Error distribution estimation based weighted least square estimation for service robot hand-eye calibration

Accurate hand-eye calibration is a very significant task in robotics. Many algorithms focus on L
₂
optimization, and Least Square Estimation (LSE) is usually employed to deal with this problem. For the reason that LSE can only be applied to occasions that errors obey normal distribution, LSE can lead to a lower precision of calibration result when gross errors occur. Different from the existing approaches, this paper introduces an error distribution estimation based weighted LSE for hand-eye calibration task. Firstly, transformation matrix is computed by traditional LSE. Error distribution is estimated on the basis of Gaussian kernel density estimation. Each data is weighted according to the density estimation and the fine result can be conducted by the weighted data. To evaluate the proposed method, an experiment was designed and the test result demonstrates the robustness of the proposed approach.