Mobile Robot Localization via Outlier Rejection in Sonar Range Sensor Data

Localization is an important ability for a mobile robot. the probabilistic localization methods become more popular because of the ability of representing the uncertainties of the sensor measurements and inaccuracies in environments. they also provide robust solutions for different localization problems. the particle filter is one of the probabilistic localization methods. in this study, sonar range sensors are used for mobile robot localization. sonar range sensors suffer from wrong reflections that may result outliers in the data set. outliers may also occur in the particle filter process. in this study, a new sensor model repealing range sensor model (r2sm) is proposed and integrated to particle filter to reduce the effects of outliers. in order to show the effectiveness of the proposed method, experiments are conducted and the results are compared with a well-known outlier rejection method, grubbs’ t-test. experiments show that results of the proposed approach are comparable to the results of the grubbs’ t-test in terms of localization success ratio (lsr) and number of iterations (noi) required for localization. the main advantage of the proposed r2sm is that it does not require any additional information such as critical value table. this provides more flexible outlier rejection approach.