Digging into key takeaways from our 2019 Robotics+AI Sessions Event

Extra Crunch offers members the opportunity to tune into conference calls led and moderated by the TechCrunch writers you read every day. This week, TechCrunch’s Brian Heater and Lucas Matney shared their key takeaways from our Robotics+AI Sessions event at UC Berkeley last week.

The event was filled with panels, demos and intimate discussions with key robotics and deep learning founders, executives and technologists. Brian and Lucas discuss which companies excited them most, as well as which verticals have the most exciting growth prospects in the robotics world.

“This is the second [robotics event] in a row that was done at Berkeley where people really know the events; they respect it, they trust it and we’re able to get really, I would say far and away the top names in robotics. It was honestly a room full of all-stars.

I think our Disrupt events are definitely skewed towards investors and entrepreneurs that may be fresh off getting some seed or Series A cash so they can drop some money on a big ticket item. But here it’s cool because there are so many students. robotics founders and a lot of wide-eyed people wandering from the student union grabbing a pass and coming in. So it’s a cool different level of energy that I think we’re used to.

And I’ll say that this is the key way in which we’ve been able to recruit some of the really big people like why we keep getting Boston Dynamics back to the event, who generally are very secretive.”

Brian and Lucas dive deeper into how several of the major robotics companies and technologies have evolved over time, and also dig into the key patterns and best practices seen in successful robotics startups.

For access to the full transcription and the call audio, and for the opportunity to participate in future conference calls, become a member of Extra Crunch. Learn more and try it for free. 


Fantasmo pivots to scooter cameras that keep them off sidewalks

GPS is too inaccurate to tell if a scooter is being driven or parked in off-limits areas. But as scooter startups compete for permits from city governments, they need a way to prove their riders play by the rules. That’s where Fantasmo’s new scooter positioning camera comes in.

The augmented reality mapping startup had been building the Camera Positioning Standard to give self-driving cars, robots and AR games a dynamically updated understanding of the real world around them. But now Fantasmo is focusing on the urgent use case of scooter accountability.

Its camera attaches to personal electric vehicles, captures video and matches that against Fantasmo’s map to reliably identify if a scooter is being illegally ridden on the sidewalk or parked in the middle of the walkway. Scooter companies could make their vehicles beep and slowly lose acceleration where not allowed, issue fines for parking in the wrong spot, notify redistribution teams to move errant vehicles or ban riders who consistently break their terms.

The tech could even make maps of available scooters more precise so you’re not wandering around searching. And scooter companies could use Fantasmo’s data to demonstrate that their riders are the most respectful.

“Scooters are under threat unless they find ways to work with cities to prevent sidewalk riding and make sure they’re parked in places the cities deem appropriate. 2D image capture can be leveraged to build out semantic, 3D maps of cities and provide a hyper-accurate position of the scooter,” says Fantasmo co-founder Jameson Detweiler. “So-called visual positioning is more precise than GPS and has centimeter level accuracy in dense urban environments — a notoriously bad environment for GPS. Visual positioning is accurate enough that a scooter can know when it is in a prohibited zone even if the zone is only as wide as a sidewalk.”

You can see in the video below how GPS can’t tell the difference, while Fantasmo shows green icons when the scooter is on the street and red ones when on sidewalks.

Originally founded in 2014 to build AR games, Fantasmo was started by Detweiler, who’d previously built startup website builder LaunchRock, and electrical engineering PhD Dr. Ryan Measel. Fantasmo has raised $2.2 million in funding, led by TenOneTen Ventures, to build decentralized 3D maps of the world. Instead of expensive LIDAR sensors like for autonomous vehicles, a simple 2D camera with the right software is sufficient for positioning.

So why wouldn’t scooter companies just launch their own camera systems? Well, beyond Fantasmo’s specialized expertise from years working on AR positioning, it benefits from network effect. Each client from across industry verticals contributes data they collect to Fantasmo’s collaborative maps. That means if construction or an event changes a street’s layout, the first Fantasmo camera that comes across it updates everyone else’s maps. An individual mobility startup might end up with less accurate maps while wasting resources far outside their core purpose. Developers and personal vehicle companies that want to work with Fantasmo can apply for beta access on its website.

The vision is to build “a next-generation Open Street Map that gets all the inputs to work together,” Detweiler explains. “Eventually you’ll have self-driving scooters to do redistribution,” he says, rather than having humans load them in trucks and place them where they’ll get rented next. Without super accurate maps, the idea of passenger-less scooters rampaging through cities is terrifying. “There’s definitely a horror movie or three in that concept right there.”

If a more open AR map like Fantasmo’s doesn’t win, we could end up with a tech giant like Google hoarding this data. “I think the crowd of all these devices will be more powerful,” Detweiler concludes. “It might take time, but that network effect would be hard to beat.