Terry KS DataMesh has unveiled DataMesh Robotics, a new embodied AI data product built on executable industrial digital twins to support real-world robotics training and validation. The solution enables robot developers to train AI systems in dynamic, rule-driven industrial environments that mirror real operational complexity.
SINGAPORE, 16 JANUARY 2026 – DataMesh, a digital twin and spatial intelligence technology provider, has launched DataMesh Robotics, an embodied AI data product designed to support industrial robotics development and deployment. Built on an Executable Industrial Digital Twin, the solution enables robot OEMs and robotics application teams to train, validate, and evaluate embodied AI systems using dynamic industrial environments, industrial-grade synthetic data, and configurable task objectives.
As embodied AI transitions from controlled research settings into real industrial operations, robotics teams face increasing challenges bridging the gap between static simulations and real-world industrial complexity. Industrial environments are governed by evolving processes, safety constraints, events, and business rules, requiring training systems that can respond dynamically rather than remain visually static.
DataMesh Robotics addresses this challenge by enabling AI training within an executable industrial world where processes evolve over time, events are triggered, and task objectives are explicitly defined and measurable. The solution is powered by DataMesh’s Executable Digital Twin technology on the DataMesh FactVerse platform.
Unlike conventional digital twins that focus on 3D visualisation and monitoring, DataMesh’s environment allows industrial objects to interact, operational processes to unfold, and logic-driven behaviors to execute during runtime. This enables the generation of training data and task feedback that more accurately reflects real operating conditions, supporting multi-step industrial tasks with safety constraints and partial observability.
The platform provides end-to-end capabilities including industrial scene modelling, physical simulation, and scalable synthetic data generation. It supports multimodal data creation with automated ground-truth labelling for robotics perception, navigation, and manipulation, while also capturing non-visual operational data such as process states and system conditions.
A key feature of DataMesh Robotics is its configuration-driven approach to defining task objectives and reward signals. By formalising goals, success criteria, and constraints, the platform helps robotics teams address one of the most complex challenges in embodied AI training, particularly in environments with strict tolerances and safety requirements.
Designed to integrate with mainstream robotics ecosystems, DataMesh Robotics supports asset and data export to platforms such as NVIDIA Isaac Sim and Omniverse, and can be deployed on-premises, in private cloud, or hybrid environments with enterprise-grade governance.
DataMesh said the solution is targeted primarily at industrial robotics use cases including factory and warehouse navigation, workstation operations, inspection and maintenance, and operations in hazardous or restricted environments. The platform has completed prototype validation and is currently running pilot programmes with enterprise partners, with plans to expand its industrial asset library, task templates, and ecosystem integrations.