Seamless Shifting of 2-speed Transmission for EV by Machine Learning

Abstract

This research aims to apply the deep reinforcement learning to the shift control of the EV two-speed transmission， develops a model for learning the shift control, and performs iterative learning to realize a seamless shift control. First， theoretical formulae are constructed to clarify the input-output relationship of the transmission. Next, it is confirmed that seamless shift control is possible based on the theoretical formula constructed. After that, a deep reinforcement learning models aiming at seamless shift control are developed, and the usefulness is verified by comparing the shift result after learning and the shift result based on the theoretical formulae.