explore and demonstrate their capabilities.
Generally, industry participants have been proactive in establishing safety principles and guidelines, which should become part of a framework that is eventually instituted by government regulators to set national safety standards for autonomous vehicles. In April 2016, Ford, Lyft, Uber, Volvo, and Waymo initiated the Self-Driving Coalition for Safer Streets. Recently, eleven companies proposed guiding principles for self-driving vehicles for the development, testing, and validation of safe autonomous vehicles.4 Those initiatives indicate that industry and academic researchers worldwide are committed to improving the safety of autonomous vehicles collaboratively and strongly suggest that their efforts will eventually coalesce to perfect those vehicles for safe use by the traveling public.
Potential Travel Improvements
Autonomous-vehicle technology has the potential to significantly improve the safety, speed, reliability, and cost of road travel in four ways. It can prevent collisions—94 percent of which, according to the National Highway Traffic Safety Administration, are caused by human error5—and thereby greatly reduce highway fatalities, serious injuries, vehicle damage, and costly insurance by gathering and reacting immediately to real-time information and by eliminating concerns about risky human behavior, such as distracted and impaired driving. It can significantly reduce both recurrent and incident delays and improve travel time reliability by reducing accidents and, when necessary, by rerouting drivers who have programmed their destinations.6 It can allow vehicles to travel closer together and at higher speeds more safely by creating a smoother traffic flow, thereby effectively increasing highway capacity without the public having to incur the enormous cost of building and maintaining additional lanes and new freeways. And it can reduce the cost of chauffeured passenger, for-hire trucking, and postal services by eliminating the need for a driver.7
Finally, by smoothing traffic flow and reducing stop-and-go driving, autonomous vehicles will achieve improved fuel economy and, in combination with the development of electric vehicles (EVs), could significantly reduce pollution and improve the environment. Companies’ development and commercialization of EVs is currently proceeding independently of the development and commercialization of AVs; it is expected that EVs and AVs will be combined eventually and that the resulting autonomous electric vehicles (AEVs) will eliminate the social costs of tailpipe emissions, which are expected to exceed the social cost of generating electricity for EVs, by replacing gasoline-powered nonautonomous vehicles.
Connectedness
The ability to drive itself, or automation, is clearly a critical feature of an autonomous vehicle. But the other critical feature is its connectedness to other vehicles and their surroundings, including pedestrians, infrastructure, and the network, which allows for the optimization of the vehicular system as a whole. Connectedness is achieved by various communication models, summarized by Mutschler (2018) as follows:
vehicle to infrastructure (V2I), which includes communications to traffic-signal timing, speeds limits, prioritization, and traffic signs (for example, changing traffic signal lights)
vehicle to vehicle (V2V), which enables collision avoidance (for example, ambulance approaching)
vehicle to pedestrian (V2P), which can transmit safety alerts to pedestrians and cyclists (for example, vehicle approaching, pedestrian crossing street ahead)
vehicle to network (V2N), which enables communication about real time traffic, work zones, routing, and cloud services (for example, traffic congestion two miles ahead)
Policymakers must engage with the autonomous-vehicle industry to determine the appropriate investments that could enable autonomous vehicles to operate safely and efficiently in their states and cities with the essential communication capabilities.
Incentives for Automakers and Technology Companies
Chris Urmson, the former director of self-driving cars at Google, argues that the private sector has a financial incentive to expedite deployment of autonomous vehicles because it can earn revenues on a per mile rather than a per unit basis. He refers to the following equation to approximate the size of the industry based on revenue:
3 trillion VMT × $0.10 per mile = $300 billion per year,
where VMT refers to annual vehicle miles traveled in the United States and travelers are charged $0.10 a mile to cover costs.
Bosch and others (2018) finds a somewhat higher cost figure, based on a review of academic empirical estimates of the cost per mile of autonomous vehicles that puts the cost in the range of $0.15–$0.20 per mile. In any case, Urmson offers a plausible hypothesis that travelers could reduce the capital costs of vehicle ownership, including insurance, by apportioning them over the number of miles driven by acquiring or sharing a vehicle through a “transportation subscription service.”8
Automakers could then improve their earnings by selling the use of autonomous vehicles that consumers share through rentals or subscriptions instead of by selling a given volume of units. Individuals and households would operate autonomous vehicles as transportation as a service (TaaS).9 For example, assuming, on average, that auto companies currently turn a profit of roughly $1,500 per car and that a car lasts 150,000 miles, then their profit is a penny per mile traveled, which is likely to be less than the profit per mile traveled that they could earn by charging consumers a per mile fee for a shared autonomous vehicle.10 Of course, the extensive collaboration in the autonomous-vehicle industry, the reinvention that autonomous-vehicle providers will have to make in the way they market and differentiate services, and firms’ extensive investments in R&D and their huge fixed costs suggest a high degree of uncertainty about the level and distribution of future profits.
Travelers’ Incentives for Adoption
Adopting autonomous vehicles would also greatly benefit travelers. Small and Verhoef (2007) reports that the average cost of highway travel time and unreliability amounts to $0.40 per mile and that the average cost of accidents is $0.14 per mile. If travelers pay $0.10 per mile, or even somewhat more, for their autonomous-vehicle transportation, those costs are likely to be strongly offset by the benefits from reduced travel times and greater reliability and from the virtual elimination of accident costs.
The large benefits from this back-of-the-envelope calculation are consistent with Fagnant and Kockelman (2015), which accounts for the benefits from the reduction in vehicle accidents and travel delays under the assumption that autonomous vehicles would reduce freeway delays from 35 to 60 percent, depending on the extent of autonomous vehicle (AV) market penetration and accounting for the additional travel that the vehicles may induce that would increase congestion.11
Technologically, the large benefits accrued from reducing congestion derive from the basic relation, traffic flow = travel speed × traffic density. Consider the case of a traffic accident that creates significant congestion, as people rubber neck to view the incident, causing traffic to slow to a crawl. Autonomous vehicles could significantly increase travel flows in such situations because they could smooth traffic flow and travel closer together without jeopardizing safety and could maintain a more consistent speed by not slowing down to view the aftermath. As noted, incident delays account for roughly one-third of all delays, so even modest improvements in traffic flows from reducing those delays could amount to significant savings in time and reliability that could yield large social benefits. Of course, autonomous vehicles could also improve traffic flows by the same type of mechanism to reduce other types of delays.
Wu and others (2017) presents simulations to examine the impact that autonomous vehicles could have on improving traffic flows even in traffic that includes nonautonomous vehicles. Small experiments involving a few dozen cars on a closed circuit show that a single autonomous vehicle could reduce traffic congestion by subtly
