Saturday, February 26, 2022

Post 26: Trends Part 5 - The Prognosis for Autonomous Vehicles

 I started this series on future trends (see Post 22) by identifying three overarching changes to our society that I thought would dramatically reshape urban living and the physical structure of our cities.  I referred to these changes as “mega trends” and I have referenced them in all of the subsequent posts in this series which have looked in greater detail at the spin-offs from these three mega trends.  This post is going to examine another future trend – autonomous vehicles – that is directly related to all three of the mega trends in almost equal measure.

The age of ideas and the power of scale.  This first mega trend might seem an odd fit for a discussion of autonomous vehicles (AVs).  After all, an autonomous vehicle is clearly a thing not an idea, and it must be manufactured in a traditional way which means it won’t scale the way an idea or a piece of software can scale.  However, viewing the trend to AVs as primarily about things that are manufactured misses the most important point – that being the shift to transportation as a service.  Think of Uber in comparison to a traditional taxi company.  Both move people from place to place, but Uber employs no drivers and owns no taxi cabs.  The genius of Uber is the idea of a computer platform that provides the service of linking a vehicle owner/driver with passengers.  

I am convinced that AVs will follow the same economic model.  While some end users might own their own AV, I think that for the next 10 years the vast majority will be owned by transportation-as-a-service companies that you will contract with (via an app) to move you or your package to the desired destination.  As for scale, there are millions of people and things that move from place to place every day giving this idea the ability to scale far faster than the actual building of autonomous vehicles.

Climate change.  I think the vast majority of AVs will end up being electric (aside perhaps from over-the-road trucks).  This means that as the sources of electricity become greener, so will the AVs that power our transportation system.  Perhaps more significant, however, is the concept of embedded resources and energy.  To begin with, it is important to understand that the typical car is parked 95 percent of the time. [1]  That means the materials and energy used to build each car are horribly underutilized.  To justify the higher cost of AVs, transportation-as-a-service companies will have to keep their vehicles busy for a much higher percentage of the day.  This higher utilization means that there should be fewer vehicles required per passenger mile traveled (or package mile traveled) which in turn means that AVs should reduce the overall energy needed to build our transportation system.

Waymo autonomous taxis
Source:  Reuters
Artificial intelligence.  Designing a computer system that makes a vehicle go, turn and stop is relatively straightforward.  Designing a computer system that can do those same things in a dynamic environment with hundreds of other cars, trucks, buses, bicycles, pedestrians, construction workers, stop lights, traffic signs, and miscellaneous debris is a much harder task.  So hard, in fact, that prediction after prediction of fully autonomous vehicles being ready “in a year or two” have turned out to be overly optimistic.  So much so that many in the industry are no longer specifying an exact date and a few are questioning whether fully autonomous driving will ever be possible.  It has certainly been the sternest test that the artificial intelligence community has had to deal with, but I’m confident it will eventually be mastered and I think it will expand and accelerate the range of tasks that artificial intelligence will be able to tackle.  But in the near term, I think we need to rein in our expectations and accept that autonomous vehicle technology will move forward in steps, not giant leaps.

The remainder of this post will focus on the three autonomous vehicle types that I think will have the biggest impact on our society and our cities over the next 10 years.  In each case, I think the technology is nearly ready and the economic rationale is compelling.  There are, however, a few important caveats.  First, state and federal legislators need to amend the motor vehicle laws to account for AVs.  This is not a trivial task but it can be done, and it needs to be completed as consistently as possible from state to state so that AV companies are not dealing with a mishmash of regulations.  Second, insurance companies need to figure out how to insure AVs and how to allocate the liability for damages when the inevitable collisions occur between AVs and cars driven by humans.  Again, this is not a trivial task.  

Finally, the general public and the media need to stop expecting autonomous vehicles to be perfect.  I think the standard we should shoot for is “significantly better than the average human driver” but there will still be collisions that result in property damage, injuries and deaths.  As long as all of those things decline over time, the shift to autonomous vehicles will be good for our society.  Unfortunately, we currently seem to treat each AV accident as a catastrophe that calls into question whether autonomous vehicles should ever be allowed.  This reaction stands in contrast to our society’s seeming indifference to the nearly 40,000 people who die each year and the 4 million people who are injured in accidents involving human drivers – many of whom are drunk, high, distracted by their phones, or emotionally distraught. 

Long-Haul Trucking

The recent supply chain disruptions have underscored the importance of trucking to the American economy.  Furthermore, the trucking industry frequently pushes out the story that our country needs 80,000 new truckers to fill open positions.  The more accurate story is that there are plenty of people trained as truck drivers but they have moved on to other occupations because of low pay and terrible working conditions. [2]  Over-the-road truck driving involves long hours of lonely, boring work.  So much so, in fact, that the industry has a significant drug and alcohol abuse problem in addition to its driver retention problem.

So it should come as no surprise that several companies are specifically developing autonomous vehicles and systems for long-haul trucking.  In fact, autonomous trucking start-up TuSimple recently completed an 80-mile test run from a railyard in Tucson to a distribution center in Phoenix using a completely automated truck with no human in the vehicle and no human intervention of any kind. [3]  At least initially, autonomous trucking is likely to be terminal to terminal, or what the industry refers to as the “middle mile.”  The “first mile” and “last mile” segments that consist largely of city driving will probably still be done with human drivers.  Even the autonomous portions of each trip will be monitored remotely by humans and supported along the way by humans that can perform whatever troubleshooting might be required, but the amount of manpower required per vehicle mile will drop significantly.

Autonomous truck by TuSimple

To begin with, states will probably authorize autonomous trucking primarily on limited access highways such as the interstate system.  Autonomous trucking companies will develop a network of routes, terminals and support stations that will eventually crisscross the nation.  Limiting autonomous trucking to major highways has the distinct advantage of drastically simplifying the amount of “learning” that the artificial intelligence systems will have to do and it allows a computer to handle the portion of the trip where human perception and intelligence add the least value.  Driving a semi hundreds of miles on interstate highways is largely boring work that leads to drivers being distracted or sleepy – something that never happens to a computer.  In fact, autonomous trucks may operate in the middle of the night as much as they do in the middle of the day since traffic will be lighter and sleep is not an issue.

Finally, there is a compelling financial argument to be made for autonomous trucking.  Not only do labor costs go down, but trucking efficiency goes up.  A typical semi can travel 1,500 to 2,000 miles on a full tank of diesel fuel and autonomous driving systems would allow that to happen without any stops (potentially 24 hours of continuous driving).  Human drivers, on the other hand, are supposed to be limited to 11 hours of driving during a 14-hour driving window which is supposed to be followed by 10 hours of being off-duty.  This means that a truck driven autonomously could have roughly double the operational efficiency of a truck driven by humans.  As systems are perfected, fuel costs and insurance costs are also likely to decline.

Hundreds of thousands of miles of autonomous driving have already been completed (in trucks with a human driver standing by to take over if needed).  The sensors and driving systems for fully autonomous operation are largely ready to be rolled out, and the logistical details of the initial routes are being finalized.  Assuming that the legislative and insurance issues are worked out, I expect production runs to begin within two to three years, a coast-to-coast network to be operational within five years, and 30 to 50 percent of long-haul trucking volume to be autonomous within 10 years.

I don’t expect that autonomous trucking will have a significant impact on most cities since the bulk of the action will happen in the rural areas between cities rather than in cities.  Still, cities should prepare for intermodal trucking facilities on the periphery of urban areas to grow in both number and size.

Autonomous Shuttle Buses

In most midwestern cities, buses are the only form of mass transit.  Even in large cities with subways, elevated trains or light rail, buses are still a vital (if less glamorous) part of the network. They are flexible and effective, with relatively low capital costs and virtually no special infrastructure costs. But buses are relatively labor intensive, with labor costs accounting for roughly 40 percent of operating costs – a number that will probably increase over time.  Consequently, autonomously operated buses are under development by a variety of companies worldwide and are already in use (at least on a test basis) in several different cities in China and Europe.

Autonomous bus being tested in Cambridge, England

Autonomous buses on a fixed route would have the advantage, of course, of having to “learn” far less than a vehicle that is expected to be able to go anywhere.  Routes could be mapped at high levels of precision and buses could communicate with key pieces of infrastructure such as bus stops and traffic lights.  Knowing the precise location and speed of each bus would enable smartphone apps that would tell users when the bus will be at a specific stop, thus minimizing wait time.

The potentially game-changing advantage, however, is that the reduction in labor costs might enable a substantial increase in the number of buses on a given route.  Bus transit is currently perceived as inconvenient largely because the gaps between buses are often 30 to 60 minutes.  If that could be reduced to 10 to 15 minutes I think that usage would increase dramatically.  In fact, I have intentionally used the term “autonomous shuttle bus” to suggest smaller buses (perhaps 15 to 30 passengers) on a shorter route (perhaps 5 to 10 miles).  In my opinion, big buses running long bus routes focused on people commuting from suburban residential areas to the city center is an outmoded approach that reinforces car-centric urban design.  Frequent, localized service would do more to get bus usage viewed as a routine transportation option for more people and a wider variety of trips.

As with long-haul trucks, the technology for autonomous buses is largely in place now.  Several manufacturers are testing a variety of purpose-built designs that are fully electric and incorporate a full suite of sensors and cameras for both guidance and traffic safety.  Buses, however, have the additional concern of being able to monitor the behavior and safety of the passengers.  Most of the pilot projects to-date have included an on-board human monitor, both to reassure passengers in case something goes wrong and to make sure that passengers don’t misbehave.  On-board camera systems, artificial intelligence and remote monitoring could fulfill much of the same functionality, but would need to be backed up with human security officers and service crews in case some type of intervention was required.  This would not be too dissimilar to current subway and light rail systems which currently do not have transit personnel present in each of the passenger cars.

Buses also have the issue of fare collection.  I have previously argued that transit service should be free to riders since it would provide an important benefit to the economically disadvantaged segments of our society and since enticing people to take the bus is probably one of the more affordable ways to reduce traffic congestion.  Assuming that not everyone agrees with me, there would need to be some type of automated fare collection, probably using e-tickets on your phone or RFID tokens purchased from vending machines at each stop.  Again, a process that needs to be worked out but not much of a departure from technology that is already available.  My forecast is that autonomous buses will move out of the pilot program phase into regular usage within 5 years, and be fairly commonplace within 10 to 15 years.

Robo Taxis

This is the currently favored terminology for an Uber-type service in an autonomous vehicle.  The Google subsidiary of Waymo is perhaps the best known provider of this type of vehicle and service, but there are several well-funded corporate players in this arena.  In the U.S. the primary competitors to Waymo are General Motors/Cruise, Ford/Argo/Lyft, and Amazon/Zoox.  Europe and Asia have their own list of giant corporations chasing this market.  All together, billions of dollars are being spent each year in the effort to find a safe, attractive and cost effective solution to the illusive dream of an on-demand passenger service that doesn’t require a driver.

The Cruise Origin Robo Taxi

The reason for all of this investment, of course, is the potential for huge profits in the future.  Uber, Lyft and others have proven that the demand for this service is real.  At its pre-pandemic peak, Uber alone had over 110 million active users and brought in revenue (from the mobility service only) of more than $10 billion.  The trick is to turn revenue into profit – something Uber has yet to accomplish.  Many feel that the biggest stumbling block is the portion of the revenue stream that Uber (et al) have to share with the driver.  Precise data is hard to come by, but some studies have estimated that the driver gets roughly half (or more) of ride revenue. [4]  Capturing the driver’s half by operating autonomously is the key to the hoped for profit stream.

In a recent interview, Bill Nash the CFO of Cruise stated that their goal was to keep costs between $1.00 and $1.50 per mile, well below the rates typically charged in major cities by Uber and Lyft.  He is hoping to have “tens of thousands” of vehicles in their fleet starting in the next couple of years.  Cruise will be using a GM-built vehicle called Origin and is currently in negotiations to start service on a test basis in San Francisco. [5]

Although early pilot projects have used traditional cars converted with sensors and automation to operate autonomously, production robo taxis are likely to be custom designed vehicles with a distinctive look.  Current prototypes tend to be about the size of a mid-sized sedan but with a boxy, van-like shape and a single, large door.  With no steering wheel, dashboard, or other operating controls, robo taxis will have a lot of space for passenger seating – enough to comfortably accommodate 4 to 6 people.  Finally, robo taxis are almost assuredly going to be all electric primarily because of simplicity and durability.  Electric cars are far easier to maintain which will be crucial for a vehicle designed to last for 500,000 miles or more.

Users will summon a robo taxi with an app on their phone, much like scheduling an Uber ride today.  The app will handle pick-up/drop-off points, timing and payment using algorithms that constantly balance supply and demand.  Users will probably be sent a QR code or similar validation token to gain access to the vehicle, although facial recognition or other biological markers might eventually make that unnecessary.  Since seating will be relatively spacious, users will have the option to share their ride to reduce the cost, provided that trip length is not significantly increased.

The primary challenge to making all of this work is the complexity of designing a system that can pick up anywhere, drive through a wide variety of street configurations and traffic conditions, and then drop off anywhere.  This is exactly what is being tested now in Chandler, Arizona, Las Vegas, and San Francisco.  Progress is being made but I don’t expect robo taxis to move beyond the pilot project level into routine operation for several more years, and I don’t think they will be commonplace (particularly in the midwest) for at least 10 years.  The technical difficulties are likely to limit robo taxis to predefined areas that have well mapped street configurations, and financial concerns are likely to focus robo taxis in areas that have relatively high demand.  Plus, the legislative and insurance issues discussed earlier will need to be fully resolved which will be more difficult than for the more limited operational scope of city-to-city trucking and fixed bus routes.

Where are robo taxis going to show up initially?  Well first, it will be in states that are willing to accept the risks of autonomous vehicles on their streets and make the appropriate modifications to motor vehicle laws.  The conservative nature of many midwestern states may push robo taxis elsewhere.  Second, it will be in moderately dense, mixed use areas where there is demand for relatively short trips at least 18 hours per day.  This means that traditional downtowns with a high concentration of office space are not likely to be good candidates because offices tend to generate trips around the morning and evening rush hours but not during the rest of the day.  Third, it will be in areas where parking is inconvenient and/or pricey.  Initially, robo taxis will be expensive to build (current estimated cost around $150,000) which means fares will not be particularly cheap.  Robo taxis will be more cost competitive where every trip in a private car includes a $20 or $30 parking expense.

Perhaps surprisingly, I think that robo taxis might flourish in areas that are well served by mass transit.  To begin with, I think that transit tends to serve a user group that is much more price conscious so they won’t always compete head to head.  Second, I think that many trips may be a combination of transit for the bulk of the trip and robo taxis for the “last mile” segment.

Ubiquity Delayed

Quite a few futurists have predicted that once the kinks of autonomous driving are worked out the advantages will be so overwhelming that virtually every driver-dependent car will be swapped out for an autonomous one within just a few years.  The rosy scenarios include swarms of autonomous vehicles using vehicle-to-vehicle and vehicle-to-infrastructure communication so advanced that both traffic jams and accidents become a thing of the past.  Traffic lanes can be narrowed (because autonomous cars will be highly precise) which will create more room for pedestrians and bicyclists.  Parking lots will shrink dramatically allowing more open space or developable land.  Urban nirvana, here we come.

Whoa Nelly.  First, the cost of the early generations of autonomous vehicles will be high enough that their appeal will be mainly for commercial purposes or the luxury car market.  Costs will eventually drop but they won’t reach the “average Joe” price range any time soon.  Second, there are a lot of specialty vehicle types and a lot of skeptical drivers who will be resistant to change.  The reality is that traffic will be a mixture of driver-controlled and autonomous vehicles for a long time, which means that reduced lane widths will not be feasible any time soon and traffic jams caused by impatient and irrational humans will remain a common occurrence.

I do think it is plausible that in 10 to 15 years private car ownership will be significantly reduced, particularly in relatively “close in” locations.  Households that now own multiple cars may drop to just one as it becomes easier to substitute robo taxis or other forms of transportation for trips that now require a private car, and as it becomes more common to work from home and have much of the routine shopping done online.  

I also think that private cars will continue to develop more and more advanced driver assistance systems that will be able to handle driving responsibilities in more and more situations, but most will fall short of true autonomous driving capability.  Steering wheels and brake pedals will be slow to vanish because developing a fully autonomous system that can handle 100 percent of all automotive trips is harder than you might think.

Urban Impacts

There actually isn’t a great deal that cities need to do to get ready for a future with autonomous vehicles.  Cities can lobby in support of the needed legislative changes and look for opportunities for local pilot projects.  Yes, being an early adopter comes with risks but being a late adopter carries risks of its own.  Urban economies are being driven by young professionals looking for cool places to live and work.  Being perceived as technologically backward might well be a death blow to any hopes of playing a significant role in the information age economy.

Second, cities need to promote zoning districts that allow mixed land uses at moderate to high densities (for example, an FAR above 2).  Midwestern cities, in particular, tend to use zoning districts that feature homogeneous uses in low-rise buildings surrounded by open yards and parking lots.  This will not be a favorable environment for either autonomous buses or robo taxis.  Dense, mixed-use development does not need to be everywhere – in fact, cities will be better off concentrating such development into reasonably compact corridors or districts – but there must be a critical mass of trip demand that can support an investment in autonomous travel.

Third, cities need to rethink parking ratios.  Parking requirements in most cities are already too high, and work-from-home and online shopping is causing parking demand to drop.  As AVs become more common, parking demand will drop further.  Cities need to plan how to harvest this excess parking and convert it to a more productive use.

Finally, cities should begin doing some outside-the-box thinking about how their street rights-of-way are being used.  Most cities assume that everything between the curbs should be either parking or traffic lanes for cars.  However, autonomous vehicles – along with the now common UPS and FedEx trucks – are going to need at least one drop-off and pick-up location on nearly every block face.  At the same time, pedestrians, bicyclists and scooter riders will be demanding space of their own.  The result will be that street design will become far more complicated and the goal of maximizing traffic throughput will have to take a backseat to other considerations.

Thoughts?  As always, share your thoughts and ideas by leaving a comment below or sending me an email at  Want to be notified whenever I add a new posting?  Send me an email with your name and email address.


  1. Paul Barter; “Cars are parked 95% of the time.”  Let’s check!; February 2013;

  2. Alana Semuels; “The Truck Driver Shortage Doesn’t Exist”; Time; November, 2021;

  3. Rebecca Bellan; “TuSimple completes its first driverless autonomous truck runs on public roads”;TechCrunch; December, 2021;

  4. David Mamaril Horowitz; “As rideshare prices skyrocket, Uber and Lyft take a bigger piece of riders’ payments”; Mission Local; July 2021;

  5. Nina Trentmann; “Cruise CFO Hammers Out Robotaxi Venture’s Cost Structure Ahead of Charging Customers”; Wall Street Journal; February, 2022;

No comments:

Post a Comment