MCSLAM : a Multiple Constrained SLAM

Datta Ramadasan, Marc Chevaldonné and Thierry Chateau

Abstract

The real-time localization of a camera in an unknown or partially known environment is a problem addressed by Structure From Motion algorithms and more particularly CSLAM algorithms (Constrained Simultaneous Localization And Mapping). In this paper, we propose a new algorithm, named MCSLAM (Multiple Constrained SLAM), designed to dynamically adapt each optimization to the variable number of parameters families and heterogeneous constraints. An automatic method is used to generate a dedicated optimization algorithm, from an exhaustive list of constraints. To our knowledge, this is the only implementation that combines flexibility and performance. Known objects are used to constrain the 3D structure of the reconstruction and a continuous-time representation of the trajectory is used to deal with motion constraints. A continuous trajectory provides a simple way to add heterogeneous constraints into the optimization framework like other unsynchronised sensors or an evolution model. Several experiments show the effectiveness of our approach in terms of accuracy and execution time compared to the state of the art on several public benchmarks of varying complexity.

Session

Poster 2

Files

Extended Abstract (PDF, 1371K)
Paper (PDF, 1549K)

DOI

10.5244/C.29.107
https://dx.doi.org/10.5244/C.29.107

Citation

Bibtex

@inproceedings{BMVC2015_107,
	title={MCSLAM : a Multiple Constrained SLAM},
	author={Datta Ramadasan and Marc Chevaldonné and Thierry Chateau},
	year={2015},
	month={September},
	pages={107.1-107.12},
	articleno={107},
	numpages={12},
	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
	publisher={BMVA Press},
	editor={Xianghua Xie, Mark W. Jones, and Gary K. L. Tam},
	doi={10.5244/C.29.107},
	isbn={1-901725-53-7},
	url={https://dx.doi.org/10.5244/C.29.107}
}