never combine linearizations around different linearization points, especially in the presence of non-linear nullspaces! It will render unobservable demensions observable, and corrupt the system.
Grisetti, G. , et al. "A Tutorial on Graph-Based SLAM."IEEE Intelligent Transportation Systems Magazine2.4(2011):31-43.
2. Kummerle, Rainer , et al. "G2o: A general framework for graph optimization."IEEE International Conference on Robotics & AutomationIEEE, 2011.
3. LOURAKIS,M. I. A. "SBA: A software package for generic sparse bundle adjustment."Acm Trans.math.softw36.1(2009):2.
参考
^Walter M R , Eustice R M , Leonard J J . Exactly Sparse Extended Information Filters for Feature-based SLAM[J]. Int.j.robotics Res, 2007, 26(4):335-359. https://deepblue.lib.umich.edu/bitstream/handle/2027.42/86031/mwalter-22.pdf?sequence=1
^Tong Q , Peiliang L , Shaojie S . VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator[J]. IEEE Transactions on Robotics, 2017, PP(99):1-17. https://arxiv.org/pdf/1708.03852.pdf
^Jauffret C . Observability and fisher information matrix in nonlinear regression[J]. Aerospace & Electronic Systems IEEE Transactions on, 2007, 43(2):756-759. https://hal-univ-tln.archives-ouvertes.fr/hal-01820468/document
^Dong-Si T C , Mourikis A I . Consistency analysis for sliding-window visual odometry[J]. Proceedings - IEEE International Conference on Robotics and Automation, 2012:5202-5209. https://intra.ece.ucr.edu/~mourikis/papers/DongSi2012-ICRA.pdf
评论(0)
您还未登录,请登录后发表或查看评论