Tech

To Make Robots Safer, Install Sensors on Humans

Until robots take over all our jobs, humans will work alongside them. But humans are unpredictable beasts, and at the moment, literally working alongside manufacturing bots can be a dangerous proposition. Imagine working in tandem with those enormous welding robots used in the auto industry, with you pointing out where you need welds as you piece together a complex part. Then—whoops—you scratch your head, and now your scalp’s been burned off.

Okay, that might be a bit of an extreme example, but the point stands. Programming a bot to know where a human is moving next, when even we don’t always know, is a difficult task. That means humans and robots have yet to be able to truly work in concert. Some new research from Georgia Tech poses a solution: By putting sensors directly on someone’s arms, robots can better predict in real-time what their human worker is going to do next.

Videos by VICE

The video above shows an interesting use case, using robotic lever is controlled by human input. Pulling backwards or forwards is easy for the bots to assess and assist with, as the human input is fairly steadily unidirectional. But holding the lever steady, which requires a fluctuating balance between both push and pulling muscles, is difficult. To you, you might simply sense that you’re holding the arm still; to a robot, it appears as a constant twitching back and forth.

“It turns into a constant tug of war between the person and the robot,” Billy Gallagher, a recent Georgia Tech Ph.D. graduate and project leader, said in a release. “Both react to each other’s forces when working together. The problem is that a person’s muscle stiffness is never constant, and a robot doesn’t always know how to correctly react.”

The sensor system reads fine muscle movements to provide robotic systems with data that’s easier to analyze than simple movement-based sensors. With that data, robotic systems can better read and predict what humans are going to do next and react accordingly, which, if successful in practice, would greatly increase the safety of human-robot work interactions.

As Evan Ackerman at IEEE Spectrum notes, the obvious bummer is that utilizing the system effectively means covering workers with sensors. But with all the interest in self-quantification these days, there are plenty of interesting form factors for wearable sensors and whatnot. Forget the latest fitness wristband, perhaps the big market for wearables is in developing sensors for industrial and manufacturing environments. 

As the above video shows, robotics have entered the workplace safely. Amazon has roboticized much of its shipping operations, and the Kiva bots it uses zip around Amazon’s shipping centers without fear of running over a human. Clearly demarcated routes and sensors that alert the bots to slow or stop for humans has made the tech safe enough to deploy.

But when it comes to manufacturing robots, especially ones swinging around heavy objects like a windshield, or those doing precise tasks like conducting surgery, humans and bots don’t yet work hand in hand very well. Separating the two adds more steps to the process and generally slows things down.

Better combining robots and human workers is potentially very lucrative, especially as the robot manufactuing space continues to grow. It’s telling that the GT lab working on adaptive robots got a $1.2 million grant from the National Science Foundation, which will help fund the development of the arm sensor system. 

“Such next-generation manufacturing is expected to improve productivity and reduce assembly time as well as physical burden of assembly line workers,” reads the award’s abstract, and it surely will. At the same time, with the pace at which automation is going, robots that work well with humans are likely to be replaced with robots that don’t need humans at all. 

@derektmead

Thank for your puchase!
You have successfully purchased.