The algorithm lets robots find the shortest route in unfamiliar environments, opening the door to robots that can work inside homes and offices.
The news: A team at Facebook AI has created a reinforcement learning algorithm that lets a robot find its way in an unfamiliar environment without using a map. Using just a depth-sensing camera, GPS, and compass data, the algorithm gets a robot to its goal 99.9% of the time along a route that is very close to the shortest possible path, which means no wrong turns, no backtracking, and no exploration. This is a big improvement over previous best efforts.
Why it matters: Mapless route-finding is essential for next-gen robots like autonomous delivery drones or robots that work inside homes and offices. Some of the best robots available today, such as Spot and Atlas made by Boston Dynamics and Digit made by Agility Robotics, are packed with sensors that make them pretty good at keeping their balance and avoiding obstacles. But if you dropped them off at an unfamiliar street corner and left them to find their way home, they’d be screwed. While Facebook’s algorithm does not yet handle outside environments, it is a promising step in that direction and could probably be adapted to urban spaces.
Two billion steps and counting: Facebook trained bots for three days inside AI Habitat, a photorealistic virtual mock-up of the interior of a building, with rooms and corridors and furniture. In that time they took 2.5 billion steps—the equivalent of 80 years of human experience. Others have taken a month or more to train bots in a similar task, but Facebook massively sped things up by culling the slowest bots from the pool so that faster ones did not have to wait at the finish line each round.
As ever, the team doesn’t know exactly how the AI learned to navigate, but a best guess is that it picked up on patterns in the interior structure of the human-designed environments. Facebook is now testing its algorithm in real physical spaces using a LoCoBot robot.
Sign up here to our daily newsletter The Download to get your dose of the latest must-read news from the world of emerging tech.
Why Meta’s latest large language model survived only three days online
Galactica was supposed to help scientists. Instead, it mindlessly spat out biased and incorrect nonsense.
DeepMind’s game-playing AI has beaten a 50-year-old record in computer science
The new version of AlphaZero discovered a faster way to do matrix multiplication, a core problem in computing that affects thousands of everyday computer tasks.
A bot that watched 70,000 hours of Minecraft could unlock AI’s next big thing
Online videos are a vast and untapped source of training data—and OpenAI says it has a new way to use it.
Google’s new AI can hear a snippet of song—and then keep on playing
The technique, called AudioLM, generates naturalistic sounds without the need for human annotation.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.