Towards Road Transport Automation: Opportunities in Public-Private Collaboration (2015)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

65 APPENDIX B: COMMISSIONED WHITE PAPER 2 Road Transport Automation as a Societal Change Agent Oliver Carsten, University of Leeds, Leeds, United Kingdom Risto Kulmala, Finnish Transport Agency, Helsinki, Finland White Paper 2 is a companion to White Paper 1, “Road Transport Automation as a Public–Private Enterprise.” White Paper 1 examines transport automation as a diverse technological and policy opportunity to systematically address the challenges of our current transport system; in contrast, White Paper 2 considers the changes involved for individuals, companies, governments, and society at large. Automation may have dramatic impact for road transport. This paper gives an overview of the potential impacts of auto- mation but also provides a critical examination of the additional costs that may be involved in the new technology. The rate of introduction of automation—and its breadth of application— will determine its overall impact on society, both positive and negative. Wide application of automation in transport could rep- resent a significant force for societal change, perhaps on a level with personal communication devices. Section 1 of this paper sets the scene, while Section 2 examines the impacts, benefits, and costs in the short and medium term, where medium levels of automation can be reached. Section 3 addresses the long-term changes that can be expected with high levels of automation. Finally, Section 4 draws conclusions and enumerates open questions. 1 setting the sCene In the near term, technological changes are likely to be incre- mental. There are two tracks of development. The first has a focus on increased assistance to the driver in motorway driv- ing with a gradual evolution toward fully assisted motorway driving in at least some situations. The second track focuses on the development and deployment of low-speed urban shuttles that have the capability for full automation (initially in circumscribed environments). Likewise, the impacts of such automated driving are likely to be incremental. Driving will be assisted in some locations on those vehi- cles that are equipped with the capability. It is likely that the motorway systems will be delivered first on high-end vehicles, that the first systems will be mainly autonomous (i.e., they will not require connection to other vehicles or to infrastructure), and that supported operation will ini- tially be on high-quality, well-regulated roads, that is, on motorways where the road and traffic situation can more easily be assessed by the system. A set of limitations on sys- tem usage may be imposed that will permit operation only under certain conditions—for example, in good visibility. The next step would be to extend capability to interurban roads, with urban capability being the final step. Precondi- tions for operation would gradually be reduced. Once full door-to-door capability under most operat- ing conditions is achieved, then true driverless vehicles will be available and will essentially provide a new mode of transport. That new mode is likely to be quite dis- ruptive in terms of socioeconomic impacts, and so the provision of this new capability or service may have revolutionary as opposed to evolutionary consequences. True driverless transport may be implemented first in shared or public transport, or both. In the near and medium term, driverless autonomous vehicles will trans- port passengers on streets and areas in urban communi- ties and transport terminals. Such vehicles are already on the market from manufacturers such as EasyMile.1 1 See http://www.easymile.com/.

66 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N Automated driving has many types of benefits, some direct and some indirect. The benefits originate at the individual level, in changes in the behavior of drivers and travelers with regard to driving and mobility, and con- clude with benefits at the social level via changes in the whole transport system and society, in which many of the current planning and operations paradigms are likely to be transformed by automated driving. There may also be disbenefits (e.g., in intensity of travel), particularly at the social level, that could result in additional congestion and increased use of natural resources. There may also be unintended consequences. For example, the impacts on public transport are not known. Driverless vehicles could provide a means to lower-cost service provision, but the availability of automated cars could lead to more car travel at the expense of collective transport. 2 near anD meDium term In this period, most of the changes are likely to be evo- lutionary, with a gradual introduction of higher levels of automation, particularly for privately driven vehicles. Additionally, urban pods may operate in limited and per- haps segregated environments. There are still some major issues in technology and design that need to be resolved, including the following: • Determining whether automated vehicles will have maneuvering capability, for example, the capability to carry out lane changes (some original equipment manu- facturers envisage such capability within a few years, but it is as yet unclear under what circumstances vehicles will have the authority to change lanes); • Maintaining driver situational awareness at medium levels of automated driving as the technology moves toward full automation, given that humans tend not to maintain attention over long periods of supervi- sory control; • Ensuring safety in mixed traffic, including when vehicles have different levels of automation; and • Ensuring the safety of vulnerable road users in interaction with automated vehicles (on motorways, interaction with motorcycles is the main issue). These issues are extensively discussed in White Paper 1. 2.1 Benefits 2.1.1 Individual Benefits For an individual, access to infotainment and the possi- bility to work or relax or just be connected while driving is likely to be the major motivation for highly automated driving. For many, this possibility will mean a major change in their lifestyle and improvement in their qual- ity of life. These changes could also make long-distance commuting by car more palatable and thus make pos- sible a wider choice of residency location. According to DFT (2015), an average driver spends 235 hours driving every year, during which he or she must concentrate on driving 100% of the time. In an automated vehicle, this journey time could be safely used however the occupant wished—working, reading a book, surfing the web, watching a film, or just chatting face-to-face with other passengers. Mohktarian (2015) has argued that the freedom to multitask is a significant factor in mode choice. It can be argued that some of the benefit of any such ability flows to an employer rather than just to the individual. However, in some countries regulatory change would be needed to allow the use of infotainment while driving. One significant issue is the interplay between levels of auto- mation and engagement in non-driving-related activities. Carsten et al. (2012) found that drivers were very willing to engage in nondriving activities while driving—particu- larly watching videos—even with Level 1 automation that provided only automated lane keeping. Admittedly, that study was carried out with a driving simulator, but it can certainly be expected that, even at the lower levels of auto- mation (Levels 1 and 2), drivers will wish to exploit the support to use their time in a more rewarding manner. For many individuals, the reduction of the risk of fines related to compliance with traffic rules and regulations may be a meaningful benefit. That benefit would apply also to current rules about using mobile phones and other devices while driving. There may be a need for legislation to change in this regard to accommodate non-driving- related activities under certain circumstances. The comfort of driving may be one of the main sell- ing points in the near term. Vehicles will be able to offer more and more automation for boring tasks; for exam- ple, long-distance driving on freeways and other high- ways will be supported by lane keeping combined with adaptive cruise control. Another individual benefit resulting from the increased level of safety offered by the new systems could be potential cost savings resulting from reduced insurance premiums. Short- and medium-term levels of automation could be attractive for elderly people, who may adopt auto- mated driving relatively quickly unless they find it too complicated to use. It is likely that manufacturers will limit their liability by setting a series of use restrictions. A major issue is public acceptance of these systems, which may limit the freedom of the driver with a variety of warnings when the driver engages in tasks other than driving. At lower levels of automation, the benefits will be restricted, as the driver will need to be prepared at all times to take control of the vehicle.

67A P P E N D I X B : C O M M I S S I O N E D W H I T E P A P E R 2 2.1.2 Social Benefits 2.1.2.1 Safety Effects on safety in the transitional period depend largely on the features of automation and on the penetration of vehicles with automated driving capability. One might expect some crashes on motorways to be avoided because of the fast reaction times of highly automated vehicles. There is also the potential for automated vehicles to have an effect on fatigue-related crashes, although operator sleepiness may be increased as a result of boredom in driving and of disengagement from vehicle control. Vehicles capable of high-level automation [Society of Automotive Engineers (SAE) Levels 3 and 4] will of necessity come equipped with an array of sensors and of crash avoidance systems. Those technologies will also be available to provide driver support and crash avoid- ance in manual driving and in driving at lower levels of automation (SAE Levels 1 and 2). Therefore, it can be expected that these vehicles will be safer in general operation. It can also be expected that the automation aspects will provide only a small additional benefit. The general safety effects of driver support systems have been estimated in a number of studies. eIMPACT examined 12 different driver support systems and estimated their fatality reduction potential to range from 1.4% to 16.6% (Wilmink et al. 2008). The systems evaluated were • Electronic stability control, • Full-speed-range adaptive cruise control, • Emergency braking, • Precrash protection of vulnerable road users, • Lane change assist (warning), • Lane-keeping support, • Night vision warning, • Driver drowsiness monitoring and warning, • eCall (an initiative to bring rapid assistance to motorists involved in a collision anywhere in the Euro- pean Union), • Intersection safety, • Wireless local danger warning, and • SpeedAlert (i.e., advisory intelligent speed alert). It was estimated that combining all 12 driver support systems together could reduce fatality by about 50% (Wilmink et al. 2008). The overall safety impact of these systems would naturally depend on their penetration into the vehicle fleet and their relative usage. 2.1.2.2 Efficiency and Capacity The reduction in shockwaves and crashes that will accompany increased driving under vehicle control should enhance capacity and efficiency. This enhance- ment is one of the major likely benefits of cooperation. However, there are also factors that could mitigate against enhanced capacity and efficiency: • Long vehicle platoons in the inner or a middle lane could act as an obstacle to lane changing and therefore inhibit overtaking. • Long vehicle platoons in the outer lanes could make merging in from an entrance ramp more difficult and could also inhibit access to exit ramps. • Dedicated lanes for automated vehicles could reduce capacity for vehicles with only manual driving capability. • In urban areas, any dedicated space for automated vehicles might be at the expense of other vehicular traf- fic. If automated vehicles require totally segregated space, then pedestrians and cyclists could also be nega- tively affected through loss of street space. The provision of vehicle-to-vehicle (V2V) communica- tion could mitigate against negative impacts on nonpla- tooned vehicles but would require (a) high penetration of V2V systems into all vehicles and (b) a consensus or set of regulations about operational rules, so that pla- toons could be broken apart to meet requests for road space from other vehicles. There is a potential need also for more general agreement or regulation concerning limitations on the operation of long platoons in weaving sections, and especially around exits and entries to the roadway. Other road sections where limitations might be needed are up gradients and places where the number of lanes reduces or is limited. 2.1.2.3 Environment Any automated driving will be more fuel efficient than manual driving because automated control is smoother than manual control and is less prone to the very late reactions often exhibited by human drivers. An auto- mated vehicle will drive in an anticipatory manner, which is at the core of ecodriving. Fuel savings will also be incurred by adherence to the speed limit in motor- way driving. Carslaw et al. (2010) found that on British motorways, there would be an overall savings in fuel and carbon dioxide of approximately 6% with even loose compliance of all cars with the standard speed limit of 70 miles per hour (112 kilometers per hour). It is also possible for vehicles under automated control to be per- manently engaged in a more elaborate ecodriving mode. However, environmental benefits are not likely to be substantial at lower levels of penetration and usage. There could also be environmental disbenefits as a result of the encouragement of long-distance car journeys

68 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N and of an increase in the attractiveness of long-distance commuting because the time spent in such commut- ing could be used more productively. An increase in long-distance commuting could promote urban sprawl, although again, the effects would be small at low levels of penetration. 2.2 Costs The socioeconomic impacts also include costs related to additional investments caused by the move to automa- tion. These costs are presented below on the basis of the stakeholder role in automated driving. 2.2.1 Individuals Additional investments will likely be required for driver training and road user education, especially at the medium levels of automation. Changes to education programs will be necessary to ensure that drivers and travelers are capable of driving an automated vehicle and are fully aware of the vehicle’s limitations and the conse- quences that those limitations impose on drivers. Drivers need to be aware of the circumstances in which they can give up control of the vehicle and when and how they should again regain control of the vehicle. Special licenses or permits to operate an automated vehicle may be needed if research, pilots, or first-use experiences indicate licensing to be useful. For instance, Level 3 automation requiring the driver to resume vehi- cle control within a specific short time period could be found to be too demanding for some drivers. 2.2.2 Vehicle Owners For the vehicle owner, automation comes at a cost. The cost of today’s technology packages, which bundle navigation, infotainment, and safety (including adaptive cruise control, lane-keeping assist, blind spot detection, and emergency braking) and provide the essential ele- ments for Level 2 automation, is in the range of $3,000 (J. D. Power 2014). Automated systems will require a degree of redundancy of safety-critical systems and com- ponents that could bring the price above this range; how- ever the price to the customer is difficult to predict, as it is heavily influenced by market factors. According to Öörni and Penttinen (2014), about half of the drivers polled in 2011 were prepared to pay for driver support systems (i.e., electronic stability control, blind spot monitoring, lane support system, advanced emergency braking system, speed alert, adaptive head- lights). Interestingly, they found that the proportion of people willing to pay for such systems had increased by 4 to 25 percentage points since 2009. The median value of the willingness to pay for a system ranged from €300 to €500. A focus group study by KPMG (2013) found that consumers were willing to pay a 15% premium for self-driving capability. 2.2.3 Infrastructure Owner–Operators Automated driving likely requires investments from the owners and operators of both road and information and communications technology infrastructure. For high- level automated vehicle performance in all conditions, there can be a need for • Potential special lanes or roads reserved for auto- mated vehicles; • Road markings and traffic signs, which need to comply with global standards and to always be kept vis- ible and in good condition; • Roadside solutions to facilitate automated driving also in adverse weather and on private roads, including forest roads; • Availability of infrastructure-to-vehicle (I2V) [and maybe vehicle-to-infrastructure (V2I)] capability; and • A system for cost recovery. Until the full-scale deployment of highly and fully automated driving, special lanes or roads could be needed to reap the full safety and efficiency benefits of automated driving. If automated vehicles need to inter- act with human-operated vehicles and vulnerable road users, the necessary functions to ensure safety will con- siderably reduce the mobility of the automated vehicles and the vehicle flow efficiency. In the transition period to full-scale automation, the building of special roads or lanes or reservation of special lane space for automated vehicles will lead to higher costs for road investment, operation, and maintenance. Therefore, such infrastruc- ture will only be built if a critical mass of automated vehicles exists. However, there are dedicated lanes that could be repurposed or designated for automated vehicle operation during certain periods. Road markings and traffic signs are necessary for the safe and efficient operation of automated vehicles. Road markings and signs need to be globally harmonized to the extent that vehicles will be able to interpret them correctly. This harmonization causes additional costs related to upgrading the markings and signs as well as to the harmonization process itself. The markings and signs should also always be kept visible and in good condition, which will also result in additional costs. For instance, in countries in which ice or snow, or both, covers roads frequently, winter maintenance costs may be doubled if

69A P P E N D I X B : C O M M I S S I O N E D W H I T E P A P E R 2 automated vehicle use is desired at all times, to ensure that road surfaces are clean of ice and snow (Innamaa et al. 2015). Roadside solutions to mark the road line will be needed to facilitate automated driving when road markings are either nonexistent or not visible. This is the case for gravel roads, narrow paved roads, and roads temporarily covered with snow, ice, or mud and also roads subject to poor visibility caused by dense fog or smoke. Even if satellite positioning is quite accurate, it tends to drift. Accurately positioned fixed objects on a digital map may be needed to maintain the accurate position of the vehicle on the road. Therefore, road operators that wish to facilitate automated driving on the road at all times should install specific landmarks such as fixed marker posts or poles alongside the road so that these will also be accurately marked in the digital maps used in automated vehicles (Dreher and Flament 2014). Such markers are analogous to reflector posts or win- ter maintenance guidance sticks placed along roadsides to provide visual guidance to human drivers in adverse conditions. Naturally, the installation, maintenance, and accurate positioning of such landmarks, posts, and poles will add to the road operator’s costs. I2V and V2I infrastructure could be a component of automated driving that improves traffic efficiency. The communication infrastructure to be provided depends on the communication solution and the road and traf- fic environment. Dedicated short-range communication (DSRC) beacons should be available at appropriate intervals to ensure full road coverage of a specific sec- tion. In urban areas, equipping signal controls at inter- sections with I2V-V2I communications could be the most cost-efficient option, as the electric power and infrastructure-to-infrastructure communications would already be available. Elsewhere, the provision of DSRC could be much more costly and likely restricted to hot spots where traffic problems would require the avail- ability of I2V-V2I communications. Cellular-based I2V-V2I communications could be the basic solution in other parts of the road networks. In the medium term, with future 5G cellular networks, no major changes in the communication infrastructure would be required, but in the existing and emerging 4G networks, some software and hardware modifications will likely be needed. To offset the increased infrastructure costs, cost- recovery mechanisms may need to be established. In the case of public infrastructure, costs can be recovered via taxation of vehicle ownership and use or fuel. Differ- ent road user charging schemes could also be set up to cover the investment, maintenance, and operation costs for infrastructure-related elements that facilitate auto- mated driving. Paradoxically, for policy reasons, higher charges might be imposed on manually driven vehicles to encourage the adoption of automated driving. Dynamic road user charging (e.g., via a distance tax) will be quite cost-effective to employ, as the necessary data collection, recording, and communication equipment may be read- ily available in the automated vehicle. Indeed, one might expect such taxation to be built into the usage fee, just as it is nowadays for taxi services. 2.2.4 Service Providers To cater to the needs of the higher levels of automated driving (Level 3 or higher), various service providers will be useful in delivering • Digital maps of sufficient quality for self-localization and environment interpretation and • High-quality, real-time traffic information, espe- cially for events, incidents, and congestion. Digital maps are useful in automated driving. Local dynamic maps are used as a central point to collect infor- mation for decision. Digital map information is used as an additional sensor to provide an electronic horizon for the automated vehicle, and map information is impor- tant in supporting positioning. Hence, automated driv- ing would benefit from highly accurate digital maps that include • Data on fundamental road features (lanes and their widths, physical and painted features), • Road malformations such as potholes and ruts, • Information derived from human drivers for humanlike automation (e.g., median trajectories, aver- age speed profile, median point of first breaking), • Specific landmarks in the street to increase position- ing accuracy (e.g., poles, shape of curbs, speed bumps), and • Information to facilitate evasive decisions (e.g., nature of the adjacent lanes, guardrails, detours). This information would be complemented by a feedback service from the vehicle to the map provider concerning detected discrepancies in the map data (Dreher and Fla- ment 2014). Provision of high-quality real-time traffic informa- tion may also add costs. High-quality information, espe- cially on events, incidents, and congestion, is needed for extended preview information outside the vehicle sen- sor range (Försterling 2014). Automated driving calls for much more accurate information than that provided by today’s traffic information services, especially with regard to event coverage, timeliness, and the location accuracy of the messages.

70 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N 2.2.5 Automotive Industry The costs for vehicle manufacturing may increase owing to the provision of the basic elements of automated driving: • Extended environmental sensing, • Accurate positioning, • Vehicle-to-everything (V2X) connectivity, • Need to preserve driver–occupant privacy, and • Need to ensure security. These costs may decrease in time with the mass produc- tion of automated vehicles, but it is likely that the rela- tive average cost of an automated vehicle will increase as automation levels increase. In addition, there will be additional costs related to standardization, vehicle dealer training, and vehicle ser- vicing, at least in the transition period to full automation. Remote diagnosis and remote software updating may be a must before more automation can be introduced. Costs for vehicle servicing will also be affected because of the capability of automated vehicles to detect wear, faults, and failures. There will be low fault tolerance, which would tend to drive up maintenance costs. There is concern as to the ability of parties other than franchised dealers to repair automated vehicles that is likely to have an impact on the costs of repair. EU legis- lation regarding access to repair and maintenance infor- mation requires that manufacturers commit to making repair information available on a nondiscriminatory basis to official dealerships and independent repairers alike, and certain minimum information must be included on websites as part of vehicle type approval. An automated vehicle is likely to be particularly complex and to utilize proprietary technology extensively, so manufacturers may not wish to permit or enable repair by other parties. They may be concerned that their intellectual property will be compromised if they reveal programing code, and they might also be concerned with the potential for those of criminal intent to gain knowledge that enables them to hack into vehicles (DFT 2015). Currently, as vehicles age, repair of the more complex and expensive systems on board can become uneconomic. If there is a problem with the automation systems, such vehicles may still be able to be used in manual-only mode. It is essential that safety be maintained, but at the same time, it would be preferable to avoid premature scrapping of vehicles, which damages sustainability and negatively affects those who cannot afford new vehicles (DFT 2015). Insurance-related costs are likely to be affected con- siderably for vehicle manufacturers if liability for driv- ing is transferred from the driver–vehicle occupant to the vehicle manufacturer at higher levels of automation, starting already from Level 3. Naturally, this change in liability will be offset by the change in drivers’ insurance. The overall change in insurance costs will depend pri- marily on the effects that vehicle automation has on the number and severity of crashes, and thereby the related insurance claims made. 2.2.6 Authorities The authorities likely need to set up regulations con- cerning automated vehicles, and doing this will require resources and investments in regulation, research, and cross-border harmonization. While the deployment of automated driving may reduce the overall level of risks related to road safety, it likely will lead to a major lia- bility shift among the stakeholders involved. The most common question raised with respect to automated vehi- cles is who would be held responsible in the event of a collision. There is a range of entities that may bear or share liability in road accidents: • Vehicle drivers, • Vehicle owners, • Vehicle operators, • Vehicle manufacturers, • Vehicle suppliers and importers, • Service providers, • Data providers, and • Road operators. Each of these parties may be found to be civilly (or in some cases criminally) liable to a greater or lesser extent, depending on the exact circumstances of the situation (DFT 2015). Concerning product liability, whether the product is judged to be defective or not is crucial. Defects include manufacturing defects, design defects, and failure to warn. Manufacturers and service providers have several liability defenses (e.g., the defect is attributable to com- pliance with mandatory requirements, such as domestic or European law; the defect did not exist in the product at the time the product went into circulation; or “the state of scientific and technical knowledge at the time when . . . the product [was put] into circulation was not such as to enable the existence of the defect to be discov- ered” (DFT 2015, 60). Contributory negligence will also be taken into account when an award of damages to a claimant is being considered. A court would need to consider whether the driver or vehicle occupant was sufficiently aware of the potential for a collision and take into account their abil- ity to avoid a collision. For example, the driver may have been taking advantage of the automated driving mode to undertake other tasks, so he or she either may not have been aware of an impending collision or may have been unable to react in time to intervene (DFT 2015).

71A P P E N D I X B : C O M M I S S I O N E D W H I T E P A P E R 2 Another liability-related issue is misuse of the vehicle. A vehicle manufacturer could argue before a court that the vehicle was being misused and therefore should not be found to have a defect. When a driver uses a vehicle in a manner that was clearly not intended or ignores a warning, the court might find that the vehicle was not defective (DFT 2015). As in the case of driver support systems, the auto- motive industry may propose a self-certification process to show that it has performed its duty of care during the development and design process of an automated vehicle and has performed all tests necessary to show that the vehicle is safe enough for operation on public roads. These tests may be carried out in simulations, driving simulators, test tracks, and, finally, on public roads. Whether there will be an additional need for a public conformance test is yet to be determined. There is also likely to be a role for independent testing, such as that administered by the Insurance Institute for Highway Safety and the European New Car Assessment Program. Such tests can effectively create pseudostandards. To settle liability issues in the event of a collision, automated vehicles are likely to be equipped with event data recorders. These recorders will indicate whether a vehicle was operating automatically or was in manual control at the time a collision occurred. They will also record how far in advance of a collision the mode of operation changed, a measure for which there may well be no other or better source of evidence than an event data recorder. These data will be a compelling source of information regarding what occurred and must be available to the relevant authorities for determination of liability and insurance responsibility (DFT 2015). There are also data protection and privacy concerns with automated vehicles. Any processing of data col- lected by an automated vehicle should, where an individ- ual can be identified, comply with data protection rules. Data are collected by the vehicle’s own electronic control units, event data recorders, and via the different sensors on the vehicle. This information can potentially be sent from the vehicle via the Internet to remote server storage. To comply with the fair processing requirements of data protection legislation, drivers and the registered keepers of vehicles should be made aware of the data that their vehicle is collecting and the uses to which it might be put (DFT 2015). The ownership of the (big) data pro- duced by automated vehicles needs to be resolved, as these data offer major business opportunities, even in the short term. Theft and security measures are also required to pre- vent vehicle theft and “hacking,” just as with nonauto- mated vehicles. Given the data that may be collected by a vehicle, such as GPS data and camera recordings, there may also be concerns that information on the move- ments of a vehicle or its location could be extracted without authorization, which would have implications for privacy issues and could potentially facilitate crimi- nal activities (DFT 2015). Certification and roadworthiness testing need to be developed for higher-level automated vehicles. In Europe, the vehicles will need type approval, and the framework for that approval needs to be enhanced to cover automated vehicles of all levels. The standardization of vehicle performance (accelera- tion, braking, time headway, response lag) as well as the warning tone or tell-tale to inform the driver of the need to take back control are issues to be resolved. Standard- ization of these items is proposed to prevent confusion among the public in moving between different vehicles. A global agreement of infrastructure requirements would be useful to clearly specify what is required to facilitate automated driving at higher levels. This agree- ment would include the global harmonization of road markings and signs to the extent than can be achieved. 3 long term In the long term, fully automated vehicles that operate door-to-door can be expected to have full freedom of movement with many of the substantial technological obstacles having been addressed. There is the potential for a wide range of vehicles to be automated—private vehicles, pods for both personal transport (individual and shared) and goods delivery, and public transport vehicles (buses and trams). 3.1 Transformational Potential As indicated earlier, fully automated driving would con- stitute a totally new mode of transport, the impacts of which are quite hard to predict (in the same way that the impacts of mass vehicle ownership and large-scale road freight on almost every aspect of social, economic, and cultural life could hardly be predicted at the onset of the 20th century). Ridesharing via automated collective transport could secure a substantial reduction in vehicle travel by reducing single-person use. However, that reduction presumes that worries about personal security and privacy can be over- come. The means to ensure personal security will be a big factor in automated collective transport because sharing unsupervised rides with strangers will probably be unac- ceptable. If concerns about personal security cannot be addressed, there could be substantial reluctance to use such services, and the availability of automated door-to-door transport at affordable costs could have substantial nega- tive environmental implications by increasing car use at the expense of walking, cycling, and collective public trans-

72 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N port. That in turn would also result in a negative impact on public health and even life expectancy. It could have the same perverse effect on journey time that large-scale vehicle use has had in the past in industrialized countries and that motorization is now having in industrializing countries: a reduction in journey times for early adopters followed by an enormous increase in congestion and reduction in travel speeds as vehicle usage in urban areas grows. Impacts on logistics operation could be consider- able, with driverless vehicles providing last-mile service for goods movement and delivery. Both on the freight side and on the personal transport side, there would be impacts on employment. In the long run, the occupations of taxi driver, delivery driver, and tram and bus driver may be threatened. In the intermediate term, regulations on driver hours might become less of a restriction on freight operation because vehicles could be used more intensively, perhaps even for 24-hour operation. That level of operation would almost certainly require some road zones in which full automation was provided. Fully automated vehicles may require dedicated road space in urban areas. That requirement has implications for the space remaining for other modes—automobiles, bicycles, pedestrians, and public transport. Less land would likely be required for parking because of increased vehicle and ride sharing and reduced vehicle ownership, thereby allowing more intensive land use in urban areas. It is estimated that currently in the United States there are up to eight parking spaces for each vehi- cle (Chester et al. 2010). Driverless vehicles could provide more accessibility to employment, particularly for low-income families who currently cannot afford a private car (or maybe a sec- ond car for the household) and who lack a variety of employment possibilities because of inadequate public transport. The positive social effects would be reduced unemployment and underemployment. However, there might also be negative environmental impacts, in that driverless vehicles might encourage long-distance com- muting and residential dispersion. Truck platooning and even limited-scale automated driving could reduce road haulage costs and thereby encourage even greater movement of freight by road. This could have wider implications in terms of harmful environmental effects and impact on other modes such as rail. 3.2 Benefits 3.2.1 Individual Benefits The highest levels of automation will provide individual mobility for people without a vehicle or driving license and for those with physical impairments. Among those who will benefit are the elderly and children. Those impaired by fatigue, illness, medication, alcohol, or drugs will also benefit. Others may simply not want to drive or be concerned about their ability to do so (DFT 2015). The individual benefits of automation will depend on how frequently the automated functions are switched on. For many drivers who enjoy manual driving and demonstrating their skills in it, automated driving may not appeal in normal circumstances. For these individu- als, the benefits of automated driving will be limited. There will be increased efficiency of time. People will get to places with greater certainty and more directly, because with full automation there is no need to find parking and to travel from parking to the actual destination. Automated driving at the higher levels will bring about various benefits to individual mobility. People will likely become less interested in owning a vehicle and instead subscribe to different on-demand services for vehicle or ride sharing. Individual mobility may become more affordable. There may also be an individual prefer- ence for procuring mobility as a service and not having to spend time on vehicle purchase, vehicle maintenance, and vehicle insurance. Service providers would presum- ably offer a variety of vehicles tailored to particular uses (e.g., commuting, family holidays, leisure activities). The cost per kilometer of vehicle use is expected to diminish with increased efficiencies in service provision. For a public transport user, autonomous driverless vehicles and people movers will likely provide smoother travel and improve possibilities for work and leisure activities during travel. It is also likely that driverless operation will shorten service intervals, thereby reducing both waiting times at stops and travel times. However, travel times may also increase if the vehicles are using the same space as vulnerable road users. 3.2.2 Social Benefits Automated driving will have major benefits for the trans- portation system in terms of the primary transportation policy objectives of efficiency, road capacity, road safety, and environment. 3.2.2.1 Use of Travel Time The social benefits of highly automated driving include the more efficient use of time, in that time spent while driving for work can be used more productively and without the safety risk of distraction. This issue has wider implications for the value of time spent in travel, a topic that is already being investigated by transport economists. There may be reduced willingness among

73A P P E N D I X B : C O M M I S S I O N E D W H I T E P A P E R 2 travelers to pay for journey time savings in driving, in that car travel will be less costly because the time can be used productively. 3.2.2.2 Safety The impacts on road safety are expected to increase with higher levels of automation, and full automation should assist in the elimination of serious road crashes, as the main risk factor of human error will be totally excluded. There is significant challenge, however, in being able to deliver interaction with drivers of nonautomated vehi- cles and with vulnerable road users (pedestrians, cyclists, and riders of two-wheeled motor vehicles). There is also a significant challenge in delivering systems with very low failure rates. In motorway driving, automated vehicles have the advantages of maintaining full attention at all times (they do not get distracted, fatigued, or impaired by alcohol and drugs) and of faster reaction times than human driv- ers. Under automation, vehicles will comply with regula- tions such as static and dynamic speed limits, and both car following and lane keeping will be enhanced because of control that is superior to human performance. Sensor limitations may, however, preclude automatic operation in challenging conditions such as snow. Safety can be further enhanced by the following technologies: • V2V communication to deliver cooperative adap- tive cruise control and smart platooning, which will help to eliminate shock waves and secondary crashes and could help to eliminate crashes in conditions of poor vis- ibility, such as fog, in which there currently are still sig- nificant multivehicle collisions that often result in serious injuries and fatalities; • Assisted lane changing to overcome failure to detect vehicles in the blind spot, which would be enhanced by cooperative V2V capability to deliver negotiated lane changes; and • I2V communication to notify vehicles of down- stream events beyond the visible horizon. 3.2.2.3 Efficiency and Capacity The effects of automated road transport on efficiency and road capacity are expected to be very high but will depend on the settings for following headway. The smaller the headways used, the higher the road capac- ity achieved. At low and medium levels of automation, shorter headways could increase crash risk because there could be a requirement for very fast driver reaction in takeover situations. The effects on efficiency and capacity also depend on the mix of vehicles at various levels of automation and on whether the automated vehicles are equipped with V2X or not. With V2X, automated driving carries much less risk of shock waves and shorter headways can be used. The U.S. DOT (2015) states, perhaps somewhat optimistically, “A fully automated automobile fleet can potentially increase highway capacity five-fold.” How- ever, there could be negative effects at lower levels of automation and in interaction with manually driven vehicles. For example, the ability of manually driven vehicles to change lanes (e.g., to overtake slow-moving trucks) could be impeded by automated vehicles driving in platoons with short headways. This scenario implies a potential need to manage the behavior of automated control and provide V2V communication to enable lane changing by nonautomated vehicles. Entrance and exit ramps might have to be managed in a similar manner so that platoons do not block intended maneuvers. Better lane keeping facilitated by automation would enable the use of narrower lanes for automated vehicles, so that more lanes could be fitted on the same carriage- way to increase road capacity. However, this is only achievable with dedicated lanes for automated driving. Interaction with motorcycle riders would have to be considered because filtering between such narrow lanes would not be possible. Better efficiency will also result if the increased use of vehicle sharing results in a reduction in vehicle miles or kilometers traveled. Such a reduction would lessen congestion and help counteract the effect of population growth on travel demand. There is also a large potential for vehicles to be used more intensively. This point is made by Schoettle and Sivak (2015). They argue that analysis of U.S. travel data indicates that there is con- siderable potential for vehicle sharing within households because trips do not overlap in location in time. Thus, if vehicles had a return-to-home capability, there would be less need for multiple vehicles within households. They conclude that ownership rates per household could be reduced by 43% and that individual vehicle travel (vehi- cle kilometers per year) could be increased by 75%. There is also the potential for operational efficiencies. The use of driverless buses and trams could lower public transport costs and thus act as a counterbalance to the use of low-occupancy door-to-door vehicles. Similarly, the costs of freight transport could be lowered with the advent of long-distance road trains (which should lead to labor effi- ciencies) and the use of automated pods for local delivery. 3.2.2.4 Environment Vehicles operating under automated control can be expected to save energy and reduce emissions because of

74 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N smoother driving (i.e., fewer harsh accelerations and decel- erations and cruising with less flutter in accelerator control than in manual driving). The maximization of such effects depends on manufacturers’ vehicle-control algorithms. Vehicle standards could provide a means to ensure such benefits. There is also the potential to use I2V communi- cation to actively manage energy consumption and emis- sions, along the lines of the programs for active emissions management already implemented on Dutch motorways. V2V communication is likely to enhance energy savings. Accident reduction would also result in energy savings by reducing network congestion resulting from incidents. Vehicle sharing would result in substantial energy sav- ings by reducing the energy consumed in manufacture. It could also reduce the land space allocated to parking because, with fewer vehicles being owned, there would be less need for parking in residential areas. However, some of that savings would be canceled out by the move- ment of empty vehicles around the network to cater to different demand patterns over the day and the week. The need to shuttle empty vehicles around has been noted in regard to urban shared bicycle schemes in cities such as London and Paris. 3.2.3 Service Providers Overall, as consumption of vehicle travel changes into use of services as opposed to ownership of one or more vehi- cles, huge economic opportunities are likely to open up for new service providers. Services such as those offered by Uber may be the precursors to that change. Other kinds of new services will likely emerge. Some big players, such as Google, Apple, and Nokia HERE, have already indicated interest in this potential. Software services for connected and automated vehicles, including the provision of info- tainment, could constitute a very large market (which may in part explain the interest of Google and Apple). The pro- vision of the software that sits on top of the basic vehicle platform, particularly for driverless vehicles such as urban pods, may be another huge market. 3.3 Costs The socioeconomic impacts also include costs related to additional investments caused by moving to automation. These costs are presented below on the basis of the stake- holder role in automated driving. 3.3.1 Individuals When the highest levels of automation are available and used, drivers of nonautomated vehicles, bicyclists, pedes- trians, and other travelers outside automated vehicles will be better off if they are aware of the behavior of automated vehicles. Information campaigns and aware- ness measures may also be required to ensure user accep- tance and uptake as well as nonuser acceptance. 3.3.2 Vehicle Owners The price of a fully automated vehicle could be much higher than that of a manually driven vehicle. However, the costs of vehicle use are likely to diminish with increased sharing of vehicles and higher intensity of use. Indeed, it is likely that there will be less actual ownership and more purchasing of vehicle use as a service. Shared ownership will likely impose some additional costs for managing the use, parking, and maintenance of the vehicle. Another possible option is that automated vehicles could be leased rather than sold to the public, thus allowing the manu- facturer to retain control and specify conditions, such as requiring that repairs or servicing be performed only by the manufacturer itself or other parties specified by the manufacturer (DFT 2015). This option would likely increase the costs of having the vehicle. In the long run, if all or most vehicles are fully automated, the urban robot vehicles or pods will likely be much lighter and perhaps also simpler than today’s vehicles, which will lower costs of vehicle ownership and use. 3.3.3 Infrastructure Owner–Operators Automated driving may require considerable investments from the owners and operators of both road infrastruc- ture and information and communications technology infrastructure. Investments for special lanes or roads dedicated to automated driving or manual driving, road markings, traffic signs, roadside solutions, and I2V-V2I infrastructure, which are useful already for Level 3 auto- mation, will increase for higher levels of automation because the coverage of the road network needs to be more comprehensive. In addition some new needs for investments arise: • Changes in road paving and repaving practices and costs, owing to narrower lanes and stricter lane keeping, and • Other changes in road infrastructure. Road paving and repaving practices may face major changes resulting from automated driving. Stricter lane keeping allows narrower lane width and therefore more lanes to be fitted on the same carriageway, which will improve road capacity. Narrower lanes will also mean that vehicles’ wheels run over the same parts of the road

75A P P E N D I X B : C O M M I S S I O N E D W H I T E P A P E R 2 cross section, which will focus pavement wear on narrow strips along the road, with the result of the formation of wear and deformation ruts on the road. Depending on the percentage of trucks with axles wider than those of automobiles, the ruts may also be wider. These ruts will necessitate shortening of the repaving cycle by perhaps 20%. Otherwise, or in addition, changes in road paving will be needed so that the narrow strips on which the vehicle wheels run will be equipped with material that better tolerates wear. This material with higher-quality aggregate for better wear resistance could be 10% to 15% more expensive to use. Furthermore, paving equip- ment could face major changes to facilitate paving of virtual rails on the road. In any case, the costs for pav- ing and repaving will be affected (J. Törnqvist, personal communication, Feb. 16, 2015). Facilitating automated driving may also mean higher asset management standards for operation and mainte- nance concerning, for instance, road pavement conditions. Automated vehicles may also make other changes in road infrastructure necessary. For instance, model- ing studies have found that, particularly at high flows, roundabouts are more efficient than traffic signals for automated vehicles with V2V communications (Azimi et al. 2013). Therefore, signalized intersections will likely be gradually replaced with roundabouts in the long term. In case of the establishment of urban zones restricted to automated public transport in addition to pedestrians and bicyclists, substantial investments could be needed. 3.3.4 Service Providers For Automation Levels 4 and 5, the quality of digital maps and traffic information must be at a very high level that will involve high maintenance and operation costs from the relevant service providers. As with nonautomated vehicles, there is a need for breakdown services to deal with broken down or other- wise stopped vehicles. On the one hand, a higher service level will likely be required for automated vehicles, so costs will increase. On the other hand, I2V communi- cations and accurate vehicle positioning may make the service more efficient. 3.3.5 Automotive Industry The automotive industry may change drastically, in that fewer vehicles will be in use (which perhaps will reduce income from servicing) but vehicles will be used more intensively. Relationships with service providers may be more important than relationships with individuals. There is a risk that the balance of power may switch to service providers, as has happened to some extent in markets such as mobile phones and television services (Internet protocol television). 3.3.6 Authorities The liability, security, harmonization, and standardiza- tion issues already addressed in the short and medium term will need even more efforts in the long term to deal with full automation. Security issues include, for instance, the use of driverless vehicles to commit crimes and as weapons. 4 ConClusions anD oPen Questions The impacts of automated driving are neither obvious nor simple, particularly with regard to fully automated door-to-door services. Nor will those impacts necessarily be favorable in all aspects. To a significant extent, out- comes will depend on how the introduction and roll-out of the new technologies and new forms of vehicles are regulated. Investments are required from a variety of stakehold- ers, and, as usual, those carrying the costs will not share the benefits of automated driving to the same degree. The benefits may also accumulate in a quite different place than the costs. These issues need to be solved, as do major institutional and legal issues. Certainly, there are many issues that should be opened up for public discussion, so that the process of arriving at the necessary political decisions can begin. Before that debate can take place, there needs to be more research that moves beyond pure technology development and human-in-the-loop or human factors studies to a deeper assessment of the wider impacts. referenCes Abbreviations DFT Department for Transport U.S. DOT U.S. Department of Transportation Azimi, R., G. Bhatia, R. Rajkumar, and P. Mudalige. 2013. V2V-Intersection Management at Roundabouts. SAE Technical Paper No. 2013-01-0722. SAE International Journal of Passenger Cars: Mechanical Systems, Vol. 6, No. 2, pp. 681–690. Carslaw, D. C., P.S. Goodman, F. C. H. Lai, and O. M. J. Carsten. 2010. Comprehensive Analysis of the Carbon Impacts of Vehicle Intelligent Speed Control. Atmospheric Environment, Vol. 44, pp. 2674–2680. Carsten, O. M. J., F. C. H. Lai, Y. Barnard, A. H. Jamson, and N. Merat. 2012. Control Task Substitution in Semi-

76 T O W A R D S R O A D T R A N S P O R T A U T O M A T I O N Automated Driving: Does It Matter What Aspects Are Automated? Human Factors, Vol. 54, No. 5, 747–761. Chester, M., A. Horvath, and S. Madanat. 2010. Parking Infra- structure: Energy, Emissions, and Automobile Life-Cycle Environmental Accounting. Environmental Research Let- ters, Vol. 5, No. 3, DOI:10.1088/1748-9326/5/3/034001. DFT. 2015. The Pathway to Driverless Cars: A Detailed Review of Regulations for Automated Vehicle Technolo- gies. Department for Transport, London. https://www .gov.uk/government/uploads/system/uploads/attachment _data/file/401565/pathway-driverless-cars-main.pdf. Dreher, S., and M. Flament. 2014. iMobility Forum Working Group: Automation in Road Transport, Digital Infrastruc- ture Subgroup. Presented in Antwerp, May 8, 2014. Försterling, F. 2014. Automated Driving: Key Application of ITS and Networked Car. Presented at ITS World Congress Board of Directors’ Meeting, Bordeaux, France, May 13, 2014. Innamaa, S., H. Kanner, P. Rämä, and A. Virtanen. 2015. Automaation lisääntymisen vaikutukset tieliikenteessä (The impacts of increasing automation in road transport). Finnish with English abstract. Traffic Research Reports, Helsinki, Finland, 2015. J. D. Power. 2014. J. D. Power Reports: Vehicle Owners Will- ing to Pay for Smartphone Functionality, but Not Con- nectivity. Press release. May 1. http://www.jdpower.com /sites/default/files/2014057_US%20_Auto_ET.pdf. KPMG. 2013. Self-Driving Cars: Are We Ready? https://www .kpmg.com/US/en/IssuesAndInsights/ArticlesPublications/ Documents/self-driving-cars-are-we-ready.pdf. Mokhtarian, P. 2015. Will Rising Trip Productivity Change Travel Choices? CTS Catalyst, Jan., Center for Transpor- tation Studies, University of Minnesota, Twin Cities. Öörni, E., and M. Penttinen. 2014. Study on Users’ Awareness and Demand for iMobility Technologies. iMobility Chal- lenge Deliverable 2.3, Version 1.0, Oct. 3, 2014. http:// www.imobilitychallenge.eu/files/studies/iMobility_Chal lenge_D2.3__User_Awareness_and_Demand_for_iMobil ity_systems_version_1.0.pdf. Schoettle, B., and M. Sivak. 2015. Potential Impact of Self- Driving Vehicles on Household Vehicle Demand and Usage. Report No. UMTRI-2015-3. University of Michi- gan Transportation Research Institute, Ann Arbor. U.S. DOT. 2015. Beyond Traffic: US DOT’s 30 Year Framework for the Future. http://www.transportation.gov/Beyond Traffic. Wilmink, I., W. Janssen, E. Jonkers, K. Malone, M. van Noort, G. Klunder, P. Rämä, N. Sihvola, R. Kulmala, A. Schi- rokoff, G. Lind, T. Benz, H. Peters, and S. Schönebeck. 2008. Impact Assessment of Intelligent Vehicle Safety Sys- tems. http://www.eimpact.info/download/eIMPACT_D4 _v2.0.pdf.