Experimental Evaluation of Kalman Filter-based Fault Detection and Fault-Tolerant Control on a Water Hydraulic Robot Arm

This paper presents an experimental evaluation of fault detection and fault-tolerant control (FTC) for a water hydraulic robotic arm driven by an Air-Hydro Servo Booster (AHSB), a new type of pneumatically operated water-based hydrostatic transmission (HST). This HST incorporates a redundant sensor system consisting of water hydraulic pressure, pneumatic pressure, and piston displacement sensors. In this study, we derive the mathematical model of the controlled plant, then employ a Kalman Filter-based residual generation approach combined with Mahalanobis distance for anomaly scoring to leverage sensor redundancy for effective fault identification and isolation. The failures in this study are assumed to be: 1) disconnection of one of the sensors; 2) fluctuation of the internal friction in the hydraulic cylinder section; and 3) disturbance torque to the arm. However, it is assumed that no multiple failures occur simultaneously. The experimental evaluation is conducted on a single-axis water-hydraulic arm focusing on detecting and managing common failures, such as sensor disconnection and the application of external disturbance torque during the position tracking task. The results demonstrate that the proposed system can detect sensor faults, and make the transition to a sensorless control mode, maintaining sustained position control. Although the scope is limited to a simple one-joint robot, this paper is the first experimental report on FDI/FTC on HST-based and/or pneumatically operated water-hydraulic robots.