High-Performance Path Tracking of a 4WD Autonomous Vehicle Using NMPC with Virtual 4WD Torque Distribution

Area of Science:

• Automotive Engineering
• Control Systems
• Robotics

Background:

• High-performance path tracking is crucial for autonomous vehicles.
• Nonlinear Model Predictive Control (NMPC) offers advanced control capabilities but can be computationally intensive.
• Optimizing torque distribution in four-wheel-drive (4WD) systems is key for enhancing vehicle dynamics.

Purpose of the Study:

• To develop a reduced-complexity NMPC framework for high-performance path tracking in 4WD autonomous vehicles.
• To investigate the impact of a virtual 4WD torque distribution law on lap time and tracking accuracy.
• To determine the optimal torque distribution gain for improved vehicle performance and stability.

Main Methods:

• A reduced-dimensional NMPC was designed, optimizing steering and rear-wheel torques while using a gain-based law for front-wheel torques.
• Offline trajectory optimization (TRO) generated reference racing lines and velocity profiles.
• Simulations were conducted on the Shanghai International Circuit to evaluate the NMPC performance.

Main Results:

• The proposed NMPC with rear-dominant virtual 4WD torque distribution reduced simulated lap times by approximately 10.3%.
• Bounded path-tracking errors (max lateral error < 0.33 m) and heading-angle errors (max < 2.95 deg) were maintained.
• Increasing the torque distribution gain (Kr) up to 0.5 improved performance, but values beyond 0.5 led to degraded tracking and stability.

Conclusions:

• A reduced-complexity NMPC framework effectively enhances path tracking and reduces lap times for 4WD autonomous vehicles.
• An appropriate virtual 4WD torque distribution gain is critical for optimizing corner-exit acceleration and overall performance.
• Excessive front-wheel torque assistance can negatively impact tracking accuracy and vehicle stability, highlighting the importance of balanced torque distribution.