Autonomous Vehicles or eVTOL: Which Will Be First?
A mile of road will take you a mile. A mile of runway will take you anywhere.
It’s a popular maxim in the aviation community, and one that is relevant more than ever with the development of autonomous vehicles (AVs) and electric vertical takeoff and landing aircraft (called eVTOL and pronounced ee-vee-tol). These new aircraft won’t actually need a full-length runway, just a landing pad, and their simplicity and low operating cost will democratize access to vertical space. Over time, this will change how we move, where we live and, in the very long-term, the layout of cities.
Just a few years ago many thought AVs would hit the market soon, and that flying cars were a thing of sci-fi fantasy. I believe the opposite is true and now see that understanding spreading. Here, I will compare making cars vs. eVTOLs autonomous to show why we likely may have partially and fully autonomous eVTOLs before we have AVs.
I encourage you to read Peter Shannon’s articles on nodal vs linear transportation (point-to-point vs defined paths along roads) and eVTOL technology and their effects on society in Radius Mobility. These will give you an excellent baseline before diving into this more nuanced discussion.
So, bottom line — will everyone have a Doc Brown flying DeLorean in their garage? Will we be able to take off from our driveways and land directly at the office? Unfortunately, no. At least not any time soon. Initially, urban air mobility (UAM) will be much more structured than that. Flights will start with existing helipads and airport infrastructure and as the tech and networks mature, dedicated skyports will be developed. There are also technological constraints, namely limited battery capacity and high aircraft noise that need to be solved before more widespread introduction.
Autonomous cars get a lot of attention (and investment), with companies like Uber, GM, Tesla, and Google’s Waymo striving to be the first to market with a fully autonomous vehicle. Imagine the productivity gain realized when you can hop in a car and knock out some work, watch Netflix, or even take a nap! And surely, it’s easier to build something that drives on the ground than flies through the air, right?
Wrong. The difference lies in the nature of the problem an AV needs to solve vs an eVTOL, the supporting sensing and perception technologies required, and the respective environments in which they will operate. Yes, Tesla has gotten pretty good with autopilot on the highway, and Waymo launched an autonomous ride sharing test program in Phoenix. But these are constrained, ideal conditions. Highways are generally easy driving and well-marked, and Phoenix is known for mild weather. Even so, Tesla instructs drivers to watch the road and be ready to take control at any moment and Waymo has safety drivers for emergencies or should the car encounter an irregular situation and the autopilot disengage.
But it’s not really operations in ideal conditions that give us pause. These cars have trouble with so-called “edge cases” such as snowy roads, construction zones, and erratic behavior of other drivers. For all the rules of the road, driving is an intensely social experience, with an innate understanding of human behavior at the core of remaining safe. So, with humans behind the wheels of other cars for the foreseeable future, will AVs have to successfully interpret things like the subtle hand wave or nod drivers give when deciding who moves first at a stop sign?
The road environment is also chaotic in other ways. An AV must be able to distinguish between a harmless plastic bag blowing across the street, and a rock or an animal. It must anticipate what cyclists or pedestrians are about to do, accommodate school buses and emergency vehicles, and navigate crowded parking lots and a host of other unforeseen events. These add up to a nearly infinite number of contingencies that a computer would need to detect, interpret, and take appropriate action upon almost instantaneously.
Now, it is possible we will achieve a level of autonomy in cars that, while not comprehensive, has real utility. There may be limitations such as geofencing, weather conditions, restriction to paved roads, etc., and AVs will likely operate more conservatively than many drivers, but that would still be useful. In fact, it’s probably pretty similar to most driving people do now, at least in developed nations. While a weakness of AVs is anticipating human behavior, its strengths are 360-degree vigilance and decreased reaction time vs a human driver. Thus, the point at which the combined relative strengths and weaknesses of AVs are safer than a human driver, is the point at which they become viable in the market.
As for a 100% autonomous vehicle with no limitations, capable of traversing all conditions a human driver could navigate- I’m talking off-road, blizzards, Black Friday at Best Buy, what have you, that seems like wishful thinking. Even John Krafick, the CEO of Waymo, has said that a self-driving car with this level of autonomy will never exist. Now, they say to never say “never” in tech, so those are his words not mine!
Until we reach sufficient autonomy, cars will require some human driving and vigilance. If we can’t take our eyes off the road, how beneficial is the autonomy? Autonomy that fails one percent of the time degrades the utility of such a system almost entirely. We see the same in voice recognition- we still have to double check a text dictated via Siri, but in transportation lives hang in the balance.
Let’s turn our attention to flight. eVTOLs will initially be human-piloted and, barring any near-term leaps in battery technology, first iterations may even have hybrid propulsion. Both factors will significantly accelerate the timeline against AVs. Because they will provide such drastic time savings, even with human pilots, these new aircraft will deliver tremendous value. Eventually autonomy will reduce costs and increase safety as skies become more crowded.
The technology to automate eVTOLs largely exists already and has a long history. The first aircraft autopilot was invented by the Sperry Corporation in 1912, a mere 9 years after the Wright brothers’ first flight. Granted, it did little more than hold a heading and pitch, but with short missions and fair-weather flying, that sufficed. Instrument Landing Systems that allow an aircraft to follow a glideslope to a runway in low visibility have been around since the 1930’s and the first autoland systems that enable an aircraft to land itself were introduced in the early 1960’s. Remember, this was analog navigation technology controlling the plane via radio signals, wires, pulleys and hydraulic systems.
Today’s digital systems, GPS navigation and fly-by-wire controls are far more reliable and accurate, allowing the average airline flight to only be manually piloted for 6 minutes, the rest being flown by the autopilot. New single-engine trainer aircraft like the venerable Cessna 172 now come with glass cockpits (screens rather than round dials) and autopilots that are just as advanced as those in airliners, save for autoland. And NASA is already operating modified MQ-9 Reaper drones that sense and avoid air traffic in its UAS Integration Pilot Program.
Cars, however, have not had widespread automation in their comparatively longer lifetime, unless you count cruise control, which wasn’t commercialized until the 1958 models of the Chrysler Imperial, New Yorker and Windsor- 72 years after Karl Benz invented the first car in 1886! Attacking the problem of automating cars really only began in 2007 with the DARPA Urban Challenge, which saw cars with cumbersome hardware bumbling around parking lots at 10 mph attempting to navigate simple orange traffic cone courses.
This is clearly a simpler problem in the air because aircraft systems have far fewer edge cases to interpret- If an aircraft encounters anything in the air, that object poses a danger and must be avoided. It’s a binary decision for a computer to make vs the infinite number of scenarios on the ground. AVs need a lot more information and processing power and the objects that they need to sense are much closer than aircraft are from each other. Traffic Collision Avoidance Systems (TCAS) already exist at varying levels in aviation, and there are a bevy of other technologies at work to increase situational awareness including radar, ADS-B, etc. Further, piloted eVTOLs can slot into today’s ATC system without modification and future air traffic could be digitally directed and deconflicted vs current inefficient voice communication. Even so, future eVTOLs will need onboard sensing capabilities like lidar or radar to replace a pilot’s eyes for rare cases such as non-cooperative traffic or birds.
The reality is that automation in the air won’t be a tech problem; it will be a societal acceptance problem. Remember, elevators had attendants for over 50 years after automated push-button elevators were first installed. With the rapid pace of technological advancement in our world today, I don’t think it will take that long for people to feel comfortable in an autonomous aircraft.
AVs would no doubt make travel cost less from a financial as well as psychological standpoint as time spent in a car would no longer be idle behind the wheel. As the ease of car travel increases, so too will road congestion and with 1.3 million people moving to cities worldwide every week, traffic is only going to get worse. It’s time to start thinking in three dimensions. It’s time to look up.
