Coauthored with Emily Hamilton
Last week, the autonomous vehicle company Waymo began testing cars in Chandler, AZ with no employees sitting in the front seat. While Waymo is busy creating systems of vehicle-mounted sensors that allow cars to safely navigate existing urban infrastructure and obstacles, some cities are pursuing plans to build “smart streets” that broadcast information about roads and potential hazards to autonomous vehicles. The American history of auto-centric infrastructure demonstrates that building specific infrastructure for autonomous vehicles may have long-lasting negative consequences.
Waymo’s cars rely on both detailed maps and car-mounted lidar sensors that “see” the world around them in order to follow their route and to to avoid collisions. While the current technology is very safe, car-mounted sensors remain imperfect. As Tim Lee points out, there are reasons that Waymo launched in Chandler: its sunny weather, wide streets and minimal pedestrian traffic. Fully autonomous vehicles will need even better sensors than those that are currently available to drive safely in snowy conditions and in places with less regular streets that may confuse a vehicle’s sensors.
Some analysts have advocated for publicly-provided smart streets and smart intersections that could limit the need for vehicle-mounted sensors and, perhaps, speed the adoption of autonomous vehicles. My colleague Brent Skorup has this view:
Car-mounted sensors often confused by road materials (a shift from dirt to gravel or asphalt), reflective buildings, bridges, or even the weather. Roadside sensors not only mitigate these problems, but also reduce the computing load on car-mounted systems, because the vehicles have to make fewer snap decisions.
There’s a way to get this information to vehicles quickly and accurately. Just as lawmakers and city planners started laying asphalt, installing streetlights, posting speed limits, and zoning property to accommodate Henry Ford’s cars, we need to design roads and infrastructure for the self-driving generation of vehicles.
Atlanta recently unveiled a “smart corridor” that pushes information about the environment to cars. The plan is for a driverless bus line to eventually carry passengers along the city’s North Avenue. Cities are just beginning a rollout of infrastructure designed for autonomous vehicles, but in the past two years Waymo engineers have increased threefold the distance that its sensor’s can see while reducing the cost of these sensors by 90 percent.
By the time driverless buses have completed testing on Atlanta’s smart corridor, driverless taxis that don’t rely on any specific infrastructure may already be commonplace in Phoenix. Google’s earlier models of self driving cars have already logged millions of miles in Washington, Pittsburgh, California, and other parts of the country with very good safety records.
As we’ve already seen with the rapid advancement in sensor technology, autonomous vehicles and the technologies that make them possible will be constantly upgraded and changed in a competitive market. The long lag between legislative proposal, vetting, and implementation necessary for public projects means that “smart” street technologies will likely be outdated by the time city governments roll them out.
Autonomous cars are likely to be better off relying on each other than on fixed infrastructure. As autonomous vehicles capture a larger share of road traffic, they will be able to crowdsource extremely-detailed, real-time maps of urban roads. Each member of the network will benefit from the information provided by other vehicles and would likely provide its own data in exchange for access.
Those who support investment in smart streets have observed that map-dependent autonomous vehicles will not function as well in lower traffic suburbs or rural areas. They’re right, but we should learn from the disastrous history of subsidizing suburbs at the expense of urban areas. Should municipalities succeed in creating infrastructure that benefits autonomous vehicles, they will be in the business of shaping where people live and how they make their transportation decisions under a new technology. Current technology may allow autonomous vehicles to work best in heavily traveled urban areas, but it should be left to entrepreneurs to overcome these challenges in less trafficked places, or suburban and rural residents should bear (reap) the full costs (benefits) of their lifestyle decisions.
Likewise, municipal investments in smart streets will have their own unintended consequences that policymakers are unlikely to foresee or respond to after they arise. If policymakers think that additional infrastructure is needed for autonomous vehicles to function efficiently, they would do better to allow competing companies to install it at their own expense. This would avoid locking in outdated technologies and make consumers and autonomous vehicle companies more likely to internalize the costs their transportation use that public provision would otherwise impose on the general public.
Supporters of smart streets have argued in favor of using them to implement demand-based congestion pricing, to optimize efficiency at intersections, and to route vehicles through cities more quickly. But there’s little reason to think that city governments would take advantage of this capability. Congestion pricing in other urban areas has proven politically toxic. Leaders in San Francisco and Washington, DC each decided to never fully activate their expensive smart parking systems grids because of their they fear of political fallout from pricing parking spots based on supply and demand.
Rather than looking to municipalities to implement congestion pricing for smart cars, driverless taxi companies may be in a better position to internalize congestion costs and charge for them. Coupled with a price system, the networked vehicles will allow their passengers to cooperate and reach their destinations more quickly and safely.
Instead of catering to the newest transportation trends, planners can focus on allowing human-centric development to flourish and let autonomous vehicle designers work within the constraints this creates. Waymo’s rapid progress in producing better sensors at lower costs demonstrates that the remaining technical challenges facing autonomous vehicles will be surmountable in a short time period.
While technological hurdles remain, the larger obstacles for widespread adoption of autonomous vehicles will be legal and cultural. Human drivers, pedestrians, and cyclists don’t follow the letter of the law. Rather they follow norms that generally allow people to get where they’re going safely. Police officers and judges enforce the rules on the books with these norms in mind. Autonomous vehicles, however, will follow the letter of their programming. An autonomous car will respect a 25-mile-per-hour speed limit on a wide avenue when all the human-driven vehicles around it are going 50. Once autonomous vehicles become prevalent, pedestrians will know that they can safely cross in front of them, legally or illegally. Autonomous vehicles will slow to a crawl when human drivers merging in front of it cause the autonomous vehicle to violate tailgating rules, and so on. Existing rules will need to be changed to better reflect actual driving practices so that autonomous vehicles can safely coexist with human drivers.
Waymo’s Arizona debut demonstrates that a policy environment of permissionless innovation is the only necessary condition for the adoption of autonomous vehicles. Cities can best accommodate the coming revolution in transportation by putting the onus on tech companies to find ways to make their products work with existing infrastructure and by allowing cities to grow organically as technologies evolve.