MATLAB implementation of Probabilistic Framework for Hand-Eye and Robot-World Calibration AX = YB
This is a MATLAB code of probabilistic solver for hand-eye and robot-world calibration AX = YB, the detailed algorithm of which is presented in the paper entitled "Probabilistic Framework for Hand–Eye and Robot–World Calibration AX=YB" (under review). The algorithm incorporates different individual noise distributions of measurements A_i and B_i, and also provides calibration uncertainty as an error covariance matrix.
The code is uploaded for review purpose now. The instruction and the code will be cleaned up once the paper is accepted.
- Scripts to plot figures in the first and second revisions have been newly included (e.g., main_fig5_R1_R2.m)
- This repository will only contain the pure algorithm once the paper gets accepted. The experiment codes will be moved to another repository.
- Codes from https://github.com/ihtishamaliktk/RWHE-Calib were used in experiments. Please allow us some time to clean up codes and properly cite them.
- See the three system noise configurations presented in the paper.
- Calibration functions are different between noise configurations 1,2 and noise configuration 3.
- For noise configurations 1 and 2, call
[X, Y] = solveAXYB_prob(A, B, X0, Y0, invSig_wN, invSig_pN, invSig_wM, invSig_pM, noiseConf, step_R, step_p) - For noise configuration 3, call
[X, Y] = solveAXYB_prob_noiselessA(A, B, X0, Y0, invSig_wM, invSig_pM, step_R, step_p)
- For noise configurations 1 and 2, call
Here X, Y are the calibration results, and A, B are the measurements pairs in size of 4 X 4 X n each (n is the number of meausurement pairs). invSig_wN, invSig_pN, invSig_wM, invSig_pM are the inverses of rotational and translational noise covariances of A and B, each of which is in size of 3 X 3 X n. step_R, step_p are stepsizes for rotation and translation, respectively.
The demo scripts are named as main_fig5.m, main_fig6.mm, ..., which generate the figures in the paper.