CES last week was full of robots and virtual assistants, many of which made use of computer vision. Though still a proof of concept, Samsung’s Bot Retail analyzes facial expressions to better interact with customers and guide them around the store floor. Ubtech’s bipedal Walker butler robot has built in facial and object recognition and U-SLAM to recognize people and avoid crashing into things, as well as visual and force perception, to maintain hand-eye coordination (useful when carrying a tray of drinks), though getting around uneven terrain beyond carpets and floors is still a challenge for two-legged robots in general . The new Ozmo Deebot 960 robovac has a camera on its front with object recognition capability for hundreds of household objects from socks to shoes to cables, enabling it to avoid those items as it sucks up dirt. Groove X’s Lovot is a cute companion robot out of Japan that besides facial and object recognition also uses a thermal camera to tell the difference between humans and objects. Among the most practical and real-world robots on display last week were those that do deliveries, especially of the last-mile variety: Unsupervised.AI’s Maryam is aimed specifically at manufacturing facilities; Tactile Robots’s TR1 is optimized for room service tasks hotels; and Robby Technologies' Robby 2 "Snackbot" will soon be delivering snacks to students at the University of the Pacific. As with the Segway Loomo we featured last week, most of the new class of delivery robots also look like file cabinets, but it’s the computer vision inside that helps them get around and interact better that counts. Look for these soon at an office, hotel, factory, or neighborhood street near you.
sign up for our newsletter