Schaeffler, Wandelbots form technology partnership on NOVA-based Physical AI

Schaeffler and Wandelbots announced a technology partnership on Oct. 22, 2026, to develop a Physical AI platform for manufacturing based on Wandelbots NOVA. The plan is to develop and validate applications for industrial and humanoid robots and deploy them at Schaeffler factories worldwide. The companies aim to reduce the life-cycle costs of automation solutions by up to 40% in both new and existing production environments.

The companies aim to develop and digitally validate automation solutions using equipment from different manufacturers and categories on NOVA, then deploy them at Schaeffler factories worldwide. The target of cutting life-cycle costs by up to 40% is a goal set by the companies. It should be assessed separately from the reported development-time reduction in an earlier packaging cell and from results after deployment across all factories.

The focus is not on installing robot hardware but on a production platform that handles different equipment through a common software layer and connects virtual validation with execution in factories. NOVA separates application logic from hardware and covers everything from industrial robots to mobile systems and humanoids. This appears designed to make it easier to reuse applications developed for individual machines when transferring them to different factories. The up-to-40% cost reduction, however, is a goal of the companies, not a reported result. The partnership does not start entirely from scratch. In an earlier robot packaging cell, the companies said they cut development time by about two weeks per application. The new element is the plan to extend that single example across Schaeffler’s production network, which has thousands of robot systems, and connect it to a reusable skills library. If a system for transferring logic validated in a virtual environment to physical equipment works across multiple sites, a key question will be how much factory-specific redevelopment and rework can be standardized. This design also underscores the need to distinguish between evaluating models or movements and providing an implementation platform that can be used repeatedly in factories. WorldOracle publishes a benchmark comparing world-model predictions with real-world robot actions examined the comparison of model predictions with real-world robot actions. By contrast, this partnership aims to validate processes using digital twins and transfer them to production equipment through the same software layer. Whether validation results can be reproduced in real-world operations will depend not only on evaluation accuracy but also on factory process conditions and quality checks. The industry implication is an effort to separate equipment selection from application development by enabling development that does not depend on control environments specific to each robot manufacturer. This is relevant to deployment across factory networks that include different equipment, but the announcement does not specify which models or processes will be supported. Questions to watch include the pace of expansion to factories worldwide, the conditions used to measure the up-to-40% target and actual results, the scope of the skills library, and alignment between virtual validation and the quality and safety of physical equipment.

NOVA separates application logic from hardware and brings AI models, digital twins, machines and manufacturing data into one software environment. It uses NVIDIA Omniverse and NVIDIA Isaac Sim to simulate, validate and optimize production processes virtually, then transfers them to actual production equipment through the same software layer. Supported equipment includes industrial robots, mobile systems and humanoid robots. Schaeffler plans to develop reusable robot skills and build a skills library covering simulation and training through factory operations. In an earlier robot packaging cell, the companies said development time was reduced by about two weeks per application. Schaeffler’s production network has thousands of robot systems.