Adaptive linear MPC for a PMSM-driven autonomous EV with a filtered third-order generalized integrator observer.

Area of Science:

• Control Systems Engineering
• Automotive Engineering
• Electrical Engineering

Background:

• Autonomous electric vehicles require precise control for trajectory tracking and motor coordination.
• Permanent magnet synchronous motors (PMSMs) exhibit nonlinearities (e.g., inductance variation) that challenge conventional control methods.
• Fixed-parameter model predictive control (MPC) struggles with dynamic transitions and operating-point-dependent nonlinearities.

Purpose of the Study:

• To develop an adaptive linear MPC (AL-MPC) strategy to enhance trajectory tracking performance in autonomous electric vehicles.
• To address the limitations of conventional MPC in handling PMSM nonlinearities during dynamic operations.
• To improve the real-time adaptability and robustness of vehicle control systems.

Main Methods:

• An adaptive linear MPC (AL-MPC) strategy integrating a flux observer, Taylor-series linearization, and an active-set quadratic programming optimizer.
• Real-time estimation of electromagnetic torque and stator reactance using a moving-average-filtered third-order generalized integrator flux observer.
• Formulation of a unified nine-state predictive model updated dynamically and optimized using quadratic programming for voltage and steering control.

Main Results:

• AL-MPC achieved significant reductions in errors: 99.9% yaw MAE and 65% lateral position RMSE compared to linear MPC.
• Outperformed adaptive nonlinear MPC with 77.7% lower yaw MAE and 94.6% lower lateral MAE.
• Demonstrated real-time feasibility with a 9.65 ms control cycle execution time in HIL testing.

Conclusions:

• The proposed AL-MPC effectively manages PMSM nonlinearities for superior trajectory tracking in autonomous electric vehicles.
• AL-MPC offers substantial performance improvements over existing linear and adaptive nonlinear MPC methods.
• Validated through simulations and HIL testing, AL-MPC is a computationally efficient and real-time feasible solution for advanced vehicle control.