Local map joining algorithms in SLAM:

Local map joining is a strategy for efficiently building large-scale maps. By first building a sequence of small local maps and then combining them together, the frequency of a big map update step is significantly reduced. Especially, we have shown that the global map information matrix is exactly sparse without any approximation, which results in significant computational saving.

Sparse local submap joining filter (SLSJF) and Iterated SLSJF (I-SLSJF) for large-scale SLAM

·        SLSJF Paper (IEEE Trans. on Robotics, Vol. 24, No. 5, 1121-1130, October 2008, with Zhan Wang and Gamini Dissanayake)

·        I-SLSJF paper (2008 Australiasan Conference on Robotics and Automation, December 2008, with Zhan Wang, Gamini Dissanayake and Udo Frese)

·        3D I-SLSJF paper (48th IEEE Conference on Decision and Control, December 2009, with Gibson Hu and Gamini Dissanayake)

·        MATLAB code for 2D I-SLSJF (code and data sets, document)

·        3D simulation data sets (small data set, larger data set)

Iterated D-SLAM map joining (I-DMJ) for large-scale SLAM

• I-DMJ paper (Autonomous Robots, accepted, to appear, with Zhan Wang, Gamini Dissanayake and Udo Frese)

 

Idea of map joining algorithms: each local map can be treated as an observation; two objects (feature or robot pose) have no link unless they share the same local map.

(Idea of SLSJF and I-SLSJF)

(I-SLSJF map obtained by joining 700 local maps, in the simulation data, 219332 observations are made from 35188 robot poses to 4202 features)

 

                                                                                                                                 

 

(Sparse information matrix by I-SLSJF)

(Computation time of SLSJF and EKF map joining)

 

(I-SLSJF result using Victoria Park data set)

(I-SLSJF result using DLR-Spatial-Cognition data set)