Tokyo-based robotics maker ugo, Inc. introduced ugo Nova, a wheeled dual-arm semi-humanoid designed and manufactured entirely in Japan, with mass production scheduled to begin in 2027.
The robot pairs a flexible dual-arm and upper-body mechanism built to operate in the same spaces as people with the autonomous-mobility technology ugo has developed across its existing product line, aiming to combine movement, perception, and arm manipulation into one machine for manufacturing, factory, and warehouse work, according to ugo. The company is also entering ugo Nova in a domestic general-purpose robot development competition run by Japan's AI Robot Association as part of a project commissioned by NEDO, the country's national energy and industrial technology development organization.
According to ugo, the machine is built to double as a data-collection platform for physical AI: head- and hand-mounted cameras capture visual data, the arms run on a 100Hz control cycle for precise motion recording, and operators can drive it through a bilateral controller or VR headset, with every teleoperated motion logged as training data. Its arms handle up to 4kg each and 10kg combined, a multi-joint waist adds torso rotation and bending for reaching both floor-level items and overhead shelves, and a Mecanum-wheel base lets it move sideways without turning, sustaining up to six hours of continuous operation. An onboard NVIDIA Jetson Thor computer runs trained AI models locally, so the robot infers and acts without a cloud connection.
The company frames the intended workflow as a loop: operators demonstrate tasks, the resulting multimodal data trains ugo's robot foundation models, and the updated models are then tested back on the physical machine. Industrial design came from GK Kyoto, and the bilateral controller used for data capture was built by Asratec. CEO Ken Matsui described ugo Nova as a ground-up redesign built for the physical AI era rather than an update to the company's existing patrol, security, and guidance robots; ugo, founded in 2018, positions those products as automation for labor-short sites such as office buildings, plants, and train stations.
