A Redundant Robot Repeated Motion Planning Method Based on Quadratic Radical Final State Attraction

Abstract

The invention relates to a redundant robot repetitive motion planning method based on a quadratic formula final state attraction property index. The target terminal trace rd(t) of a robot tail end actuator is provided in the Cartesian space, and the expected returning angle theta d(0) of each joint is provided as well. The property index enabling the quadratic formula final state attraction property of the deviation of a current joint vector from an expected joint vector is designed. The quadratic optimization problem of trace panning of a redundant robot can be converted to solving a time-varying matrix equation, and a finite value final state neural network can work as a solver. The method achieves finite-time convergent repetitive motion planning of the redundant robot under the condition of initial position deviation. The redundant robot repetitive motion planning method has the advantages of finite-time convergence, high calculation precision and easy achievability.

Description

technical field [0001] The present invention relates to industrial robot motion planning technology, specifically, it provides a redundant robot repetitive motion control method based on finite value final state neural network under the condition of finite time convergence performance index and initial position deviating from expected trajectory. Background technique [0002] Industrial robots can move any object according to the desired spatial posture, so as to complete a specific job requirement. The number of joints whose mechanism can move independently is called the degree of freedom of movement of this mechanism. Six degrees of freedom is the minimum number of degrees of freedom with the ability to complete spatial positioning, and redundant robots have more degrees of freedom than are required for the task. Utilizing the redundancy feature, the flexibility and fault tolerance of redundant robots can be enhanced, avoiding obstacles in the working environment, avoidin...

AI Technical Summary

Problems solved by technology

In addition, the neural networks with finite time convergence in the published literature mostly use linear activation functions, or have infinite value activation functions. In actual implementation, due to limited energy, there are inherent difficulties in the implementation of infinite value activation function neural networks.

Images