How to make a less creepy robot? Simple, just add data

Disney’s research arm has solved a problem that you probably didn’t even know robots have — their inability to accept objects from people in a natural way. The Disney Research team, working with funding from the International Center for Advanced Communication Technologies (interACT) at Carnegie Mellon and the University and Karlsruhe Institute of Technology (KIT), believe that robots who can’t naturally accept “handoffs” of objects from people are creepy. In a paper presented this month, Disney and its partners detailed how they used several motion-sensitive cameras, a database of gestures and some fancy algorithms to solve this handoff problem.

From the press release announcing the findings:

“If a robot just sticks out its hand blindly, or uses motions that look more robotic than human, a person might feel uneasy working with that robot or might question whether it is up to the task,” Katsu Yamane, Disney Research, Pittsburgh senior research scientist explained. “We assume human-like motions are more user-friendly because they are familiar.”

RecMo_image-1024x174

Despite the robot pictured on the Disney page touting this research looking like the mechanical, blue-haired skeleton that haunted my childhood nightmares, its attempts to grab the purse from the person do seem reactive to the human’s gestures, as opposed to the robot just sticking his arm out there and the person having to accommodate it. And that sort of naturalism will be important as we bring more robots into our homes and workplaces.

For example, an MIT group used a dancer’s motions to build a robotic bartender in a quest for naturalism — even though that robot doesn’t interact with people.

Today, designers try to endear robots to us with quirky noises (like R2D2) and maybe light displays or LED faces — anything to help anthropomorphize them. But as robots become more human-looking they can also become more sinister — achieving that same uncanny valley that Disney and other content companies have struggled with in animation. Remember the dead-eyed stars of the Polar Express that you probably couldn’t empathize with? The jerky movements of a home health robot might engender similar feelings — or worse — they may scare people.

Building the natural gestures of the Disney robot took the creation of a hierarchical gesture database that the robot can access as it detects the person passing something to it. In the Disney paper research, the robot is not only able to reach for the handbag, but when the human attempts a fake pass to the robot, the ~~blue-haired monstrosity~~ robot is able to adapt. From the release:

To enable a robot to access a library of human-to-human passing motions with the speed necessary for robot-human interaction, the researchers developed a hierarchical data structure. Using principal component analysis, the researchers first developed a rough estimate of the distribution of various motion samples. They then grouped samples of similar poses and organized them into a binary tree structure. With a series of “either/or” decisions, the robot can rapidly search this database, so it can recognize when the person initiates a handing motion and then refine its response as the person follows through.

Even if you don’t have an opinion on how naturally robots should move, this research brings home the awesome amount of work it takes to build computers and robots that mimic the capabilities of a person. Much like computer visualization, the science of robotic interaction takes a problem the size of a mountain and has to chip it down into grains of sand using a toothpick to find solutions. It’s a testament to human curiosity that people are willing to try.

Also, I expect Disney might be lured by the idea of natural-looking robots roaming its theme parks. My only question is would they be dressed up as characters or working the cash register at the gift stores.