Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Research & technology· 日本語で読む

Panasonic Holdings and others jointly develop T-Rex AI for bimanual tasks responsive to touch

Panasonic Holdings and others reportedly jointly developed T-Rex, an AI for robot manipulation. The report says it features bimanual tasks that respond quickly to fingertip touch. The robots used, methods and performance could not be confirmed from the article text provided.

What happened

automation-news.jp reported on November 15, 2026, that Panasonic Holdings jointly developed T-Rex, an AI for robot manipulation, with Panasonic R&D Company of America, the University of California, Berkeley, NVIDIA and others. The headline and excerpt describe it as featuring bimanual tasks that respond quickly to fingertip touch.

Why it matters

The reported T-Rex is a development example that incorporates processing of contact information into robot manipulation AI, targeting bimanual tasks responsive to fingertip touch. The specific tasks, performance and applicable robot platforms could not be confirmed from the information provided, so it is not possible to determine what operational challenges it addresses.

Analysis

The features described in the report combine responsiveness to fingertip touch with bimanual tasks. This suggests an approach to robot manipulation that treats it not only as a task of reproducing predefined motions but also as one of acting in response to information received at the moment of contact. However, the article excerpt provided is brief, and it does not specify how tactile information is collected or incorporated into control, or which tasks were evaluated. Therefore, what can be gleaned at this point is limited to the functions described as development targets. NVIDIA introduces SOMA Retargeter to adapt human movements to robots, covered by this newspaper, addressed the use of human motion demonstrations as a starting point for teaching robot movements. The reported tactile responsiveness of T-Rex points to a different operational issue. However, the information provided does not indicate whether the two technologies are connected or use common robots or learning methods, so they cannot simply be treated as part of a single development system. A report on NEXCOM and NexCOBOT showcasing RoboWIZ and four technology partners at GTC Berlin introduced computing platforms for robots and multiple technology partners. NVIDIA also appears in the T-Rex report as a co-developer, but it does not describe any connection to computing platforms, application to robots made by other companies, commercialization or supply arrangements. The participants in a joint development alone are not enough to infer implementation or business relationships. The next points to verify are the configuration of the tactile sensors, integration with other inputs such as vision, the target tasks and success rates, and the conditions under which the system was evaluated on real hardware. The specifications of the robot used for bimanual tasks, the training data, operational safety and available computing resources would also help determine its scope of application. None of these details appears in the article excerpt provided, and it is not yet possible to determine whether T-Rex is a research-stage result or is intended for real-world operation.

Sources

Translated from the Japanese original: /news/panasonikku-hd-nado-shokkaku-ni-sokuou-suru-ryoute-sagyou-ai-c7ad36. The Japanese version is authoritative if they differ.