Online fault detection of a mobile robot with a parallelized particle filter

Fault diagnosis is one of the most challenging problems, which have to be solved if one considers real-life applications of mobile robots. In this paper, we present a particle filtering-based approach combined with the negative log-likelihood test to address the fault detection task. The major disadvantage of the method is its high computational burden closely related to the number of particles used, which can be computationally too expensive to be processed online by the onboard computer of the robot. In order to address this problem, a solution, in which a part of computations are delegated to an external parallel computing environment such as a computer cluster, is presented. The proposed methods of parallelizing particle filters are aimed at improving their performance in terms of efficiency, estimation error and execution time, which are vital factors in an online setup. To depict the performance benefits of the presented methods, they are confronted with some other existing approaches in a series of experiments.