Deep-Neural-Network-Based Positioning Error Compensation Method And Its Application Of Industrial Robots

Industrial robots are widely used in intelligent manufacturing industries due to high efficiency,high flexibility,and low cost.However,their low absolute positioning accuracy has limited their development and application in the field of aerospace manufacturing.Therefore,the research on robot precision compensation method is the key to promote the application of industrial robots and to improve the level of aviation manufacturing.In order to solve the problems that the traditional off-line precision compensation methods depend on the kinematic models and limit the compensation precision,which can not further improve the absolute positioning accuracy of the robots,this paper puts forward a robot positioning error compensation model based on deep neural network.This model is based on the Latin hypercube sampling planning method,and can significantly improve the absolute positioning accuracy of robots.The main work of this paper is as follows:(1)In view of the existing sampling point planning methods are difficult to apply to the non kinematic model calibration methods and attitude sampling of robots,a Latin hypercube sampling point planning method for robots is proposed.Based on this method,the influence of robot attitude on positioning error is analyzed,which provides theoretical support for establishing the error compensation model.(2)A method which can optimize initial network weights and thresholds based on genetic particle swarm optimization is proposed.Based on the Latin hypercube sampling,a prediction model of robot positioning errors based on deep neural network is established,and a compensation method of robot positioning error is proposed.(3)The simulation optimization of the positioning error prediction model is realized.The feedforward compensation method of the robot positioning error based on the depth neural network is verified by the experiment,which realizes the compensation of the robot positioning error.The comparison test results with other precision compensation methods show that the method proposed in this paper is feasible and superior in improving the absolute positioning accuracy of robots.(4)The hardware and software composition,work flow of the automatic drilling system of mobile robot are described,the compensation method of drilling accuracy is verified.The test results show that the hole positioning accuracy of the automatic drilling system for the mobile robot is increased from 1.879 mm to 0.227 mm.