In recent years, many robot-centric SLAM approaches have been proposed by the research community. Several authors claim that this approach allows to achieve algorithms with better consistency and convergence properties.
Dear Bruno: thank you for setting up such an interesting topic. I, for one, am particularly interested in following this discussion - my main topic of research in the past few years has been focussed on an egocentric approach for robotic multisensory perception, so I'd really like to understand how the SLAM community would feel about using such a solution.
You can find the answer in this article:
Javier Civera, Óscar G. Grasa, Andrew J. Davison, J. M. M. Montiel
1-Point RANSAC for EKF Filtering: Application to Real-Time Structure from Motion and Visual Odometry
Journal of Field Robotics, vol. 27(5), pp. 609-631, October 2010.
you can get from here:
http://webdiis.unizar.es/~jcivera/publications.html
Thank you for the reference, Mohamed Heshmat. I just browsed through the paper and was wondering if you could you tell us how the 1-point RANSAC is useful in a robot-centric approach. With this question I was looking for insight on the observability and convergence properties that a robot-centric approach may provide, that a earth-centric approach might not (see, for instance, the link below, where global stability for a robot-centric SLAM is demonstrated). The use of many earth-centric EKF-SLAM approaches has been shown to have pitfalls in terms of observability, and thus, convergence and statistical consistency.
http://dx.doi.org/10.1109/TRO.2013.2273838
- Badges
- Science topic
- Similar topics
- Methodology
- Research Methodology
- Research Papers