Motion planning of cleaning robot based on 3D vision.

Aiming to address the current problems in the industrial cleaning field, including water waste, blind spots, and poor equipment compatibility, this study proposes a three-dimensional vision-based cleaning robot design. The proposed design is primarily developed for cleaning workpieces in the manufacturing of rail transit equipment. It integrates both high-flow, low-pressure cleaning and low-flow, high-pressure cleaning, using the proportional-integral-derivative (PID) control technology to precisely adjust the cleaning pressure and water temperature, allowing robots to automatically adapt to the characteristics of various materials and stains. The proposed design employs a template matching algorithm and an improved CIoU-YOLOv7 object recognition algorithm to achieve rapid workpiece recognition, stain detection, and accurate positioning. Compared to the baseline algorithm, the proposed method increases the detection accuracy by 5-10%, reaching 62.34%. The genetic algorithm optimizes the cleaning path, reducing the total path by 64.7% compared to the original path while increasing efficiency by 10.5% over the basic genetic algorithm. These improvements enable efficient automatic cleaning of various workpieces, enhancing both cleaning efficiency and quality. The proposed method not only saves water but also significantly enhances the intelligence level of cleaning systems by reducing the number of generations required to achieve an optimal solution by 95%. Thus, the proposed design can effectively address the problems of low efficiency in traditional cleaning methods.