Mobile Scanning

D. Borrmann, J. Elseberg, K. Lingemann, A. Nüchter, J. Hertzberg
"Globally Consistent 3D Mapping with Scan Matching" 25.04.2007

A globally consistent representation of a robot’s environment is crucial for basic robot tasks such as localization and navigation. Equipped with a laser scanner, many mobile systems gather information about their local environments. These local representations have to be matched to build a global map. Iterative application of pairwise matching algorithms leads to inconsistencies due to errors in laser scans and the matching procedures itself. To avoid these problems, global matching algorithms are needed, taking correspondences between all scans into account. The common methods for merging all scans are based on probabilistic information. Lu and Milios presented a solution using a network of relations between laser scan poses. A single linear equation system, built from all error measurements, yields optimal estimations for all poses. Limited to 2D laser scans, this approach does not fulfill the requirements for building a correct map of outdoor environments. In this article the linear estimation algorithm is extended to work with 3D scans and 6 DoF. To run the algorithm automatically, the network is built using a distance heuristic. Furthermore, the global optimization into a system consisting of odometry extrapolation and the well-known iterative closest point algorithm (ICP) is integrated. After describing the problem formulation and the algorithm its functionality is documented by experimental results.