Versatile Approach to Probabilistic Modeling of Hokuyo UTM-30LX

Abstract

The objective of this paper was to analyze several approaches to the probabilistic modeling of Hokuyo-type laser rangefinder (LRF). These models, which reflect mostly the mixed pixels, beam width, signal, and impulse noise, were compared mutually. Motivation for creating these models was the failure of existing LRF models to faithfully represent the features of sensed environment. Moreover, the characteristics of each particular model were highlighted, and their suitability for various applications was determined. This paper also dealt with data fusion by selected model and its representation in the occupancy grid. This was based on the classical Bayes’ rule, which was adapted for the desired characteristics of data fusion. Finally, in order to improve data from the sensor, data pre-processing was proposed. Mentioned data pre-processing was proposed with the emphasis on the elimination of impulsive noise. The overall concept of data processing from the LRF can be regarded as versatile and robust.