Facebook is developing haptic technology for robots
Facebook is developing haptic technology for robots

Facebook has announced that it has developed a number of touch technologies that can give bots a sense of touch.

According to the company's Artificial Intelligence Research Laboratory, the next generation of robots should make better use of touch technology.

To seize the opportunity in this relatively new field of artificial intelligence and robotics, the company and its partners have developed a new type of cheap and durable electronic skin and fingertips that provide the basic and reliable tactile sensation of robots.

Although the sense of touch is not useful for judging whether an object is a picture of a cat or a dog, or who is talking in the room. But if robots or AI systems want to interact with the real world, they need more.

“We understand pixels and appearance very well,” said Roberto Calandra, a researcher in the company's Artificial Intelligence Research Laboratory. But understanding the world is much more than that. We need a physical understanding of things to determine this.

Although cameras and microphones are inexpensive, there are many tools that can effectively process this data. But touch does not know.

Facebook is developing shapes and fingertips for more sensitive bots

Precision pressure sensors are not a common consumer product. As a result, useful equipment is often left behind in laboratories and industrial environments.

The DIGIT touch sensor was published as an open source design in 2020. It uses a tiny camera for missions to produce detailed images of the affected object. Very sensitive when touching different things.

The ReSkin project was born in 2009. In 2014, a project called GelSight appeared at the Massachusetts Institute of Technology.

The company has grown like bamboo shoots after rain and is the production partner for this documented tactile method. Basically, magnetic particles hang on the surface of the Softgel. The magnetometer below can detect the motion of particles and convert those motions into accurate force maps of the potentials they cause.

One advantage of GelSight is that the hard components - chips with magnetometers, logic, etc. - are separated from the flexible components, which are just a flexible washer filled with magnetic dots.

So the surface can get dirty or scratched and can be replaced easily, while fragile parts can be safely hidden under it.

In the case of the ReSkin, this means you can connect an array of chips of any shape, place the elastomers, and then combine the signals to get tangible feedback from everything.

Although you need to calibrate it. But this system is much simpler than other synthetic leather systems that can operate at a scale greater than a few square inches.

With such a pressure-sensitive surface, the robot can more easily sense the presence of objects and obstacles. No additional joint force is applied in this direction. This can make the robot more sensitive to touch.

Previous Post Next Post