CMU and Facebook AI Research use machine learning to teach robots to navigate by recognizing objects

Brian Heater

Carnegie Mellon today showed off new research into the world of robotic navigation. With help from the team at Facebook AI Research (FAIR), the university has designed a semantic navigation that helps robots navigate around by recognizing familiar objects.

The SemExp system, which beat out Samsung to take first place in a recent Habitat ObjectNav Challenge, utilizes machine learning to train the system to recognize objects. That goes beyond simple superficial traits, however. In the example given by CMU, the robot is able to distinguish an end table from a kitchen table, and thus extrapolate in which room it’s located. That should be more straightforward, however, with a fridge, which is both pretty distinct and is largely restricted to a singe room.

"Common sense says that if you're looking for a refrigerator, you'd better go to the kitchen,” Machine Learning PhD student Devendra S. Chaplot said in a release. “Classical robotic navigation systems, by contrast, explore a space by building a map showing obstacles. The robot eventually gets to where it needs to go, but the route can be circuitous.”

CMU notes that this isn’t the first attempt to apply semantic navigation to robotics, but previous efforts have relied too heavily on having to memorize where objects were in specific areas, rather than tying an object to where it was likely to be.

More From

  • Apple boots Fortnite from the App Store after Epic adds direct payments

    After its creator Epic Games implemented a workaround to duck Apple's hefty developer fees, Fortnite has vanished from the App Store. The popular game's disappearing act came the same day that Epic added a new direct payment option for in-game currency on mobile, offering an enticing 20% discount for players who pay the company for its virtual V-Bucks rather handing that money to intermediaries Apple or Google. "Currently, when using Apple and Google payment options, Apple and Google collect a 30% fee, and the up to 20% price drop does not apply," Epic wrote in a blog post introducing the new option.

  • Impossible Foods gobbles up another $200 million

    Impossible Foods has raised $200 million more for its meat replacements. The new round was led by Coatue, a technology-focused hedge fund, another New York-based hedge fund, XN, also participated in the round. Since its launch the company has raised $1.5 billion from investors including Mirae Asset Global Investments, Temasek.

  • Instagram wasn't removing photos and direct messages off its servers

    A security researcher was awarded a $6,000 bug bounty payout after he found Instagram retained photos and private direct messages on its servers long after he deleted them. Independent security researcher Saugat Pokharel found that when he downloaded his data from Instagram, a feature it launched in 2018 to comply with new European data rules, his downloaded data contained photos and private messages with other users that he had previously deleted. Instagram said it takes about 90 days for deleted data to be fully removed from its systems.

  • Starting with Michigan, Sidewalk Infrastructure is looking to build roads specifically for autonomous cars

    Sidewalk Infrastructure Partners, which spun out of Alphabet's Sidewalk Labs to fund and develop the next generation of infrastructure, has taken the covers off its first big project -- the launch of a subsidiary called Cavnue to develop roadways for connected and autonomous vehicles. Starting in Michigan, Cavnue will be working with partners including Ford, GM, Argo AI, Arrival, BMW, Honda, Toyota, TuSimple and Waymo on standards to develop the physical and digital infrastructure needed to move connected and autonomous cars out of pilot projects and onto America's highways, freeways, interstates and city streets. The starting point for Cavnue is a 40-mile corridor between downtown Detroit and Ann Arbor, Michigan that will be dedicated to autonomous vehicles.