My job at Hyperloop One is to lead the teams building our software platforms and safety systems. It’s one of the more challenging tasks I’ve ever taken on, and there’s a lot of future-proofing to do. I just got back from a whirlwind tour of some future-of-transportation conferences with my mind blown, but a healthy amount of skepticism confirmed. The overlap of automotive, robotics, artificial intelligence, sensors and Internet-of-things is one of the most complex spaces in tech right now, but it’s partly my job to interpret what other companies are saying and handicapping the chances of any of this stuff happening. Here’s my take on a few trends:
Conventional wisdom holds that self-driving cars will be vastly safer than human driving. Yes, but. While fully autonomous cars will be on the road in a few years, I do not believe they will be safer than humans on public roads for many more years. There are several reasons for this. The compute, network, and storage resources, as well as the algorithms, are all in their infancy and that creates areas of concern. What about the “school’s out” problem? When the school bell rings, and everyone just pours into the street. Autonomous vehicle processors cannot handle that kind of problem in a quick fashion. The answer in that situation, as it is all too often for autonomous vehicles, is just to stop the car. Humans can edge forward and make decisions for what to do.
One big issue is that autonomous vehicle systems are only at a level of sophistication of drawing 3D blocks around objects. I think it will be crucial for the technology to improve the algorithms so that bodies become more like bodies and less like stick figures in the “eyes” of the computers driving the cars. Given my gaming background and that most technology is built on gaming technology (hardware like Nvidia, software like Havoc, etc.), I feel like autonomous is still in the age of Pong and we need at least Space Invaders tech to make this work. Luckily for Hyperloop One, we don’t have to deal with a heterogeneous environment. We build and monitor the equivalent of the cars and the road and the city and the signals. Our set of variables is far smaller. We can get away with a Pong--for now.
The great debate: Will there be more cars or fewer cars? I have heard arguments for both. On the one hand, more autonomy and more direct-to-destination requests will put more cars on the road. On the other hand, there could be a lot fewer cars if more people take up ride-sharing. A compelling argument for there being way fewer cars on the road came from this recent RethinkX report that estimates a drop in at the number of passenger vehicles on American roads from 247 million in 2020 to 44 million in 2030. As a person who is not a millennial, I love my private space. I won’t be contributing to that trend. Even though uberPOOL and Lyft Line are fast-growing services, there are a lot more of us non-millennials not sharing, so I think there will be more cars.
You can plan on autonomous cars impacting the urban infrastructure in unplanned ways. An autonomous car may drive down the street in the same location over and over again and wear down the road in one place. Self-driving electric cars will likely be traveling many more miles than traditional vehicles, as they will never need to park until they need a recharge. If AVs do excessive damage to our roads, we’re going to need more creative policies to pay for any additional upkeep. Chris Urmson, CEO of Aurora Innovation and former director for self-driving cars at Google, suggests charging a user fee per mile driven. Charge 2 cents per mile and suddenly the three trillion miles driven in the U.S. amounts to $60 Billion in revenue per year.
Also, there are many random unanswered questions about seat belts. What happens if you unbuckle? Does the car stop? Does someone come on the PA? What if it's a shared vehicle? How do you do maintenance? Who pays for the electricity? One solution for that latter problem could be turning roads themselves into refueling stations. Israel is currently testing technology that would allow electric vehicles to recharge while they drive using inductive charging. The real value proposition is to make wireless charging a two-way street and capture back the energy generated from braking. The technology is on its way, but the cost is critical here.
Cities in the not-so-distant future will be swarming with connected vehicle “clouds,” but in between those metro areas, we are facing an information and service gap. That’s where we come in. Hyperloop One plans to build the "range extender" for autonomous urban mobility. Let’s say you’re in your office in Berlin and it’s your wife’s birthday and you want to take her to this great new gastropub in Hamburg for dinner. Six o’clock rolls around, and you tap your phone, an Uber picks you up at your office, drives home to get your spouse, and off you both go to the Hauptbahnhof Hyperloop One portal (our term for the station). Your self-driving Uber has already contacted the portal via Hyperloop’s “autonomous vehicle interface” or AVI, and just as you arrive at the portal, your Hyperloop pod is waiting for you or, better yet, your Uber slips right into the Hyperloop itself. Security and authentication and payment are all settled in the backend via the AVI. Your pod accelerates gradually to airline speed, and you arrive at the Hamburg portal in 20 minutes. Your Uber drives right out of the Hyperloop portal and straight to your dinner. It’s direct, autonomous, ultrafast intercity travel and the doors only open twice, once to let you in and once to let you out. It’s the closest thing to teleportation or a private jet experience for the masses.
That’s the long-term vision. We obviously have a ton of engineering work to do. Our plans for the Hyperloop AVI and platform also incorporate other data services such as vehicle-cloud supply/demand forecasting, rebalancing of connected vehicles between cities (imagine Ubers or Lyfts taking themselves between Chicago and St. Louis when there’s a big rock concert or baseball game), pricing information sharing and emergency/urgent care. As FedEx taught all of us, the information around the package is as important as the package itself. Same goes with autonomous vehicles and their passengers.
I think what’s going on around autonomous vertical takeoff and landing (VTOL) vehicles is pretty cool. The vehicles are in development now--Uber is targeting 2020 for its first passenger tests--but I think there’s a growing consensus that they won’t get past regulatory approval for passenger flights at least 10 to 15 years. Package delivery by drones will come much sooner. Check out this concept video UPS is showing about how drones might work launched from the top of their trucks to extend a driver’s reach per trip. Amazon Prime Air has already begun limited drone delivery trials in Cambridge, England. DHL has run three tests for its parcelcopter in Germany.
For passenger service, the issue is safety—things fall out of the sky at high speed—and the need to coordinate a lot of regulatory and software systems among the airlines, consumer drone manufacturers, and the new autonomous VTOL systems makers. I think autonomous ground-based vehicles will get full regulatory approval long before passenger VTOLs will be able to fly in the same airspace as everything else.
Transportation and urban planners dream of connecting their mobility grids at strategic hubs in the center of town. Transit systems lose roughly 30% of ridership any time passengers are forced to switch modes. That puts a premium on smart station and multimodal hub design to achieve more fluid movement. The complexity will go up once we introduce AVs to the mix of human-driven cars, commuter rail and metros, bikes, Hyperloops, and everything else can come together as a transfer point to create a better passenger experience.
Multimodal stations aren’t new — think Penn Station in N.Y., Union Station in D.C., or San Francisco’s new Transbay Terminal, which will connect 11 different Bay Area systems. New transit hubs are emerging as economic development opportunities, as well, to stand alone as attractions during off-peak hours. The Oculus, pictured above, has sparked controversy for its price tag but is a prime example of how cities are integrating multimodal centers into city design.
Why do we need bottleneck intersections if vehicles are talking to other vehicles or the infrastructure (lights, road)? AV networks can accelerate and decelerate cars as they head into an intersection to keep everything flowing in both directions continuously and without collisions, cutting down on pollution and delay. A group from MIT, the Swiss Institute of Technology and the Italian National Research Council came up with an idea for a slot-based approach called Light Traffic. Their system would use sensors to keep driverless cars at a safe distance from each other and put the cars in virtual slots as they approach the intersection.