Tactile Sensors for Manufacturing: How Robots 'Feel'
Watch this video to see Howie Choset, Lu Li, and Victoria Webster-Wood — from the Manufacturing Futures Institute — explain their work. It creates specialized sensors that allow robots to "feel" the world around them.
Transcript
00:00:10 Howie: One of the things that make us, or all animals, successful are sensors. But sensing is just a hard problem. And we see with our eyes, obviously, but we also feel. We can also sense force. And building those kinds of sensors has proven to be very difficult for industry. Lu Li, he sort of had this little hobby on the side for building four sensors on this one project. Lu: And so the goal for this project is to build smart, intelligent tactile sensors to get robots sense of touch so that we can use the sensor for challenging applications such as medical, assembly, disassembly, and manufacturing applications. Howie: When you go and, say, grab that doorknob or put the key in the keyhole, you are not perfectly positioning your joints just right, you're approximating it, but you're feeling your way
00:01:06 through. And to do that, you need force feedback, force sensing. The problem with the force sensors that exist today are they're very expensive, and they're finicky, and they break easily. Vickie: Effective, low-cost tactile sensing is an ongoing challenge in manufacturing and recycling robots. Without the sense of touch, these robots rely on expensive camera systems with high-power computers to know where objects are in their environment. In contrast, animals and humans are able to explore much of our world via our sense of touch. So part of our ongoing work is to try to bring some of these capabilities from human touch into robotic sensors. Lu: We are able to use magnetic field to indirectly measure tactile and force information. So this is novel because magnetic field is one of the special signal that we can take advantages to convert contact information directly into electrical signal. And
00:02:06 this is super valuable since we now can understand what is the magnitude and direction of contact. Vickie: In this work, we've taken inspiration from human fingers in order to create a low-cost and robust tactile sensor for robotics. We've mimicked the different layers of tissue that you see in the human finger, as well as fingerprint ridges on the surface of the sensor. And together these features allow our sensor to sense both pressure and shear as well as vibration signals if we pull the sensor across the surface, and so this lets us do things like surface classification and material classification. Because the sensor is low-cost, it can be easily integrated and replaced as needed in demanding manufacturing environments. Lu: Tactile sensors are integrated into a parallel gripper, which in our case is a grasper that can touch and grasp object to apply assembly and also disassembly operations.
00:03:06 Vickie: Many of the ongoing challenges in incorporating this type of tactile sensing into manufacturing robots are really around cost and robustness. You need the sensors to be low enough cost that they can be readily deployed on fleets of robots, and they need to be robust enough that you can trust the data and you're not having to replace them frequently. And so this work really starts to address many of these challenges by bringing the cost down in such a way that these sensors are more accessible for manufacturing partners. Howie: And this sensor, you can take a hammer and mallet it if you want, you can rub a feather over it and it will detect the force interactions that are having. And that's necessary when you're performing a variety of tasks such as assembly. Lu: In the Center for Manufacturing Future Institute we build a rockstar team
00:03:54 combining roboticist, biological inspired group, and also engineers to build a unique sensor that can help promoting better intelligence in medical, manufacturing, and service industry.