YASKAWA Electric Corporation (Head Office: Kitakyushu, Fukuoka, Japan; Representative Director, Chairman of the Board and President: Hiroshi Ogasawara, “Yaskawa”) and SoftBank Corp. (President & CEO: Junichi Miyakawa, “SoftBank”) have validated a deformable object manipulation system developed with the AI robot MOTOMAN　NEXT of Yaskawa by using the “AI Data Center GPU Cloud”*1 that SoftBank is developing as a development platform for Physical AI. The two companies are accelerating their collaboration to support the social implementation of Physical AI, beginning with use cases in office environments.*2

　In this validation, the companies confirmed that the system can stably manipulate deformable objects such as wire harnesses, which are generally difficult for robots to manipulate. By using the Physical AI development support tool that runs on SoftBank’s “AI Data Center GPU Cloud,” they significantly improved the efficiency of the development process, from collecting robot operation data to AI training and evaluation, and then feeding the results back to the robot. As a result, they demonstrated that Physical AI can be deployed more quickly and more easily.

*1 For details about the “AI Data Center GPU Cloud” refer to SoftBank’s press release dated May 25, 2026 (SoftBank to Launch “AI Data Center GPU Cloud” Powered by “Infrinia AI Cloud OS” as Part of its Neocloud Business in October 2026)

　 https://www.softbank.jp/en/corp/news/press/sbkk/2026/20260525_01/

*2 Yaskawa Electric and SoftBank Corp. Begin Collaboration on the Social Implementation of “Physical AI”

 　https://www.yaskawa-global.com/newsrelease/news/178574

　In recent years, manufacturing and logistics sites have needed broader use of robot automation to address labor shortages and improve productivity. However, deformable objects such as strings, cloth, bags, and wire harnesses remain difficult to handle. Their shape and position change during each task, and they can easily bend, tangle, fold, or vary in grasping position. As a result, stable handling is difficult with conventional robot control alone.

　Yaskawa has been developing a manipulating system that can recognize, grasp, and manipulate deformable objects by using VLA (Vision-Language-Action), which generates robot actions based on visual information and task instructions, to expand robot applications into tasks that are difficult to automate with conventional teaching playback control.

　Applying Physical AI to real robot systems requires an efficient development environment for the entire workflow, including robot motion data collection, AI model training, simulation-based evaluation, and deployment to Physical robots.

　Yaskawa and SoftBank have validated both the effectiveness of Physical AI in deformable object manipulating and improvements in development efficiency by combining Yaskawa’s deformable object manipulation system with the Physical AI development support tool that runs on the “AI Data Center GPU Cloud” being developed by SoftBank.

　The ” deformable object manipulation system ” developed by Yaskawa is a system that uses robots to grasp and manipulate objects. This system uses VLA based on visual information from cameras and task instructions to recognize the state of an object.

　Deformable objects change their shape and position every time they are handled. As a result, stable handling is difficult with conventional rule-based control, which assumes predefined positions and orientations. To address this challenge, Yaskawa separated tasks that can be handled reliably by conventional robot control from tasks for which Physical AI is particularly effective, such as object state recognition and grasp point selection. In addition, Physical AI was integrated into the existing robot system as a functional module.

　In this validation, the system was applied to a task in which wire harnesses with varying shapes and positions were placed into a box. The AI can provide appropriate instructions to the robot by learning visual information and other inputs in real time. As a result,　the robot can  recognize the condition of each wire harness and select appropriate grasping and handling actions. The validation confirmed that the system can stably manipulate deformable objects.

　Through the combination of Physical AI and the robot control technologies that Yaskawa has developed over many years, Yaskawa aims to expand the use of robots into task areas that have traditionally been difficult to automate.

Please watch the video of the deformable object manipulation system at the following URL

https://www.youtube.com/watch?v=NFWMibbtJn0&feature=youtu.be

　Yaskawa will continue to advance the development of AI Robotics and expand its application to more complex tasks. Through these efforts, Yaskawa aims to contribute to the social implementation of Physical AI.

＜Reference＞

For SoftBank’s press release on July 13, 2026, please refer to the page at the following URL.

https://www.softbank.jp/corp/news/press/sbkk/2026/20260713_01/(Japanese)

For inquires, contact:

Corporate Communications Dept.

YASKAWA Electric Corporation

E-mail: pr@yaskawa.co.jp