My photographThe aim of this project is to exploit a power of self-organization for intelligent reconstruction of 3-D shapes from laser range finder. We apply artificial neural networks that moves the task to three dimensions.

This project is focused on research of new methods for three-dimensional shapes reconstruction using computational intelligence, mainly artificial neural networks. The reconstruction is spit into four phases:

  1. decomposition of input data to clusters,
  2. construction of SOM submaps within each cluster,
  3. joining of the submaps,
  4. final reoptimization of joins and the final mesh using SOM algorithm.

The algorithm overcomes a problem of fitting one planar SOM mesh to complex non-continuous shapes found in building interiors. The method is easily paralelizable. Our aim is to implement a real time incremental variant of this algoritm, which can be directly usable for mobile robot navigation and real time collaborative interior map construction.

Check out the compressed VRML model, which visualizes all phases of the algorithm (the compressed model is 16MB large, loading may take a while and it's source may be displayed before the VRML browser (e.g. Cortona) starts).


  • Koutnik J., Mazl R., Kulich M.: Building of 3D Environment Models for Mobile Robotics Using Self-organization. In: Parallel Problem Solving from Nature - PPSN-IX. Heidelberg, p. 721-730, Springer, 2006. ISBN 3-540-38990-3 BibTex, PDF

    In this paper, we present a new parallel self-organizing technique for three dimensional shape reconstruction for mobile robotics. The method is based on adaptive input data decomposition, parallel shape reconstruction in decomposed clusters using Kohonen Self-Organizing Map, which creates mesh representation of the input data. Afterwards, the sub-maps are joined together and the final mesh is re-optimized. Our method overcomes a problem of fitting one mesh to complex non-continuous shapes like building interiors. The method allows to process unordered data collected by mobile robots. The method is easily paralelizable and gives promising results.