Research On The Manipulator Simulation Platform Based On Augmented Reality

With the maturing and improvement of manipulator technology,manipulator become more and more precise and versatile,related research is moving toward integration,systematization and intelligence.In the field of manipulator simulation technology,simulation platform has become an indispensable method,which can be used to assist in the research of manipulator kinematics,dynamics and other properties.However,the simulation platform which based on virtual environment has many problems,such as large amount of modeling work in the early stage of construction,lack of comparison with the real environment and lack of immersion.Therefore,the research on manipulator simulation platform based on augmented reality is carried out to improve the performance and shorten the development period of the simulation platform,which in line with the current development needs and has important economic and practical value to the development of manipulator simulation technology.This paper studies the manipulator simulation platform based on augmented reality and focuses on three-dimensional registration and collision detection issues.The main contents are as follows:Firstly,design of overall architecture and function module of manipulator simulation platform based on augmented reality is carried out.The overall architecture and each functional module are designed based on the characteristics and requirements analysis of the simulation platform.On this basis,the design of work flow and data flow of the simulation platform is completed.It laid the foundation for the later research work.Secondly,marker-less three-dimensional registration algorithm of simulation platform is studied.This paper studies marker-less three-dimensional registration algorithm based on statistical learning.A marker-less three-dimensional registration algorithm of the simulation platform is realized by constructing a random forest classifier.Aiming at the poor robustness of rotation variation,illumination variation and viewing change of the existing algorithms,the method of constructing feature vector in training set is improved.The texture feature with rotation invariance and illumination invariance is added to the feature vector of the feature point,which improves the accuracy and robustness of the marker-less three-dimensional registration algorithm.Then,collision detection algorithm of simulation platform is studied.This paper studies collision detection algorithm based on hierarchical bounding volume.Based on the preprocessing of scene object and manipulator model,an AABB binary tree based on point cloud model is constructed and the collision detection algorithm of the simulation platform is realized.Aiming at the problems of single detection process and low detection efficiency of the existing collision detection algorithms,the process of collision detection algorithm is improved according to the distance of the collision object,which improves the efficiency of the collision detection algorithm.Finally,experimental verification of manipulator simulation platform based on augmented reality is carried out.In order to validate the practical application of manipulator simulation platform and related algorithms,a virtual simulation environment was set up based on Power-Cube modular manipulator experimental platform.Aiming at the marker-less three-dimensional registration algorithm and the collision detection algorithm,corresponding experimental tasks are designed to verify the accuracy and effectiveness of the algorithm.