- Pre-trial of performance and usability based on digital human-device twins
- Expected to reduce the burden of clinical trials and cut development time and costs

Korean researchers have succeeded in developing a technology that can verify performance and usability of wearable robots1)Wearable robot: A robot worn directly on the body to assist muscle strength and movement or enhance physical function. during the development process without requiring a person to physically wear the device. The technology developed this time is a wearable robot evaluation technology based on a “digital human2)Digital human: A virtual human model that faithfully reproduces biological characteristics such as human body structure, movement, and responses in a virtual environment.-device twin3)Device twin: A digital replica of a real device (robot, equipment, or product) faithfully reproduced in a virtual environment.,” and it is expected to dramatically transform the paradigm of wearable robot development in the future.

Electronics and Telecommunications Research Institute (ETRI) announced that it has developed “digital human-device twin-based integrated evaluation technology for wearable robots” that can pre-verify the performance and usability of wearable robots even without actual users wearing them.

Conventional wearable robot development required repeated wear tests with actual users following prototype fabrication, resulting in significant time and cost burdens. The process of creating a device, having users wear it, and conducting tests had to be repeated multiple times, with any issues necessitating redesign and additional testing.

The validity of the technology developed this time was verified through joint experiments with the Glocal Clinical Trial Center at Pusan National University Hospital. Reliable evaluation results were derived by comparing and analyzing the results of clinical evaluations conducted with actual patients wearing wearable robots to perform muscle strengthening, rehabilitation therapy, and five types of basic functional tests with digital twin-based simulation results.

A key feature of this technology is that it can verify the performance of wearable devices and user experience (UX) at the design stage, before actual wearing, by precisely simulating in a virtual environment a diverse range of users who need neurological and musculoskeletal assistance. As a result, it is expected to overcome the limitations of conventional development methods that have inevitably relied on recruiting large numbers of subjects and conducting repeated wear tests.

ETRI researchers built an integrated evaluation framework for wearable robots using digital twin technology that links a physics-based neuro-musculoskeletal digital human twin with a wearable device twin.

This enables the design and performance of wearable robots to be comprehensively verified through software by combining a digital human, representing an actual person, with a device twin, representing the actual device, in a virtual environment.

The research team secured four core technologies through this research. First, there was neuro-musculoskeletal digital human twin generation technology4)Digital human twin generation technology: A technology that quantitatively models the bodily, neural, and musculoskeletal characteristics of a person to generate an individualized virtual human.. By quantitatively modeling the physical and cognitive characteristics of users who need neurological and musculoskeletal assistance, it can create personalized digital human twins for each individual. The team used this approach to create virtual users that precisely reflect the characteristics of diverse clinical target groups.

Second, there was physics-based device twin generation technology. By building an all-in-one software framework capable of generating digital twins that reflect the dynamics and statics structure, control algorithms, and sensor characteristics of wearable devices, scalability to diverse forms of wearable devices was secured.

Third, there was digital human-device linked simulation technology. By precisely simulating the interaction between the digital human and the device twin in a virtual environment, wearability, usability, and interactivity could be quantitatively evaluated. This can significantly reduce the number and cost of clinical trials while deriving user-optimized designs and control algorithms.

Fourth, an integrated system for evaluating wearable robot performance and usability was secured. The team established an evaluation framework that quantifies the performance and usability of wearable devices and can immediately reflect them in the design by integrating simulation results. In particular, this approach enables user experience (UX) metrics such as wearability, usability, and interactivity that had previously been assessed only qualitatively, to be verified against objective indicators.

ETRI announced that it has secured a validated correlation coefficient of 0.65)A correlation coefficient of 0.6: A figure indicating a statistically significant similarity between the actual experimental results and the digital twin-based evaluation results. The closer it is to 0, the less relationship there is; the closer it is to 1, the more similar it is. 0.6 indicates a moderate-to-high level of similarity. or higher for the integrated assessment system, demonstrating a level of reliability comparable to that of wearable robot evaluations conducted with actual patients.

Kim Woojin, principal researcher at ETRI’s AI Robot User Experience Research Section, explained, “Previously, wear tests with actual users were essential for verifying the performance of wearable devices. Using the digital twin-based technology developed this time, diverse user characteristics can be virtually combined and verified, making it possible to derive optimal device specifications and control algorithms in advance with only minimal clinical trials.”

Yoon Daesub, director of the AI Robot UX Research Section, said, “We plan to expand the application of the digital twin core technology secured this time to robot UX fields in which user experience is important, such as rehabilitation robots, walking assist devices, and industrial wearable robots.”

Professor Park Jonghwan of the Department of Convergence Medicine at Pusan National University Hospital said, “The developed robot evaluation and verification technology is expected to draw attention as essential software technology not only for wearable robots but also for the development of various other robots in the future. This technological development is expected to further invigorate related research.”

The research team plans to transfer this technology to wearable robot manufacturers and specialized robot manufacturing companies, pursue commercialization through follow-up R&D projects, and further enhance the completeness of the technology.

The research was conducted as part of the core source technology development project supported by the Ministry of Science and ICT and the Institute of Information & Communications Technology Planning & Evaluation (IITP).

Kim Woojin, Principal Researcher
AI Robot UX Research Section
(+82-42-860-0622, wjinkim@etri.re.kr)

- Overcoming GPU memory capacity limitations...disaggregated memory sharing across servers and accelerators to innovate performance
- Develops OmniXtend...validating technological leadership at RISC-V events in Europe and North America
- Optimizes AI and big data infrastructure using Standard Ethernet, without proprietary interconnects or dedicated hardware

Korean researchers have successfully developed a core technology that can fundamentally resolve “memory shortages,” a chronic bottleneck in large-scale artificial intelligence (AI) training. This technology is a next-generation memory expansion technology1)Memory expansion technology: A technology that enables memory distributed across multiple devices to be used as if it were a single large memory. Its goal is to flexibly expand memory capacity and utilization without degrading overall system performance. based on Ethernet2)Ethernet: A standard communication technology that connects computers, servers, and network devices to enable data transmission. It is the most widely adopted wired networking technology in data centers and enterprise networks., which is expected to drive infrastructural innovation across the entire AI and big data industries in the future.

Electronics and Telecommunications Research Institute (ETRI) announced that it has developed “OmniXtend,” a new memory technology that overcomes GPU memory capacity limits and data movement overheads, which are regarded as the biggest problems in large-scale AI training.

Recently, as demand for large-scale AI models and high-performance computing (HPC)3)High-performance computing (HPC): A computing paradigm that leverages multiple processors and accelerators in parallel to rapidly process large-scale computations. It is widely used for scientific research, AI training, large-scale simulations, and more. has surged, the volume of data to be processed is growing exponentially. However, no matter how much GPU4)GPU (Graphics Processing Unit): A processor specialized for massively parallel computation. It serves as a primitive computing engine in AI training and high-performance computing workloads. performance improves, the problem of the “Memory Wall5)Memory wall: A bottleneck phenomenon in which memory access speed fails to keep up with the rate of processor performance improvement, resulting in overall system performance degradation.”—where computational efficiency drops sharply due to memory capacity—has remained an unresolved challenge.

OmniXtend, developed by ETRI, addresses this by leveraging standard Ethernet as a memory interconnect fabric, enabling memory sharing across servers and accelerators, effectively treating distributed resources as a single, massive “Memory Pool”.

In other words, memory resources that were traditionally tightly coupled and locally constrained are now disaggregated and exposed over the network, allowing dynamic and scalable allocation of memory capacity for AI workloads.

As AI models continue to scale in size and the amount of memory they require keeps increasing, a “scalable architecture that shares and utilizes memory in the form of a pool” is gaining attention as a key technology for next-generation AI infrastructure.

ETRI’s OmniXtend demonstrated this scalable shared-memory architecture over Ethernet, achieving performance, scalability, and cost-efficiency of hyperscale AI training at the same time.

First, by minimizing data movement latency, it improved AI training speed. In addition, memory capacity can be expanded without replacing servers, leading to reduced data center deployment and operational costs.

In particular, conventional architectures based on high-speed serial interfaces such as PCIe had limitations in inter-device connectivity distance and system scalability. In contrast, OmniXtend leverages conventional Ethernet switches to aggregate multiple physically distributed devices into a unified memory pool, making it well-suited for highly scalable, large-scale AI system environments.

ETRI researchers developed key enabling technologies including ▲a Field-Programmable Gate Array (FPGA)6)Field-programmable gate array (FPGA): A semiconductor device whose hardware configuration can be changed by the user in the field. It is commonly used in systems requiring high-speed processing and flexible function implementation.-based memory expansion node and ▲an Ethernet-based memory transfer engine7)Ethernet-based memory transfer engine: A core hardware and software component that enables remote memory to be accessed and transferred over an Ethernet network with semantics similar to local memory access., and verified the stable operation of the system.

In an actual demonstration, they successfully showed multiple devices in an Ethernet environment forming a shared memory pool and accessing each other’s memory in real time.

Furthermore, through a computational workload test8)Computational load test: A test methodology that artificially imposes heavy computational workloads on a system to verify its performance and stability. It is used to evaluate processing capability in actual settings. using a large language model (LLM)9)Large language model (LLM): An AI model capable of understanding and generating natural language through training on vast amounts of data. It requires high computational performance and large-capacity memory., they confirmed that the OmniXtend architecture contributes to performance improvement even in actual AI training environments. Experimental results showed that in environments with insufficient memory capacity, LLM inference performance degraded significantly, whereas when memory was expanded using Ethernet, performance recovered by more than twofold. This demonstrates that it can maintain processing performance at a level similar to that of a conventional environment with sufficient memory.

ETRI drew significant attention by unveiling the technology at “RISC-V Summit10)RISC-V Summit: An international conference for sharing technologies and standards in the RISC-V ecosystem, an open instruction set architecture. It is attended by global semiconductor/system companies and research institutions. Europe 2025,” held in Paris, France, in May 2025, and “RISC-V Summit North America 2025,” held in Santa Clara, the United States.

In addition, ETRI is leading the Interconnect Working Group11)Interconnect Working Group: A specialized consortium that discusses and develops standards for interconnection technologies between systems and semiconductors, including high-speed data transfer and scalable architectures. under the CHIPS Alliance12)CHIPS Alliance: A semiconductor technology consortium under the Linux Foundation that promotes open-source chip design and the adoption of open interconnect standards. of the Linux Foundation, contributing to the establishment and global dissemination of open-source standards for AI networking and memory expansion.

ETRI plans to promote commercialization of this technology in the future by transferring this technology, particularly to data center hardware and software companies. This technology is expected to be applied to AI training and inference servers, memory expansion devices, and network switches, thereby generating tangible industrial impact in the next-generation AI infrastructure market.

Furthermore, ETRI plans to pursue follow-up research to expand this technology into a large-capacity memory interconnect for high-reliability embedded systems13)High-reliability embedded system: A dedicated computing system designed for environments requiring high levels of safety and reliability, such as automotive, ships, and aerospace systems. These systems are engineered for fault tolerance and long-term stable operation. such as automotives and ships, and to advance a shared-memory architecture across heterogeneous accelerators such as NPUs, GPUs, and CPUs.

Kim Kang Ho, Assistant Vice President of the Future Computing Research Division at ETRI, said, “We plan to significantly expand research on memory interconnect14)Memory interconnect: A communication architecture that enables data transfer among processors, accelerators, and memory. It is a critical factor in determining system performance and scalability through high-bandwidth, low-latency data movement. technologies centered on neural processing units (NPUs)15)Neural processing unit (NPU): A specialized processor designed for artificial neural network computations. It improves AI training and inference performance by efficiently handling massive parallel computations. and accelerators through new project initiatives,” adding, “We will continue to advance the technology and strengthen international collaboration to ensure its adoption in next-generation systems of global AI and semiconductor companies.”

This research was conducted as part of the “Research on Memory-Centric Next-Generation Computing System Architecture” project under the SW Computing Industry Original Technology Development Program, supported by the Ministry of Science and ICT and the Institute of Information and Communications Technology Planning and Evaluation (IITP).

Seung-Jun Cha, Principal Researcher
Future Computing Research Division
(+82-42-860-4826, seungjunn@etri.re.kr)

- Wins the most prestigious global cloud-native open source award
- Recognized for contributions to the cloud-native ecosystem through sustained international open-source engagement

A Korean researcher has become the first Korean to receive one of the most prestigious awards in the global open-source ecosystem, presented by an open-source software foundation under the Linux Foundation.

Electronics and Telecommunications Research Institute (ETRI) announced that Dr. Seokho Son, Principal Researcher (Special Fellow) at its Artificial Intelligence Computing Research Laboratory, has received the CNCF Community Award 20251)CNCF Community Awards: Awards presented to individuals and organizations that have made significant contributions to the cloud-native technology ecosystem. from the Cloud Native Computing Foundation (CNCF)2)CNCF (Cloud Native Computing Foundation): An open-source software foundation under the Linux Foundation that promotes the development and adoption of cloud-native technologies., a global open-source foundation.

The award ceremony took place at KubeCon North America 20253)KubeCon North America 2025: One of the largest global events in the cloud-native open-source ecosystem., one of the world’s largest cloud-native conferences, held in Atlanta, USA in November 2025.

The CNCF Community Awards are among the most prestigious honors in the global open-source ecosystem, presented to standout contributors from approximately 270,000 cloud-native open-source community members worldwide who have made exceptional contributions across the ecosystem, including technology development, documentation, and community engagement. Dr. Seokho Son became the first Korean to receive this award.

Dr. Son was named the winner of the Lorem Ipsum Award, in recognition of his sustained and impactful open-source contributions to numerous CNCF cloud-native projects, including Kubernetes, and his role in strengthening the long-term growth and sustainability of the community.

Beginning in 2025, CNCF elevated the former Top Documentarian Award to a higher level of recognition and renamed it the Lorem Ipsum Award.

This award further recognizes the open-source contributions of ETRI and Korea across the cloud-native ecosystem, following Dr. Son’s receipt of the Kubernetes Contributor Award4)Kubernetes Contributor Awards: Awards presented to a select few from more than 30,000 Kubernetes contributors in recognition of outstanding contributions to the ecosystem. in 2022.

In particular, by attending KubeCon North America 2025, Dr. Son engaged with more than 10,000 cloud and open-source experts from around the world. He also actively built relationships with the CNCF CTO, the Technical Oversight Committee (TOC)5)CNCF Technical Oversight Committee (TOC): The highest technical decision-making body within CNCF, responsible for setting the technical vision and direction of the open-source project ecosystem and making key technical decisions such as approving new projects and coordinating alignment among projects., key CNCF staff, and Kubernetes maintainers while serving as a CNCF Global Ambassador. Through these efforts, he is widely credited with significantly enhancing the international profile and standing of both ETRI and Korea’s open-source technologies.

Dr. Son said, “Cloud-native computing has established itself as a core software infrastructure for enabling large-scale services and AI workloads. Behind the widespread adoption of these technologies lies not only technical maturity but also a global open-source community and ecosystem built over many years. R&D can no longer move forward without engaging with open source. Korea also needs to recognize open source as an essential activity for learning rapidly evolving technologies and fostering collaboration. I hope this award will inspire Korea’s outstanding experts to take greater interest in the global open-source ecosystem and begin participating in it.”

Building on this award, ETRI plans to view open source not merely as a tool, but as a core strategic asset that supports national competitiveness and technological sovereignty, and to further strengthen related R&D and global contributions.

In particular, ETRI plans to prioritize developing open-source-based foundational technologies and advancing global standards and ecosystem leadership in key national strategic areas, including cloud-native, AI computing, and next-generation networks, while continuing to contribute research outcomes to international open-source communities and strengthening its leadership.

In addition, aligned with the Korean government’s national digital and AI strategy, ETRI plans to establish a responsible open-source contribution model as a national research institute and build an R&D framework that supports the global adoption of Korean technologies and strengthens technological sovereignty through collaboration with industry and academia.

ETRI President Bang Seung Chan said, “This CNCF Community Award recognition is a symbolic achievement that goes beyond an individual’s accomplishment, demonstrating that a Korean national research institute has established itself as a sustained and responsible contributor to the global open-source ecosystem. ETRI will continue to expand R&D and international collaboration to drive growth in the fields of cloud native, AI computing, and open-source-based core software.”

ETRI announced that it will continue to play a pivotal role in establishing Korea’s ICT and AI technologies as a core pillar of the global digital ecosystem, through close collaboration with world-leading open-source foundations and global communities.

This achievement stems from open-source activities conducted through the SW Computing Industry Original Technology Development Program, supported by the Ministry of Science and ICT and the Institute of Information and Communications Technology Planning and Evaluation (IITP). The source code developed under this project aims for official registration as a CNCF open-source project by 2028, supporting the global adoption of Korean software.

Seokho Son, Principal Researcher
Cloud Computing SW Research Section
(+82-42-860-5286, shsonkorea@etri.re.kr)