DARPA Creates a Robot Arm That Can Really Feel

Science fiction has long promised us prosthetic limbs that can match or exceed the capabilities of their flesh-and-blood counterparts. Video games and movies such as Deus Ex and I, the Robot, squeezed a lot of mileage out of this trail. For years, though, the idea has been something of a sick joke for real amputees, who often get stuck with simple mechanical claws.

Finally, technology began to catch up with the vision of science fiction. This week DARPA announced a new breakthrough that brings a natural sense of touch to these mechanical limbs.

Hi-tech

For legs, replacements are already quite difficult. Here’s prosthetist Hugh Herr. showcasing its sophisticated robotic legs that allow users to walk, run, jump and even dance in a completely natural way.

Unfortunately, the weapon proved to be more of a challenge. The legs serve a well-defined function: keep the user upright while moving. This goal is so simple that a limb can usually determine what it needs from the context using just a few sensors. On the contrary, we use weapons for all sorts of things, which makes them much harder to control.

Early robotic limbs had only a few degrees of freedom—maybe an elbow, a rotating wrist, and a gripping mechanism. They can be controlled by, say, buttons inside the shoe. This works for some things — holding an object or even shaking hands — but not for more unusual or delicate tasks. As robotics becomes more sophisticated, the control methods of these weapons were forced to improve.

One of the main innovators in this area was DARPA, a DoD affiliate that researches futuristic technologies, develops prosthetics for injured veterans. Here is one of their most recent prototypes that is as strong and almost as agile as a real human cock.

Obviously this cannot be controlled with the buttons in your shoes! The two main modern methods of control involve direct interaction with the nervous system.

Targeted Energy Recovery works by tearing the nerves that lead the missing limbs and sending them to the muscle tissue of the chest. Then, when the user tries to move the various joints in the hand, tiny pieces of the pectoral muscle flex instead. These curves can be detected and used to control the hand. For users who have a significant amount of remaining arm, it may also be possible to fix the flexion of the remaining muscles without surgery.

Here is PBS’s Miles O’Brien trying out the hand of DARPA using this technique:

Another variant — brain electrodes . For this procedure, the user undergoes surgery to implant an array of electrodes a few millimeters long into the area of the brain involved in controlling your hands. The electrode array has hundreds of individual electrodes, each recording a different pattern of neurons. Machine Learning Algorithm Algorithms can be trained with this data so that the hand can interpret the will of the user and act accordingly.

Sensor Technology

Unfortunately, by itself, this is not enough for natural limb control. Humans rely on their sense of touch and how their limbs (proprioception) are positioned to perform nearly every task. Being forced to rely only on visual information makes our limbs clumsy and unwieldy.

So how do you add a sense of touch to the robot’s hand?

The sensors are simple, but getting the information back into the body is quite difficult. One technology being developed is attaching electrodes to nerve bundles in the remaining limb. Unfortunately, this limits the number of sensor channels to a few — the feedback is relatively rough. Here is Miles O’Brien again, interviewing an amputee who had these electrodes implanted.

Another approach that DARPA has recently made a breakthrough with is to go back to the brain and use electrodes to stimulate it, «closing the loop» so that you can control the hand and feel its sensors through a direct neural interface. It’s much more natural and much more detailed than was possible before.

According to a DARPA statement, the user, a 28-year-old man with quadruple paralysis, was able to sense and control a robotic limb with great precision using an electrode array implanted in his brain. The blindfolded user was able to identify which finger was being touched with near-perfect accuracy and reported that he felt his own organic fingers being stimulated. It was the first time he could feel his fingers in over ten years.

He was even able, when the researchers touched two fingers, to notice that something was wrong and asked if they were trying to play a trick on him. This is a measure of how accurate this synthetic sense of touch is. In the not so distant future, more sophisticated brain implants can allow users to perform complex tasks such as folding origami or playing a musical instrument.