MPC based preview control for dog clutch-based electric vehicle transmission

Abstract

Recently, the global electric vehicle market has been rapidly growing, and the importance of the technology for powertrain electrification is rising. In particular, the development of electric transmission has become important. This is because by using a transmission, the vehicle can use a smaller motor and inverter to lower the costs while maintaining high performance. However, conventional automatic transmissions equipped with friction clutches lead to increased cost, complexity, and loss of efficiency. One way to resolve these issues is to use a light and high-efficiency dog clutch for EV transmission. Despite these advantages of the dog clutch, it has a critical weakness, which is a torque hole, during the shift. This paper proposes a model predictive control (MPC) based preview gear-shift strategy coupled with a two-motor architecture that can potentially eliminate the torque holes. The optimal control problem is formulated so that the torque holes and the number of gear-shift are minimized while maximizing the system efficiency. The system utilizes not only past information but also available future information to predict the vehicle’s trajectory. Long short-term memory network is utilized to predict the future information, which is input to the MPC controller. Simulation results show that the MPC-based preview gear-shift strategy can dramatically reduce torque holes compared to an existing rule-based gear-shift method, which opens up new opportunities in developing EV transmission using dog clutches.