Comprehensive Analysis and Control of Front and Rear Electric Drive Units in a Dual Motor Battery Electric Vehicle Through Improved Regenerative Braking and Clutch Utilization

Abstract

Electrification is reshaping the transportation sector, and dual-motor battery electric vehicles (BEVs) have emerged as a prominent architecture for improving traction capability and enabling flexible powertrain operation. However, the added degree of freedom can introduce efficiency penalties, increased actuator activity, and driveability challenges if torque allocation and braking control are not managed carefully. This dissertation investigates how powertrain architecture, braking strategy, and energy management design can be leveraged to improve the energy efficiency of independently driven axle (IDA) dual-motor BEVs while respecting driveability and component durability constraints.

The research first reviews the state-of-the-art in dual-motor energy management systems (EMSs), highlighting practical challenges, control objectives, and learning-based trends. It then develops a modeling and evaluation framework to quantify how single-motor and IDA dual-motor BEVs differ in energy use across representative driving cycles and payload conditions, clarifying when the dual-motor powertrain yields net efficiency gains and when it can erode driving range. Next, an experimental characterization of braking control in a production dual-motor BEV is presented, followed by an analysis of how commonly used braking force distribution constraints can affect regenerative energy recovery. Building on these results, a regenerative torque limit curve is introduced to enhance low-speed recuperation without extensive electric machine parameter identification.

Finally, the dissertation proposes a deep reinforcement learning–based EMS for a clutched IDA dual-motor BEV that internalizes clutch synchronization dynamics and energy cost. In simulation, the learned policy achieves near-optimal energy consumption (approximately 0.6\% above a dynamic programming benchmark) while reducing clutch toggling by approximately 75\% relative to a LUT baseline. Real-time model-in-the-loop validation in a high-fidelity driving simulator further demonstrates implementability and how transient harshness can be mitigated.

This dissertation advances dual-motor BEV control by quantifying architecture- and braking-driven trade-offs and proposing methods that improve efficiency and driveability, guiding future EMS development and deployment.