M.Sc. Thesis: Design and Implementation of Model Predictive Control for Improving Stability of a Vehicle Teleoperation System

• Collaborators: Ali Nahvi,
• November 2019 - February 2022

In autonomous and semi-autonomous vehicles, the driving path is typically computed by an external system, which then sends the desired steering angle to the vehicle. However, occurrence faults in the steering motor, particularly deviations from its optimal operation and the resulting asymmetry in its behavior, can reduce path-tracking accuracy and compromise vehicle safety. Classical controllers, such as PID, usually fail to maintain reliable performance under these conditions. Furthermore, driver behavior under such faults can be complex and nonlinear, and existing models do not fully capture these effects. These challenges highlight that controlling the steering system under motor faults while accounting for realistic driver behavior is a critical and practical problem for ensuring the safety of autonomous vehicles.

To address these challenges, we proposed a Hybrid Model Predictive Control (MPC) approach, incorporating separate modes for clockwise and counterclockwise motor operation to effectively handle asymmetry. We then suggested modeling motor faults as parametric uncertainties rather than external disturbances, which eliminates the need for online disturbance computation and allows the controller to maintain fast and robust performance under varying conditions. For realistic simulation of vehicle dynamics, a comprehensive driver model was integrated, combining neuromuscular (agonist and antagonist muscles), perceptual (otolith and semicircular canals), and physical dynamics components. The control system was implemented on a real remote-controlled vehicle, where one operator acted as the external path-planning system while the vehicle followed the desired steering commands under faulty conditions. Additionally, to overcome computational limitations of real-time MPC, an explicit MPC (eMPC) and a hybrid online-offline MPC strategy were employed. Finally, the performance and stability of the proposed control strategy were rigorously validated through comparisons with classical PID control and through Lyapunov-based and cost-function-based stability analyses. The results showed that, unlike PID, the MPC controller could stabilize the system. Moreover, the eMPC method achieved a significantly better tradeoff between performance and computational cost compared to other methods.