The Digital Twin Consortium® announced a liaison agreement with the National Aerospace Research & Technology Park (NARTP) to establish a technical collaboration to design, develop, and validate multi-agent digital twin systems for aviation.
This collaboration integrates AMD edge HPC, Rowan University DEHub, and DTC composability frameworks to field on-premise AI intelligence for air traffic management, cybersecurity, and autonomy.
The core engineering problem: aviation systems will require AI inference at the edge with deterministic latency and on-premise data sovereignty, yet must aggregate insights across enterprise and cloud infrastructure without compromising security.
The architecture solves this through local LLM inference on AMD Ryzen AI NPU/GPU hardware for edge-to-cloud agent coordination, utilizing DTC’s dual Digital Twin and Agent CPT composability frameworks, with DEHub’s Pythia HPC supercomputer providing the physics-informed AI backbone for component-level digital twin validation.
Application Domain Collaboration
The convergence of conventional IFR/VFR traffic, eVTOL urban air mobility, high-density UAS operations, and integrated autonomy demands an Air Traffic Management (ATM) architecture that current tooling cannot provide. Multi-agent digital twins may be useful to managing complex Traffic Flow Management (TFM) decision-making, optimizing airport resource availability, maintaining aviation cybersecurity, or any number of potential applications. The goals of the collaboration will be to determine the metrics and identify architectures that will unlock these capabilities for the aviation industry.
NARTP’s co-location with the FAA’s William J Hughes Technical Center for Advanced Aviation (WJHTC) and live aviation testbeds provides an ideal operational validation environment that anchors this work. The NARTP’s Strategic Innovation Center bridges FAA research programs with industry and academic product development, providing member organizations with structured access to testbed validation, technology integration, and data collection capabilities without having to independently maintain the full infrastructure stack.
The DTC will provide standards requirements and working groups’ expertise, with Aviation as a new focus area of the Mobility and Transportation Working Group, while DEHub’s Pythia will close the loop between academic research and operational deployment, with DTC Ambassador and Manufacturing Working Group Co-Chair Dr. Antonios Kontsos, who will coordinate testbed and research output connections to drive development in this effort.
Technical Stack
- AMD Ryzen AI — Hybrid NPU + integrated GPU edge compute. Lemonade Server enables local LLM deployment on-device; frontier data-center models with specialized edge models under NPU acceleration.
- XMPro’s MAGS (Multi-Agent Generative Systems) runs on this substrate as the operational agent orchestration layer: MAGS deploys, manages, and coordinates the AI agents defined under the AIA CPT framework, executing digital twin workflows in real time against live aviation data streams.
- The combination of Ryzen AI hardware, and XMPro MAGS delivers a complete on-premise agentic execution environment — from raw edge inference to coordinated multi-agent decision-making — with sensitive aviation data remaining on-premise and cloud scaling applied selectively and only when operationally authorized.
- Rowan University DEHub / Pythia — On-premise HPC with physics-informed AI, real-time sensor integration, and robotics testbed. Enables terabyte-scale digital twin validation of aviation components — turbine blades, airframe structures, advanced materials — under Dr. Antonios Kontsos, Co-chair of DTC’s Manufacturing Working Group.
- DTC Digital Twin Capabilities Framework + AIA CPT — DTC is concurrently advancing two complementary frameworks: the Digital Twin Capabilities Framework, which defines twin fidelity levels, data integration depth, autonomy tiers, and lifecycle management for operational digital twins; and the AI Agent Capabilities Periodic Table (AIA CPT), led by Pieter Van Schalkwyk (CEO, Co-founder of XMPro) and Sean Whiteley (CEO, Founder of AXOMEM), which catalogs and classifies the agent capabilities — perception, reasoning, planning, actuation, and coordination — that elevate a digital twin from a passive model to an active decision-making system.
- Together these frameworks define Multi-Agent Generative Systems (MAGS) architecture: digital twins as the operational environment, AI agents as the intelligence layer acting within and upon those twins. Aviation is the proving ground where both frameworks are validated simultaneously under real operational constraints.
Standards Requirements, Engagement, and Workforce Development
The collaboration will integrate aviation data with DTC’s composable digital twin architecture, and working groups will be used to address (for example) multi-agent orchestration, NPU-accelerated edge inference validation, zero-trust frameworks, and modular component standards spanning sensor systems, aircraft platforms, and ground infrastructure. Member organizations can engage through use-case development, testbed access, technology integration, participation in standards requirements, and jointly sponsored research with Rowan University. Workforce development will also operate under NARTP’s mandate, including graduate research programs, industry certifications, and hands-on multi-agent training.
