The motion control system, as the core executive component of the automatic
hierarchical framework, directly determines whether autonomous vehicles can
reliably and stably follow planned trajectories, making it crucial for driving
safety. This article focuses on steering lock faults and proposes a cross-system
fault-tolerant control (C-FTC) algorithm based on dynamic model reconstruction.
The algorithm uses a classic hierarchical collaborative architecture: the
upper-level controller employs an MPC algorithm to solve lateral velocity and
yaw rate reference values in real-time, while the lower-level controller,
designed based on the reconstructed dynamic model, uses an MPC algorithm to
adaptively adjust actuator control quantities. In cases where four-wheel
steering vehicles lose steering ability due to locked steering axles, the locked
axle’s steering angle is treated as a state variable, and healthy actuator
outputs are used as control variables to dynamically reconstruct the vehicle
dynamic model. The required lateral force for steering is then allocated to
healthy actuators to achieve fault-tolerant control. To verify the algorithm’s
effectiveness, validation combines hardware and real-vehicle testing, conducting
high-speed obstacle-avoidance tests under three fault conditions: front axle
lock, rear axle lock, and both axles locked. Results show that under all three
conditions, the proposed algorithm keeps lateral trajectory tracking errors
within 0.35 m, ensuring vehicle safety even with steering system faults.
