Using the microsoft kinect for 3D map building and teleoperation

This paper describes the use of the Microsoft Kinect for building three dimensional maps for use in teleoperation. Though these maps are being used for teleoperation in this paper, these map building techniques could also be applied in localization for navigation or robot path planning. The Kinect is a relatively new depth sensor that has become popular in the field of robotics, often replacing significantly more expensive systems like tilting laser range finders and stereoscopic vision systems. Two main sources of error in the map making are investigated, random and systematic error from the depth sensor and uncertainty in the navigation solution of the vehicle. Systematic and random error in the Kinect has previously been described in the literature [1] and this paper looks at how these errors affect the users ability to do mapping and offers some practical solutions. The teleoperation system design around this map making process is presented, and it builds on existing work [2] using octrees as the storage for the maps.