In response to the low accuracy, poor deployment and high calibration costs of vision-based manipulator systems, this paper proposed the adaptive multiple-elites-guided composite differential evolution algorithm with a layered optimization mechanism(AMECoDEs-LO) and a joint system calibration strategy. Firstly, it integrated the kinematic model of the robotic arms and the hand-eye calibration external model systematically. Then, based on AMECoDEs algorithm, it performed principal component analysis on the phased data in the population. It divided parameter optimization prioritization according to the degree of dominance of each dimensional vector in the current generation, and achieved implicit guidance on the accuracy and speed of population convergence with the idea of parametric dimensionality reduction optimization. It was validated against existing first-class evolutionary algorithms in simulated and real environments. Finally, it verified the system's robustness by adding Gaussian white noise with different intensities to address the sensitivity of the vision sensor to environmental noise. Experimental results show that the algorithm has the advantages of high calibration accuracy, fast convergence, good robustness and no additional calibration instruments are required for the calibration of the vision-based manipulator system, which can be used for the rapid deployment of vision-based manipulators.

[2022-07-05] Accepted Paper /doc/2022.03.0187/v1

[2022-11-05] Printed Article /doc/2022.03.0187/v2