Correspondence finding for indoor navigation using landmarks

Mobile robot positioning requires absolute sensing strategies to periodically update the robot´s pose and to reset the unbounded growing of dead reckoning cumulated errors. Geometric map matching techniques using natural or artificial can be used for absolute localization. The success of such techniques depends on the fidelity of the models being considered. This paper analyzes the various sources of errors involved, and includes them in the estimation of pose and in the criteria for selecting candidates and hypothesis validation. In addition, results using those techniques with landmarks detected with a monocular vision system are presented.