[campus icon] Accesskey [ h ] University of Paderborn - Home
DE deutsch
Die Universität der Informationsgesellschaft


Autonomous 3D Mapping and Exploration in Interior Environments based on a Next-Best-View Approach
Date: 2012/02/08
Time: 16:30 h
Place: P1.4.17
Author(s): Marc Östermann

Am Mittwoch, den 08. Februar 2012, hält Marc Östermann um 16:30 Uhr im Raum P 1.4.17 einen Vortrag über seine Materarbeit mit dem Titel:

Autonomous 3D Mapping and Exploration in Interior Environments based on a Next-Best-View Approach


The exploration of the three-dimensional environment is a common problem, arising in the field of autonomous mobile robots. The acquisition of data is often performed by a laser rangefinder and provides a set of three-dimensional points, forming a point cloud. While the scanning devices have only a limited sensing range, more than one scan has to be taken to map large scenes of the environment. After the first scan is acquired, the robot has to move to its next scan position to continue the environment covering process.

Before the robot can move to its new scan position, it has to determine this position. The position determination process is typically combined with quality requirements and constraints that should hold for the new scan position. Because the robot has a real mass and size it cannot reside at all positions in the environment, for instance. Considering certain requirements and constraints, the determination of a next-best-view position turns into an optimization problem.

This work presents algorithms and concepts that form the basis to solve this next-best-view problem. In particular, the process of environmental modeling and the registration of different scans is addressed. The modeling process covers different stages, starting with the three-dimensional data acquisition of the laser rangefinder and moves on to the enrichment of the collected data with information extracted from the model.

The second part of the work introduces the iterative closest point algorithm. This algorithm is used to find the correct alignment of different scans. It is applied due to the fact that pure odometry data, collected by the robot, is not sufficient for the precise alignment of the single scan models.