Mixed production is a practical solution to realize the idea of mass product customization. The production system needs to adapt to raw materials of product variants appearing in the production line, and thus to properly complete the manufacturing task. Such system autonomy is a core competence in smart manufacturing. This paper presents a smart manufacturing solution for shoe sole gluing in mixed production environment. The solution automatically guides a 5-axis CNC dispensing machine to perform the gluing task of a sole randomly placed on the worktable without use of fixturing. A single RGB-D camera is used to capture the color and depth data of the work environment. A two-stage algorithm is implemented to provide the intelligence required by the solution using the captured data as input. The LINEMOD method first determines a rough pose from training data of all shoe soles at pre-defined orientations. A trimmed interactive closest point (ICP) method then refines the pose using the rough one as the initial solution. Pre-determined working paths are transformed into the correct location around the sole with the refined pose. The corresponding motion commands drive the dispenser controller to complete the gluing task without collision. Test results of real shoe soles validate the effectiveness of the proposed solution. This work contributes to smart manufacturing in the shoe making industry, which has been less studied in the past.

中文翻译：

---

鞋底混合生产中多轴点胶机运动计划的智能制造解决方案

混合生产是实现批量产品定制想法的实用解决方案。生产系统需要适应生产线上出现的产品变体的原材料，从而正确完成生产任务。这种系统自治是智能制造的核心能力。本文提出了一种在混合生产环境中用于鞋底胶粘的智能制造解决方案。该解决方案会自动引导5轴CNC点胶机执行粘合操作，该粘合操作是将随机放置在工作台上的鞋底完成而无需使用夹具。单个RGB-D摄像机用于捕获工作环境的颜色和深度数据。实现了两阶段算法，以使用捕获的数据作为输入来提供解决方案所需的智能。LINEMOD方法首先根据所有鞋底在预定方向上的训练数据来确定大致姿势。然后，修剪后的交互式最近点（ICP）方法使用粗糙的姿势作为初始解决方案来改进姿势。预定的工作路径将以精致的姿势转换成围绕鞋底的正确位置。相应的运动命令驱动分配器控制器完成胶粘任务，而不会发生碰撞。真实鞋底的测试结果验证了所提出解决方案的有效性。这项工作为制鞋业的智能制造做出了贡献，过去很少对此进行研究。预定的工作路径将以精致的姿势转换成围绕鞋底的正确位置。相应的运动命令驱动分配器控制器完成胶粘任务，而不会发生碰撞。真实鞋底的测试结果验证了所提出解决方案的有效性。这项工作为制鞋业的智能制造做出了贡献，过去很少对此进行研究。预定的工作路径将以精致的姿势转换成围绕鞋底的正确位置。相应的运动命令驱动分配器控制器完成胶粘任务，而不会发生碰撞。真实鞋底的测试结果验证了所提出解决方案的有效性。这项工作为制鞋业的智能制造做出了贡献，过去很少对此进行研究。