Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Product launches· 日本語で読む

RP Hand releases specifications and development materials for five-finger robotic hand

Materials for RP Hand on GitHub list specifications for a five-finger hand with 11 degrees of freedom. Six axes are actively controlled, and the hand connects to RP1 via CAN-FD. The materials also include an SDK, a ROS 2 interface, and CAD, URDF and MJCF models.

Key facts

  • RP Hand has five fingers and is listed as having 11 mechanical degrees of freedom: six actively controlled axes and five passive or linked degrees of freedom.
  • It measures 147.54 × 73.53 × 26.35 mm, weighs 379 ± 10 g and uses a worm gear transmission.
  • It supports CAN-FD communications with RP1 and is described as supporting internal wiring.
  • The SDK includes C++, Python and a ROS 2 interface; CAD, URDF and MJCF models are also provided.

What happened

The GitHub repository “rp_dexhand” contains product and development materials for RP Hand, a five-finger hand for small and medium-sized humanoids. It has 11 mechanical degrees of freedom: six axes are actively controlled, while five degrees of freedom are passive or linked. It connects to RP1 via CAN-FD.

Why it matters

The RP Hand materials provide not only the five-finger hand’s dimensions and weight, but also information on its CAN-FD connection, left- and right-hand models, and development environments for C++, Python and ROS 2. Engineers considering integration into a robot can review both the mechanical design and control software, though the materials do not indicate performance in tasks such as grip strength or durability.

Analysis

A notable feature of the materials is that they bring together information for integrating the hand into a robot and developing applications, in addition to the hand’s mechanical specifications. Dimensions, connector pin layouts and left- and right-hand CAD models help with mechanical and electrical design, while the SDK and ROS 2 interface support control software integration. In other words, the materials are structured to help adopters proceed through installation, communications and control, rather than simply provide product information such as grip-performance figures. Of the 11 degrees of freedom, six axes are actively controlled, while the remaining five degrees of freedom are described as passive or linked. This differs from an approach that treats every finger movement as an independently controlled axis. However, the materials do not specify the exact range of motion for each finger, grip strength or performance manipulating different objects. The number of degrees of freedom therefore cannot be taken on its own as an indicator of dexterity or task capability; the movements that can be controlled must be verified using the physical hardware and control specifications. For connectivity, the materials say the hand uses CAN-FD with RP1 and support internal wiring. The SDK provides entry points for C++, Python and ROS 2. Having materials for both hardware connections and software development may reduce the effort needed to find information when integrating the hand into an existing robot. However, the information covers use with RP1 and does not indicate compatibility with other robots or whether modifications would be needed. The next points to verify are the conditions under which the specifications were measured, how the active axes and passive or linked degrees of freedom are handled in control, and durability and safety functions during continuous operation. The materials also point to a configuration app and a gesture-recognition demo, but do not state the demo’s recognition accuracy or operating conditions. Evaluations must distinguish these points to avoid conflating the breadth of the development materials with actual task capability.

Details

The hand measures 147.54 × 73.53 × 26.35 mm and weighs 379 ± 10 g. It uses a worm gear transmission. The materials describe quick-change mounting on RP1 and a CAN-FD connection that supports internal wiring. Development materials include SDKs for C++ and Python, a ROS 2 interface, and CAD models for the left and right hands, along with URDF and MJCF models. The materials also point to a configuration app for Windows and a gesture-recognition demo for Windows and Linux.

Sources

Translated from the Japanese original: /news/rp-hand-5-yubi-no-robottohando-shiyou-to-kaihatsu-shiryou-wo-3cffc5. The Japanese version is authoritative if they differ.