Thanks to people like Wyatt Newman, Ph.D., the science of robotics will soon blend nicely with the science of material handling. Newman is professor of electrical engineering and computer science at Case Western Reserve University in Cleveland, and he's working with ABB Automation Products to find ways to make robotics more acceptable in real-world material handling.
MHM interviewed Dr. Newman at the opening of ABB's R&D center in Cleveland. As you'll learn, he's convinced that MHM's readers will have realistic job openings for the new breed of industrial robots that will soon be entering the market.
MHM: What are the main roadblocks to broader robotic applications in industry?
Newman: Robots are bad with feeling. Most of the interesting tasks that we do that change our environment involve dynamic interactions. You have to push and pull things, insert parts, polish a surface, and that requires a sense of feeling. Robots grew up from CNC machines. They interact with their environment by redefining that environment. If something is in the way, it carves it out of the way. We'd like to make our robots more creature-like so they act more cooperatively with the environment. We need to add more feeling to robots.
MHM: The robotic programming challenges to do that must be daunting.
Newman: We don't have any programming language that uses the sense of feeling. Because of the lack of this capability, robots are clumsy at what they do. If you structure your environment very carefully you can get away with it. But we want to make these robots more aware of their environment, more capable of using senses, and make them easier to talk to.
MHM: How can robots have senses?
Newman: We're creating the underlying dynamics that allows a robot to have a sense of feeling and respond to its environment, then we add motivation with programming that communicates schemas to the machine. We're trying to emulate what we think people are doing when they evoke procedural knowledge. If you ask someone on an assembly line what they're doing, they'd tell you, but they'd probably be wrong. We don't have a conscious awareness of our procedural knowledge. It comes through practice, like tying your shoes. We want robots to learn from examples and get better through experience. So we're programming them in schemas, where we say, here is a set of reactions. If you get this type of stimulus, this is the response we should get from you. If you build a table full of these reactions, that would be the way you program it.
MHM: How will this translate into practical work for robots?
Newman: We want to get robots into the types of jobs they're not good at now. Our motivation comes from ABB's customers. We work with ABB's engineers to turn their input into products. Getting information directly from customers helps focus our research. We're a large group of researchers; we can bring novel ideas in and help introduce innovation in your research. This is a way to get academic research into industrial practice much faster.
MHM: How about material handling applications? Right now there's a lot of buzz about radio frequency identification technology. Will RFID play a role in robotic applications as more shipments contain RFID labels?
Newman: One of the difficulties is to make a machine smart enough to recognize something. That's a tremendous programming problem. RFID sidesteps that. Instead of dealing with the ambiguity of what you're handling you can focus more on what you're supposed to do with it. A robot will affiliate an object with its RFID tag. It can then associate what you say with what it can detect from unambiguous sensing. Thereafter you can say I want to put this pallet over here and the robot will affiliate coordinates with that. So when that part comes in the robot will know what that is, you showed me where it belongs. The challenge is to make RFID tags smaller, cheaper and more powerful.
MHM: Do you foresee robots eventually fulfilling orders?
Newman: They will be like AGVs, but they'll go off the wire and navigate by GPS and machine vision. They'll get mile markers from RFID tags posted around the facility. This will require less user programming. It will make fewer mistakes if the robot already has an idea of what it needs to do. It will have the ability to fill out the details of the implications of what you're saying, and you won't have to program all that in.
(Ed. Note: If you'd like to ask Dr. Newman a few questions of your own, you can e-mail him at [email protected].)
