Robot posture control method, apparatus, device, medium, and program product

By acquiring RGB-D images of the garment surface and tactile sensor signals, visual and tactile feature vectors are constructed. Combined with a neural network inverse dynamics controller and a PID corrector, joint torque is dynamically adjusted, solving the problems of lack of material perception and insufficient real-time performance in existing technologies. This enables precise control and safety protection for different materials.

CN122401433APending Publication Date: 2026-07-17SHENZHEN PINKUO INFORMATION TECHNOLOGY CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SHENZHEN PINKUO INFORMATION TECHNOLOGY CO LTD
Filing Date
2026-06-11
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing robot posture control technology cannot accurately distinguish the mechanical properties of different materials, resulting in damage to thin materials or insufficient control gain for thick materials. Furthermore, it involves large computational loads, insufficient real-time performance, and lacks the ability to predict potential risks of exceeding limits.

Method used

By acquiring RGB-D images of the garment surface and tactile sensor signals, visual and tactile feature vectors are constructed. Combined with a neural network inverse dynamics controller and a PID corrector, joint torque commands are dynamically adjusted to achieve automatic sensing and real-time control of different materials.

Benefits of technology

It enables automatic sensing and differentiation of different materials, dynamically adjusts joint torque, improves real-time performance and safety, and reduces the risk of clothing damage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122401433A_ABST
    Figure CN122401433A_ABST
Patent Text Reader

Abstract

本申请适用于机器人控制技术领域,提供了一种机器人姿态控制方法、装置、设备、介质及程序产品,方法包括:采集服装表面的RGB‑D图像并构建视觉特征向量,通过触觉传感器采集力响应时序信号并构建触觉特征向量,确定材质类别并计算动态软阈值和动态紧急阈值;将当前关节角度向量、关节角速度向量与目标接触力向量输入神经网络逆动力学控制器执行推理,得到推理输出,并通过目标接触力与上一控制周期的实际接触力计算补偿力矩后与推理输出叠加,得到关节力矩命令;基于动态软阈值和动态紧急阈值对关节力矩命令进行调整,并将调整后的关节力矩命令下发至关节驱动器执行,本发明将安全响应从事后被动制动提升为事前主动预判,降低了服装损坏风险。
Need to check novelty before this filing date? Find Prior Art