An Enhanced Step-Size Gaussian Damped Least Squares Method Based on Machine Learning for Inverse Kinematics of Redundant Robots

Xiaoqi Wang,Xing Liu,Heyu Hu,Lerui Chen,Jianfu Cao

doi:10.1109/access.2020.2986421

Xiaoqi Wang, Xing Liu + Show 3 more

Open Access

https://doi.org/10.1109/access.2020.2986421

Copy DOI

Abstract

An existing problem in the robotic field is to solve the inverse kinematics (IK) problem of redundant robot with high speed and high precision. A novel IK optimization method based on the Gaussian Damped Least Squares (GDLS) is proposed in this paper. A significant contribution of this method is to make the iteration converge in a faster and more accurate way by introducing an optimal enhanced step-size coefficient. The machine learning model can be trained with 10 6 data points in the reachable region of the robot, and the optimal enhanced step-size coefficient in each solving process can be predicted by the model. The accuracy and stability of the algorithm proposed are verified through an example of an arbitrary 7R redundant robot. The average number of iterations is less than 10, with super high solving speed. Furthermore, the algorithm also has better convergence, which can reach 96.23% when the error threshold is 0.01 mm. The common IK methods are evaluated in this paper, and the results show that the optimized method has good performance in convergence, accuracy and speed.

Highlights

The robot inverse kinematics (IK) solves joint velocities with a given robot endeffector velocity
In [2], [3], the singular control method based on damped least square(DLS) method is proposed, which is known as Levenberg-Marquardt method
Gaussian Damped Least Squares (GDLS) method is used to solve the inverse kinematics of each data point, and the k_optimal value of each solution can be obtained by manually adjusting k with the step size of 1 in the range of [1, 100]

Summary

INTRODUCTION

The robot IK solves joint velocities with a given robot endeffector velocity. Jacobian matrix pseudo-inverse method [1] is a basic and general method. The optimal enhanced step-size coefficient in each solving process is predicted by the models of machine learning This method is a new optimization idea of DLS-series methods to improve the solving speed. The damped least squares (DLS) method has been well-known as stabilizer of pseudo inverse for near-singular points It considers the balance between the precision of the solution and the increase in joint velocities to achieve smooth motion when the robot approaches the singular configuration. It creates continuity in transition between singular configuration and non-singular configuration, and involves selection of the damping factor to approximate solution of the Jacobian near-singular points. Compared with methods such as DL-DLS, etc., G-DLS can save the time cost of calculation

ENHANCED STEP-SIZE COEFFICIENT

PERFORMANCE EVALUATIONS

CONCLUSION