Risk-Quantified Autonomous Maintenance of Robotaxis Using Self-Tuning Model Predictive Control

For robotaxis, or complex unmanned autonomous systems, prognostics and health management (PHM) technology is significant to support their condition-based maintenance (CBM). This study aims to innovatively propose a risk-quantified autonomous maintenance framework to extend the operation of robotaxis and enhance their task completion and safety in the presence of component degradation. Specifically, taking the critical component, the traction motor of the electric robotaxi, as an example, first, a nonlinear degradation model based on the Wiener process is established to accurately predict the motor’s remaining useful life (RUL). Next, to achieve a satisfactory balance between control performance and minimum motor degradation, a closed-loop controller considering motor degradation is designed based on model predictive control (MPC) for the speed tracking problem. Finally, this study constructs a risk evaluation function to quantify the travel risk in an emergency when the RUL of the motor is less than the time required for the travel task. The controller’s weight matrix will be self-tuned based on the risk value to extend the motor’s RUL further. The effectiveness and applicability of the proposed framework and controller are verified by co-simulation and hardware-in-the-loop (HIL) tests. Results show that adapting the controller parameters to the risk value can extend the motor’s lifespan further and improve robotaxi task completion.