Scholarlink

<p>To address the challenges of task allocation, collision-free trajectory planning, and consistent multi-view 3D reconstruction in multi-robot collaborative measurement systems, a normal vector optimization algorithm based on principal component analysis (PCA-NOA) is proposed. This method enhances the accuracy and robustness of normal estimation for complex surfaces under sparse point cloud conditions. Experimental results indicate that, compared to conventional principal component analysis, the proposed algorithm improves normal consistency by 0.011, increases normal uniformity by 0.255, and reduces curvature estimation error by 0.009. Furthermore, a model-driven adaptive viewpoint trajectory planning method (MD-VTAP) is introduced, enabling the hand-eye system to autonomously generate optimal 3D execution trajectories in the presence of occlusion and significant curvature variation. Applied to an eye-on-hand grinding system, the framework accurately guides robotic arms in positioning and grinding highly curved surfaces. In a dual-arm eye-on-hand cooperative 3D scanning setup, the system achieves an average reconstruction error of 0.106&#xa0;mm with a standard deviation of 0.032&#xa0;mm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A model-driven and normal-optimized viewpoint planning method for collaborative 3D measurement of robots

  • 期刊论文

  • 发表日期：

    2026年2月13日

查看全文
  • Tenglong Zheng,
  • Xiaoying Feng,
  • Siyuan Wang,
  • Haozhen Huang,
  • Shoupeng Li

摘要

To address the challenges of task allocation, collision-free trajectory planning, and consistent multi-view 3D reconstruction in multi-robot collaborative measurement systems, a normal vector optimization algorithm based on principal component analysis (PCA-NOA) is proposed. This method enhances the accuracy and robustness of normal estimation for complex surfaces under sparse point cloud conditions. Experimental results indicate that, compared to conventional principal component analysis, the proposed algorithm improves normal consistency by 0.011, increases normal uniformity by 0.255, and reduces curvature estimation error by 0.009. Furthermore, a model-driven adaptive viewpoint trajectory planning method (MD-VTAP) is introduced, enabling the hand-eye system to autonomously generate optimal 3D execution trajectories in the presence of occlusion and significant curvature variation. Applied to an eye-on-hand grinding system, the framework accurately guides robotic arms in positioning and grinding highly curved surfaces. In a dual-arm eye-on-hand cooperative 3D scanning setup, the system achieves an average reconstruction error of 0.106 mm with a standard deviation of 0.032 mm.