Jafari, Reza;
Sarhadi, Pouria;
Paykani, Amin;
Refaat, Shady S;
Asef, Pedram;
(2024)
Optimal Torque Allocation for All-Wheel-Drive Electric Vehicles Using a Reinforcement Learning Algorithm.
In:
2024 13th Mediterranean Conference on Embedded Computing (MECO).
IEEE: Budva, Montenegro.
Preview |
PDF
MECO paper final version.pdf - Accepted Version Download (2MB) | Preview |
Abstract
A novel reinforcement learning-based algorithm is proposed in this paper for the optimal torque allocation among the four wheels of an all-wheel-drive (AWD) electric vehicle (EV) through a direct yaw moment control approach. A hierarchical structure was utilized for the control procedure, in which a linear quadratic regulator (LQR) controller is exploited for the high-level controller to generate yaw moments and a novel deep deterministic policy gradient (DDPG) algorithm is employed for the low-level controller. The DDPG agent possesses the ability to interact with the environment and learn to optimally split torque among four wheels. The vehicle is modeled via a nonlinear model with seven degrees of freedom (7 DOF), while the reference signals are generated by a bicycle model with two degrees of freedom (2 DOF). For enhanced precision in vehicle modeling, the tire model is characterized by the Pacejka Magic Formula (MF), which offers a rigorous and empirically validated representation of tire behavior. The proposed model is verified through a scenario of the response of the vehicle while circumnavigating a curve on a slippery road. The obtained results depict improved performance and enhanced dynamic stability compared to the conventional model with the average torque distribution method. Control over the yaw behavior and increased dynamic stability are achieved, while the understeer and oversteer are avoided. Index Terms—Direct yaw control, Electric vehicle, Reinforcement learning, Torque vectoring, Vehicle dynamics.
| Type: | Proceedings paper |
|---|---|
| Title: | Optimal Torque Allocation for All-Wheel-Drive Electric Vehicles Using a Reinforcement Learning Algorithm |
| Event: | 2024 13th Mediterranean Conference on Embedded Computing (MECO) |
| Location: | Budva, Montenegro |
| Dates: | 11 Jun 2024 - 14 Jun 2024 |
| Open access status: | An open access version is available from UCL Discovery |
| DOI: | 10.1109/meco62516.2024.10577860 |
| Publisher version: | http://dx.doi.org/10.1109/meco62516.2024.10577860 |
| Language: | English |
| Additional information: | This version is the author accepted manuscript. For information on re-use, please refer to the publisher’s terms and conditions. |
| Keywords: | Direct yaw control, Electric vehicle, Reinforcement learning, Torque vectoring, Vehicle dynamics |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Mechanical Engineering |
| URI: | https://discovery-pp.ucl.ac.uk/id/eprint/10195071 |
Archive Staff Only
 |
View Item |
