Real-time optimization of gear shift trajectories and robust control strategy for electric vehicles with dual clutch transmission

Abstract

This dissertation deals with electric vehicle’s gear shift control strategy with a dual clutch transmission (DCT). DCT transmits power to the vehicle body using two clutches alternately. Gear shift proceeds by controlling two clutches and a power source. Gear shift performance is determined by the control strategy of each clutch and power source. Electric vehicles have a motor as a power source, and themotor torques can be controlled faster and more precisely than conventional engine torques. Therefore, we intend to improve the gear shift performance by actively utilizing the characteristics of the motor during gear shift control. The gear shift control strategy is divided mainly into upper- and lower-level controllers. The upper-level controller generates gear shift trajectories, and the lower-level controllertracks the gear shift trajectories. The upper-level controller determines the gear shift performance in an ideal case, and the lower-level controller plays a role in coping with model uncertainty and disturbance. Since they are model-based strategies, a driveline model that can accurately depict the movement of the driveline is required. Therefore, a control-oriented model with the lumped driveline transmission efficiency is proposed in this dissertation. Since the lumped driveline transmission efficiency is an unknown value, it must be estimated through testbench experiment data. A method to estimate the lumped driveline efficiency using recursive least square estimation is proposed.The upper-level controller generates gear shift trajectories with optimal shift performance through a real-time optimization process with the model. In shifting performance, the crucial important point is that the driver does not feel shift shocks. Therefore, jerk, a physical quantity representing shift shock, is set as the objective function of the optimization problem. In addition, the driveline model, modelswitching condition, and driver’s pedal demand are set to equality constraints. The maximum jerk limit, the maximum torque limit of each hardware, and torque rate limit of each hardware are set to the inequality constraints. The problem is converted into a quadratic programming problem, and an optimal solution is obtained by solving the problem in real time through a numerical solver. Finally, optimal gear shift trajectories and feedforward input are generated through the algorithm. The performance of the generated gear shift trajectories is verified through testbench experiments.A model-based torque observer and controller should be designed to track the given gear shift trajectories with the lower-level controller. In the case of the torque observer, it must estimate each clutch torque and the output shaft torque during the gear shift process. The observer is designed as a reduced-order observer to obtain good torque estimation performances. In addition, the observer’s transfer function is analyzed in the frequency domain to ensure robust torque estimation performance against disturbance. The gain of the observer in each gear shift phase is set using this analysis. In the case of the lower-level controller, it consists of a feedforward and feedback controller to obtain good tracking performance against model uncertainty and disturbances. A disturbance observer-based decoupling controller is used as the feedback controller. The disturbance observer and decoupling controller are designed considering the testbench’s time delay and each input’s characteristics. The robust performances of the proposed observer and lower-level controller are verified through testbench experiments.Finally, the robustness of the entire gear shift strategy, which combines upper- and lower-level controller, must be guaranteed. In general, combining two control strategies designed independently cannot guarantee robustness. For the robustness analysis, maximum uncertainty boundary analysis of the system according to the driveline model parameter variations and disturbances must be performed.We try to generate conservative gear shift trajectories by tightening the constraints of the upper-level controller algorithm in consideration of the maximum uncertainty bounds. In addition, a feasibility analysis of the algorithm is conducted. The robustness of the revised gear shift strategies with the above contents is guaranteed and verified through testbench experiments.