The Optimization of RBFNN Gearshift Controller Parameters for Electric Vehicles Using PILCO Reinforcement Learning

Abstract

In order to improve the efficiency of gearshift controller parameter optimization and obtain a good gearshift controller control effect, this article proposes an optimization method for electric vehicle transmission gearshift controller, and selects dual clutch transmission as the research object that establishes an 11-degree-of-freedom gearshift dynamics model and a feedforward-feedback gearshift control model. The feedforward control is chosen from the target trajectory given by the Legendre pseudo-spectral approach, and the feedback controller is a Gaussian kernel radial basis function neural network controller. The feedback controller performs parameter optimization by the Probabilistic Inference for Learning Control (PILCO) reinforcement learning algorithm to obtain a control strategy that matches the actual gearshift conditions. By comparing how well the main/secondary moving disk can follow the target trajectory during various optimization iterations, it is verified that the algorithm requires only a few experiments to complete the optimization, and the optimized Radial Basis Function Neural Network (RBFNN) control has a better control effect by comparing the results of different iterations. Applying the learned controller to various slope and load circumstances yields data that demonstrate that all can have an obvious optimization effect with good robustness. Additionally, the reinforcement learning technique suggested in this research can be used for various gearshift controller parameter optimization to assist engineers and technicians in increasing the effectiveness of their Research and Development.

Keywords

• Electric vehicle
• dual clutch transmission
• gearshift control
• reinforcement learning