connected cars navigating the roads

How self-driving car subsidies could carry us through the ‘dark age’ of deployment

A game-theory approach identifies which policy could support autonomous vehicles’ market penetration—and environmental benefits

Among the many promises of autonomous vehicles is a more energy efficient transportation sector, which would be a major boon in the nation’s efforts to reduce greenhouse gas emissions. The transportation sector is responsible for roughly a third of US carbon dioxide emissions, according to the EPA.

Autonomous vehicles could clean up how we move people and goods. They could enable vehicle platooning on highways and advanced traffic management in cities. And they could lead to more shared trips and carpooling. But these efficiency gains hinge on market penetration. What’s more, society is likely to see the needle go the wrong way in the earliest stages of adoption.

In a recent paper, researchers used a dynamic games approach to determine which policies might carry us quickly through what they call this “dark age” of early deployment. Policy levers could provide incentives for automakers to continue innovating, and consumers to increase their use of self-driving cars. In this Q&A, University of Michigan Engineering doctoral student Qi Luo discusses the research. Qi works with Romesh Saigal, a professor in the U-M Department of Industrial and Operations Engineering.

Your paper describes how the environmental benefits of autonomous vehicles won’t be realized until we reach a tipping point in market penetration. Why is this? And what is that tipping point?

Although we are still a long way from fully autonomous vehicles (AVs) becoming available on a broad scale, researchers are studying the profound impacts they will have on our society. What we studied is how AVs will impact road capacity, which is a measure of how much traffic road networks can handle. We, and other research groups, have found that very early adoption of automated vehicles can potentially decrease road capacity, and increase traffic congestion. The reason is human driving behavior. In a mixed traffic flow of many human-driven cars and few automated vehicles, automated vehicles are likely to keep a larger safe distance than necessary because humans are very unpredictable drivers. But when AVs saturate traffic, they can stay very close to one another.

This ‘tipping point’ of AV adoption is estimated to be between one-half and two-thirds of traffic flow, according to a previous simulation study. Of course, it also highly depends on how AVs are controlled, how infrastructure is designed, and how humans actually react to AVs. Our study focuses on designing transportation policy for this transition period.

If we stay status quo, how long would it take to reach this critical market penetration?

There is no simple answer, as we need to specify which scenario we are looking at.

(Let me clarify that our work focuses on deployment of fully autonomous vehicles, rather than those with automated features.) It depends on whether we’re adding a new shared mobility service or replacing the entire vehicle fleet with personally-owned AVs. People give different perspectives. John Zimmer, the co-founder of Lyft, envisions widespread use of AV fleets in the next five years. More conservative estimates from academia and government suggest that the mass production of AVs won’t occur until 2025-2030.

Our study applies once we achieve mass production of AV technology—after it has passed the ‘driving tests’ and regulations are all set. Our model depicts what an optimal subsidy policy would look like from the day one of mass production to market saturation. However, when will the ‘day one’ come is beyond the scope of our paper.

We found that if AVs are used as shared taxis such as the Uber model, we can expect this critical level to be achieved, under the status quo, within five years. But if AVs are privately owned, it could take twice that long.

You lay out some policy options that might get us there quicker. What are they?

We studied subsidies in this work. The main problem of the early deployment of AVs is that the negative impacts are on every road user, not only the AV users. We briefly discussed other options such as taxation. We conclude that those are less effective than subsidies because subsidies directly shorten the low market penetration period. It is possible that other policies could be used in tandem to address this problem.

More broadly, we propose a new technique to design this type of policy for new technologies. This work studies the effects of the policy on the dynamics of the market and the responses of the entities affected by the policy.

In an optimal approach, who would the government subsidize? How much? And when?

We found that the answers change as the market penetration evolves. Optimally, at an early stage of AV deployment, government would give priority to subsidizing customers (which is the same for electric vehicles). But when the market gets more saturated, more budget could be allocated to manufacturers, or OEMs. The optimal amount is sensitive to the input data—the market size, and the environmental benefits of AVs, for example. But we prove a consistent trend in the optimal policy: The amount decreases and then increases over the planning horizon. We measure the time points by the market share, or the number of AVs on the road. When market share reaches 15%, the government could eliminate AV subsidies, then restart the subsidy with 65% market share. Again, these numbers are based on a couple of survey results so we need to conduct more studies to confirm the predictions of AV market.

What’s next? How would society move this forward?

This is a great question. There are different visions for the future of AVs. Will they be shared, or privately-owned, for example. And AV technology itself still faces a lot of uncertainty. So this is not a case where our derived policy is implementable at this point. What we offer is a methodology to evaluate the long-term impact of AV subsidy policy, and when we get to the appropriate point in time, we believe it should be considered to be a standard way to assess and compare different subsidy policies.

Other co-authors are: Yafeng Yin, U-M professor of civil and environmental engineering, and Zhibin Chen, an assistant professor at NYU Shanghai. The paper is titled, Accelerating the Adoption of Automated Vehicles by Subsidies: A Dynamic Games Approach. The research was funded by the National Science Foundation.