This study discusses the environmental impacts of large-scale autonomous vehicle (AV) adoption in order to explore desirable directions of AV technologies and policies from the environmental perspective. It is expected that AVs will solve problems in many areas including the environmental field, but this optimism requires continued efforts of various fields including environmental science, social science, and urban and transportation planning to keep the development of AV technologies and policies environmentally sustainable [1]. The technologies of AV are in the final stages, and experts predict that AVs will be available to the public by 2030 at the latest [2–7]. While mixed forces of increasing or decreasing mobile emissions caused by AVs are expected, there has been little in-depth discussion about their potential impacts [1]. Automobiles are one of the biggest sources of air pollution and demand for them is increasing [8–10]. Air pollution caused by excessive use of fossil fuels is one of mankind’s most daunting challenges. Nonetheless, most countries’ socio-economic systems are still fossil fuel-dependent, and the development of clean energy to replace them has not come to fruition. Transportation contributes approximately 28% of greenhouse gas (GHG) emissions generated by human activity [11]. Given the increasing trends of interregional and international trade [12], the GHG emissions from the transportation sector are expected to increase further. Moreover, the increase in traffic congestion is intensified in populated areas due to the imbalance in traffic demand and supply, and this has caused more immediate exposure of mobile pollution to people. In order for AV to become a solution to such problems, multidisciplinary discussions should be made on the way of developing and adopting AVs to make larger environmental benefit [1].

As part of this context, this study analyzed the impacts of road capacity increase due to platooning and speed limit increase on mobile emissions. AVs will have incredibly improved string stability which is based on vehicle-to-vehicle(V2V) technology that is essential element of AVs because it enables cooperative operation between vehicles, and managing extremely narrow inter-vehicle distances will be possible [2, 13]. As the inter-vehicle distance decreases, the transportation network can accommodate more vehicles, resulting in an increase in traffic capacity [14]. As market-penetration level of vehicles with this technology increases to 50%, 80%, and 100%, traffic capacity increases by 18%, 49% and 80%, respectively, when inter-vehicle distance is set about 0.7 s [15]. Our concern is the environmental impact when the inter-vehicle distance (in seconds or meters) becomes smaller. In addition, AVs will be able to manage faster vehicle speed without compromising safety with the help of V2V and vehicle-to-infrastructure (V2I) technologies [1, 16, 17]. V2I technology is a type of Intelligent Transportation System (ITS) that provides real-time traffic information between vehicles and road facilities [18, 19]. V2I technology research originally started with the aim of improving safety by providing driving information to drivers, but has developed into a study for fuel optimization through optimal gear shifting [20] and minimizing operation of brakes and excels [21]. Talebpour and Mahmassani [22] verified the effect of AVs on improving the traffic flow stability by simulating different vehicle types and distinct communication capability situations. However, improved string stability and traffic flow stability can increase vehicle speed [1, 16, 17, 23]. In congested city centers, mobile emissions may decrease due to increase in traffic speed, but mobile emissions may seriously increase when high-speed driving occurs on highways. In areas where just-in-time services are important, such as cargo transportation, shortening travel times is a very important issue. Some people may argue that AVs cannot be allowed to run beyond the current speed limit due to safety issue. Nevertheless, it is practically difficult to limit the maximum speed of an AV if it is owned by an individual because the purpose of a travel is to reach to a destination quickly and safely. If safety increases significantly, a traveler will choose to go faster while sacrificing safety to some degree [1, 23]. Therefore, this claim may be possible when AVs are adopted as a form of the semi-public transportation system. However, whether this restriction is environmentally preferable is an important issue that determines the ultimate direction of AVs. Our concern is the environmental impact when mitigating the speed limit.

Likewise, the prospects for AVs are mingled with positive and negative environmental impacts, and platooning and increased speed are one of them [1, 17]. It is unclear whether the adoption of AVs in large-scale with both of these technologies will have a positive impact on the environment and will be a solution to the problems raised above. Previous studies argue that AVs will have positive effects in improving the level of service in the congested area and smoothing traffic flow [24], and the fuel economy would be significantly improved [25]. These studies, however, are the results of measuring the effects at the individual vehicle level by testing on a driving circuit without traffic or conducting a mathematical calculation, and did not take into account the complicated mechanism within the traffic system. For example, traffic flow of a congested route becomes smoother when inter-vehicle distance decreases, but it becomes worse to some extent because other vehicles begin to use the improved route to reduce travel time. Also, some vehicles will choose a longer route instead of a shorter but congested route if high-speed driving become feasible, and this will improve the traffic flow of the congested route.

In this study, we investigated how the adoption of AVs works in a traffic system, and what is the consequent environmental outcome. We especially focused on the effects of road capacity increase due to platooning and speed limit increase on mobile emissions. Three scenario simulations including the null scenario (Scenario 1) simulation were conducted using a conceptual modeling approach (CMA). The null scenario simulation result is a snapshot of a situation where the road capacity is fixed and the speed limit is 100 km/h, which is a normal level. The result of Scenario 1 was used as a baseline scenario to illustrate the current situation to be compared with the results of Scenario 2 and Scenario 3. Scenario 2 simulates situations where the speed limit is unchanged and the road capacity increases from 0% to 100%. Scenario 3 simulates situations where the road capacity is fixed and the speed limit changes from 60 km/h to 140 km/h. The maximum speed limit was set based on the average speed of German Autobahn [26]. The Methods section summarizes CMA and assumptions adopted in this study. The simulation results and their implications were summarized in Results and Discussion section.

The development of platooning technology is environmentally desirable, but mitigating speed limit on motorways is environmentally unfavorable. According to Scenario 2, it is expected to reduce mobile emissions when the market share of AVs is less than 80%. For example, when the market share of AVs is 50%, the road capacity is increased by 18% [15]. In this case, when the inter-vehicle distance is decreased less than 20 m, or 0.7 second when vehicle speed is 100 km/h [15], the road capacity is increased more, and the mobile emissions reduction effect is expected due to the improved string stability (Fig. 1(a)). On the other hand, when the market share of AVs reached to 80%, the road capacity is increased by 50% [15], and mobile emission reduction rate starts to stagnate from this point. It is difficult to expect the environmental benefit from this point on. However, the 3rd quartile and max lines of Fig. 1(a) shows that increased road capacity is effective in mitigating mobile emissions in high-emission areas regardless of the market share of AVs.

The main reason for the decrease in the total mobile emissions is due to the increase in the average link speed (Fig. 1(b)). As link capacity increases, links with slower initial link speeds approached link speed around 65 km/h, which emits the smallest mobile emissions per distance. In particular, the lowest speed link with an initial speed of only 5 km/h increased to 28 km/h when link capacity increased by 100%. Since the traveling speed was limited to 100 km/h, links with high link speeds converged to 100 km/h with increasing link capacity. Therefore, there was no rapid increase in link emissions due to high-speed driving.

According to Scenario 3, the change in total mobile emissions shows a decreasing and increasing concave trend as the driving speed limit increases (Fig. 2(a)). The inflection point is at the speed limit of 95 km/h, and this is the result of the offset between the increasing links and the decreasing link. The emissions of the highest emission link have declined gradually, but there are links that show a rapid increase in emissions as the 3rd quartile trend line shows.

A plausible conjecture for the existence of links with rapidly increasing emissions is due to the traffic volume that bypasses the congested links at high speeds as the travel speed limit increases. This conjecture is based on the fact that the trend of maximum values in Fig. 2(b) is equal to the travel speed limits. As the travel speed limit increases, the average link speed showed an increasing near-linear trend but the variance of link speed has also increased. It is interpreted that although some traffic flowed at high speed by bypassing the congestion links and resulting higher emissions, the dispersed traffic volume was insufficient to solve the congestion.

Fig. 3 shows four cross-sectional results of link-level mobile emissions for Scenarios 2 and 3, respectively. Fig. 3.1 confirms that the emissions from the high-emission links such as (6, 8), (5, 9), (11, 12), and (13, 24) decreased sharply as the link capacity increases. In addition, Fig. 3.1(c) and 3.1(d) confirm that the emission reduction trend was stagnated when the link capacity increased by more than 50%. Fig. 3.2 confirms that there are links that show a steep increase in emissions as the speed limit increases. In particular, the increasing trend of emissions from links (1, 2), (1, 3), (3, 12), and (12, 13) demonstrates the previous conjecture that some vehicles generated more emissions by bypassing congested links with high speed.

In order to maximize the positive effect of AVs on the environment, continuous development and application of the platooning technology is necessary (Fig. 4(a)). Apart from the market share of AVs, the reduction of the headway distance makes it possible to increase the road capacity additionally, which makes it possible to increase the road capacity beyond 100%. However, easing the speed limit of motorways is not desirable from the environmental point of view because mobile emissions are increased due to vehicles that bypass at high speeds. These technical policy directions are effective not only in reducing the total amount of mobile emissions (Fig. 4(a)), but also in improving the areas experiencing extreme emissions (Fig. 4(b)). For example, if we follow the expected future direction, about 10% reduction of mobile emissions is expected, but if following the desirable future direction, about 30% reduction of mobile emissions is expected. To sum, it is clear that the vehicle platooning technology [7, 44] implemented at reasonable speed limits is one of the AV technologies that are very encouraging while increasing speed limit is not desirable from the environmental point of view.

Some people may argue that mobile emission will not be a concern anymore in the era of AVs because electric vehicles (EVs) will become widespread. Furthermore, the combination of shared vehicles (SVs) and AVs will be the ultimate form of future transportation systems [45]. It is foreseen that in such an era, individuals will not own cars anymore but will consume vehicle services [46]. It will be a form of semi-public transportation system that vehicles move around without drivers, and people will use smart devices to call vehicles and pay as much as they use. According to this prospect, we cannot help but agree that the ultimate form of AV is a solar-powered EV. However, there are still at least two important reasons why we should pay attention to the effects of AVs on the reduction of air pollution.

First, there may be a considerable time gap between the widespread of AVs and EVs. While AVs are expected to be adopted in large-scale in various sectors by 2030, EVs have numerous obstacles to be tackled such as technology and market competitive issues before large-scale adoption [47–49]. Second, GHG reduction is a very urgent global issue, and AV technology can lead to immediate GHG reduction in the transportation sector. According to a recent study, if the share of renewable energy is increased by 2% a year, the tragedy caused by climate change cannot be stopped after 2035 [50]. In the Intergovernmental Panel on Climate Change Sixth Assessment Report [51], only the reduction of the fuel-carbon intensity by increasing the share of renewable energy is considered as the GHG reduction strategy of the transportation sector. This is because the improvement of the engine technology has not lead to substantive fuel efficiency improvement due to the increase of the functions and the weight of vehicles [52]. However, the recent development of the ITS and the AV technology are expected to improve the fuel efficiency and reduce the GHG emissions accordingly.

It is not easy to predict all the changes that will occur caused by the large-scale adoption of AVs. Of the many predictions currently being discussed, the following three factors must be considered in subsequent studies. First, it is necessary to predict the increase in traffic demand due to the introduction of AVs. While the current study conducted with the fixed traffic demand assumption, AVs will increase traffic demand by providing mobility for the under-served population [2, 16, 53, 54], by shifting mode of transportation [17], and by providing easier travel [55–58]. Since the degree of congestion varies according to the traffic volume, the simulation results may change if the traffic demands changes. Second, we need to consider the system of combining AVs and car sharing services [59]. The AV taxi service was first introduced by Waymo in December 2018 in Phoenix, Arizona, USA [60]. Under such a system, AVs will no longer stay in parking lots, but will drive more distances to generate more profit. In this case, mobile emissions can be increased not by technological development, but by the way of people’s utilization or operation systems of AVs. Therefore, the operational aspects of AVs should be investigated along with the technical aspects.

Lastly, the development directions of AV technology and policies discussed in this study should be considered together with safety issues. The simulation results show that narrowing the inter-vehicle distance has an excellent effect in reducing the mobile emissions and spatial deviation. But the narrower the inter-vehicle distance, the higher the risk of a collision in unexpected incidents. Also, the higher driving speed is, the greater the damage to people and property in accidents. Although the current study shows that increasing the speed limit is not environmentally desirable, increasing demand for long-distance transport, including the cargo transport sector, can lead to a discussion of easing speed limits on freeways to reduce transportation cost.

This study assumes that the average link speed of the null scenario is 60 km/h in order to assume the general situation. This is the level of service at the suburban level [42]. However, the results of this study cannot be generalized to urban areas in metropolitan cities whose average vehicle speed is less than 30 km/h or to suburban motorways with average vehicle speed approaching the speed limit. In order to fill this gap, a study to measure the effect of introducing AVs with varying average speeds or to compare urban, rural, and suburban conditions should be conducted in subsequent studies. In addition, this study is useful for showing the overall trend because it uses the conceptual simulation model and simplified parameters, but it is inappropriate to predict the future that will occur in a specific area. Therefore, further studies that combine a more sophisticated model such as an agent-base model and the composition of the local vehicle fleet are required.