Prox Industries and Shimizu Corporation Collaborate on AI Robot Development

Prox Industries Inc., headquartered in Bunkyo, Tokyo, and led by CEO Yuji Kamei, has officially announced its collaboration with Shimizu Corporation. The partnership aims to advance the research and development of Physical AI technologies, focusing specifically on the behavioral learning of humanoid robots. This initiative is set to unfold in 2025 and signifies a significant step toward the integration of AI in construction.

Background

As labor shortages become increasingly severe in the construction industry, AI robots are being recognized as a promising solution for this issue. Traditionally, robots used in construction have been specialized to perform specific tasks, executing pre-programmed actions. However, the potential of AI robots lies in their ability to recognize and adapt to their surroundings, allowing a single robot to perform multiple tasks with versatility.

In the complex and ever-evolving environment of construction sites, Physical AI stands out as a vital technology that can enhance productivity and safety while also addressing labor shortages. This adaptability is crucial in a field where conditions fluctuate day by day.

Overview of Initiatives

In its support of Shimizu Corporation's AI robot research, Prox Industries has played a vital role in building the fundamental infrastructure necessary for the behavioral learning of humanoid robots. This includes the establishment of a simulation environment using NVIDIA Isaac Sim, which is essential for experimenting with various learning algorithms.

Additionally, the collaboration involves collecting motion data through teleoperation and expanding the dataset required for robot behavioral learning. This comprehensive approach to research and development encompasses all necessary processes for creating an effective learning environment for humanoid robots.

An essential component of this initiative is the exploration of the viability of incorporating the Vision-Language-Action (VLA) model within the robot framework. Prox Industries has provided technical advice based on their expertise, ensuring that the project remains at the forefront of innovation.

The technical support provided has been instrumental in generating and accumulating efficient learning data for humanoid robots, thus establishing a robust research and development foundation that will facilitate continued verification and enhancement of AI-driven learning.

It's worth noting that specific details regarding the research content and technical specifications are classified and not publicly disclosed.

Future Prospects

Looking ahead, Prox Industries plans to leverage its technological expertise in AI and robotics to promote the research and implementation of Physical AI within various industrial sectors, including construction, manufacturing, logistics, and infrastructure.

About Prox Industries

Prox Industries is a startup that aims to implement physical intelligence from Japan, merging AI with robotics to bring intelligent automation to tasks traditionally performed by human labor. Comprised of researchers and engineers specializing in core Physical AI technologies, Prox leads multiple projects centered around Physical AI for enterprises in Japan. Utilizing a range of robotics, from humanoids to quadrupeds and collaborative robots, Prox addresses advanced technical challenges, including the Sim2Real and Real2Sim transitions.

By connecting cutting-edge academic technologies with the on-ground needs of industries, Prox facilitates everything from foundational research and development of robotic framework models to the evaluation and improvement necessary for real-world deployments. This dedication to progressing Physical AI is critical for supporting the establishment of a new industrial base in Japan.

- Company Name: Prox Industries, Inc.
- Location: 4F Hongo K&K Building, 3-30-10 Hongo, Bunkyo, Tokyo
- CEO: Yuji Kamei
- Company Website: prox-industries.jp

For ongoing updates, follow Prox Industries on social media: