At the Physical AI Summit held in Tokyo, the AI Robot Association (AIRoA) unveiled its development results. Prototypes of dual-arm mobile manipulators from 7 companies, including ugo, were exhibited, and ugo unveiled ugo Nova for the first time.
At the Physical AI Summit held in Tokyo on 16 September 2026, the AI Robot Association (AIRoA), a general incorporated association, unveiled prototypes from the domestic general-purpose robot development competition. The 7 exhibiting companies were ugo, THK, 不二越, Keigan, HatsuMuv, Enactic and EmplifAI, while Telexistence was excluded. Each company displayed a prototype dual-arm mobile manipulator designed for teleoperation, with an emphasis on acquiring real-world data and accumulating training data.
Among them, ugo announced its new semi-humanoid, “ugo Nova.” It has 22 degrees of freedom across the body, 7 degrees of freedom in each arm, a maximum payload of 4 kg per arm and a combined payload of 10 kg for both arms. Its features include a wheeled mobility mechanism and 6 hours of operating time. Nova is said to have been designed from the ground up in both hardware and software, rather than being an improvement on an existing model, and mass production is scheduled to begin in 2027. The design’s direct link between teleoperation and AI training data collection was also highlighted.
It was emphasized that the prototypes unveiled this time are not finished products, and that the program is designed to progressively improve and move toward mass production of units that meet the specified requirements and evaluation criteria. Mass-produced units are also expected to be used to collect real-world data for the robot foundation model being developed under a NEDO commissioned project, with the aim of establishing a domestic improvement and mass-production framework.
Each company is pursuing a policy of prioritizing teleoperation to collect and train data, emphasizing hardware design based on imitation learning. The information targeted for collection was said to include facial and neck movements as well as dialogue data, with a vision of linking this data to future AI model development, evaluation, field operation and continued learning.