A 6-axis robot, such as you might see in an automotive plant, is an incredibly versatile machine, capable of an almost infinite number of movements and actions. But most 6-axis robots spend their lives doing only a single specific action over and over, limited by the directions specified in their programming. In a sense, until now, robots have been capable of “learning” only what the programmer teaches.
Heather Kerrick is trying to change that by giving robots the tools to learn on their own. She’s a senior research engineer in the Applied Research Lab at Autodesk’s Pier 9. Kerrick’s project, called Supervised High Rate Metal Printing (SHRMP, for short), is experimenting with 6-axis robots that weld metal and learn on the job. She and her team are equipping robots with cameras and other sensors to observe their work and the world around them. Then they use machine learning so that the robots can make decisions and adjust their actions based on what they observe.
Kerrick is working toward a future where robots become not just doers, but learners—entities that constantly get better at the tasks they’re assigned. As she explained in her Innovation Zone presentation at AU 2016 in Las Vegas, “[We can] turn 6-axis robot arms into the kind of versatile machines that they’ve had the potential to be for 30 years.”
Check out her Innovation Zone presentation: