Vision–Language Model-enabled Cyberphysical Link: Emerging Possibilities for Production and Operations Management

Abstract

We introduce vision–language model (VLM)-enabled cyberphysical linking in production and operations management (POM). Drawing on four implemented cases that combine robotics, digital twins, and computer vision (CV), we propose how novel VLMs, together with more conventional CV and Internet-of-Things (IoT) technologies, enable new vision-based forms of cyberphysical linking across production and operations, during the beginning-of-life component manufacturing and end-of-life waste handling. We explore and explain the possibilities that emerge as the digital and physical become perceptually and operationally connected through VLMs. The research identifies four key design dimensions: persistent versus ad hoc digital counterparts, symbolic naming versus spatiotemporal identification, pre-planned versus interactive resource use, and passive versus self-controlling physical objects. VLM-based solutions are found to be potentially most impactful when integrated with persistent digital counterparts, interactive resources, and active physical objects. Further, a layered solution architecture that incorporates these design dimensions is proposed that provides a practical pathway for extending manufacturing execution systems. The findings advance cyberphysical operations and point to new directions for theory and practice in POM.