Research On Robot Multi-style Calligraphy Based On Force Sensing Signal Learning

Combining force information and machine learning algorithms to study the calligraphy robot can better preserve the calligraphy style,while providing the robot with higher performance in other complex tasks.Reference significance.At present,the application of robots is mainly teaching.In order to improve the learning ability of robots,the paper proposes a robot calligraphy scheme that combines force sensing technology and machine learning algorithms,and is verified and analyzed through experiments.The thesis first collects the contact force characteristics of the Chinese characters written by the demonstrators through force sensors,obtains the trajectory coordinates in the vertical direction of calligraphy,learns the calligraphy style through the neural network,and finally combines the refinement algorithm and the contact force characteristics to obtain the trajectory in the process of robot calligraphy Combined with the above technologies,the robot calligraphy cloud simulation platform was developed.The main research contents of the paper are as follows:(1)Propose a method to deal with the signal of calligraphy contact force of the demonstrator.This method first designed a writing brush fixture fixed on the force sensor,used the force sensor to cooperate with the fixture to collect force signals,and used multiple fitting equations to summarize the contact force characteristics.After the parameter comparison,the sine equation was used to fit the data.The elasticity of the brush strokes was analyzed,and finally the trajectory coordinates in the vertical direction were obtained.(2)According to the more complex features of the calligraphy style data set used in the paper,the traditional Le Net-5 convolutional neural network is improved,and the effect is compared through experiments,which proves that improved Le Net-5 can learn the calligraphy style better.And based on Cycle GAN,a transfer model between different calligraphy styles is constructed,and the trained generation model is used to transfer the style of Chinese characters.(3)Based on the thinning algorithm,the pixel coordinates of the Chinese character skeleton are extracted,and then converted into actual coordinates according to a certain ratio.Based on Open CV,performed operations such as binarization,denoising and filtering on Chinese calligraphy characters,and the thinning algorithm was used to extract the trajectory of the Chinese characters and converted into X,Y plane trajectories in the Cartesian coordinate system of the robot.The contact force characteristics are used to obtain the robot end trajectory.Finally,the robot's kinematics and coordinate system calibration are combined to generate the corresponding control program.(4)Build a cloud platform for robot calligraphy simulation.The robot calligraphy system proposed in the paper is integrated into the platform,the front-end interface is built based on React,the back-end server is built based on Flask,and the effectiveness of the robot calligraphy scheme proposed in the paper is verified through experiments.