View profile

AutonoMusings - AI Drone Racing, Anyone?

AutonoMusings
AutonoMusings - AI Drone Racing, Anyone?
By Joshua Ouellette • Issue #18 • View online
Welcome to AutonoMusings!
This week we got confirmation that Alpha<InsertTopicName> is the new default shorthand for AI projects in the wake of AlphaGo. Ouster pushed an update to their LIDARS unlocking new super powers. Lastly, the data shows AI might truly make for safer drivers than humans, but will SkyRyse show that AI can make for safer pilots, too?

Autonomous Flight
Working with the Drone Racing League and NVIDIA, Lockheed Martin announced a challenge this week called AlphaPilot, in which participants will create AI/ML framework to fly drones autonomously through several racing courses better than human-piloted drones. No navigational preprogramming allowed!
If any of you would like to form or join a team, email josh@raven-ops.com and I’ll share the list with everyone interested!
www.cnet.com  •  Share
SkyRyse plans to launch AV air ambulance service (with pilots) in January
Silicon Valley startup SkyRyse is testing a fleet of small helicopters equipped with AV technology in Tracy, CA, in anticipation of eventually providing pilotless service. Between the launch of the service in January 2019 and to going pilotless, there will be a long period of autonomous flying with a safety pilot on board.
SkyRyse’s approach makes sense to me. As an ex-professional pilot – there’s no such thing as an ex-pilot – I’ve always believed separating “autonomous flight” from “newfangled electric-drive multi-rotor contraption” is the faster way to FAA approval.
www.nytimes.com  •  Share
Sensor Tech
Ouster updates its LIDAR to provide low-light panoramic photos
San Francisco-based Ouster has updated its OS-1 LIDAR to enable it to simultaneously function as a camera, providing 2D panoramic imaging capabilities. This has several implications:
  • The imaging takes place in near-infrared with strong low-light performance, making it a distinct addition to the vehicle’s sensing options.
  • Having the imaging and ranging functions performed by a single unit eliminates the need for cross-sensor calibrations.
  • The alignment of the coordinate frames makes it easier to include combined image/depth data in existing computer vision models.
CEO Angus Pacala has written before about Ouster’s strategy for simple, managed over-the-air firmware updates. It will be interesting to see what other improvements they deploy to their existing products.
arstechnica.com  •  Share
Autonomous Vehicles
Who’s to blame for most accidents involving AVs in California?
In the vast majority of California car accidents involving a self-driving vehicle in autonomous mode, humans were at fault. How vast, you rightly ask? Of 62 accidents reported since 2014, all but one accident is attributed to humans — that’s a staggering 98.4 percent of accidents.
Some of this may be that people have not adapted to how the AVs around them will drive, but such lopsided numbers speak well for the promise of self-driving cars making our roads safer.
www.axios.com  •  Share
Drive.AI brings AVs to Arlington, Texas
A suburb of Dallas will put three self-driving vans on the road in October. The city of Arlington is collaborating with Drive.ai for a one-year pilot program that will shuttle passengers in and around the region’s entertainment district. A $343,000 federal grant is covering 79 percent of the project’s initial costs.
At first I was confused by the large flat panels on the van. It turns out those are displays for communicating with pedestrians and other vehicles. It might be a bit dorky, but an “I’m waiting for you to cross” sign makes a lot of sense.
www.arlington-tx.gov  •  Share
Mike sits on the boards of both Fiat Chrysler and autonomous vehicle company Aurora, giving him a unique view into the space. Some quotes:
Since each piece [of the autonomous vehicle stack] is evolving rapidly but at a different pace and because it must be loaded onto a car with a fifteen-year life cycle, the readiness is not close to being where it needs to be.
In places such as Phoenix [with simple driving environments], I would say that timeframe [to deployment] is 18 months to roughly three years, in Boston, I would say at least four years, and places such as India are even further out than Boston.
Disagree with Mike’s predictions? Send us an email at contributions@autonomusings.com!
www.forbes.com  •  Share
Did you enjoy this issue?
Joshua Ouellette

This newsletter was made with love by RavenOps in San Francisco

If you don't want these updates anymore, please unsubscribe here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue