Continuous localization in changing environments

Abstract

Continuous localization is a technique that allows a robot to maintain an accurate estimate of its location by performing regular small corrections to its odometry. Continuous localization uses an evidence grid representation, a common representation scheme that is used by other map-dependent processes, such as path planning. Although techniques exist for building evidence grid maps, most are not adaptive to changes in the environment. In this research, we extend the continuous localization technique by adding a learning component. This allows continuous localization to update the long-term map (evidence grid) with current sensor readings. Results show that the addition of the learning behavior to continuous localization allows the system to adapt to changes in its environment without a loss in its ability to remain localized. This system was tested on a Nomad 200 mobile robot.