Analysis of robotic calligraphy copying techniques based on linear regression models

Abstract

Abstract A new robotic calligraphy system is constructed to address the problem of robotic calligraphy systems that only consider robotic writing in the straight state of the brush pen, enabling robots to write Chinese characters in the tilted state. A linear regression algorithm-based stroke model suitable for robotic brush writing is developed. Then a genetic algorithm is used to obtain trajectory information from the stroke images, and rules for starting, walking, and closing strokes are added to generate virtual stroke images. Finally, the B spline algorithm was used to realize the robot path planning and coordinate transformation to complete the pose calculation and perform the actual writing. The results show that the width of the pairs of strokes deepens as the stroke depth increases and the stroke speed decreases, corresponding to six sets of experimental speeds from 20mm/s to 120mm/s at stroke depths of 3mm and 7mm, respectively. From this, it can be seen that the relative error of most points is smaller than 10%, and the maximum relative error reaches 79.08%. The higher error was due to its lower benchmark value after comparison, which led to a huge error. In particular, as the independent variable increases in the experiment, making the benchmark value increase, its relative error decreases significantly, indicating the accuracy of the fitted model derived. It can be seen that the linear regression model-based robotic calligraphy copying technique helps to improve the robot’s purely repetitive trajectory in the writing process, adding its writing style to the original one and making it closer to humans.