Abstract
This paper deals with the sensor fusion problem in mobile robot localization using particle filter. Sensor fusion has been a major technique to estimate robot's state from noisy sensor observation. Generally, the sensors are assumed to be in the nominal state of work. In realistic contexts, however, sensor characteristics may change online depending on sensor functioning conditions, and deteriorate estimation accuracy. For such situations, faulty sensors should be identified and the fusion rule should be updated accordingly to achieve robust localization. These processes should be done online without access to ground truth. The authors propose a new sensor fusion method for mobile robot localization utilizing the increasing and diversifying onboard sensors thanks to the technological development in recent years. Under the strong assumption that the majority of similar signals provide the truth and any dissimilar signal is the result of a sensor fault, faulty probability distributions computed from sensor observations by a particle filter are detected and isolated based on K-L Divergence, a measurement for difference between two probability distributions in information theory. The proposed method is examined on several experiments, showing the possibility of generating robustness to the mobile robot localization.
