一种基于迭代的质心逆运算的方法

Inverse Kinematics for Center of Mass Through Iteration

通过已知质心精确反解计算仿人机器人各关节的角度是一个经常遇到的问题。在双足行走,平衡控制等领域都很常见。但对于自由度高的仿人机器人系统,质心逆运算比较困难,尤其在双足支撑情况下,问题变为一个多自由度的并联机构,此时需要额外的约束和限制条件,使得计算非常复杂。本文基于Levenberg-Marquardt算法来解决复杂关节的逆解问题,研究在给定踝关节的情况下,用假定质心固定身体上的简化模型来使得真实质心逼近目标点,然后通过重复逼近缩小误差。我们通过NAO仿人机器人模型上的模拟验证了该算法实现了较高的准确性和计算效率。

In various humanoid robot research areas such as biped walking and balance maintenance,control of Center of Mass(COM) is frequently used.However,it becomes a complex problem to directly control the COM for a high degree of freedom humanoid robot,especially in case of double feet support.In this paper,an iterative method that is easy to implement is proposed.Based on the approximated assumption that COM is fixed in local coordinate system,the control angle of the robot leg can be solved to drive the COM toward the target position.This process can be iterated to achieve high precision.Simulation tests on a NAO humanoid robot have demonstrated that the algorithm is efficient and precise.