GOALI: Computer-Aided Design (CAD) Guided Sensor Planning for Surface Inspection in Manufacturing

本赠款机会学术联络与工业（GOALI）研究的目标是开发一个通用的计算机辅助设计（CAD）为基础的方法，在制造业中满足约束的机器人运动规划。 特别是，研究将集中在基于CAD的检测规划，即，检测传感器的自动规划，或由机器人设备携带的相机，根据给定的CAD模型的检测部件。首先，将开发一种基于CAD的传感器规划方法，该方法利用来自CAD模型和传感器模型的零件几何信息来生成满足约束的传感器配置。 其次，为了提高检测系统的效率和运动学性能，提出了两个优化问题：一个是寻找最小视点集，另一个是观察视点的最佳运动学。 通过使用离散化方案，前者被渲染为一个集划分问题，后者作为一个加权集覆盖问题。 将开发新的算法来解决这些问题。 最后，研究了零件检测中的机器人运动规划问题。 这被表述为集群旅行推销员问题（CTSP），并且将开发一种新的分层算法来获得次优解。 此外，理论结果将实施和实验测试。这项工作的潜在贡献包括一个新的方法，用于检测大面积的零件表面的最佳传感器规划。 该方法可以方便地提取零件表面的全局几何信息。 在这项研究中开发的方法也可以受益于许多其他基于CAD的规划问题，如喷漆和CNC零件编程。这项研究将有助于制造自动化领域的技术和人力资源的发展。

The objective of this Grant Opportunities for Academic Liaison with Industry (GOALI) research is to develop a general Computer-Aided Design (CAD)-based method for constraint-satisfying robot motion planning in manufacturing. In particular, the research will focus on CAD-based inspection planning, namely, the automatic planning of the inspection sensor, or camera carried by a robotic device, based on given CAD model of the inspected parts. First, a CAD-based sensor planning approach will be developed that utilizes the part geometric information from the CAD model and the sensor model to generate constraint-satisfying sensor configurations. Second, to improve the efficiency and the kinematics performance of the inspection system, two optimization problems are formulated: one is to find the minimum set of viewpoints, while the other is the optimal kinematics to observe the viewpoints. By using a discretization scheme, the former is rendered as a set-partitioning problem, and the later as a weighted set-covering problem. New algorithms will be developed to solve these problems. Finally, the robot motion planning problem in the part inspection will be investigated. This is formulated as a Clustered Traveling Salesman Problem (CTSP) and a new hierarchical algorithm will be developed to obtain suboptimal solutions. In addition, the theoretical results will be implemented and experimentally tested.The potential contributions of this effort include a new methodology for optimal sensor planning for inspecting large areas of part surfaces. It can be used to easily extract the global geometric information on the part surfaces. The methodologies developed in this research could also benefit many other CAD-based planning problems such as spray painting and CNC part programming. This research will contribute to the development of technologies and human resources in the area of manufacturing automation.