An autonomous spraying method for indoor spraying robots based on visual assistance

Abstract

A vision-assisted autonomous spraying method is proposed for enhancing the autonomy of indoor spraying robots. Firstly, an improved YOLOv5s network model, YOLOv5-CD is proposed to detect and locate the indoor non-sprayable area. Then a four degree of freedom spraying robotic arm is designed, and a spraying operation map is constructed. The optimal spacing between spraying stations is obtained by combining the wall model and spraying operation map. Finally, the path of spraying gun is planned by using quintic polynomial interpolation. Experimental results indicate a 2.7% increase in model accuracy and a 14.18% reduction in parameters compared to the baseline model, and the detection speed is shortened by 2.9ms. Meanwhile, simulation tests on the spraying site assignment, robot path planning and spraying path achieve the autonomous spraying operation under this method.