"Where Am I? - Using Vision to Localize a Robot."

Tina Kapur, 1992.

Abstract

One of the challenging issues in Computer Vision is to use information from a robot's environment as it moves so that its motion can be directed by knowledge of the objects surrounding it. The robot needs to steer its way around the environment, avoid any obstacles along its course, and determine its location at a given time in order to navigate successfully. Occasionally a robot moving through its environment will get "lost", i.e. it ends up in a location that it neither expects to be in nor recognizes. The goal of this research is to investigate the problem of determining the location of the robot at a given time using Computer Vision. This problem is referred to as the "Where am I?" problem.

This thesis presents a system that, under certain circumstances, enables a mobile robot to determine its location within a partially known environment. This system is an initial step in the exploration of a solution to the "Where am I?" problem. The claim of this work is that a mobile robot can reduce the possibliities for its location at a given time in an office-type environment that has sufficiently distinct floor plans for each room, by comparing the visual clues it gets from its sensors regarding the sections of the floor plan that are in its field of view against the actual floor plans of the various rooms. This ability of the robot, when developed further, can be of considerable use in autonomous mobile robot navigation which is an active research field in Computer Vision.