Autoencoding a Soft Touch to Learn Grasping from On‐Land to Underwater

Abstract

Underwater Grasping Thanks to a breakthrough in soft robotics and AI, robots can now “feel” their way through underwater environments by learning from grasping on land. Using a high-frame-rate camera and a soft, omnidirectional adaptive finger, this pioneering work demonstrates a robot learning system to precisely grasp objects underwater, even in low-visibility conditions. This new “touch” technology unlocks exciting possibilities for underwater exploration and scientific discovery, allowing robots to interact with delicate objects and navigate challenging environments with much-enhanced reliability and robustness. For further details, see article number 2300382 by Fang Wan, Chaoyang Song, and co-workers.