------------------------------------------------------------------------------------------------------ Algorithm Description A detailed description of the algorithm used can be found in the file algorithm_description.pdf, containing several descriptive figures about the experiment. ------------------------------------------------------------------------------------------------------ Output of the algorithm The raw output of the algorithm can be seen in bs_output.txt, in the format Origin for the absolute position is the first camera location. ------------------------------------------------------------------------------------------------------ Absolute Trajectory Error (ATE) mandatory performance measure ATE = [1.3812 0.2234 0.7111 2.0513] in meters The file XY.txt contains the translational components of the robot poses, reconstructed by our algorithm, evaluated in the timestamps when the GT is available. To compute this performance measure we use the function "get_ATE" in Matlab v7.4, available in the file get_errors.rar that we provide. The interpolation is carried out with the Matlab function "interp1", in linear mode. ------------------------------------------------------------------------------------------------------ Rough Estimate of Complexity (REC) mandatory performance measure The file rec.txt contains the computational time for each timestamp with the format . CI-Graph SLAM is an approach that has a constant time during exploration steps. Its cost can be linear in the number of map features when loop closures are performed. ------------------------------------------------------------------------------------------------------ Relative Pose Error (RPE) recommended performance measure T-RPE = 3.5468 m^2 R-RPE = 0.6002 rad^2 The file with the relative transformationsfor the GT used was provided by Michael Ruhnke from http://www.informatik.uni-freiburg.de/~ruhnke/rawseeds/relative-pose-error-relations.tar.gz To compute this performance measure we use the function "get_RPE" in Matlab v7.4, available in the file get_errors.rar that we provide. The interpolation is carried out with the Matlab function "interp1", in linear mode. ------------------------------------------------------------------------------------------------------