Supernumerary Robotic Limbs (SRLs) are robotic limbs that, when worn, give you more limbs than you'd normally have. In other words, they're not robotic limbs designed to replace biological limbs that you might be missing, but rather robotic limbs designed to augment the number of limbs that you have already.
MIT researchers have been developing SRLs that can help you do stuff that would be annoying, uncomfortable, or impossible to do on your own. Today at the IEEE International Conference on Robotics and Automation (ICRA) in Hong Kong, they presented their latest SRL prototypes, with one model featuring a pair of limbs that spring from your shoulders and another with limbs that extend from your waist.
MIT's shoulder-mounted SRL is designed to assist in tasks that take place over your head, or in situations where your other two arms are busy and you need a hand (literally) with something. One example, shown in the video below, would be in a construction context, where anything that needs to be attached to a ceiling has to be held up and hammered or screwed into place at the same time.
Another example (that's probably a bit more common for most of us) is trying to open a door when you're holding something with both hands.
The SRL shoulder robot uses two arms mounted on your shoulders such that the reaction forces on them are aligned with the spine. Each arm has five degrees of freedom, with interchangeable and customizable end effectors, and the complete systems weighs about 4.5 kilograms (10 pounds).
What's tricky about having a pair of shoulder arms is getting them to do what they're supposed to do without having to control them with your other arms, which would kind of defeat the purpose of the entire setup. Instead, the SRL watches what you're doing with your arms to decide how to move. It does that by monitoring two inertial measurement units (IMUs) that the user wears on the wrists. A third IMU sits at the base of the robot's shoulder mount, to track the overall orientation and motion of the SRL.
The SRL uses the gyro and accelerometer data from the IMUs to make a prediction (based on a model that's been created by demonstration learning) about what would be the most helpful, proactive position for its own arms. If you put your arms up above your head, for example, the SRLs raise above your head too, because it figures you're trying to hold something up. Using their SRL prototype, the researchers are testing different "behavioral modes" to program the limbs to do what they want.