"Bayesian estimation and Kalman filtering: A unified framework for Mobile Robot Localization "
Decision and estimation theory are closely related topics in applied probability. In this paper, Bayesian hypothesis testing is combined with Kalman filtering to merge two dif ferent approaches to mapbased mobile robot localization; namely Markov localization and pose tracking. A robot carries proprioceptive sensors that monitor its motion and allow it to estimate its trajectory as it moves away from a known location. A single Kalman filter is used for track ing the pose displacements of the robot inbetween different areas. The robot is also equipped with exteroceptive sen sors that seek for landmarks in the environment. Simple feature extraction algorithms process the incoming signals and suggest potential corresponding locations on the map. Bayesian hypothesis testing is applied in order to combine the continuous Kalman filter displacement estimates with the discrete landmark pose measurement events. Within this framework, also known as Multiple Hypothesis Track ing, multimodal probability distribution functions can be represented and this inherent limitation of the Kalman fil ter is overcome.