Large-scale map-making

  • Authors:
  • Kurt Konolige

  • Affiliations:
  • SRI International, Menlo Park, CA

  • Venue:
  • AAAI'04 Proceedings of the 19th national conference on Artifical intelligence
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Current mapping algorithms using Consistent Pose Estimation (CPE) algorithms can successfully map areas of 104 square meters, using thousands of poses. However, the computation to construct the map grows as O(n log n), so larger maps get increasingly difficult to build. We present an abstraction method for postponing the growth in computation. This method solves a much smaller problem in the space of the connection graph of the map.