Behavior analysis of grid-map based sensor fusion algorithms

Many algorithms for sensor fusion exist; some algorithms have better performance than others. The purpose of this paper is to present a procedure to analyze the behaviour of a sensor fusion algorithm. The analysis reveals the necessary behaviours for performing successful sensory fusion. The analysis methodology was developed for a sensor fusion framework, which assumes three basic concepts: Logical sensors, grid map paradigm and performance measures. Behaviour analysis of the sensor fusion algorithms is conducted by analyzing transition matrices. Three types of behaviours are characterized: Normal, dissimilarity and hysteresis. Five algorithms were compared: Three logical algorithms (AND, OR and MOST) and two adaptive algorithms: Adaptive Fuzzy Logic algorithm and adaptive Dempster Shafer algorithm. The analysis indicated that the MOST, OR and adaptive Fuzzy Logic algorithms are all normal, but the adaptive Fuzzy Logic algorithm has better performance. The adaptive Dempster Shafer has dissimilarity and therefore in some cases diverges.