The technology and automobile industries have been promising self-driving cars, formally termed autonomous vehicles (AVs), for decades. At last, it seems that the necessary technology is finally veering into the home stretch. Many elements of autonomous vehicles have already been incorporated into popular consumer products: GPS, cruise control, parking assist, lane-keeping and crash-warning systems. Over thirty major corporations, from Google, Intel, and IBM to Tesla, Ford, Nissan, and BMW are working independently or jointly on Connected and Autonomous Vehicle technology. Tesla models currently on the road are fully equipped with the necessary self-driving software suites. Nearly every other corporation has announced that their own autonomous offerings will be for retail within the next five years, legislation permitting.
The main potential benefits and potential harms of this new technology are, for the most part, obvious. The biggest draw is that humans will by definition no longer need to drive, which will allow more free and productive time for current drivers in our commute-oriented culture. Additionally, road trials thus far show that AI drivers are much safer than human drivers. In the US, an average of 1.3 million drivers die in road accidents each year.
Detriments include a fundamental distrust of AI technology, justified by the considerable danger of cyber-attacks, as well as driving jobs that would be replaced by robots. Moreover, the creation of autonomous vehicles that can surmount weather and varying road terrain while working within current transportation infrastructure, particularly without ballooning cost, is still a considerable challenge. Nevertheless, as billions of dollars of private funding, unending media hype, and the National Highway Traffic Safety Administration can attest, autonomous vehicles are considered the biggest thing in transportation since the popularization of the personal vehicle with Ford’s Model T.
Of course, a consequence of such a major shift in transportation habits must be a similar shift in the production of carbon emissions. Globally, 14% of total greenhouse gas emissions are produced by the transportation sector, and 27% is the equivalent number in the United States. Meanwhile, 76% of transportation sector greenhouse gas emissions around the world are the products of car and truck transportation, which are overwhelmingly carbon emissions resulting from burning fossil fuels. As we can at this point take as a truism that carbon emissions are a major contributor to global warming, which has its own host of attendant negative consequences on the equilibrium of the world, any possible change to car transportation, in particular, would have a notable potential impact on the journey towards the theoretical tipping point at which climate change will become irreversible. Yet this is a question that has been largely ignored by the technology industry as their innovation continues apace.
At this point it may be helpful to take a moment to introduce levels of automation as defined by the Society of Automotive Engineers (SAE) International, currently in use also by the NHTSA. This is a simple scale ranging from zero to six. Level zero is defined as no automation whatsoever; level one indicates driver assistance, which places most cars currently on the road as level one automation. Level two specifically is defined as the AI possessing control of both steering and acceleration/deceleration, while the driver must continue to monitor surroundings. Tesla’s Model S and Model X are already equipped with this capability. Level three cars fundamentally fulfil the dream of autonomous vehicles, as the vehicle monitors its surroundings as well as controlling its own movement, but still require human drivers to stay alert for and take control of the car in an emergency, limiting the time-save benefits of self-driving cars. Level four is most developers’ current goal, that is, a car that is completely autonomous given controlled and well-mapped environments, such as urban spaces or national highways in non-inclement weather. Most autonomous vehicles expected to be for retail within the next five years are level three; outliers such as Tesla and Ford have decided to leapfrog level three directly to level four, which is in some ways technologically simpler, as it no longer requires a transition between AI and human driving. Level five is defined as a car that is completely autonomous in any and all conditions or surroundings, which, while a long-term goal of developers, will not affect the general population of consumers enough to perform a significant role in this investigation. Whether government legislation and consumer interest allow level two, three, or four cars to dominate the market, however, will have a considerable effect on the nature of the usage of autonomous vehicles and thus the consequences on their carbon impact.
Studies have been conducted modelling how individual elements of the adoption of CAV technology are likely to affect carbon emissions, although no paper thus far has attempted to make a confident declaration about how the eventual usage of autonomous vehicles will cause these factors to play out. Wadud et al come closest, postulating scenarios based on the level of penetration to model potential impacts on carbon emissions. The paper uses the ASIF framework to organize the carbon impacts of autonomous vehicles. The formulation is summarized as Emissions = Activity Level * Modal Share * Energy Intensity * Fuel Carbon Contact, wherein activity level refers to the general amount of transportation, modal share refers to the mode of transportation (in our case, of course, AVs), and energy intensity is defined as average energy consumption per kilometer. Though this framework is a simplification, as it does not take into account interdependence between factors, it can help organize the specific elements that will affect carbon emissions and allows a simple formula for deriving the overall change in carbon emissions using available studies. A similar framework in use is the Kaya Identity, which calculates carbon emissions based on use intensity, fuel intensity, and liquid intensity.
First, we will address those factors that can come about without requiring automation level three, which is generally considered the threshold for true autonomy. One such factor is the potential for congestion mitigation through more efficient routing, as well as accident prevention, which of course also would aid in preventing congestion. Congestion accounted for 56 billion additional pounds of carbon dioxide pumped into the air in 2011. In the short term, Advanced Driver Assistance Systems (ADAS) already offer benefits in terms of accident reduction and improvements in traffic efficiency. Kesting et al claim that relatively low penetration of Adaptive Cruise Control (ACC) can eliminate certain types of congestion entirely; although other studies are less optimistic, they are in agreement that widespread use of ADAS will have a positive impact on congestion, reducing the likelihood of accidents by 60%. Shrank et al estimate that 2.6% of fuel used in 2020 will be wasted on congestion and 4.2% in 2050; should CAV technology eliminate congestion entirely, carbon emissions could logically be reduced as much as 4%.
Another factor that will likely improve carbon efficiency is that of eco-driving, that is, changes in driver habits that do not depend on vehicle design. Humans are not always motivated to perform eco-driving and at any rate, cannot consistently perform it with the same efficiency as an autonomous system. These practices include maximizing the efficiency of the drive cycle and driving to plan out minimal brake-acceleration cycles, as both braking and acceleration waste fuel. The drive cycle refers to the speed of a vehicle at a given point in time, which leads to greatest carbon reductions when optimized for a stop-and-go urban environment. Berry (2010) has shown that humans taught to eco-drive show maximum energy intensity reductions of 20%, which may be considered the potential maximum benefit of eco-driving in AVs.
Thirdly, automated vehicles will allow platooning, or the practice of car and truck fleets travelling in closely-packed groups in order to reduce drag. Such reduction in following distance is perilous for human drivers, who must account for human reaction time, but autonomous vehicles would be networked and thus experience less limitation in driving in formations. Platooning may also increase roadway capacity and aid in reducing congestion. This would, for the most part, be a benefit seen in highway driving, as platoons would be unlikely to successfully roam crowded urban streets. From a combination of factors from the literature—estimates made using platoon formation, vehicle following distance, vehicle type, total highway travel, and typical highway driving speeds--fully implemented platooning could reduce energy intensity between 10% and 15%.
Other potential benefits of AVs depend on level three or higher automation levels, and a significant level of penetration of the technology into society. One common misconception about automated vehicles is that early models would be intended for popular consumer use. While certainly, models will be available—Tesla, of course, is marketing its luxury self-driving cars—major promoters of and investors in AVs thus far have largely been ride-share companies such as Uber. This partiality can be seen also in many of the early real-world trials of AVs. One of the biggest and most successful trials has been conducted with Singapore’s Nutonomy, a Southeast Asian Uber equivalent; the road tests currently occurring at Greenwich in London also use shuttles and pods meant to ferry multiple passengers rather than more traditional single-family sedans. The consequence of level four AVs interacting with such ride-share models is the prophesied transformation in transportation infrastructure, which, with sufficient penetration, would culminate in the end of the personal vehicle. Such a world, envisioned, would consist of AVs on call at any given time and in almost continual use, networked together into a collaborative hive mind.
Taking this future into account, more carbon emission reductions could be realized on a network level. On the most extreme end of automation and penetration is the possibility that accident rates could become so low that the thorough and expensive safety equipment currently present in all vehicles could be removed, leading to both less weight and carbon emission. This could supposedly reduce carbon emissions, through a simple calculation of weight and fuel usage, by 5.5%.
In a less distant timeframe, such communal usage of transportation would lead to electrification, and reduction of time spent looking for parking. Electrification refers to the deployment of the vehicle matched to user trip need, that is, cars would be used by passengers when needed and the size of car could be matched to the number of passengers, thus reducing the efficiency of the current model, where the typical four-passenger sedan holds only 1.47 passengers on average. This would lead to less demand for cars in general, as it has been shown that through ride-sharing, Singapore’s population could be served by only one-third as many cars as currently are operated within the city. This could conceivably reduce energy consumption by as much as 75%, at the generous end of the estimate. Moreover, as individuals would not need to keep their cars with them as they conducted non-transportation business, parking would become a more centralized and organized business than in the current system. Currently the Texas Transportation Institute reports that the average American driver wastes 19 gallons of fuel looking for parking each year. On the rough assumption that an AV could cut that number in half, AVs would be providing a 4% reduction in car usage in general.
Finally, the ownership of cars in large and organized by businesses rather than individuals would expedite the adoption of more recent carbon-saving technologies such as new fuel mixes, which currently remain unpopular due to the cost and complication of refuelling. This also includes electric vehicles, although, in the US, where much of the electricity is still gained from burning fossil fuels, electric cars do not provide much net reduction in carbon emissions as a whole. At this point, however, it is still difficult to provide numbers for the potential carbon reductions from burning, say, natural gas rather than oil.
The detrimental effects of automated vehicles on carbon emissions is more straightforward than the potential benefits and essentially scale with the level of automation and penetration. There is only one downside inherent to self-driving cars—at worst they would be exactly as current cars and drivers are, except with the potential of a higher safe maximum speed, which would increase energy intensity by as much as 22% if typical highway speeds were increased without limit and humans drove at the maximization of their time and money—projected to be between 10 and 20mph higher than current speed limits of 65-70mph, a prediction that aligns with Germany’s unrestricted Autobahn speeds.
Rather, the main downside of AVs is the possibility that they will grant too much quality of life improvement to car travel in general, thus massively increasing the modal share apportioned to cars. For one, level four AVs, which require no human supervision, would open up car travel to populations currently unable to use road vehicles, such as the disables, the elderly, and children. Estimates of exactly how much these underserved populations would actually use CAV technology vary wildly, from 10% to 50%, but undoubtedly would result in a net gain. Moreover, people would be more willing to spend more time in vehicles and would be more likely to take long trips or commute long distances, would prefer car trips over other forms of transportation, and might even increase the weight of vehicles through adding luxury and entertainment facilities. The increase in car usage must, at this point, be estimated using a function of the worth of an individual’s time. At level two and three automation, this difference would be less impactful, as the driver must still be aware. However, at level four automation, a net balance in opportunity cost of an individual’s time and cost in fuel creates—in purely theoretical terms—a 40% to 50% gain in total car usage.
So, it can be seen that the actual net effect of all the energy savings and losses from automated vehicles depends on a combination of factors. These include the extent to which energy-saving design and practices are used, the degree to which automation penetrates and creates the transportation revolution towards the car-sharing system, and the degree to which automation allows time to be saved through increased usage of AVs. These, in turn, depend on public adoption, private sector motivations, and government legislature.
The United States government, although proceeding much more cautiously than the private sector, is preparing for a widespread adoption of level four autonomous vehicles, although its current efforts at legislation are more well-intentioned than comprehensive. Thirteen states have passed legislation related to autonomous vehicles, out of 33 total that has introduced it; trials are currently taking place in countries the United Kingdom and Singapore who are also promoting research into and trials of self-driving cars. The UK, in particular, is attempting to acquire a pole position in the CAV technology industry, as unlike the rest of the EU it never ratified the Vienna Convention, which requires that a driver is in control of a given vehicle at all times. The spread of AVs has been compared to that of the proliferation of cellphones or smartphones, wherein the lion’s share of profits were taken by the earliest creators, i.e. Apple, which grew on the basis of its iPhones to one of the largest corporations in the world. It has been estimated that each AV on the road will bring total savings of nearly $5000 a year. Thus, both the private sector and each nation has a significant stake in not falling behind in developing and introducing CAV tech.
An additional factor is the degree to which carbon-efficient techniques and processes are implemented by the automotive industry. As CAV technology is a process driven almost entirely by the private sector, said implementation is likely to be at the more affected by profitability than any altruistic dream of reducing carbon emissions of car transportation. Fortunately, this does not necessarily indicate a death knell for carbon efficiency. For one, the government still plays an active role in controlling carbon emissions. The Environmental Protection Agency ‘s Corporate Avery Fuel Economy (EPA CAFE) standards, a long-term plan to reduce carbon emissions of road vehicles, has called for an average of over 50 miles per gallon for all cars and light trucks by 2025; such standards, admittedly, are imperiled by the Trump administration, which called for a review of CAFE in March 2017. For another, fuel economy is at a current record high while carbon emissions are at a record low, trending at a stable downwards rate, which is motivated by consumer interest and economic expediency as much as government intervention. Especially considering unstable fossil fuel prices and their associated politics, both everyday consumers and private industries have often been finding it more cost effective to invest in carbon reduction friendly technology. This is additionally influenced by the fact that early adopters of AVs are most likely to be ride-share companies as previously discussed, who can reduce costs and complications of widespread implementation of carbon-efficient technology by organizing the infrastructure around their fleets. This is also opposed to the common consumer, who is more likely to drive unsustainably out of personal impatience or love of speed.
A collective blind spot in the movers and shakers of the autonomous vehicle industry seems to be customer reaction. That is, despite widespread consumer unease with the concept of self-driving cars, those in tech and business have taken the eventual acceptance and popularity of autonomous vehicles as pat. This view is not entirely unfounded. For one, car culture has been on the decline in the United States since the 1980s. Only 28% of sixteen-year-olds received their driver’s license in 2010, down from 46% in 1983, while services such as Uber and car-share apps have become popular. Americans are not as enamored with the idea of the personal vehicle as they once were. Additionally, there have been claims that the technology adoption curve has been compressing—although the only conclusive example of this is the smartphone. The spread of new technology is known as diffusion—only at a certain point of diffusion is it possible for this new technology to make an impact on society. Considerable research has gone into the rates of said diffusion, which is the aggregate of a series of individual choices to adopt said new technology. Although nearly as many factors go into predicting the rate of technology adoption as go into predicting the usage of self-driving cars specifically—the general trends are obvious: old technology essentially never replaces new technology, and new technology is adopted at an S-curve rate which begins as a trickle but accelerates until the population is nearly saturated. Which is to say, useful new technology is the genie in the bottle: although rate of penetration is a complicated function, once out, the genie can never be put back in. However, no comprehensive study has been performed of projected penetration by AVs, an urgent next step for more accurately predicting the effects of carbon emissions as well as all other consequences of CAV technology.
As level four automation becomes more widespread and the diffusion of CAV tech saturates the population, automated vehicles will likely enable a modest reduction in carbon reductions. Although the automobile industry will likely not devote their full resources into creating efficient vehicles, current trends show that energy efficiency remains a significant priority in new designs. Due to the many potential energy savings made possible by AVs, coupled with monetary savings, carbon emission intensity will almost certainly be reduced. However, it will be a challenge for said savings to balance the increase in car travel as a whole, which will become more attractive as costs decrease and drivers become unnecessary. Although the allure of profit and government support cautiously indicate that level four penetration is all but inevitable, it is difficult to say in what timeframe this shift will take place. If the technology is implemented slower than anticipated, perhaps the fundamental assumptions regarding automated vehicles will need to be reevaluated.
Anderson, James M., Nidhi Kalra, Karlyn Stanley, Paul Sorensen, Constantine Samaras, and Tobi A. Oluwatola. "Self-Driving Vehicles Offer Potential Benefits, Policy Challenges for Lawmakers." March 22, 2016. http://www.rand.org/pubs/research_reports/RR443-2.html.
Berry, I.M., 2010. The effects of driving style and vehicle performance on the real-world fuel consumption of US light-duty vehicles (Masters thesis, Massachusetts Institute of Technology).
Brown, Austin, Jeffrey Gonder, and Brittany Repac. "An Analysis of Possible Energy Impacts of Automated Vehicle." Road Vehicle Automation Lecture Notes in Mobility, 2014, 137-53. doi:10.1007/978-3-319-05990-7_13.
Dresner, Kurt, and Peter Stone. Sharing the Road: Autonomous Vehicles Meet Human Drivers. Proceedings. International Joint Conference on Artificial Intelligence.
Fagnant, Daniel J., and Kara Kockelman. "Preparing a Nation for Autonomous Vehicles: Opportunities, Barriers and Policy Recommendations." Transportation Research Part A: Policy and Practice 77 (2015): 167-81. doi:10.1016/j.tra.2015.04.003.
Hall, Bronwyn, and Beethika Khan. "Adoption of New Technology." 2003. doi:10.3386/w9730.
IPCC, 2014: Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp.
Sivak, Michael, and Brandon Schoettle. "Recent Changes in the Age Composition of U.S. Drivers: Implications for the Extent, Safety, and Environmental Consequences of Personal Transportation." Traffic Injury Prevention 12, no. 6 (2011): 588-92. doi:10.1080/15389588.2011.605817.
Spieser, K., Treleaven, K., Zhang, R., Frazzoli, E., Morton, D., Pavone, M., 2014. Toward a Systematic Approach to the Design and Evaluation of Automated Mobility-on-demand Systems: A Case Study in Singapore, Forthcoming in Road Vehicle Automation, Springer Lecture Notes in Mobility.
Wadud, Zia, Don Mackenzie, and Paul Leiby. "Help or Hindrance? The Travel, Energy and Carbon Impacts of Highly Automated Vehicles." Transportation Research Part A: Policy and Practice 86 (2016): 1-18. doi:10.1016/j.tra.2015.12.001.