Soft suction cup with light sensors measures how hard a robot presses on kiwifruit

Researchers led by Romina Adhami, with Cosimo Della Santina and Barbara Mazzolai among the authors, have studied in an arXiv paper a soft suction cup with built-in optoelectronic sensing, designed for kiwifruit, that estimates the forces between a robot gripper and delicate fruit.

On fruit not used in training, the best small neural network estimated contact force with an R-squared of 0.968 and a mean absolute error of 0.138 newtons, and even a simple mapping from the average optical signal reached 0.925. Sorting fruit into soft and firm from the same signals worked only moderately, at about 0.705 balanced accuracy, and finer three-way grading performed poorly.