MSc Dissertation: The Future of Mobility Part Two

Introduction

For many people the daily commute has become something of a ritual. Children go to school and adults go to work in adherence to society’s expectations that humans fulfil their productive capabilities by being in certain places at certain times. From this collective need has risen the transportation industry in all of its vehicular manifestations. The mass adoption of the personal automobile in particular has given the individual much greater freedom to travel within cities, regions, and even long-distance as they see fit. Those who can afford to do so drive daily, while applications on mobile devices allow the sharing of rides and vehicles with increasing regularity. This has resulted in many vehicles in daily operation on roads and highways around the globe, particularly in urban areas. This level of current use may prove unsustainable, with major drawbacks concerning public safety, traffic congestion, and environmental impact. Proponents of the Highly Autonomous Vehicle (HAV) have put forth a number of theories suggesting these problems could be greatly reduced with the widespread use of HAVs on public roads.

It is likely the benefits of HAV technology are already partially in effect, as various functions and systems within modern vehicles are increasingly being automated. This is described as “driver assistance” while the HAV would become “the ultimate driver assistance system” of the future (Neads, 2017). Examining current forms of driver assistance and investigating the sentiment of human trust that surrounds their use allows this research to gain further insight and understanding as to what barriers will need to be overcome on the road to fully autonomous driving. To do this, a mixed methods approach was used to collect background data from interviews, expert opinions, and observations, and a survey was developed from this information. This survey was used to interpret large scale barriers to HAVs in terms of individual human feelings of trust as well as usage behaviour. These were then related back to strategic demonstrations made by transportation industry leaders to gain a better overall sense of the barriers themselves and the timelines leading to their solutions in what could potentially be the next revolution in transportation.

Background Study of Autonomous Vehicles

Driver Assistance and Autonomy

               The evolution of the driverless car can be traced back to early innovative enhancements made to automobiles. Charlotte Bridgewood patented the automatic windshield wiper in 1917 (Professional Safety, 2011). Allen Breed patented the first sensor and safety system allowing airbags to deploy on impact in 1968 (Bellis, 2017). Since then, many incremental improvements have been made to further assist and protect automobile drivers, their passengers, and others who share the road. Virtually every vehicle on the road today has modern versions of these systems intended for the purpose of “driver assistance.” Modern cars have been described as computers on wheels (West, 2016). With this rapid pace of technological improvement, specifically the increasing level of computerisation found in modern automobiles, has come an increase in the amount of complexity these driver assistance systems are able to handle. 

               In 2016 the National Highway Traffic Safety Administration (NHSTA) of the US Department of Transportation (USDOT) called for “standardisation to aid clarity and consistency” in the definitions given for driver assistance systems found in vehicles (NHSTA, 2016), expressed in “levels of automation”. In an effort to do so they adopted the SAE International (SAE) definitions, which “divide vehicles into levels based on ‘who does what, when.’” (NHSTA, 2016) as follows (all quotes are directly from the policy guidance):

Level 0 (L0) – “The human driver does everything.”

Level 1 (L1) – “An automated system on the vehicle can sometimes assist the human driver conduct some parts of the driving task.” Examples of L1 include anti-lock brakes (ABS) and cruise control.

Level 2 (L2) – “An automated system on the vehicle can actually conduct some parts of the driving task, while the human continues to monitor the driving environment and performs the rest of the driving task.” An example of L2 is automatic cruise control, which keeps the vehicle a certain distance behind the vehicle in front of it, while the human driver steers (Neads, 2017).

Level 3 (L3) – “An automated system can both actually conduct some parts of the driving task and monitor the driving environment in some instances, but the human driver must be ready to take back control when the automated system requests.” An example of L3 is lane-keeping assist, which combines L2 automatic cruise control with vehicle-controlled steering to detect and remain in its current lane on the road (Neads, 2017). Bear it in mind that at any time should the vehicle have difficulty detecting where it needs to be the human is expected to immediately or quickly assume control, depending on the circumstances. To this end, these vehicles are also equipped with systems to monitor driver attentiveness.

Level 4 (L4) – “An automated system can conduct the driving task and monitor the driving environment, and the human need not take back control, but the automated system can operate only in certain environments and under certain conditions.” For example, a vehicle of this type could drive itself through urban environments in normal conditions, but could not be taken beyond a certain area; NHSTA describes this is as “geofencing”.

Level 5 (L5) – “The automated system can perform all driving tasks, under all conditions that a human driver could perform them.” There are no L4 or L5 vehicles on public roads at this time.

               SAE and NHSTA consider L3, L4, and L5 vehicles to have “highly autonomous vehicle systems” while defining such a system as “a combination of hardware and software (both remote and on-board) that performs a driving function, with or without a human actively monitoring the driving environment.” (NHSTA, 2016).

A Note on Terminology

Moving forward, this research shall follow the SAE & NHSTA standard of referring to any vehicle equipped with L3, L4 or L5 autonomous systems as a “highly autonomous vehicle” (HAV). “Self-driving vehicle” (SDV) will only be used in reference to test vehicles and prototypes. This shall provide distinction between vehicles operating under highly supervised conditions and/or in closed environments (SDVs) and those available to consumers for unrestricted operation on public roads (HAVs). The terms “driverless vehicle” and “driverless car” are ambiguous  and will be avoided, while the use of “driverless” will be reserved as a general reference to the entire ecosystem of development in this area.

The Unsustainability Problem

There are more vehicles on this planet today than ever before. The current number of vehicles in operation (VIO) registered in the US is 264 million (an all-time high) while the average age of a VIO in the US is 11.6 years (IHS Markit, 2016). That’s one VIO for every 1.2 Americans. Meanwhile in the UK, 2.6 million new vehicles were registered in 2016 (another all-time high), marking the fifth consecutive year of that figure’s growth (SMMT, 2017).

With more vehicles come more accidents. In its Global status report on road safety 2015 the World Health Organisation estimated 1.25 million road traffic deaths globally, 90% of which occur in developing countries (half of these involve motorcycles, cyclists, and pedestrians) (Bruun & Givoni, 2015). In 2015, an estimated 34,000 traffic deaths occurred in the US and another 1,800 in the UK (World Health Organisation, 2015). For the same year the National Highway Traffic Safety Administration reported that 94% of automobile crashes in the US were due to human error, with another 2% caused by critical vehicle failure (2015).

Most vehicles are concentrated in urban areas, where space is limited and expanding existing infrastructure is difficult and expensive. This means cities are becoming increasingly congested. Traffic congestion increases travel time while decreasing reliability (Karpilow & Winston, 2016). Meanwhile, idle drivers could be using their time differently, with the average driver spending 111 hours per year stuck in congested traffic (about 25 minutes per business day), costing $100 billion a year globally in lost productivity (Lukaszewicz, 2017). One study found that “highway congestion has significantly reduced the growth rates” in GDP, employment, and wages in several California counties (Karpilow & Winston, 2016).

This all-time high VIO figure has a tremendous impact on air quality as well. Most modern automobiles burn fossil fuels and release carbon dioxide (CO2), a processes acknowledged by many major scientific organisations  including the World Health Organization as contributing to climate change. Furthermore, carbon emissions also contain the pollutant gas Carbon Monoxide and pollutant nanoparticles of particulate matter (PM) pollution. In its Review of the UK Air Quality Index, the Committee on the Medical Effects of Air Pollutants (COMEAP) stated that in determining the toxicity of PM “studies have revealed that increased daily deaths, increased hospital admissions of patients suffering with heart and lung disorders, and worsening of asthma may occur” (2011). In its Ultra Low Emission Zone: Report to the Mayor, Transport for London (TFL) states “emissions from road transport are a major contributor to poor air quality in London” (2017).

Benefits of the HAV

As with all revolutionary technologies, there is much speculation about the impact of HAVs once their use becomes widespread (Litman, 2017). They have the potential to operate in a much more coordinated manner, unimpeded by the dangerous unpredictability of human error. Many theoretical use cases and scenarios exist for HAVs, including those where they remain in operation as a shared resource to improve mobility for people with limited access (Gruel & Stanford, 2016).

Human errors occur while driving for a variety of reasons. Conversing on a mobile phone while driving has been linked to failing to detect traffic signals (Strayer & Johnson, 2001). Human fatigue caused by sleepiness has been associated with accidents in technology-rich societies (Dinges, 1995) and car accidents (Powell et al., 2001). Drivers sometimes wilfully violate traffic laws or social codes of driving behaviour, which can lead to accidents (Parker et al., 1995). HAVs are not susceptible to distraction and fatigue, nor do they deviate from obeying the systemic rules which govern the environments in which they operate, such as right of way and speed limits (Russell-Carroll, 2014). This removal of human error would result in far fewer traffic accidents and the collateral damage they cause, potentially culminating in so-called “vision zero” of totally eliminating traffic deaths (US Burden of Disease Collaborators, 2013).

A privately operated vehicle only operates when it’s individual owner drives it. This requires extra time for parking at the destination, followed by the vehicle remaining parked and inactive for a high proportion of time relative to the amount it was driven. Alternatively, HAVs could be used in networked fleets (Musk, 2017), operating in a way that their movements relative to one another are coordinated through space and time (Burns, Jordan, & Scarborough, 2013). This would allow one HAV to provide rides for multiple individuals (“ridesharing”) which could also be done along a shared route toward a similar destination (“rideshare pooling”). Once the rideshare trip is complete, the HAV remains in active operation to satisfy the next logistically efficient demand request. One study simulated a city-wide replacement of private cars with HAV taxis in Berlin, concluding that one such vehicle could replace the demand served by ten conventionally driven vehicles (Bischoff & Maciejewski, 2016). This would obviously have a tremendous impact on reducing the amount of vehicles on urban roads and lessening the associated negative side effects.

A mix of HAV sizes could be deployed to meet a variety of purpose-specific passenger capacities (Simpson et al., 2017). This could be a low-occupancy “pod” that covers a less than a kilometre, a “robo-taxi” that carries one to seven passengers, or larger HAVs such as vans or busses. This has the potential to lower cost of use to consumers (Burns, Jordan, & Scarborough, 2013) thus increasing individual freedom of mobility. This is particularly beneficial to the 15% of the population whose disability or age prevents them from driving a vehicle (Simpson et al., 2017).

Some concern has been raised over whether or not HAVs will cause unemployment to rise, as many people are employed as commercial drivers. One study looked at 144 years of UK census data and concluded that technological innovation which occurred caused more jobs to be created than were lost during this time period (Stewart, De, & Cole, 2015). This study argues “the role played by technology in boosting employment often goes overlooked because of its more conspicuous destructive effects” (Stewart, De, & Cole, 2015).

The benefits of HAVs truly become transformative when this technology is adopted en masse (Howard & Dai, 2013). HAVs will be able to communicate from one vehicle to another (V2V), creating a sort of “mobility internet” where locally relevant data (road conditions, obstructions on the road ahead, traffic congestion, etc) could be shared for the benefit of all vehicles traveling in the vicinity (Burns, Jordan, & Scarborough, 2013). V2V could allow HAVs to better anticipate emergency response vehicles and detect them earlier (Lukaszewicz, 2017). Using V2V, HAVs could share the data they collect, enabling them to learn from one another while operating in a more cooperative fashion (Fleetwood, 2017).

Testing Driverless Technology

Road traffic deaths per 100,000 population is lowest in high-income countries at 9.2% (World Health Organization, 2015). It is on these safe and well-maintained roads where the first wave of driverless testing is happening. The US, UK, Sweden, and the Netherlands all have SDV testing underway, as do China, Japan, and South Korea, among other countries (West, 2016). Only 30-40 manufacturers in the world currently have the resources and capability to develop SDVs (Reed et al., 2017), so often testing them involves highly collaborative efforts between OEMS, component manufacturers, software companies, national and local government agencies, academic institutions, and private businesses.

               In 2010, Google announced it had a fleet of seven SDVs that had driven a total of 140k miles on public roads in California (Brown, 2011). Since then, US states have had varying responses to the potential that SDVs could be driving on their roads. Driverless testing in the US varies by state due to each state’s right to determine its own automobile regulations. Florida, Nevada, Arizona, Michigan, and Pennsylvania were among the first to pass legislation allowing SDVs and HAVs to be driven on public roads with minimal restrictions (Mitchell, 2017). Companies such as Uber, Tesla, Waymo (Google’s autonomous vehicle arm), Volvo, and Ford were quick to set up tests in these states. The University of Michigan has plans to operate a driverless shuttle (made by Navya, a French manufacturer) that moves slowly through its campus on a 2 mile loop (Slagter, 2017).

The UK is leading SDV testing in Europe (Shanahan, 2017) through the government’s Centre for Connected and Autonomous Vehicles (CCAV), which manages a £100 million Intelligent Mobility fund. These resources are invested in three major R&D projects (GOV.UK, 2017). Other funding goes towards feasibility studies and sponsoring several R&D competitions to stimulate business, energy, and industrial growth (GOV.UK, 2017). By demonstrating permissive regulations, innovative infrastructure test beds, and an active domestic automotive sector, the UK is making a strong effort to prove it can compete with other nations in its ability to advance autonomous technology (GOV.UK, 2017).

               While European auto manufacturers have been keen to promise increased economic prosperity should the EU prove to be globally competitive for driverless development, they have been met with slow reactions from EU countries: only two have conducted law reforms related to driverless (Jee & Mercer, 2017). Sweden is allowing testing in limited areas, while Germany has decided not to permit SDVs or HAVs on public roads at all (Jee & Mercer, 2017). Meanwhile, Asian countries have been more cautious; China has a national ban on all testing on public roads (China Daily, 2016) and Asian automakers such as Toyota and Honda have yet to make any major moves with HAV deployment, preferring small pilot projects for their SDV R&D (West, 2016).

State of the Industry

The move towards HAVs has begun. Manufacturers are pouring resources into research and development. As startups receive investment and software and component makers join the larger auto manufacturers (Hook & Bradshaw, 2017), the capabilities of driver assistance systems continue to increase as this technology approaches achieving high level autonomy.

In a 2015 interview, Tesla CEO Elon Musk said, "I almost view [the HAV] as a solved problem. We know what to do, and we'll be there in a few years.” (Roberts, 2015). The Tesla Model S sedan currently on public roads has L2 systems, with further driver assistance capabilities potentially available via software updates (Kessler, 2015). Audi recently released the first L3 equipped production car, the 2017 A8 (Ross, 2017). Ford is targeting 2021 for releasing its L4 HAV, which would make it the first manufacturer to deploy an HAV to the general public (Lukaszewicz, 2017). Based demonstrated R&D efforts by OEMS in addition to increased investment activity in automotive capital markets, one report forecasts 600K HAVs will be sold by 2025, followed by a decade of “substantial growth” resulting in a total of 21 million units sold by 2035 (IHS Automotive, 2016).

The American Auto manufacturing industry is currently in its 7th month of decline in sales after several consecutive years of all-time high sales (Marketplace, 2017), which has resulted in an the decision to cut overall production by US automakers (as opposed to lowering prices or increasing attractiveness of auto loans in an effort to stimulate demand). This is aligned with the forecasted shift towards manufacturing and selling less vehicles while also expecting them to be used more frequently, as evident by the ever-increasing use of ridesharing and carsharing. This trend towards shared use rather than outright vehicle ownership has ushered in the rise of what is called “mobility as a service” (MaaS). It is anticipated that MaaS shall experience tremendous growth over the next decade while also facilitating greater adoption of HAV technology (Thomas, 2017).

A note about this version:

This version of The Future of Mobility: Trust & Preferences in Highly Autonomous Vehicles by Sean Duggan (submitted to Goldsmiths University London in partial fulfillment for MSc Management of Innovation) has been slightly shortened and edited to produce a more condensed representation of the original document. Certain appendices and the reference list have been omitted but can be found in the primary document upon request, although the reader will notice that the original reference annotations remain within the text of this document. The references included 26 news articles, 5 books, 10 research papers, 17 journal articles, 6 government policy documents, 6 lectures/discussion panels, and 3 interviews with subject matter experts conducted by the author.