Tuesday, October 9, 2018

Driverless Vehicles: Two Radically Different Visions

I concur with Bob Poole's commentary published as follows:

Surface Transportation Innovations
By Robert W. Poole, Jr.
Searle Freedom Trust Transportation Fellow and Director of Transportation Policy
October 2018

There is no question that personal transportation will undergo significant changes in coming decades. Three such changes will be the advent of affordable electric vehicles, fully autonomous vehicles, and mobility as a service (MaaS) in which people opt to rely on shared vehicles rather than individually owned vehicles. These are separate changes, which may well arrive on different time scales and with different degrees of market penetration.
Several times in recent months, various people have sent me a report that links all three together via a dramatic scenario. The report comes from RethinkX and is called “Rethinking Transportation 2020-2030,” released in May 2017. Its headline claims include the following:
  • Fully autonomous vehicles (presumably SAE Level 5) will achieve regulatory approval and be on the market in 2020.
  • By 2030, those AVs will provide 95% of all [surface] passenger miles of travel.
  • Those 95% will all be in shared vehicles (Mobility as a Service), rather than in personally owned AVs.
  • These AVs will all be electric, and will last 500,000 to 700,000 miles on their initial battery pack.
These assumptions are shared by virtually no one actually working on AVs, whether at technology companies or traditional auto companies. The past year has seen a growing number of articles explaining that full autonomy (on all kinds of roads, in all kinds of weather, etc.) is turning out to be a much harder problem than many researchers expected. Most expect gradual introduction of AV features in the next decade, with full Level 5 not being likely until at least 2035 or beyond.

As I wrote in a recent column for Public Works Financing, there is no necessary connection between electric propulsion and autonomy: neither one depends on the other. The current generation of EVs costs nearly twice as much as comparable non-EV vehicles, seriously limiting mass-market appeal.

Likewise, as of now, autonomy itself requires a large array of costly sensors and very complex artificial intelligence software, Hence, RethinkX’s idea that electric AVs will be cheaper than conventional cars by 2020 looks to me like a pipe dream. In addition, the idea that the original battery pack will last 500,000 to 700,000 miles (a key to Rethink’s lower ownership cost estimate) is unproven. (The Toyota Prius battery pack has a 10-year or 150,000-mile warranty, while the Tesla Model 3 warranty is for 8 years of 100,000 miles.)

A far more realistic assessment of future mobility was released in May 2018 by S&P Global Ratings, “The Road Ahead for Autonomous Vehicles.” S&P’s analysts conclude that “mass adoption of driverless autonomous vehicles (AVs) [is] still decades away.” By contrast, they expect a faster penetration rate of electric vehicles (EVs), especially if there continue to be government “incentives” (subsidies) for those purchasing them. (S&P’s EV projections are somewhat exaggerated by including plug-in hybrids.)

S&P developed three scenarios (low/medium/high) for AV penetration, depending on a array of assumptions about technology, the price premium over conventional cars, extent of government “incentives,” growth in ride-sharing/ride-hailing (Mobility as a Service), etc. For the 2020 to 2030 period, the fraction of AVs in the total light-vehicle fleet by 2030 is projected at <1 2="" and="" av="" be="" fleet="" fraction="" high.="" in="" low="" medium="" of="" p="" phase="" scenario="" the="" vehicle="" would="">
I find the assumptions underlying the three scenarios to be reasonable, and a number of implications for highways and travel emerge. First, even in the high (“disruptive”) scenario, only 35% of the light vehicle fleet will be AVs by 2040. So that means our roadways and highways are going to have to deal with a mixed fleet for many decades. That is far different from popular media visions of a near-term all-AV future. Second, S&P suggests that the early impacts of Level 5 AVs will be felt most by transit agencies and parking enterprises. Between 2020 and 2030, S&P expects an increase in urban traffic congestion, due partly to the continued growth of ride-hailing. (Incidentally, a new paper by Alejandro Henao and Wesley E. Marshall, “The Impact of Ride-Hailing on Vehicle Miles Traveled,” projects that “ride-hailing leads to approximately 83.5% more VMT” than would have existed had ride-hailing not emerged.) As connected AV market penetration increases beyond 2030, S&P expects “lane capacity could increase by 5% to 7% by 2030-2035 [due to] an increase in platooning.” That would partially offset the impact on highways from increased VMT due to ride-hailing and increased personal travel by those who cannot drive today (very old, very young, and disabled).

These are still early days for EVs, AVs, and MaaS. The sober analysis from S&P is a far better guide to thinking about the implications of these developments than the blue-sky vision of RethinkX