In small- and medium-sized enterprises, machine tools generate significant quantities of various types of dust, including plastic, iron, and other materials. To facilitate efficient cleaning or recycling of these materials, workers are required to manually vacuum the machines, directly targeting the accumulated dust. This paper introduces a novel approach for automated cleaning in industrial settings, diverging from the common practice of systematic surface coverage. Our method, mobile arm vision-based adaptive cleaning, employs a two-stage adaptive strategy that necessitates only a coarse representation of the machine. First, an offline process optimizes the positions of the robot’s mobile base and arm configurations to maximize the visibility of machine sections from the camera’s perspective at the end effector. Then, an online process detects dust accumulations and adjusts its trajectory based on their locations. We demonstrate the capability of our approach by cleaning three machines with intricate geometries. Our results indicate that our method can clean approximately 84% of the dust heaps present on the machines.
Dupoiron et al. (Thu,) studied this question.