Industrial quality inspection plays a critical role in modern manufacturing by identifying defective products during production. While single-modality approaches using either 3D point clouds or 2D RGB images suffer from information incompleteness, multimodal anomaly detection offers promise through the complementary fusion of crossmodal data. However, existing methods face challenges in effectively integrating unimodal results and improving discriminative power. To address these limitations, we first reinterpret memory bank-based anomaly scores in single modalities as isotropic Euclidean distances in local feature spaces. Dynamically evolving from Euclidean metrics, we propose a novel Geometry-Guided Score Fusion (G2SF) framework that progressively learns an anisotropic local distance metric as a unified score for the fusion task. Through a geometric encoding operator, a novel Local Scale Prediction Network (LSPN) is proposed to predict direction-aware scaling factors that characterize first-order local feature distributions, thereby enhancing discrimination between normal and anomalous patterns. Additionally, we develop specialized loss functions and score aggregation strategy from geometric priors to ensure both metric generalization and efficacy. Comprehensive evaluations on the MVTec-3D AD and Eyecandies datasets demonstrate the state-of-the-art detection performance of our method, and detailed ablation analysis validates each component's contribution. Our code is available at https://github.com/ctaoaa/G2SF.

End-of-line product dimensional quality assurance is crucial in multistage assembly processes (MAPs). Active control strategies involve the deployment of controllable, programmable tooling machines (PTs) to adjust part positions for in-process dimensional error compensation. In MAPs, multiple PTs may be deployed within a single stage or across various stages. To enable the normal operation of deployed PTs within a specific stage, an associated supporting platform (SP) is necessary. Consequently, the lower PT-level and the upper platform-level define a two-level hierarchical structure in MAPs. This paper aims to propose an optimal placement strategy of PTs, focusing on the final dimensional quality of the product and the total cost related to the number of accommodated PTs and required SPs. Based on the stream-of-variation (SOV) model of MAPs, we develop a novel sparse learning framework along with the corresponding parameter estimation algorithm to achieve the optimal placement. The case study demonstrates the effectiveness of our proposed method for the optimal placement of PTs in reducing dimensional variation in MAPs. Note to Practitioners—This paper was motivated by the widespread use of programmable tooling machines (PTs) in multistage assembly processes (MAPs), such as controllable fixtures and computer numerical control (CNC) machines. These PTs automatically compensate for in-process dimensional errors and enhance the final assembly quality. Multiple PTs can be allocated within the same stage or across various stages, but an associated supporting platform (SP) is needed to install, maintain, and manage the PTs at a specific stage. Given the high costs of PTs and SPs, it is crucial to find an optimal placement of PTs that minimizes the number of PTs and required SPs while ensuring assembly products acceptable by industrial quality standards. To address this issue, we propose a novel methodology that formulates the problem into a sparse learning framework, which can be efficiently solved using our developed algorithm. This research provides valuable insights for practitioners working on the dimensional quality control of final products in MAPs and similar manufacturing processes. By implementing our proposed optimal placement strategy of PTs, practitioners can assure product quality while minimizing the associated total cost.

When a transportation service accommodates both people and goods, operators sometimes opt for vehicles that can be dynamically reconfigured for different demands. Motivated by air service in remote communities in Canada’s north, we define a pickup-and-delivery problem in which aircraft can add or remove seats during a multi-stop trip to accommodate varying demands. Given the demand for people and cargo as well as a seat inventory at each location, the problem consists in finding a tour that picks up and delivers all demand while potentially reconfiguring the vehicle capacity at each location by adding or removing seats. We develop a total of six models using three different approaches: constraint programming, mixed integer programming, and domain-independent dynamic programming. Our numerical experiments indicate that domain-independent dynamic programming is able to substantially outperform the other technologies on both solution quality and run-time on a set of randomly generated instances spanning the size of real problems in northern Canada.

Due to the high positioning accuracy and relatively low prices, SCARA robots are widely used in industrial fields. The objective of this paper is to propose a hand-eye self-calibration algorithm for SCARA robots which could consider both accuracy and computational cost. The previous global optimal hand-eye calibration algorithms based on branch-and-bound (BnB) optimization is limited by their expensive computational cost. The speed of these algorithms depends on the volume of the search space to a large extent, which is the main concern in this paper. Instead of searching over the 3-dimensional parameter space corresponding to the rotation component of hand-eye pose, a new 2-dimensional search space is defined by separating and coupling some calibration parameters by means of the special structure of SCARA robots, which have 4 degrees of freedom (DoFs) including three translation DoFs and only one rotation DoF. The simulation and real experiments show the similar accuracy but much faster speed of the proposed algorithm compared with previous optimal algorithms based on BnB.