Research And Design Of Writing Robot Based On Attitude Information

With the development of artificial intelligence,self-learning writing robots have always been the focus of contemporary research.This is a leap from simply accepting instructions to self-learning.At present,the mainstream of the writing robots are generally based on the vision of writing robots,because of its image data acquisition and entry have greater difficulty,the ever-changing images,the quality of writing is completely dependent on the results of image preprocessing,so to the characteristics The extraction has brought great difficulties.And the hardware with image processing is generally expensive and the cost is relatively high,which brings difficulty to large-scale practical applications.This thesis uses low-cost development chips to produce low-cost writing robots which are suitable for low-end market needs,and solves the problem that traditional vision-based writing robots to require high image processing costs.The main research contents of this article are as follows:(1)The low-speed six-axis accelerometer and gyro were used to obtain the attitude information data in writing,and then the RBF radial basis neural network was used to classify the acquired attitude data using a preprocessing algorithm to normalize the eigenvalues.Then,the manipulator send the character information data that has been learned to the lower computer controller and write the characters.This makes the characters written by the robot arm only depend on the learned characters,so that only the controller with the accelerometer and the gyroscope can control the robot arm to write,which reduces the cost of the traditional writing robot.(2)The RBF neural network is used to classify the character data that needs to be written.The RBF neural network is optimized from several aspects to improve the recognition rate.In addition,several groups of comparative experiments are designed to observe the recognition rate before and after the optimization.By analyzing the comparison of the line graphs of the recognition rate before and after optimization,the conclusion that the accuracy rate of the method proposed in this paper was higher.(3)The Neuromorphic Memory classification technology is studied,including KNN classifier and RBF neural network classifier.Using its excellent characteristics,the neuron learning classification is performed,and the results of the learning classification are written by the lower machine driving robot arm.Finally,it is experimentally verified that the improved accuracy of the RBF neural network classifier is higher than that of the traditional RBF neural network classifier,and the recognition result is transmitted to the robot arm by the lower machine.Achieve better results.