Real-Time Acquisition of Compact Volumetric Maps with Mobile Robots

Christian Martin and Sebastian Thrun

This paper describes an online algorithm for generating compact 3D maps of mobile robot environments. Maps generated by our approach consist of small numbers of planar surfaces, which are augmented by fine-grained polygons for non-flat environmental features. Our approach builds on the expectation maximization (EM) algorithm, but develops a new, incremental version that can be executed in real-time. Experimental results obtained in corridor-type environments illustrate that compact and accurate maps can be acquired in real-time, from range and camera data collected by a mobile robot.

The full paper is available in gzipped Postscript and PDF