A study on variable curvature trajectory planning method for robotic weld grinding based on line-structured light vision guidance

The service life and dynamic performance of structural welds are significantly influenced by the grinding precision and surface consistency. Currently, welding seam grinding is predominantly performed manually or through teaching methods, with the quality of grinding heavily reliant on the experience of the workers. Additionally, due to the complex shape and uneven curvature of weld seams, it is challenging to effectively track the weld seam contour, often resulting in workpiece scratches or inadequate weld seam removal during the grinding process. Based on the above problems, a trajectory planning method of robot weld grinding with variable curvature based on linear structured light vision guidance is proposed. Firstly, the effects of measurement angle ($\alpha$) and forward-looking distance ($l$) on measurement error are analyzed based on the principle of pinhole imaging. This analysis yields the optimal measurement parameters ($\alpha ={90}^{°}$, $l=200\mathrm{mm}$). Subsequently, the topological relationships between the weld point clouds are constructed, the three-dimensional topography is reconstructed, and the weld features (width, height, and centre position) are extracted. Next, the variable curvature-based grinding path planning method is described in detail. The proposed method can adaptively adjust the target points of discrete trajectories according to the threshold of curvature change, thereby enabling efficient and accurate tracking of the weld contour paths. To verify the effectiveness and superiority of this method, a comparative experiment on weld grinding was carried out with a curve weld of a pump truck as the experimental object. The experimental results show that the weld surface after grinding by this method is smooth, with no apparent ripples in the curvature transition zone, the minimum Ra value can reach 0.412 μm, and the average residual height is 0.151 mm.