Road hazard detection such as potholes has been actively attempted so far, usually focusing on methodology using vision systems (e.g., cameras, LiDAR) and vertical vibration signals measured by accelerometers mounted on suspension systems, because significantly uneven road surfaces (e.g., potholes) generate the vertical and lateral force on the contact patch of the tire during driving maneuvers and they are transmitted to the vehicle steering system. It usually acts as an unknown disturbance input, causing control performance degradation and it often leads to the failure of tires and automotive components such as suspension systems. Therefore, the research objective is to explore the new research on road hazard detection such as potholes based on rack force estimation of electric power steering (EPS) system (or motor driven power steering, MDPS). In this study, a robust pothole detection algorithm is explored based on Kalman filtering of EPS to estimate the unknown disturbance input (i.e., rack force in EPS). The Kalman filter (KF) has been recently adapted for the discrete-time domain, which enables the implementation of real-time embedded control systems. The improved KF-UI (Kalman filter with unknown input) algorithm, which combines the existing Kalman filter with disturbance observer (DOB), is designed based on a torque angle sensor (TAS) information and rack displacement sensor of the steering system. The estimated rack force is then used to detect potholes using signal processing and high-pass filtering. MATLAB/SIMULINK (estimated) and CARSIM software (assumed to be true) will be used to evaluate the estimation performance of rack force and detection accuracy of potholes. The performance of the proposed algorithm also should be evaluated by comparing it to the actual measured rack force. Thus, the effectiveness of the proposed algorithm will be experimentally validated through the HILS (Hardware-in-the-Loop simulation) test along with real measurement of the rack force. During the Kalman filtering, the covariance matrix P, Q, and R must be optimally tuned in advance. In addition, the system parameters of EPS should be precisely identified because the KF is a model-based optimal estimator. The steering system may also be affected by the dynamics of other vehicle systems (e.g., suspension system) under road hazards. Because it may degrade the estimation performance, it is necessary to decouple it. To the best of our knowledge, we report for the first time that it is possible to achieve a new means of detecting potholes based on the estimation of the unknown disturbance input of EPS (i.e. rack force) which is different from existing suspension model-based methods. Furthermore, the proposed study can be extended to autonomous vehicle driving systems (e.g., path tracking control). The proposed approach using steering dynamics of EPS appears to offer promising results for alternative road hazard monitoring and shows a potential for improving the vehicle steering control system such as path tracking control of autonomous vehicles, although there are technical problems that remain and need to be further examined. We further demonstrate the benefits of our approach by comparing the different estimation algorithms. In addition, further studies include the study on the robustness analysis, in-vehicle test, and the improvement of the pothole classification algorithm, etc.
Prof. Gi Woo Kim, professor, Inha University