A robot grasping any objet with 99% accuracy
Grasping objects with irregular forms is a great challenge for most robots. Dex-Net (Dexterity Network) is a robot consisting of a neuronal network, a Microsoft Kinect 3D sensor and a robotic arm. It is capable of identifying any object and determining the best way to grip it, 99 times out of 100. Dex-Net artificial intelligence was fed with data on thousands of objects. So, faced with a new or unknown object, Dex-Net can study its shape and determine the best grip.
The Berkeley University Researchers behind the project hope that Dex-Net will quickly be adopted in industrial production lines. Today, robots play important roles inside retail giants like Amazon and manufacturing companies but they are programmed for very specific tasks. Dex-Net is one step closer to multi-skilled robots with adaptive intelligence.
Spotted in 2018
By HEC PARIS (France)
- United States