Research On Trajectory Planning Of Six-Degree-of-Freedom Manipulator Based On Target Positioning

Mechanical arms are high-tech products that integrate advanced technologies from multiple disciplines.They can replace humans to work in extremely harsh environments and have high production efficiency,making them a hot research topic for scientists.Traditional robotic arms can only carry out a single,repeated operation,which cannot meet the needs of modern production.Therefore,researchers have combined computer vision with mechanical arms.By using visual computing to detect the position of target objects and then controlling the mechanical arm to reach the target position,the operation process is simplified and made more intelligent.This article analyzes the motion trajectory planning and other issues of the AUBO-i5 robotic arm and focuses on the following research topics:(1)Firstly,the article briefly introduces the structure and characteristics of the AUBOi5 six-degree-of-freedom robotic arm,analyzes the basic theoretical knowledge of kinematics,and establishes the kinematic model.Secondly,the article describes the pose information of the end effector of the robotic arm and the transformation methods of various coordinate systems.Finally,to analyze the accuracy of forward and inverse kinematics,this study verifies the results using the Matlab simulation software.(2)Secondly,the grasping environment of the robot arm is complex and changeable.For example,when the robot arm works in the fog and haze environment,features will be lost in the images collected by the camera.To solve this problem,a single image defogging network based on Cycle-dehaznet is proposed.This method combines the Cyclegan and Dehazenet networks,uses Dehazenet architecture as the feature extraction module of the generator in the Cyclegan network,and combines existing assumptions and prior knowledge to improve the dehazing performance of the network,which effectively reduces the adverse effects of haze on image quality and improves the grasping and detection accuracy of the robotic arm.Furthermore,the object localization technology,the article analyzes the rectangular box grasping detection algorithm based on convolutional neural networks.The acquired image is subjected to rectangular box grasping detection operation.By using the feature extraction ability of convolutional neural networks,the two-dimensional position relationship of the robotic arm grasping can be learned from the image data output.Finally,the two-dimensional grasping parameters are transformed into spatial coordinates,and the spatial position information of the target object is output,preparing for the next trajectory planning and optimization.(3)Thirdly,the article addresses a problem of acceleration mutation that occurs during the operation of the robotic arm,which can cause the motor to exceed the driving range.To solve this problem,the article selects a combination of third and fifth-order polynomial interpolation algorithms to optimize the motion trajectory of the robotic arm.In the trajectory planning process,the driving device of the robotic arm must meet the actual load requirements.Therefore,the selection of joint speeds and accelerations will be relatively conservative,resulting in a long time required to complete a set of actions,which means that the continuity and smoothness of the robotic arm cannot be fully utilized.To solve the problem of joint speed and acceleration optimization selection,the article proposes a novel method of time-optimal trajectory planning for the robotic arm based on improved butterfly algorithm.By introducing the Levy flight and sine-cosine algorithm into the butterfly optimization algorithm,the time of the robotic arm’s operation can be optimized,the improved run time was reduced by nearly 36%.It greatly improved the working efficiency of the mechanical arm.