speed error between No. 1 following vehicle and the leading vehicle is 0.167 m/s, and the average speed error between No. 2 following vehicle and No. 1 following vehicle is 0.153 m/s. Meanwhile, the spacing between the platoon vehicles is always kept within a reasonable range. These will ensure good following and driving safety of the platoon. In the lower-level controller, compared with dynamic programming (DP), the average fuel consumption per 100 km of the vehicle with the Q-learning algorithm is increased by 7.67%, but the offline calculation time is reduced by 23%. The results indicate the proposed strategy can not only adapt to random condition but also be realized online for HEV platoon.