Enactic Begins Japan's First Practical Validation of NVIDIA GR00T N1.7 in Nursing Care

Enactic (HQ: Shinjuku, Tokyo; CEO: Yasutoyo Yamamoto), a leader in humanoid robot development, has commenced Japan's first practical validation project in the nursing care domain using 'NVIDIA Isaac GR00T N1.7,' the latest humanoid robot foundational model provided by NVIDIA. Through this collaboration, Enactic is advancing the implementation of humanoid robots for diverse tasks in living spaces, including nursing environments.

This initiative maximizes the use of Enactic's open-source dual-arm robot 'OpenArm 2.0,' globally recognized as an evaluation platform for robot foundational models, and incorporates insights accumulated through partnerships with over 90 care providers. By leading the optimization of these models for next-generation architectures, the project aims to improve autonomy in complex, non-routine tasks, thereby accelerating the societal implementation of Physical AI to tackle the acute nursing labor shortage.

Background

Japan faces one of the world's most acute aging populations, leading to a critical gap between the supply and demand for nursing labor. Enactic is dedicated to the social implementation of humanoid robot technology to solve these issues. Nursing environments are uniquely challenging for AI, as environments, materials, and human movements vary significantly by facility.

To bridge the gap between these societal challenges and the difficulty of real-world implementation, both companies decided to initiate this effort.

Project Overview

Enactic, in collaboration with NVIDIA, will develop the project across three main axes:

### 1. Practical Validation and Optimization of Foundational Models in Nursing Tasks Representative nursing tasks will be evaluated to systematically measure performance. The project will identify challenges related to success rates, generalization, and safety, optimizing the models through additional training and data augmentation.

### 2. Practical Testing on Company Platforms Optimized models will be integrated into Enactic’s humanoid platforms for functional testing. By bridging the gap between simulation and real-world performance, the project will refine the hardware-model combination for real-world reliability.

### 3. Practical Validation in Nursing Facility Environments Testing will occur in cooperating facilities that mirror real-life living spaces. Starting with limited, safe tasks, the team will cautiously verify platform behavior under human-shared conditions, incorporating feedback from on-site staff and experts.

Utilizing NVIDIA's Full-Stack Physical AI Technology

This project actively utilizes various NVIDIA technologies for AI humanoid development:

- Foundational Models: NVIDIA Isaac GR00T N-series - Simulation Frameworks: NVIDIA Isaac Sim and Isaac Lab - World Models: NVIDIA Cosmos, DreamZero - Model Training: NVIDIA DGX H100 - Edge Inference: NVIDIA Jetson Thor

The goal is to build a high-performance system covering data augmentation based on real-world collection, large-scale training, and low-latency edge inference.

NVIDIA GR00T is an open foundational model family aiming for general-purpose intelligence in humanoid robots. The commercially available Isaac GR00T N1.7 features large-scale pre-training using first-person human perspective video, action space design for cross-embodiment transfer, and an enhanced vision-language backbone, allowing humanoids to adapt to diverse tasks with minimal data.

Comment from Yasutoyo Yamamoto, CEO of Enactic

'Since our founding, we have focused on realizing humanoids that are genuinely useful in the spaces where people live. Collaborating with NVIDIA accelerates this journey significantly. We are committed to delivering our products to nursing care professionals and users as quickly as possible.'