This is the second in a two-part series by Brent Skorup and Emily Hamilton discussing the future of autonomous vehicle infrastructure.
Autonomous vehicles (AVs) promise many potential benefits in terms of safety and increasing access to transportation. In an effort to speed the arrival of AVs, some cities are pursuing plans to build "smart streets" that broadcast information about roads and potential hazards to autonomous vehicles. Rather than attempting to design cities to promote AV adoption, cities should recognize that the market for AVs is innovating rapidly and that municipal investments in the industry are likely to be outdated by the time that they’re built. But leapfrogging isn’t the worst potential outcome of smart street investment. Municipal smart streets could further entrench government intervention in transportation rather than taking advantage of an opportunity to rely on more voluntary transactions in transportation.
Atlanta has already built a “smart corridor” designed to work with city-owned autonomous shuttles. The corridor runs along 2.3 miles of the city’s North Avenue and has been used in test rides for autonomous municipal buses. Other cities and states have also planned expenditures on smart streets in an attempt to lure AV’s to their jurisdictions. The City of Madison in Wisconsin has developed plans for a “connected corridor” designed for AV testing. In Ohio, plans are in place to install fiber optic cable and sensors along a 35-mile stretch of I-33 that connects Dublin and Marysville.
Some proponents of autonomous vehicle technology suggest that cities should install lidar sensors in their streets that could reduce the amount of information that autonomous vehicles’ onboard sensors need to process and allow for widespread AV adoption faster than private industry alone might achieve. Supporters of building smart streets for AVs argue that this infrastructure will make regions attractive to car manufacturers, speed the adoption of the technology, and improve upon the safety of onboard sensors.
In the evolution of AV technology, developing a system of cameras, radar, and lidar to “see” the world around vehicles presented limitations for fully autonomous cars. Historically, these systems haven’t been able to detect obstacles far ahead, and they’ve been confused by precipitation. But today, lidar sensors can “see” 200 meters ahead, and companies are rapidly developing sensors that can operate in rain and snow without problems.
A potential use for smart streets is to push information about weather and road conditions to vehicles to compensate for information the onboard sensors can’t detect, but the most advanced AV technology is progressing without any special in-ground infrastructure investments from city governments or partnerships between cities and the industry. In Phoenix, driverless taxi service Waymo is now accepting paying customers who are part of its trial program while operating entirely on standard “dumb” streets. General Motors plans to begin testing cars without steering wheels or pedals in congested urban neighborhoods in 2019. Autonomous vehicle manufacturers are not waiting on smart streets before deploying technology that’s designed to work with existing infrastructure.
To date, the accidents that have occurred as part of AV test programs have been the result of human failure rather than the failure of technology. When an Uber test vehicle killed a pedestrian in Phoenix, it was because executives decided to operate test cars without their emergency braking systems enabled. In a recent crash involving a Waymo test car in San Francisco, a safety driver overrode the AV system in an attempt to avoid a collision, but analysis following the accident showed that human intervention prevented the AV from successfully avoiding the collision.
Supporters of building smart streets have also argued that onboard sensor systems are cost-prohibitive and that building fixed in-street sensors would be more cost-effective. However, the cost of onboard sensors is falling rapidly. The primary sensor on the original Google car cost $75,000 and the car had about $150,000 worth of sensor technology on it. Today, sensors are available for 90% less, and manufacturers are targeting a price of less than $500 for future models.
But the history of government subsidies to private automobiles should serve as a cautionary tale rather than a blueprint for AVs.
As cities are just beginning to develop plans to install sensors that allow for vehicle-to-infrastructure (V2I) communications, private industry is already circumventing the obstacles smart streets are designed to address. By the time cities build smart streets, AV companies will likely have leapfrogged municipal technology.
Proponents of either municipally-managed smart streets or public-private cooperation for AV infrastructure have argued that the adoption of private cars depended on government-financed and managed infrastructure. Government road-building and parking requirements certainly played a part in the ascension of the private car as the primary mode of transportation in the United States. But the history of government subsidies to private automobiles should serve as a cautionary tale rather than a blueprint for AVs.
The twentieth-century history of government transportation is largely a story of enormous cross-subsidies. Drivers’ tax dollars subsidized expensive rail systems that they might have never used. Non-drivers still subsidize a large portion of road building and maintenance that isn’t covered by fuel taxes or other user fees. Beyond these fiscal effects, policy has dramatically shaped transportation options across cities.
Top-down infrastructure planning and cooperation between the auto industry and government have had long-lasting effects that policymakers likely did not foresee. Road building and parking requirements have made it difficult to choose not to drive in many cities across the country, reducing opportunities for people, particularly low-income people, to economize by forgoing car ownership. We can’t foresee how smart streets might shape the transportation options of the future, but the American history of determining how we can move through our cities provides reason to be wary of engineering around a new technology.
Rather than further entrenching top-down decision making in transportation, autonomous vehicles present policymakers with an opportunity to reduce the level of cross-subsidies. In addition to forcing AV users to internalize the cost of road infrastructure by not building smart roads, there will new avenues for pricing roads. As AVs become commonplace, there may even be opportunities for market-driven congestion pricing that will reduce the cost drivers impose on each other when they contribute to time-wasting traffic. If autonomous vehicles are adopted in large part under a ridesharing model, consumers will be paying for scarce car availability and naturally curtailing the overconsumption of “free” roads. Additionally, AVs may create an opportunity for charging road users based on their vehicle miles traveled. Transportation economists point out that if consumers pay a vehicle-mile-tax as part of the total price of ridesharing, political opposition to road pricing would likely fall.
If AVs are as successful as their proponents hope they will be, they promise huge benefits in terms of saving lives and allowing people to put their transportation time to good use rather than staring at the road ahead of their cars. Rather than furthering the system of cross-subsidies under which taxpayers are required to bear the cost of others’ transportation choices, policymakers should take the opportunity to allow AV firms to compete to offer the best product. We can’t predict how autonomous vehicles will change transportation. Still, their adoption provides reason to be optimistic about future private sector opportunities to improve transportation safety while moving to a system where users pay for more of their own costs, and where we waste less time in traffic—so long as municipal governments don’t try to design our cities around how their vision of how AVs should function.