A linear approximation for pose graph optimization

 

L. Carlone, R. Aragues, J. Castellanos, and B. Bona, "A fast and accurate approximation for planar pose graph optimization", International Journal of Robotics Research (IJRR), 2014.

 

 

lago

AbstractIn this work we investigate the problem of Simultaneous Localization And Mapping (SLAM) for the case in which robot measurements are modeled as a network of constraints in a pose graph. We combine tools belonging to linear estimation and graph theory to devise a closed-form approximation to the batch SLAM problem, under the assumption that relative position and relative orientation measurements are independent. The approach needs no initial guess for optimization and is formally proven to admit solution under the SLAM setup. The resulting estimate can be used as an approximation of the actual nonlinear solution or can be further refined by using it as initial guess for nonlinear optimization techniques. Experimental analysis demonstrates that such refinement is often unnecessary, since the linear estimate is already accurate in practice. Furthermore, we discuss how the approach allows to mitigate the orientation wraparound problem which is known to prevent convergence in state-of-the-art techniques.

 

 

 

Supplementary material

 

 

FR079

fr079

 

CSAIL

csail

 

FRH

frClinic

 

INTEL

intel lago_map

 

M3500

M3500 eg2o

 

M10000

M10000