Robot geometric error modeling and compensation method based on double feedback recursive dynamic network

CN122299663APending Publication Date: 2026-06-30SUZHOU UNIV OF SCI & TECH
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SUZHOU UNIV OF SCI & TECH
Filing Date
2026-05-21
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing technologies are insufficient to effectively compensate for dynamic errors such as nonlinear residual errors, joint flexibility deformation, and temperature drift in the end-effector positioning accuracy of industrial robots. Traditional methods lack a unified closed-loop collaborative mechanism, resulting in insufficient modeling accuracy and robustness.

Method used

A dual-feedback recursive dynamic network is used to construct a model of robot joint angles and nominal Cartesian coordinates. An extended Kalman filter is introduced for explicit estimation of context state. By adaptively adjusting the multi-inspiration length and self-feedback factor through the innovation energy ratio, a closed-loop collaborative recursive mechanism integrating state, weight, and feedback factor is constructed to achieve geometric error compensation.

Benefits of technology

It significantly improves the absolute positioning accuracy of robots, reducing the maximum positioning error by 63.29% and the root mean square error by 74.77%. It is simple, efficient, and easy to deploy in engineering projects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122299663A_ABST
    Figure CN122299663A_ABST
Patent Text Reader

Abstract

This invention relates to a robot geometric error modeling and compensation method based on a dual-feedback recursive dynamic network. It constructs a dual-feedback recursive network model with robot joint angles and nominal Cartesian coordinates as inputs and end-effector position residuals as outputs. The dual context layer outputs are constructed as augmented state vectors, and a noisy nonlinear state-space model is established. The context states are explicitly estimated using extended Kalman filtering, and the multi-innovation length and dual self-feedback factors are synchronously and adaptively adjusted based on the innovation energy ratio. A closed-loop collaborative recursive mechanism for state estimation, weight update, and feedback factor adjustment is constructed, forming an integrated state and parameter estimation mechanism. The residual estimates are superimposed and compensated onto the nominal kinematic output, effectively suppressing geometric errors and significantly improving the absolute positioning accuracy of industrial robots.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Multi-innovation OIF Elman network modeling method based on measurement data

    CN120654737A