Augmenting Sparse Laser Scans with Virtual Scans to Improve the Performance of Alignment Algorithms

In book: Cutting Edge Robotics 2010
Source: InTech


The presented implementation of an extension to the FFS alignment process using Virtual Scans containing hypothetical mid level real world structures could significantly improve the results for the alignment task. The implementation proves the applicability of the presented concept for the combination of LLSC and MLSC processes. The detection of simple elements (lines, rectangles) based on weak real world assumptions could improve the performance. We are aware that adding domain knowledge certainly enhances the risk of wrong inferences. The proposed system handles errors caused by premature belief in mid level features by implementing the feedback principle, which evaluates a single hypothesis. It is known that single hypothesis systems introducing higher knowledge tend to be not robust. Under certain circumstances this behavior could also be observed in experiments with our system, which needed manual parameter adjustment to steer the influence of the hypotheses. It can be embedded into a multiple hypotheses framework, e.g. particle filters, which will be part of future work.
Additional future work also has to determine the (geometric) level of elements which are meaningful enough to improve the alignment process, yet not too dominant. The current elements were chosen to model assumptions of indoor disaster areas. Future research about assumed real world elements will adjust to outdoor disaster settings.