Voting based fault isolation of in-vehicle multi-sensors

Abstract

This paper describes fault isolation of dead reckoning and laser range sensor (LRS) for mobile robot. Our mobile robot is equipped with three LRSs for omnidirectional sensing. The robot estimates its own linear and turning velocities based on the information of each LRS as well as dead reckoning. The iterative-closest-point based scan matching is applied for estimating the robot velocities with LRS in usual indoor environments. Fault isolation is based on a voting approach; the velocity estimates with the four sensors (three LRSs and one dead reckoning) are compared with each other, and the sensor whose velocity estimates do not match the others is decided as the faulty one. Experimental results validate our voting based fault isolation of multi-sensors.