Localization and Mapping for Mobile Robots using Correlation

Kurt Konolige
Artificial Intelligence Center
SRI International
[Joint work with Ken Chou (SRI) and Steffen Gutmann (U. of Freiburg)]

Abstract

In this talk I will give a general overview of recent work related to probabilistic methods for localization (Markov Localization and Scan Matching) and mapping (Expectation Maximization and the method of Lu and Milios) in the domain of mobile robots. While these basic theoretical approaches are well-founded and yield good results, in practice the computational issues of representing and updating large distributions often dominate the problem. In joint work with Ken Chou (SRI), I have developed correlation methods that are orders of magnitude faster than standard update methods. We have used these methods to perform localization at video rates from laser range information. Another application of correlation is in the very difficult problem of "closing the loop" in mapping. Here, in joint work with Steffen Gutmann (U. of Freiburg), we have succeeded in automatically and efficiently generating globally consistent maps with very large loops.
Eyal Amir
Last modified: Fri Apr 2 11:50:05 PST 1999