RobotBank Steps Up Development of Its Own Physical AI Robots, Starting With a Quadruped

RobotBank announced on November 9, 2026, that it will continue developing robots using physical AI under its own brand. Its first project is a quadruped intended for use in manufacturing, logistics and infrastructure. The company has already worked on developing robot products, software and systems, and describes this initiative not as its first product development effort but as an expansion of the focus areas for its own product development.

RobotBank cited inspection support for factory equipment and social infrastructure as potential uses for the quadruped. In its trials, the company says it will focus not only on task success rates but also on how often human intervention is needed, safety and maintenance costs. For prospective users, the plan is relevant because it will assess whether the robot can be relied on for ongoing inspection work, not just whether it can move around.

The announcement centers not only on the selection of a quadruped model but also on RobotBank's stated intention to develop products under its own brand on an ongoing basis. The company has already been involved in developing products, software and systems, so this is not an entry into the business from scratch. What is new is its decision to make embodied AI robots using physical AI a priority and to determine product specifications through field trials. The first quadruped is intended for patrols, situation checks and inspection support, with the company considering its potential to handle steps and uneven terrain. At the same time, the announcement lists sensors, remote operation, maintainability, integration with existing systems, safety and deployment costs as elements of the overall design. This means evaluation will extend beyond the robot's mobility to whether it can be used continuously in facility operations. Including task success rates, intervention frequency and maintenance costs among the trial metrics is one concrete example of those criteria. Our report, [Panasonic Holdings and others jointly develop robot-operation AI “T-Rex,” with a 65% average success rate across 12 tasks](/en/news/panasonic-holdings-partners-develop-t-rex-robot-manipulation-ai-65-ave-275dc6), covered success rates and training data for manipulation tasks involving touch. The current plan concerns uses involving movement, such as patrol inspections, so the two cannot be compared using the same performance metrics. However, both manipulation and mobility raise the question of how to measure success in real-world settings and the need for human intervention. RobotBank's evaluation criteria can be seen as an approach to setting metrics for assessing field deployment from the product-development stage. The development plan combines prototyping and trials at RobotBank Lab with potential future integration with RobotBank Brain, which is under development. However, the company has not said when software integration might be achieved or how trial results will be reflected in product specifications. It also cited mobile manipulators, industrial work robots and robot hands as possible projects after the quadruped, but has not committed to commercialization or mass production. Key points to watch are the results achieved on each trial metric, the robot's specific specifications, and the timing of commercialization, mass production and sales.

Potential uses for the quadruped include patrol inspections of factories and manufacturing equipment, situation checks at logistics and large facilities, and inspection support for social infrastructure. The company plans to use the robot's potential to handle steps and uneven terrain and determine its specifications by testing conditions at individual sites. It will consider the robot's design, locomotion control, sensor configuration, remote operation and maintainability. Functions such as environmental recognition and autonomous task execution will be implemented in stages as technical validation progresses. At its research and development center, RobotBank Lab, the company will identify challenges, validate technology, prototype, conduct trials and consider commercialization. It also plans to explore future integration with its software platform, RobotBank Brain, which is under development. In trials, it will focus on task success rates, operating time, frequency of human intervention, safety and maintenance costs as evaluation metrics. The company says it will provide details on product specifications and the timing of commercialization, mass production and sales in line with development and testing progress.