"Industrial Visual Anomaly Detection in Robotics: Methods, Datasets, and Deployable System Architectures," by Thanh-Hai Tran and Dr. Xuan-Bach Le, was published at IEA/AIE 2026 in Kuala Lumpur, Malaysia, July 2026. As collaborative robots move into unstructured environments, there is growing demand for intelligent safety systems that can spot hazards without relying on large volumes of labelled data. Unlike static quality inspection, anomaly detection in robotics has to run in real time, cope with moving scenes, and catch open-set anomalies. The survey groups existing approaches into three families — reconstruction-based, embedding-based, and vision-language model methods — and reviews the benchmarks and robotics-specific datasets available. One finding stands out: strong benchmark scores do not carry over to deployed robots, where dynamic scenes, latency limits and semantic risk are real constraints. The authors argue for treating this as a safety-critical system component, one that demands edge efficiency, low latency and integration with middleware such as ROS2. Paper: https://lexuanbach.github.io/publication/IEA2026b.pdf Slides: https://lexuanbach.github.io/slides/IEA2026b_slides.pdf DOI: https://doi.org/10.1007/978-981-92-2888-1_9