Though creating an autonomous robot that can tidy a room seems like enough of an achievement, Tokyo-based Preferred Networks goes one step further. By integrating natural language processing into their technology, their robots respond to commands and adjust their actions.
Jun Hatori, a software engineer at Preferred Networks, spoke with AI Podcast host Noah Kravitz about the company’s latest developments.
To create robots that can understand how to clean up a room and respond to demands, Hatori described two main obstacles.
“I started to realize that robots can’t do as much as we can instruct,” he said. While NLP technology allows robots to understand the commands being given, their hardware isn’t always advanced enough to carry out the tasks.
The second challenge is crafting a robot that can understand the nuances of human language. “If you’re going to give it a command — like, ‘Pick up that white stuff’ — then the robot basically has to know what kind of items are there, how they’re placed, and what the word ‘white’ means,” Hatori said.
Preferred Networks has overcome these challenges to craft a robot with computer vision and object detection technology, as well as human-robot interaction capabilities such as gesture recognition and spoken language interpretation.
Their robot first assesses the room and creates a task list based on the objects that are out of place. Using a “paragrip” — a pinching hand — the robot grasps objects and puts them away.
By integrating NLP capabilities, users can instruct the robot to put objects elsewhere.
Preferred Networks has also applied such deep learning based computer vision and human-robot interaction technologies in the biohealth, industrial and automobile domains.
But their focus is still on personal robots. “Everyone knows it has huge potential if someone can build something actually usable,” Hatori said. “In the coming years, I think there’s going to be very big competition among many companies and research groups.”
You can see Preferred Networks’ cleaning robot in action along with their other projects at their website.
How to Tune in to the AI Podcast
Get the AI Podcast through iTunes, Google Play, Castbox, DoggCatcher, Overcast, PlayerFM, Pocket Casts, Podbay, PodBean, PodCruncher, PodKicker, Soundcloud, Spotify, Stitcher and TuneIn. Your favorite not listed here? Email us at aipodcast [at] nvidia [dot] com or fill out this short listener survey.