Factory-Helper Robot Anticipates Your Next Move
Researchers have programmed a robot to learn people
CREDIT: nokhoog_buchachon | Shutterstock.com
Researchers have programmed a robot to learn how a mechanic likes to work, then to help the mechanic without breaking his or her rhythm. The researchers hope to see robot assistants working side by side with people someday, helping with small tasks.
An aeronautics researcher who worked on the robot is thinking in particular about airplane manufacturing plants. "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," said Julie Shah, one of the team of robotics researchers at the Massachusetts Institute of Technology.
Shah and her colleagues wrote an algorithm that lets robots watch and learn people's work habits and predict their next move. Her team tested the algorithm in a lab robot that worked with human mechanics who assemble spars, the main structural component of airplane wings.
Normally a mechanic lines up two pieces of a wing that have pre-drilled holes in them, then brushes sealant onto the holes, hammers bolts into the holes to hold the pieces together, and wipes off the extra sealant. Some mechanics like to go through the whole process hole by hole. Others like to add sealant to all the holes before bolting them down.
In Shah's test, a robot equipped with the new algorithm watched and learned the sequence each mechanic preferred. It then brushed on sealant and hammered in bolts alongside the mechanic following the same sequence.
Shah imagines these assistant robots someday will go through training with the humans they'll work with, learning each person's work rhythm. Once each robot is trained enough for the factory floor, it would recognize different mechanics and call up the right work plan for each person. Many workers in manufacturing plants now wear radio-frequency identification tags, so robots could use the tags to identify people, Shah said.
Robots will need to learn people's wants and work habits if they're to work alongside humans, Shah said. "Unless the robot really develops an almost seamless understanding of how it can help the person, the person's just going to get frustrated," she said.
She and her MIT colleagues plan to present their work July 12 at the Robotics Science and Systems conference in Sydney, Australia. They were funded in part by Boeing Research and worked with ABB, a Swiss company working on factory-assistant robots.