home · mobile · calendar · colloquia · 2003-2004 · 

Colloquium - Huang

Topological Exploration and Mapping using Sensing-limited Robots
Wes Huang
Rensselaer Polytechnic Institute

The basic ability to explore, map, and subsequently navigate through an unknown environment is a fundamental problem in mobile robotics. Many applications, such as hazard assessment for a contaminated building, call for cooperative teams of small, inexpensive (and therefore potentially disposable) robots. To this end, we are developing algorithms to explore, create, and merge maps for robots with limited sensing capabilities -- our prototype hardware uses only five short range (10-80 cm) infrared range sensors. This talk will present some of our most recent results on this problem.

Our approach uses topological maps defined in terms of robot behaviors, primarily hall-following and wall-following. The two main contributions of this work are an extension of topological maps for mapping open areas, long considered a deficiency, and an evidence-based approach (using Dempster-Shafer theory) to the problem of "closing the loop" in a map. We also address merging two topological maps for robots without a common reference frame. Our approach is a combination of maximal subgraph matching and image registration techniques. I will present experimental results, both from simulations and on our robot hardware.

Hosted by Gregory Grudic.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)