Position Estimation Using Principal Components of Range Data

1 sensors is to construct a structural description from sensor data and to match this description to a previously acquired model [Crowley 85]. An alternative is to project individual range measurements onto a previously acquired model [Leonard and Durrant-Whyte 91]. It is also possible to fuse range measurements directly using occupancy grids [Elfes 86], [Schiele 94]. Recently it has been shown that raw range data from nearby scans can be registered using a technique known as scan correlation [Weiss et al 95]. The correction vector from this technique provides a correction to position estimation. A thorough review in the state of the art in position estimation is provided in [Borenstein et al 96]. This paper describes a new approach to mobile robot position estimation based on principal component analysis of laser range data. An eigenspace is constructed from the principal components of a large number of range data sets. The structure of an environment, as seen by a range sensor, is re...

• Type: Conference
• Year: 1998
• Where: ICRA
• Authors: James L. Crowley, Frank Wallner, Bernt Schiele
• Keywords: ICRA 1998, Position Estimation, Range Data, Range Measurements, Robotics