A Stochastic Global Optimization Algorithm for the Two-Frame Sensor Calibration Problem (IEEE TIE 2016)

Paper

MATLAB implementation of Stochastic Global Optimization Algorithm for the Two-Frame Sensor Calibration Problem AX = YB

Overview

This is an implementation of a fast and numerically robust optimization algorithm for the two-frame sensor calibration (or equivalently, hand-eye and robot-world calibration) AX = YB, which is a two-phase stochastic geometric optimization algorithm for finding the global minimizer. Using coordinate-invariant differential geometric methods that take into account the matrix Lie group structure of the rigid-body transformations, the algorithm makes use of analytic gradients and Hessians, and a strictly descending fast step-size estimate to achieve significant performance improvements.

Instruction

To run the calibration, call

[X, Y] = solveAXYB_SE3(A,B,alpha,param)

where

• A, B are the measurement datasets, each of which is in size of 4 X 4 X n (for n measurements).
• alpha is weight factor of translation error; e.g., transration error of 1.0mm is equally weighted to 1 radian rotational error given alpha = 1.0 (when length unit in data is mm).
• param is a struct of algorithm parameters. Declare by param = defaultParam() and set param.globalOptMethod = 2 for stochastic global optimization. See instruction.docx for more details.
• X, Y are the calibration results.

Demos

Demo codes are included that generate sythetic datasets and run calibrations on the generated datasets.

• To generate a synthetic dataset, run main_dataGeneration_SE3.m.
• To run calibration on the generated dataset, run main_solve_AXYB_SE3.m.

SO(3) case is handled in main_dataGeneration_SO3.m and main_solve_AXYB_SO3.m