Positional error compensation for aviation drilling robot based on Bayesian linear regression

The quality of aircraft components is directly affected by the positional accuracy of aviation drilling robot, which may further affect the assembly accuracy, flight performance and service life of aircraft. In this work, positional error compensation based on Bayesian linear regression (BLR) is proposed to improve the positional accuracy of aviation drilling robot. Firstly, the stochastic model of robot positional error is constructed by utilizing kinematic error model involving with random uncertainties. It proves that the positional error of robot joint space follows the Gaussian process (GP) and can be modeled through a linear regression model. Secondly, BLR is utilized to construct the positional error model of aviation drilling robot, which can provide the predicted positional error and the corresponding confidence interval (CI) of target position. These predicted values and CIs are used to realize error compensation. Experimental results show that, after error compensation, the average/maximum absolute positional error of the self-developed aviation drilling robot is decreased from 1.393mm/1.795 mm-0.081mm/0.167 mm. This indicates that the BLR-based positional error compensation is conductive to ameliorating the positional accuracy of aviation drilling robot. This study lays the foundation for the positional error compensation of industrial robots in aircraft assembly.

Automated summary

This paper proposes a positional error compensation method based on Bayesian linear regression to improve the positional accuracy of aviation drilling robots. By constructing a stochastic model for robot positional error and utilizing Bayesian linear regression, predicted positional errors and confidence intervals can be provided for error compensation. Experimental results show that this method significantly reduces the positional error of aviation drilling robots.