This paper presents a new approach to obtaining nearly complete coverage paths (CP) with low overlapping on 3D general surfaces using mesh models given or reconstructed from actual scenes. The CP is obtained by segmenting the mesh model into a given number of clusters using constrained centroidal Voronoi tessellation (CCVT) and finding the shortest path from cluster centroids using the geodesic metric efficiently. We introduce a new cost function to harmoniously achieve uniform areas of the obtained clusters and a restriction on the variation of triangle normals during the construction of CCVTs. The obtained clusters can be used to construct high-quality viewpoints (VP) for visual coverage tasks. Here, we utilize the planned VPs as cleaning configurations to perform residual powder removal in additive manufacturing using manipulator robots. The self-occlusion of VPs and ensuring collision-free robot configurations are addressed by integrating a proposed optimization-based strategy to find a set of candidate rays for each VP into the motion planning phase. CP planning benchmarks and physical experiments are conducted to demonstrate the effectiveness of the proposed approach. We show that our approach can compute the CPs and VPs of various mesh models with a massive number of triangles within a reasonable time.

翻译：本文提出了一种新方法，用于在网格模型（通过实际场景给定或重建获得）上实现三维一般曲面中低重叠的近乎完整覆盖路径（CP）。该方法通过受约束质心泰森多边形法（CCVT）将网格模型分割为指定数量的聚类，并利用测地度量高效地从聚类质心寻找最短路径。我们引入了一种新的代价函数，以在CCVT构建过程中和谐地实现所得聚类的均匀面积并限制三角形法线变化。所得聚类可用于构建视觉覆盖任务的高质量视点（VP）。在此，我们将规划好的视点用作清洁构型，以在增材制造中利用机械臂执行残余粉末去除。通过将所提出的基于优化的策略（为每个VP寻找一组候选射线）集成到运动规划阶段，解决了视点自遮挡及确保无碰撞机器人构型的问题。我们开展了覆盖路径规划基准测试和物理实验，以证明所提方法的有效性。结果表明，我们的方法能在合理时间内处理包含海量三角形的多种网格模型，并计算出覆盖路径和视点。