Research On A Robot Intelligent Compliant Assembly System

With the development of artificial intelligence and robot technology,the working environment of the manipulator has gradually changed from a relatively fixed production line to a complex production and life scene,which puts forward higher requirements for the working ability of the manipulator.In order to adapt to the flexible working environment,the intelligence and flexibility of the manipulator are gradually becoming the key performance of the manipulator.In order to improve the intelligence and flexibility of the manipulator,the following research is carried out in this paper.An object 6D estimation algorithm based on deep convolution neural network.This paper studies and analyzes a variety of object 6D estimation algorithms,selects the most suitable ssd-6d algorithm in combination with the actual application scenario of the manipulator,uses OpenGL to make sample data,and uses the migration learning method to train the network model suitable for this paper.The loss function of the model on the training set converges from 1.4 to about 0.005,which decreases by 99.49%.The loss function on the verification set converges from 1.2 to about 0.02,decreasing by 99.01%.Based on the six axis force sensor,the precise force sensing of the end of the manipulator is realized.Considering the factors of load gravity,sensor weight,sensor zero point and force coefficient,the force sensing model of manipulator end is established,and a method of model parameter calibration is given.In this method,the gyroscope is used to compensate the attitude error of the manipulator,and the force coefficient,load gravity and other model parameters are obtained through multiple sets of data.Finally,three groups of experiments are carried out.The results show that the error of sensing external force of six axis force sensor is less than 0.23%of load gravity,the error of sensing external torque is 0.79%of load to sensor torque,the calibration error of load mass is within 0.089%of load mass,and the calibration error of load center of gravity is within 1.245%.Impedance control based on force error.Based on the position control as the inner loop,the characteristics of traditional impedance control are studied,and its shortcomings in unknown contact environment are analyzed.In order to adjust the desired position of the manipulator in real time,an impedance control law based on force error information is designed in this paper.The expected position function is constructed with force error as independent variable,and its stability is proved by Lyapunov stability theory.Finally,three conditions are designed,and the simulation results of the two control laws are compared and analyzed by MATLAB.Under different working conditions,the rise time of the impedance control system based on force error information is reduced by 28%-66%,the steady-state error in some environments is eliminated,and the fluctuation amplitude is reduced by 89.46%,which is better than the position based impedance control in all aspects.Finally,the experimental verification of intelligent compliant control system is carried out.The robot arm is endowed with intelligence by the object 6D estimation algorithm based on the deep convolution neural network and the contact force information at the end of the robot arm is collected by the six dimensional force sensor.The flexibility and intelligence of the robot arm are improved by integrating the force sensing algorithm and the impedance control algorithm.The experimental platform is built,the control software is written,two experimental schemes are designed,and the grabbing and assembling experimental tasks are carried out on the experimental platform.The effectiveness of the proposed algorithm is verified by experiments.