Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Industry· 日本語で読む

JianZhi Robotics’ DAS 2.0 measures finger movements and tactile data, leaderobot.com reports

leaderobot.com reported on November 10, 2026, on JianZhi Robotics’ data collection system, “DAS 2.0.” According to the report, the glove-type “Dex 2.0” has 21 DoF and joint-angle resolution of 0.2°. The company’s design aims to collect contact information in addition to movement, with the goal of improving the consistency of training data.

What happened

leaderobot.com reported on November 10, 2026, on JianZhi Robotics’ data collection system, “DAS 2.0.” According to the report, the system comprises the glove-type “Dex 2.0” and the first-person-view headset collection device “Ego 2.0.” Dex 2.0 has 21 DoF, joint-angle resolution of 0.2° and fingertip position accuracy of less than 1 mm. It is equipped with 16 tactile sensors and 980 sensing units on each hand, and samples tactile data at 100 Hz, the report said.

Why it matters

According to leaderobot.com, Dex 2.0 captures tactile data in addition to joint angles, allowing it to record movement and contact at the same time. This is relevant to collecting training data for which contact information can be difficult to supplement using movement footage alone. However, the specifications in the report do not establish data quality or training effectiveness.

Analysis

The significance of the figures cited in the report is that the system is intended to handle movement and contact records within the same collection setup. According to leaderobot.com, Dex 2.0 measures the state of the hand with joint sensors positioned along its sides and also incorporates tactile sensors. The report describes this configuration as a way to collect data that includes contact, which can be difficult to capture through video or movement alone. The figures of 21 DoF, 0.2° and fingertip position accuracy of less than 1 mm are measurement specifications for the state of the hand, while 980 sensing units and 100 Hz are specifications for the tactile system. These figures do not directly prove data quality; they indicate what information the collection device is designed to record. We previously covered reports that the company would use funding to expand its data capabilities and platform in JianZhi Robotics raises about $100 million, leaderobot.com reports. The latest report shifts the focus from the broader data infrastructure to equipment that captures hand movements and tactile data. However, both articles are reports by leaderobot.com, and the information provided does not confirm how the collection equipment is integrated with the platform or how this device is connected to the use of the funding. The concept is to combine sensors that measure hand movements, a tactile component that records contact, and a first-person-view device that records video to assemble data for robot training. As the report notes, movement shown on screen may not capture whether contact occurred or how force changed. The practical value of this setup would therefore depend on the reproducibility of measurements when the same movement is repeated and on how closely the timing of movement, video and tactile data is synchronized. The report also says tactile responses were consistent during a one-month collection test in which the same movement was performed under the same conditions, but the excerpt does not specify the test conditions or evaluation method. It remains unclear under what working conditions the reported specifications were achieved, how calibration and repairs are handled during long-term use, what formats and quality assessments are used for the collected data, and how effective the data is in actual training. In particular, the specifications of “Ego 2.0” and how data is integrated across body-free collection and teleoperation modes cannot be determined from the article provided. The relationship between data collection and model training needs to be assessed alongside the equipment’s performance figures.

Sources

Translated from the Japanese original: /news/kan-satoru-kiki-nin-wo-meguru-deeta-shuushuu-kiki-no-kiji-wo-a076dd. The Japanese version is authoritative if they differ.