Prox Industries has provided technical support to Shimizu Corporation for its research and development work on humanoid robots, helping build a platform for behavioral learning as part of its Physical AI efforts. The support was provided during fiscal year 2025.
The initiative is about a growing issue in construction, where lack of labor is making companies look for a more flexible kind of mechanized work. Usual construction robots tend to be made for particular jobs and they move through actions that are already pre-programmed is set in stone. But AI-based robots that can actually sense their environment and then decide could, in theory, manage several tasks through one single platform.
As part of the project, Prox Industries helped establish the research processes needed for humanoid robot behavioral learning. This included building a simulation environment using NVIDIA Isaac Sim, collecting motion data through teleoperation and expanding training data.
Also Read: Green AI in Japan: How Enterprises Are Reducing AI Energy Consumption Without Sacrificing Performance
The company also provided technical guidance on potential applications of robot platform models, including VLA, or Vision-Language-Action models.
The resulting platform is designed to support the efficient generation and accumulation of learning data while enabling ongoing verification for future robot behavioral learning. Specific research details and technical specifications will not be disclosed.
Prox Industries plans to continue supporting Physical AI research and implementation across construction, manufacturing, logistics and infrastructure


