Mobile Robot Diagnosis with Bayesian Filters

Abstract

In the following paper we consider a problem of fault detection for a mobile robot. The robot which our work is related to, is based on a new type of the steering principle [1]. The crucial part of the steering system are axle position sensors. A failure of one of them might result in an interruption of operation and/or serious damages to hardware and environment elements. To avoid the risk of such events, a reliable fault detection system has to be implemented. Fault detection is facilitated by incorporating measurements from various sensors located on board of the robot (incremental encoders, absolute encoders, sonar, cameras, compass, GPS). In our work we consider the Bayesian approach (Kalman filter [2,3] and particle filters [4,5,6]) to create a diagnostic system of the robot. Due to the limited resources of the computing unit it is necessary to strongly optimize the efficiency of applied algorithms. In our work we plan to perform simulations to find the best suited algorithm for our vehicle. Accordingly, we build a numerical tool in MATLAB to simulate mobile robot navigation and fault diagnosis tasks. We also present the construction of our robot and explain how a reliable fault detection system is important for the proper functioning and safety of the mobile robot. We introduce the fault detection algorithms which we plan to apply in the described hardware solution. At the end, we summarize our work and provide an outlook on our future research plans.