v12

Researchers at ETH Zurich have developed a soft, sensorized skin for robotic hands. They used a fast prototyping method to optimize the design and tested it for grasp strength, range of motion, and speed. They also mounted pressure sensors under the skin to classify objects. The final design was created using multi-material 3D printing and casting. In this video, we present our work on a soft skin for dexterous robotic hands performed in the Soft Robotics Lab of ETH Zurich. Human hands having flesh and highly innervated skin around their rigid bone structures help us in manipulating objects in daily life. Our soft skin provides ideal contact with a wide range of objects. Our sense of touch also enables us to manipulate without visual cues. Even though it has been shown that tactile sensing can enable similar capabilities for robotic hands, current humanoid hand platforms only mimic our kinematics, but rarely provide soft skins and tactile sensing. In our work, we developed a soft, sensorized skin for the five robotic hands. Our main contributions are the following. We developed a method for fast prototyping and optimizing of a soft skin for our dexterous robotic hands, the five hands. We evaluate the skin through both quasi-static pulling tests and dynamic tests and show that it enhances grasp strength without limiting range of motion or speed. We then mount custom-made pressure sensors under the skin and show that we can utilize them in classifying objects through graphs. Our fast prototyping pipeline used multi-material 3D printing to optimize the parameters of our complex origami structure made of dragon skin tan. After 11 iterations, we arrived to a design that does not buckle or limit joint motion. The final versions of our finger designs were cast together with a palm in 3D printed molds. After one hour of casting and four hours of cure time, we spent one hour demolding the gloves using isopropanol.