Optimal Realtime Toolpath Planning for Industrial Robots with Sparse Sensing

Non-contact surface processing does not involve direct contact between the tool and a worksurface. An industrial robot mostly uses preplanned toolpaths to perform non-contact surface processing. A preplanned toolpath may work well in repetitive conditions but may easily become inaccurate and unsafe if the tool needs to follow unknown worksurface variations. Many industrial processes, e.g., painting, coating, and sandblasting, typically involve worksurfaces with unknown variations. This study proposes an optimal toolpath planning method for an industrial robot equipped with end-of-arm distance sensors to automatically guide its tool motion along unknown worksurface variations. The distance sensors facilitate sparse sensing to acquire sparse data that is just enough for the quick and adequate perception of unknown worksurfaces by requiring fewer measurements and less computing. Optimization facilitates the optimality of multi-objective toolpath planning with a customizable value function, where the multiple objectives comprise adapting to unknown worksurface variations and traveling between known tool targets. To validate the proposed toolpath planning method, this study conducts a simulation experiment on a virtual robot with four end-of-arm distance sensors and a workpiece with unknown surface variations. The experimental results indicate that the proposed method is accurate and near-optimal even in the presence of sensor noises.