A new sensor enables robotic arms to pick up and manipulate sensitive, fragile and slippery objects. The sensors measure the pressure that the robot arm's gripper exerts on the object and the distribution of this pressure.
A new sensor enables robotic arms to pick up and manipulate sensitive, fragile and slippery objects. The sensors measure the pressure that the robot arm's gripper exerts on the object and the distribution of this pressure.
If we as humans pick up a sensitive or slippery object, the nerves in our fingers tell us whether we have a good grip on the object. For example, if we feel that an object is about to escape from our grip, we can apply more pressure to grip the object more firmly. For robotic arms to have the same ability, they too must be able to gather such feedback.
Elastic silicone skin with beads
A sensor developed by researchers at ETH Zurich offers this capability. The sensor consists of an elastic silicone 'skin', which has coloured plastic beads on the underside. A camera records how these beads are manipulated; if the robotic arm applies pressure, the pattern of the beads changes. Based on these changes, it is possible to calculate the distribution of pressure on the object. The researchers point out that the design of their sensor is simple, making its production cheap.
"Conventional sensors record applied pressure only at a single point. Our robotic skin lets us distinguish different forces on the measured surface and calculate them with both high resolution and accuracy," said Carlo Sferrazza, a PhD student member of a research group led by Professor of Dynamic Systems and Controls at ETH Zurich Raffaello D'Andrea. "We can even identify which direction pressure comes from."
Machine learning
The researchers use machine learning to recognise the patterns in the beads. By applying different degrees of pressure to objects at a large number of locations, the researchers visualised how this manipulates the beads.
The thinnest sensor developed by the researchers so far is 1.7 centimetres thick and 5 by 5 centimetres in size. In time, the researchers hope to reduce the thickness to just 0.5 centimetres. They are also working on applications where they will use the same technology to measure a larger area. In doing so, the researchers monitor the pattern changes of the beads using multiple cameras. This also makes it possible to recognise objects with complex shapes.
Author: Wouter Hoeffnagel
