Although there are many optical systems that provide a sense of touch to robotic hands, most of these systems are planar and are limited to the fingertips. But MIT’s new system works across the entire length of each finger. The installation, known as GelSight Svelte, is based on MIT’s existing GelSight technology.

In the second system, one side of the transparent silicone plate is covered with an opaque metallic paint layer. When the painted side is pressed against a surface, it deforms according to the shape of that surface. When you look at the unpainted side of the silicone, you can see the fine lines of the surface pressed into the paint.

Using a camera and computer algorithms, the system can transform these contours into 3D images that capture details less than one micrometer deep and about two micrometers wide. These visual details translate into the equivalent of pressure points that our nerves will feel.

MIT first incorporated the GelSight sensor into one of the robotic gripper’s finger pads, allowing the device to “sense” the cables it was gripping. The next GelSight EndoFlex hand, which can identify three-dimensional objects by sensing their shape, has two sensors placed on each of three fingers (six total for the entire hand).

GelSight Svelte also expands sensory capabilities beyond the finger pads, but in a much more refined and effective manner. It is already integrated with a three-fingered robotic hand.

Each finger of this hand contains only one inner chamber located at the base of the finger. There are also two rows of internal LEDs (one greedy, one red) and two mirrors running along the finger; A long, curved mirror extends across the finger, and a shorter, flat mirror fits at an angle. under the camera. .

When looking at a flat mirror, the camera can see the entire reflection of the curved mirror, thus seeing all reflected finger surface deformations. by him. At most, the amount of pressure exerted by the bottom surface on the silicone can be determined by analyzing the green and red light saturation of each deformation.

Additionally, GelSight Svelte continuously monitors the curvature of each finger using sensors in a flexible frame that runs along the back of the finger. The system knows that the more the finger bends, the more pressure it will put on the object it is holding.

“Because our new sensor is shaped like a human finger, we can use it to make different types of grippers for different tasks, instead of using grippers for everything,” said mechanical engineering graduate student and lead author Alan (Jialang) Zhao. from the research paper. “You can do a lot with a parallel jaw grip. Our sensor really opens up new possibilities for different manipulations we can do with robots.” Source