**1. Introduction**

Robotic arms are usually used in automated production lines to perform repetitive or hazardous operations to ensure high manufacturing quality. For instance, a wafer handling robot is usually used in the automated production line of semiconductor components for sending or taking wafers between cassettes at etching, deposition, and photolithography stations with demanding accuracy, repeatability, and reliability [1]. However, potential induced stress or deformation during continuous robotic arm operation may cause the end effector position and orientation to deviate from the position specified, which would result in the collapse of the processing unit with the

**Figure 1.** *Wafer damage caused by positioning deviation of robot end effectors during manufacturing.*

wafer scrapped or crushed, as shown in **Figure 1**, thus incurring significant production losses. To avoid such an undesired scenario, this study proposes a robust and precise monitoring method for performing the 6-degree-of-freedom (6DOF) localization of robotic arms end effectors.

Recent studies have investigated the problems related to position uncertainties of robotic arms end effectors and analyzed the life cycle of the robotic arm to improve the reliability of the wafer handling robot [2, 3]. However, in general, planar translation of end effectors has been measured using only 2D imaging; with depth information missing, 6DOF localization of end effectors cannot be realized.

To obtain accurate variations in both position and orientation of the robotic arms end effector, this study developed a 3D machine vision probe with point clouds measured for achieving 6DOF localization of the end effector. Point cloud registration is crucial to determining precise object 6DOF transformation between different locations in the 3D space and involves the coarse alignment stage and the refined model alignment stage. During coarse alignment, the object point cloud aligns only approximately with the target (reference) point cloud. Methods used include spin image [4], point signature [5], and regional surface area descriptor [6], with different levels of accuracy, efficiency, and robustness achieved. This initial matching between the object and the target model is then followed by fine model alignment for more precise matching of the two. The most widely adopted algorithm is the iterative closest point (ICP) [7], which uses singular value decomposition (SVD) to find the set of closest point clouds between the object and the target model. It refines the deviation between the corresponding points until the least-squares error is minimized.

This chapter presents a novel 6DOF detection method that uses a developed structured-light 3D scanner [8] to obtain the 3D information of the robotic arms end effector and then performs point cloud alignment to detect the 6DOF variations of the robotic arms end effector. The process of the proposed method is illustrated in **Figure 2** and described in detail in Section 2.
