!! Localization
MINERVA: Carnegie Mellon's Robotic Tourguide Project Minerva's Image

Localization: Where Am I?

Localization is the problem of how Minerva determines her position in the environment. The problem is relatively easy in static environments, or if the environment is equipped with special-purpose localization equipment (such as radio beacons). Minerva's environment, however, is densely populated, and is not modified in any way to facilitate navigation. Thus, Minerva must rely on the natural features of her environment to localize itself.

Minerva maintains a sense of orientation using a high-resolution Markov localization algorithm. Markov localization represents beliefs using probability distributions, enabling her to handle ambiguities and recover gracefully from sensor errors. This module periodically compares camera images and laser range scans with Minerva's ceiling map and the occupancy map. A unique feature of Minerva's localization component is a filter for identifying sensor data corrupted by people--an essential component for successful navigation in populated environments. Below is a typical example of a range scan, projected into the occupancy map at the most likely position:

More information on Minerva's localization module can be found in the following papers:

Home The Java-Based Image Stream About this project