Researchers design robots to learn alongside human workers

Thursday, 14 June 2012 10:27

June 14, 2012

The manufacturing sector has come a long way in the past few decades, as more and more factories have adopted some form of automation. But for the most part factories remain largely segregated, with robots handling their own individualized tasks in relative seclusion, being operated and maintained by a small number of specialists.

Researchers at the Massachusetts Institute of Technology hope to change this pattern, bringing them into contact with regular human employees by creating robots that could prove more helpful to individual workers, rather than forcing them to adapt to the machines.

Difficulty with detailed work

Julie Shah, an assistant professor of aeronautics and astronautics at MIT, works with aerospace firm Boeing, a company where the complicated and detailed construction process makes it difficult to use robots extensively.

"It’s really hard to make robots do careful refinishing tasks that people do really well," Shah explained.

And often robot assistants can get in the way of human workers who sometimes take different approaches to completing the same task.

One example Shah and her team investigated was spar assembly, the construction of the main frame of an airplane's wing, which involves applying sealant to pre-drilled holes and then hammering in bolts. As a simple difference in worker preference, some people prefer to assemble spars one bolt at a time, while others will apply sealant to multiple holes before beginning to install bolts.

Helping where they can

Robots do not need to take over the process to contribute, however.

"If the robot can provide tools and materials so the person doesn’t have to walk over to pick up parts and walk back to the plane, you can significantly reduce the idle time of the person," explained Shah. "... providing robotic assistants to do the non-value-added work can actually increase the productivity of the overall factory."

In response, Shah and her group began some fascinating engineering research and development for a robot that could observe the habits of human worker and adapt to preferences of specific individuals.

The team used machine arms dubbed FRIDA from Swiss company ABB, but developed a system of decision trees in which the robot would break down the possible actions by a worker at each step. DiscoveryNews reports that learning process required workers to dress in motion-capture suits similar to those used for green screens in movies. Less intensive motion-capture systems, like Microsoft's X-Box Kinect, are not sensitive enough yet for the initial programming.

Once the machine has a sense of how individual will act, the system will then assign preferences for an observed worker through the tree so that it might be able to anticipate their next actions. Many factories already make use of radio-frequency identification tags, which the robots would be able to use to distinguish between workers.

"We have hardware, sensing and can do manipulation and vision, but unless the robot really develops an almost seamless understanding of how it can help the person, the person’s just going to get frustrated and say, 'Never mind, I’ll just go pick up the piece myself,'" said Shah.

Robots beyond the factory floor

While robots have proven to be extremely useful construction and engineering tools, they remain relatively uncommon in the home. But Appliance Magazine reports that The Bosch Group sees extensive application for robotics in the consumer market. While Shah's research has been mainly targeted toward improving factory productivity, systems that can learn to adapt to human tendencies could help to dramatically expand the market for consumer robotics.

Shah and her team will present their work at the Robotics: Science and Systems Conference this July in Sydney.