Sensor fault detection and isolation for a powered wheelchair

This paper presents a model based fault detection and isolation (FDI) of internal sensors for a powered wheelchair. Three fault modes of five internal sensors (two wheel-encoders, two steering-potentiometers and one yaw-rate gyro) are handled: hard fault, noise fault and scale fault modes. On the hard fault the sensor output is stuck at a constant value. The noise fault causes the sensor output with large noise. The scale fault changes the sensor parameter from the fault-free. The hard and noise faults are diagnosed based on their mode probability estimated with the variable structure interacting multi-model (VSIMM) estimator. The scale fault is diagnosed based on the wheelchair velocity estimated with a laser scan matching. For fault-tolerant operation of the wheelchair even in the presence of sensor faults, output estimates of faulty sensors are exploited to control the wheelchair. Experimental results validate the effectiveness of our FDI method.