Humanoid Robotics Technology
ugo Inc. has announced the domestically produced semi-humanoid robot “ugo Nova” on September 16, 2026. ugo Nova is a domestically produced robot, with design and manufacturing carried out entirely in Japan, and mass production is scheduled to begin in 2027. This robot will be proposed by the AI Robot Association (AIRoA) as part of a project commissioned by NEDO (New Energy and Industrial Technology Development Organization) to develop a domestically produced general-purpose robot. It can convert human work directly into high-quality learning data, which can then be used in the development of robot base models for physical AI.
ugo Nova is a wheeled, dual-arm semi-humanoid robot developed to handle diverse tasks in human work environments. Combining a flexible dual-arm and upper-body mechanism that allows access to the same workspace as humans with the stable autonomous movement technology cultivated in the ugo series, it enables integrated movement, recognition, and arm operation in manufacturing, factories, warehouses, and other workplaces.
It can be used not only as a platform for physical AI research and development, data collection, and model verification, but is also designed with actual business implementation and operation in mind, making it a robot that connects research and development to social implementation in a single unit.
Cameras are standardly mounted on the head and hands to acquire visual data necessary for learning physical AI, allowing for immediate data collection after setup. Arm control supports a high control cycle of 100Hz, enabling high-speed and smooth motion recording with high precision. Furthermore, it supports teleoperation using bilateral controllers and VR controllers, allowing the operator’s movements to be intuitively and faithfully reflected in the robot, while recording the series of operations as training data. collection for physical AI can begin immediately after setup without the need for additional dedicated equipment.
Dual-arm maximum payload of 10kg, dual-arm multi-joint mechanism for working in a human-like posture. Equipped with dual arms that offer a degree of freedom similar to human arms, and a multi-joint mechanism in the waist that allows for rotation and flexion of the upper body. This allows for work at a wide range of heights and postures, from handling items near the floor to accessing overhead shelves. payload capacity is a maximum of 4kg per arm, and a maximum of 10kg with both arms, making it applicable to a variety of tasks such as transport, picking, and handling in manufacturing sites and warehouses.
Omnidirectional movement mechanism that balances stability and long operating time. The movement mechanism employs a Mecanum system, achieving high mobility that allows for on-the-spot direction changes as well as lateral movement without rotation. It can move flexibly even in narrow passages and environments with densely packed equipment, and maintains a stable posture even when applying force with both arms. Furthermore, it supports continuous operation for up to 6 hours, enabling long-term operation in manufacturing sites and warehouses.
Edge Inference Powered by “NVIDIA Jetson Thor™”
The robot is equipped with the latest robotics AI computer, “NVIDIA Jetson Thor™”. By utilizing its high computing power, AI models can be inferred and trained in real time on the robot itself without relying on the cloud. This enables the implementation of physical AI that processes sensor information such as cameras on the spot and performs recognition, decision-making, and action with low latency.
Demonstration of Work and Data Collection:
Operators use bilateral controllers or VR devices to operate ugo Nova and demonstrate the movements they want to train. Multimodal data such as video during operation, joint information, torque, end effector position, and operation commands are collected in time synchronization.
AI Model Training:
The collected data is quality-controlled and compiled into a dataset, which is used for additional training of the robot base model.
Application to Actual Machine and Verification:
The trained AI model is applied to ugo Nova, and its performance is evaluated by performing inference and operation in a real environment.
A key feature of ugo Nova is that it can handle not only “collecting data” but also “running the model” with a single device. It can be used not only for research purposes but also to perform tasks based on AI inference in actual work environments.
Future Outlook
With ugo Nova, ugo will accumulate and utilize data collected in the field, and through imitation learning and the development of robot base models, will expand the range of tasks that robots can handle and promote gradual autonomy.
Furthermore, beyond simply providing hardware, we will build a system that comprehensively supports everything from research and development of physical AI to on-site implementation by integrated services including data collection, Starting and utilization consulting, and AI model development, evaluation, and implementation support.
Source: ugo Inc.
