MIT's RoboRaise bot will mimic your bicep curls
Robots are pretty good at carrying out specific actions, provided we give them specific instructions first. But could robots and humans work together more seamlessly if the machines were able to take their cues from human movement, say like a furniture removalist might be guided by the motion of their co-worker? Scientists at MIT believe that indeed they might, having developed an impressive robotic arm that is controlled by the movements of a human counterpart.
Dubbed RoboRaise, the machine is the latest from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), and was developed as an exploration into a new kind of robotic command system. Rather than explicitly programming a robot to perform a series of movements, the researchers wondered if having it mimic a human's actions might make for a more intuitive way of doing things.
The team built the RoboRaise system into its Baxter humanoid robot, and used non-invasive EMG sensors attached to a human's biceps and triceps to track muscle activity. More specifically, it monitors the electrical signals coming from the muscles as the user flexes and relaxes, and moves upwards or downwards.
A purpose-built algorithm then converts these signals into control commands for the robot, which is then able to replicate the movements of the human arm in close to real time, to within a few inches of the same height. The team tested this out using a human and RoboRaise to pick up and assemble pretend aircraft components, along with rigid and flexible objects, and found it responded correctly to around 70 percent of the gestures.
"Our approach to lifting objects with a robot aims to be intuitive and similar to how you might lift something with another person, roughly copying each other's motions while inferring helpful adjustments," says graduate student Joseph DelPreto, co-lead author on the paper. "The key insight is to use nonverbal cues that encode instructions for how to coordinate, for example to lift a little higher or lower. Using muscle signals to communicate almost makes the robot an extension of yourself that you can fluidly control."
The researchers say it takes only a few tensing and relaxing cycles for RoboRaise to calibrate itself with a previously unknown arm. And as an aside, it is also designed to recognize subtle up-and-down hand gestures for finer control over certain movements, such as hold a position or move further away.
Looking forward, the researchers have some interesting ambitions for their new robotic system. They imagine adding more muscles and sensors to track new kinds of movement, and even factoring in things like exertion and fatigue so the robot can increase its level of assistance as the human tires. They hope that it can one day can find applications in manufacturing, construction or even around the house.
"We aim to develop human-robot interaction where the robot adapts to the human rather than the other way around," says co-lead author Daniela Rus. "This way the robot becomes an intelligent tool for physical work."
A research paper describing the work can be viewed online here, while the video below shows RoboRaise in action.