What could be easier than picking up a strawberry, an egg, a blueberry? we do this kind of things many times a day without a second thought.
As a matter of fact there is quite a lot going on behind the scene of our perception. Our eyes pick up the image of the object we want to pick up and our brain convert this image into an estimate of the weight of the object and adds to it the knowledge about the objects surface and “firmness”. Apply too little force and we won’t be able to pick it up, apply too much force and our fingers will cash it. Also it is important to know “how” to pick up the object so that it won’t slip away, notice if it has some parts that may be harmful (like picking up a rose you know it should be done from the stem, not from the flower or you may ruin it, but you should also be watchful for thorns to avoid pricking your fingers…).
What comes natural to us requires quite a bit of reasoning (“processing”) by a machine, a robot. It also requires some “pinchers” that are flexible enough to adapt to a variety of situation (from picking up a ball of steel to picking up a blueberry).
This is what several robot developing companies are trying to do. So far robots have been built to perform a specific task and did not require a lot of flexibility. Besides, in most application environment they are programmed to perform a very specific (and repetitive) action. If you want a robot that can operate in a variety of environment, particularly environment where the context and activity may change in a blink of an eye you need to have a robot that can take decisions on the spot and change its behaviour. In the clip you can watch a soft gripping robots that can pick up fruits and vegetable to speed up packaging. This is a very specific activity and what is needed is a robot equipped with “soft” grippers (fingers) progrsmmed to apply a minimal pressure.
At the National University of Singapore a team of researchers have developed a hybrid robotic system inspired by the human hand. The real breakthrough is not in the hardware part (the soft yet sturdy and flexible fingers) but in the software. A video camera pick up the image of the object creating a 3 dimensional model (as the robotic hand approaches the object the camera picks up different images that allow the construction of a 3D model) and the software captures hints on the surface texture that helps in deciding how to grip, and where, the object. If the hardware replicates the structure of the human hand the software replicates the image recognition capabilities of our brain and the decision processes that result in directing the hand muscles for the appropriate manipulation of the object.