Harvesting Robots Train to Pick Fruit Using a Silicone Raspberry

Harvesting fruit such as berries and grapes is more complicated than it seems. To prevent easily damaging the fruit, berries are harvested by gently grasping them between the thumb and forefinger and pulling until it detaches from the stem. As one might imagine, harvesting by hand takes an astronomical amount of time and using farm equipment can damage a large percentage of crops, which is reflected in higher price tags.

To help ease harvest damage, engineers from EPFL have developed a silicone raspberry that teaches robots how to grasp the fruit without too much pressure. The fake fruit houses a fluidic sensor outfitted with a soft silicone tube that measures the compression force exerted by the robot. A pair of magnets generate a pulling force that holds the fruit and receptacle together, simulating the raspberry connected to the stem.

The simulated raspberry is coupled with a machine-learning program that teaches the robot how to apply the appropriate amount of force to detach the fruit from the stem without damaging the skin. “So far, we’ve been using a very simple feedback system in our robot," the engineers explain. “The next step will be to design and build more complex controllers so that robots can pick raspberries on a larger scale without crushing them.”

The team is currently looking at ways to develop a camera system that would allow robots to see and feel the raspberries and determine when they are ready for harvest.