Sensor Fault Detection for Wheel-Legged Robot with Sliding Window Detector

As a ground-based robot with high mobility, the wheel-legged robot incorporates a variety of sensors for environment sensing. However, when sensor faults occur, the unmanned robot is unable to sense the environment correctly. This may lead to yawing and collisions with obstacles, resulting in danger. Therefore, we need to detect and identify sensor faults to ensure the safety of unmanned robots. In this paper, we propose a sliding window (SW) detector based on an Extended Kalman Filter for detecting sensor faults. The skid steering dynamics of the wheel-legged robot in wheeled mode are established and an Extended Kalman Filter (EKF) is built for state observation. We also establish two sensor fault models. For sensor fault detection, we present the principles of our proposed SW detector and compare it with the chi-square, cumulative sum (CUSUM) and multivariate exponentially weighted moving average (MEWMA) detectors. We use Matlab for simulation to verify that our proposed detector has better detection performance compared to conventional detectors in the case of minor faults in the position sensor.