Planning for Autonomous Driving

In the United States over the next ten years, governments may spend some $1.5 trillion on their roadways, consumers may purchase vehicles worth nearly $3 trillion, property owners may develop millions of acres of rural land, and the US Postal Service may drive its cars and trucks approximately 12 billion miles (with FedEx alone adding 10 billion miles more). How might these massive numbers—and others like them—be harnessed to smooth the deployment of self-driving vehicle technologies?

While dramatic changes to infrastructure, vehicles, or legal regimes may be warranted, this blog post instead considers some low-hanging fruit: modest steps that public and private actors might take now to reduce the costs of adopting and adapting to these emerging technologies down the road.

These actors would do well to consider the potential for increased vehicle automation when they make decisions about infrastructure and procurement projects. Numerous formal mechanisms in the public sector already provide analogies (or, in some cases, opportunities) for such an exercise, including administrative delegation, cost-benefit analysis, value engineering, environmental assessment, smart growth criteria, aesthetic appropriation, socioeconomic preference, and design guidance. Key questions might include:

  • How does the project timeline correspond to plausible timelines for vehicle automation?
  • Could self-driving vehicles change assumptions that underlie the project?
  • Could the project be used to create demand for self-driving vehicles?
  • Could the project mitigate technical or economic barriers to vehicle automation?
  • Could the project generate outputs that would be useful to self-driving vehicles?

Take physical infrastructure. The next ten years could see the acquisition of tens of billions of dollars in right of way and the construction of hundreds of miles of carpool, toll, bus, and possibly truck lanes; modifications to these projects might enable the separation of closely-spaced vehicle platoons from other traffic. Similarly, tens of millions of road signs and billions of dollars in traffic signal equipment may be purchased over the next decade; particular design specifications might help self-driving vehicles communicate with or recognize these control devices. Conversely, self-driving vehicles could impact demand and capacity assumptions for current transportation projects that may not be finished for decades.

Or take digital infrastructure. In the coming years, states will likely conduct or commission numerous roadway inventories and invest in construction, congestion, and incident management systems, and private companies will likewise collect vast amounts of roadway data. The accessibility and integrity of these data may be crucial to the performance of certain self-driving vehicle technologies. In addition, existing fleets—such as the 213,881 vehicles operated by the US Postal Service—might also be well-positioned to collect data for detailed mapping, machine learning, and scenario testing.

Finally, consider the vehicles themselves. A car or truck that is sold in 2022 might still be on the road in 2032—and will incorporate thousands of dollars in component electronics. The ease with which such a vehicle might be converted to driverless operation could depend on how its electronic systems, particularly its CAN bus or other internal communications network, have been designed. As regulator, the federal government already influences system design: It mandates certain systems, like electronic stability control and on-board diagnostic test equipment, and may soon mandate others, like event data recorders and external communication equipment. As customer, the public sector could also influence design by choosing what vehicles to buy and when to buy them. For example, the federal government (including the US Postal Service) owns or leases almost half a million vehicles, and New York City’s “taxi of tomorrow” concession is expected to generate at least $1 billion in sales for Nissan.

There are downsides to this kind of forecasting: Governments sometimes get ahead of themselves or otherwise make mistakes. (So does the private sector.) Ultimately, however, this complexity may demand that relevant actors perform more rather than less analysis of the potential risks and opportunities that self-driving vehicles present. Now might be a good time to begin.

Photo Credit: Neff Conner

Comments

Bryant,
While I agree that analysis will need to take into account the changing mix in manual vs autonomous vehicles once adoption begin, without knowing more about how we will initially see the systems licensed, insured, and utilized, we aren't asking the right first questions:
Will autonomous vehicles operate without passengers?
How will electrification coupled with autonomy effect driving habits?
Will car sharing being a dominant model since cars won't need to stay with the owner?
Will FLEET autonomy come first?
Will Federal/states/cities/municipalities have reduced workforce costs due to workforce reductions of drivers?
The answer to those very practical technical and logistical problems will inform policy on infrastructure. Sensor data collected (& hopefully anonymized) on seemingly inconsequential things like road condition can be collected for other uses (road sensing, road repair, traffic, behavioral research) which will inform new methods of formulating policy.
Thank you for providing great information and look forward to fleshing out the potential policy considerations for autonomous vehicles.
Thom

Hi Bryant - another great article.
I think it is appropriate to carry out due dilligence on any public trasnport projects with a likely implementation time of five years or longer. Such projects often have extremely large capital costs and/or a disruptive construction phase - whereas an equivalent autonome system may well prove to be considerably more rapid and cost efective to implement. This is especially true if we predict that the rising costs of vehicle ownesrhip combined with the advent of autonomes will actually result in fewer, but more efficient, vehicles on the roads.
Paradigm shift means 'a radical change in uderlying beliefs or theory' - and we are facing such an event in my opinion with this technology. Many traffic forecasters and planners who are constantly predicting cumultaive percentage growth in vehicle numbers may need to reassess their fundamental view of growth on the road network.
In most developed cities I suspect that we already have more than enough road space to accommodate (congestion free!) all of our future traffic levels in 10 years time - but it will probably take at least another 5 years for the market penetration of this tech for the numbers to start falling.
Will our governments be prepared to bet on this tech and take bold decisions to change policies now?.....

Hi Bryant, thoughtful article. I think you raise some great points about the cost-benefit ratio with regards to infrastructure. We've been thinking that autonomous driving will probably make fixed public transport like trains obsolete within a few decades, and building new rail can be an expensive investment.
On the government regulation, absolutely agree that governments can get it wrong, even with good intentions. Though, as a driverless car enthusiast it is hard not to advocate pro driverless regulation. We're already seeing big companies like Rio Tinto buying driverless technology and I suspect a lot of logistics companies will too, probably influencing what the standard driverless technology on the overall car fleet will be. If there was regulation though, given Google's influence in this area I would be a little worried they might distort the process to their advantage.
On a slightly different thought, do you think that perhaps there will be a degree of industry self regulation in terms of standardising driverless technologies interact or are installed on older cars?

Add new comment