Student ID: U5569470

Lecturer: Chris Browne

Tutor: Vi Kie Soo

Tutorial: W09

Course Code: ENGN2226

Autonomous Braking A Comparative Study of Human, Hybrid and Autonomous Braking Control for Cars

U5569470

1

Table of Contents 1. Abstract .............................................................................................................. 2

2. Project Context – A Lethal Problem Overview of the severity of motor accidents as a social problem ........... 2

3. Project Scope – Vehicle Brakes Describing auto-safety and levels of intervention ................................ 3

4. Braking Systems – Stopping Distances and Autonomous Braking Key braking concepts ..... 4

5. Man versus Machine – Safety Performance Analysing safety performance through reaction times ........ 4

6. Saving Lives – A Socioeconomic Perspective Quantifying the costs and benefits of autonomous braking ... 6

7. Resource Efficiency – Energy & Materials Assessing the energy benefits and developing an experiment to

develop a better understanding of brake lifetime ..................................................................................... 10

8. Importance of the Hybrid Control Case (Case 3) Discussing the importance of the transition phase ... 12

9. Conclusion & Recommendations Summary of key findings ................................................. 15

10. Future Considerations What we need to be concerned about and what we can look forward to ................... 16

Appendix A – Back-of-the-Envelope Calculations: .................................................... 18

Bibliography: ............................................................................................................ 19

U5569470

2

1. Abstract Road traffic injuries are currently one of the world’s leading causes of death. The subsequent costs

to society are significant, both from an economic and a social standpoint, not to mention the

significant emotional impacts. Human error plays a key role in a substantial proportion of motor

vehicle accidents, creating a need to explore new ways to mitigate the safety risks posed by human

drivers – even if this disrupts the status quo of human drivers.

This study uses a systems approach to analyse the impact of employing automatic braking systems

(ABS) in cars – a system where the car uses sensors to identify hazards and automatically activates

the brakes. It was found that removing humans from the car-braking control loop could more than

halve the number of road accidents, leading to significant socioeconomic benefits, while also

improving the efficiency and material lifetime of the brakes. The importance of the hybrid human-

autonomous control case was also identified as a necessary step to improve human trust in

autonomous technology.

2. Project Context – A Lethal Problem Road traffic injuries are currently one of the world’s top ten leading causes of death (WHO, 2014)

and are the leading cause of injury-related deaths globally (Zhang et al., 2012). In fact, it is the third-

leading cause of deaths in China (GBD, 2014), causes an average of 11 deaths per 100 000 people in

the USA (CDC, 2013) and accounts for ~25% of deaths among persons aged 15-24 in Australia

(Risbey et al., 2010). A problem of such scale generates significant costs, with conservative estimates

at ~1.7% of GDP for nations like Australia and the United States (BITRE, 2009) (Blincoe et al., 2015).

But what exactly are ‘road traffic injuries’? The World Health Organisation defines a road traffic

injury as “a fatal or non-fatal injury incurred as a result of a collision on a public road involving at

least one moving vehicle” (WHO, 2015). From this, we can see that the serious issue of road traffic

injuries stems from vehicular collisions, with motor vehicles playing a role in the majority of injuries

(GBD, 2014) (Wang et al., 2008) due to their much higher velocities increasing their hazard potential

compared to pedestrians and cyclists. Thus, one of the most effective ways to reduce road traffic

injuries, along with their associated socioeconomic costs, is through reducing motor vehicle accidents.

Human error plays a major role in a significant proportion of accidents, being a critical pre-crash

event in 94%±2.2% of accidents studied in the National Motor Vehicle Crash Causation Survey

conducted in the US (NHTSA, 2015) (NMVCCS, 2008). This data agrees with other findings where

~95% of accidents were attributed to human error from both drivers and pedestrians (Sabey & Taylor,

1980) and, more recently, data from the United Kingdom that found ~90% of road accidents could

be attributed to human error on the part of either driver or pedestrian over the past four years from

2011 to 2014 (Dept. Transport, 2015). Other causes have a far less prominent role, with the next

most significant being environmental factors such as road surface or defective traffic signals (<10%)

and vehicle malfunction (~3%) (Dept. Transport, 2015) (NHTSA, 2015). Note that for the UK Dept.

Transport dataset, multiple contributing factors were attributed to some accidents. In order to avoid

U5569470

3

double counting, a conservative estimate of human-related contributing factors was made by

assuming non-human factors like road surface were completely independent of human error. This

was most likely an underestimate as factors like slippery roads were often paired with human errors

like travelling too fast for conditions. Nonetheless, even with this underestimate, human error was

still found to be a contributing factor in ~90% of accidents.

3. Project Scope – Vehicle Brakes There are two main ways to reduce the damage caused by motor vehicle accidents, thereby improving

human safety – i) crash avoidance, and ii) crash protection (SafetyNet Project, 2009). The latter is

concerned with minimising damage in the event of an accident, including technologies such as

seatbelts and airbags. However, this is essentially a form of damage control that protects the highest

priority element in the system – the human user, while failing to account for other damages such as

vehicle damage and traffic congestion. Crash avoidance is a more preferable way to improve vehicle

safety as it focusses on preventing the event of an accident altogether, thereby keeping people safe

while also avoiding the other negative effects of motor accidents.

As already mentioned, human error plays a leading

role in causing the serious problem of road injuries.

According to the hierarchy of risk control (see Figure

1), a standard tool for assessing the effectiveness of

risk control methods (Safe Work Australia, 2011), it

is clear that the best-case solution would be to

remove road traffic-related errors altogether. This is

in line with Elimination-level solutions, which are

the most reliable and provide the highest protection

by completely removing the hazard from the system.

However, this is an unrealistic target entailing

perfect driving, perfect vehicle performance and perfect conditions.

Instead, noting the significance of human error in causing road accidents and the major role of motor

vehicles in causing road injuries, the scope was narrowed to solutions based on control of the motor

vehicle. Collisions occur when a vehicle’s motion takes it into contact with another road traffic

element, whether it be another vehicle, a person, or an inanimate object. From the perspective of a

driver controlling a vehicle, the main ways of avoiding this are either changing direction or stopping

the vehicle through deceleration. This project chose to focus on the second option for the following

reasons:

1. The probability of collision between two vehicles is reduced to zero if neither is moving

whereas a steering vehicle can still collide with other objects on its new path

Figure 1 Hierarchy of risk control (Source: CDC, 2015)

U5569470

4

2. Stopping the vehicle is easier to implement as a solution based on steering would require all

paths of all moving objects be considered to avoid collisions on the new path

3. Deceleration will reduce damage if a collision cannot be avoided, whereas changing direction

will not necessarily decrease vehicle momentum and potential damage caused

Having chosen to focus on collision avoidance based around controlling the vehicle’s ability to stop,

it became logical to focus on the vehicle’s braking system.

4. Braking Systems – Stopping Distances and Autonomous Braking Braking systems control the deceleration of a vehicle, meaning they have the potential to avoid

collisions completely, while having the added benefit of reducing collision damage should one occur.

One of the most important metrics used to measure the performance of a braking system (and the

driver’s use of the system) is stopping distance, defined as:

The sum of the distance travelled by the vehicle while the driver reacts and the distance travelled by

the vehicle while decelerating to a stop.

The second distance mentioned in the definition is the braking distance, and it depends largely on

external factors beyond the project scope, such as road conditions, weather and the condition of the

actual brakes. However, the reaction time is within the scope, relying on driver performance.

Despite being only on the order of a few seconds, the high velocities of motor vehicles means that

even slight changes in reaction time can greatly affect stopping distance. For instance, using simple

back-of-the-envelope estimation we find that for a car moving at 100km/h, one additional second

required to react translates to nearly 30 additional metres travelled and increases impact velocity by

~23.4km/h (based on the typical deceleration for a car travelling at ~110km/h) (UK Govt, 2015).

Calculations for the distance estimate are shown in Appendix A.

This means that human error can have a significant impact on the effectiveness of the braking system,

with even slight mistakes meaning the difference between causing and avoiding a road accident. This

has driven a number of leading car manufacturers such as Audi, Mercedes and Volvo to invest in the

development of Advanced Emergency Braking Systems (AEBS) (ADAC, 2012), a new vehicle safety

technology that operates by using sensors to monitor the environment and activates the brakes if a

threat is imminent and the human driver doesn’t react (Thatcham, 2013). Referring back to the

hierarchy of risk control (Figure 1), this solution operates at a substitution level (the most effective

after elimination) by replacing the risk posed by human error with the risk posed by machine error.

5. Man versus Machine – Safety Performance We explored this concept of autonomous braking further by comparing three methods of braking.

Case 1 was the baseline case, based on the current-day situation with the brakes operated solely by

humans. Case 2 was a situation where brakes are operated entirely autonomously, with no control

U5569470

5

from the human driver. Case 3 considered a hybrid situation similar to the Google self-driving car

where the vehicle operates in autonomous mode unless the driver activates a manual override (Google,

2015). These three cases are summarised in the control feedback systems below:

As discussed earlier, the performance of the safety system relies heavily on the reaction time to

activate the vehicle’s brakes. Based on the control systems reaction time performance for each case

was estimated (see Figures 5 to 7). Note that the braking time was not considered as this varies with

vehicle velocity and the conditions of the road and brakes, all of which are independent of the brake

controller and hence beyond the scope of this project. Furthermore, for Case 3 we assume the case

where the human needs to manually override autonomous control. This is because this transition

stage sets Case 3 apart from the other two – when it operates only in either human-controlled or

autonomous mode it would simply behave like Cases 1 and 2 respectively.

Figure 7 PERT Chart for Case 1: Human-Controlled Vehicle Braking

Figure 6 PERT Chart for Case 2: Machine-Controlled Vehicle Braking

Figure 5 PERT Chart for Case 3: Hybrid Human/Machine-Controlled Vehicle Braking

Figure 4 Feedback loop for Case 3 – hybrid human-machine vehicle control (adapted from Isa & Jantan, 2005; drawn using Draw.io software)

Figure 4 Feedback loop for Case 1 - human-only vehicle control (adapted from Isa & Jantan, 2005; drawn using Draw.io software

Figure 4 Feedback loop for Case 2 - machine-only vehicle control (adapted from Isa & Jantan, 2005; drawn using Draw.io software

U5569470

6

Upper and lower reaction total reaction times, as well as distance travelled during the reaction time

assuming a vehicular velocity of 100km/h (27.8m/s) for context, are summarised below:

Table 1 Summary of Total Reaction Times and Distance Travelled During Reaction Time (at 100km/h) for each Case

Case # 1 2 3

Metric s m s m s m

Upper Estimate 3.5 97.3 0.2 5.56 5.2 145

Lower Estimate 1.3 36.1 0.011 0.306 2.62 72.8

Note: For 2 cars travelling headlong at 100km/h, the distance travelled toward each other during the reaction time for the lower estimates of Case 1

and 3 are 72m and 145m, respectively, compared with just 11m for the upper (slower reaction) estimate of Case 2

In terms of data reliability, it is reassuring to note that the brake reaction time for a human determined

in Case 1 agrees with a number of studies that have established a 95th percentile reaction time of

approximately 2.5 seconds (TRI OSU, 1997) (Taoka, 1989) (Sivak et al., 1982), with the unusually

high upper estimate of 3.5 seconds for Case 1 taking into account external factors like driver fatigue

and distraction rather than being a reflection of slower standard reaction times (NSW CRS, 2011).

This more conservative upper estimate is preferable to help simulate realistic scenarios where drivers

are under external influences that slow their reaction times beyond the normal distribution for

standard human reaction times, such as drunk drivers.

From a time perspective, autonomous brake control (Case 2) significantly outperforms the other two

cases, likely due to the significantly higher speeds of modern computers compared to human

processing (Hars, 2010) (Eno, 2013). Case 1 is next best, while Case 3 performs surprisingly poorly.

The reason for this is apparent in the feedback loop structure, with an additional step in Case 3 due

to the process of the human driver reviewing the decision of the machine controller and deciding

whether or not it is necessary to activate manual override.

Considering the best case for Case 3 (autonomous mode for hybrid) would perform the same as Case

2 while its worst case was worse than both Cases 1 and 2, it was not further analysed in the context

of safety, socioeconomic cost or resource consumption. The decision to exclude Case 3 from the rest

of the main study, along with its key role in the transition to autonomous vehicle technology, are

further addressed in Section 8.

6. Saving Lives – A Socioeconomic Perspective Clearly autonomous braking can reduce accidents and save lives. But by how much?

Based on US crash cause statistics (NHTSA, 2015), the following human errors were judged to be

affected by a fully autonomous braking system:

Driver inattention Inadequate surveillance

Driver fatigue Misjudgement of speed and distance

U5569470

7

Causes such as illegal manoeuvres and poor directional control were deemed to be outside the control

of an autonomous braking system’s ability to avoid collisions. However, these causes still accounted

for at least 55% of accidents (‘Other’ human errors, accounting for ~8%, were conservatively

assumed to be outside the influence of autonomous braking systems).

The impact of autonomous braking in reducing the effect of these errors was determined by

comparing braking controller reaction times with the time-to-collision (TTC) for motor vehicle

accidents at a point 3 seconds before collision. The mean TTC was found to be 2.56s, with a minimum

of 1.19s, while the mean TTC in near-crashes was 3.05s (Lee et al., 2007) which suggests that a

reaction improved by 0.5s can convert a crash into a near-crash event, greatly reducing damage and

cost. Comparing with the reaction times of Cases 1 and 2, we find that this means fully autonomous

can virtually eliminate those accidents caused by the errors it can influence. This is because even the

upper reaction time estimate of 0.2s for Case 2 allows for braking to be initiated more than 1s before

the lower estimate of 1.3s for human drivers (Case 1) – an improvement almost double the 0.5s

difference in TTC between a crash and near-crash event. Furthermore, the minimum TTC of 1.19s

for crash events does not leave enough time for human reaction, with the lower estimate being 1.3s

for Case 1. However, even the upper estimate for Case 2 would still allow braking to be initiated at

a TTC of ~1s compared to no braking in Case 1 – a potentially lifesaving difference considering 0.5s

can be the difference between a crash and near-crash. Based on mean TTC values we therefore find

that fully autonomous braking virtually eliminates those accidents caused by errors it can influence,

and so we assume it completely eliminates the accidents caused by these errors.

This reduced number of accidents using the autonomous braking system was then compared with US

data from 2010 (Blincoe et al., 2015), representing the human-controlled baseline case.

Table 2 Comparison of road accident statistics between baseline model (USA 2010) and autonomous braking system control model

Scenario Baseline 2010 USA Fully Autonomous Brake Control

Number of accidents (thousands) 13600 6057

Number of fatalities (thousands 33 14

Number of non-fatal injuries (thousands) 3900 1737

Number of damaged vehicles (thousands) 24000 10689

Total Cost ($billion USD) 836 372

The data shows that the fully autonomous braking control is much better for human safety, likely

more than halving the number of accidents and possibly having an even greater impact on reducing

serious road injuries and fatalities. However, in order to achieve the full benefits of autonomous

braking systems it is necessary for the public to first adopt this technology.

Cost is a significant barrier to the widespread use of autonomous vehicular control technology,

including autonomous brake control. The main costs associated with the fully autonomous vehicle

brake control system is the actual technology required to be fitted to the car. In fact, current devices

like the Google self-driving car rely on a vast array of sensors to create a map of its surroundings (see

U5569470

8

Figure 8) rather than relying on any external infrastructure (Undercoffler, 2014) (Guizzo, 2011). We

therefore consider the upper and lower estimates of the costs associated with installing the

technologies required for one autonomous vehicle, then scaled up to the number of light vehicles in

the USA during 2010 as a baseline reference.

Table 3 Cost of Sensory Technology for Autonomous Vehicle Control

Device Lower Estimated

Cost ($)

Upper Estimated

Cost ($)

Central computer 295 29180

GPS 80 6000

Lidar (light detection and

ranging) 90 8000

Odometry sensors 80 120

Radar sensors (short and

long range) 175 250

Ultrasonic sensors 15 20

Video cameras 150 200

Total (1 car) 885 43770

Total (2010 LV fleet USA)* 204 billion 10087 billion

*2010 light vehicle fleet size was estimated at 230.4 million (BTS, 2013)

Note that in Table 3, the upper and lower estimates are quite extreme and that the upper and lower

cost for one vehicle will likely lie closer to ~$10000 and ~$2000 respectively (BCG, 2015), which

would bring total fleet cost estimates to ~$2 trillion USD upper and ~$460 billion lower. Nonetheless,

the original estimates will be maintained to illustrate the full range of possible costs. In order to better

understand the overall cost-benefits, the baseline case was compared with the upper and lower

estimate scenarios for the fully autonomous case, taking into account the reduction in costs associated

with the autonomous technology (see Table 4 below).

Note that the ‘Total Societal Cost of Accidents’ of $836b USD for 2010 was based on the figure for

total value of societal harm from motor vehicle crashes in 2010 in the US determined by the NHTSA

(Blincoe et al., 2015). This figure considered productivity and workplace losses, in addition to costs

associated with medical, legal, emergency services, insurance administration and property/vehicle

damage. Some costs were considered using necessary assumptions, such as lifetime economic cost

society for each fatality ($1.4m USD), lifetime comprehensive cost (which includes economic and

quality of life valuations) to society for each fatality ($9.1m USD), cost of medical services and lost

productivity for critically injured survivor ($1.0m USD), and congestion costs being related to travel

delay, increased fuel consumption and subsequent negative environmental impact. Note that the

figure of $836b USD is a comprehensive cost figure – both economic and quality of life valuations

were considered. Quality of life valuation was determined by considering aspects such as the physical

Figure 8 Google self-driving car visual mapping (Inset: Human vision looking forwards) (Source: Knight, 2013)

U5569470

9

pain, impairment and emotional impacts that might reduce the quality of life for survivors (Blincoe

et al., 2015). It should be noted that this methodology agrees with the road crash cost components

identified by Risbey et al. in their approach to costing human losses (Risbey et al., 2010). The cost-

benefit analysis also assumed the cost of the base vehicle excluding sensors would be constant for all

scenarios, so this was not considered in the cost-benefit analysis below:

Table 4 Cost-Benefit Analysis of Autonomous Control versus Baseline, Considering Extreme and Moderate Versions of Upper and Lower

Estimates

Scenario Baseline Autonomous

Extreme Upper

Autonomous

Upper

Autonomous

Lower

Autonomous

Extreme Lower

Total Sensor Cost (2010 Light Vehicle fleet USA)

($b USD)*

0 10100 2304 461 204

Total Societal Cost of Accidents (Fleet)

($b USD)**

836 372 372 372 372

Net Comparative Cost (Fleet)

($b USD)

836 10472 2676 833 576

% of Baseline Cost 100 1250 320 99.6 69

Savings relative to baseline ($b USD) 0 -9636.35 -1840.35 2.65 259.65

*2010 light vehicle fleet size was estimated at 230.4 million (BTS, 2013); **Blincoe et al., 2015

From this result we can see that the extreme upper estimate has very low feasibility, increasing costs

by an unsustainable amount (nearly $9 trillion, which would have accounted for two thirds of the US

GDP of 14 trillion in 2010). However, we see that both lower estimates would reduce costs, with the

moderate lower estimate doing slightly better than breaking even while the extreme lower estimate

generates savings of ~$260b USD. From a purely financial perspective, it appears that adopting

autonomous braking will likely increase overall net costs. However, four additional issues should be

considered when looking at the cost of implementing autonomous braking.

Firstly, much of this technology is required for fully autonomous driving, meaning that on a longer

term this investment can significantly reduce the investment required to achieve the higher levels of

productivity and accident prevention afforded by fully autonomous vehicles.

Secondly, the technology costs will decrease over time, with systems such as autonomous valet

parking and highway autopilot technology predicted to decrease to less than a third of introductory

prices within ten years after introduction (BCG, 2015). This would reduce the moderate upper case

to only $300b losses and would increase the savings in the moderate lower cost case to a significant

$300b savings compared to the baseline.

Furthermore, this does not represent a yearly cost but rather an overall fleet analysis. Realistically,

these costs will be spread over a number of years as they entire fleet will not instantly change to

autonomous braking – it will instead be phased in, at best at the rate of new car sales, as this would

mean every car sold that year had autonomous braking. Using an average of ~7 million new light

U5569470

10

vehicles sold or leased (US Census, 2012), this translates to $292b and $55b in losses for the extreme

and moderate upper cases, with the more realistic moderate $55b matching investments such as the

US Government’s allocation to the Federal Emergency Management Agency. There has, however,

been a decreasing trend in new car purchases/leases, with only 5.7 million in 2010 compared to over

9 million in 1990 (US Census, 2012), indicating potential over-supply that may slow down the phase-

in of vehicles with autonomous braking systems.

Finally, it must also be remembered that these numerical cost benefits cannot quantify intangibles

like the improved emotional condition from avoided collisions, meaning the savings listed above can

be considered an ‘underestimate’ in that sense. In terms of ethics, this new ability to save a significant

number of lives may be sufficient to precipitate a change in the status quo by permanently increasing

overall investment in vehicle technology to continue to improve road safety.

7. Resource Efficiency – Energy & Materials Aside from the more well-known benefits regarding human safety, autonomous braking systems

also have the potential to improve the efficiency of material and energy use.

In fact, the potential of autonomous braking to improve fuel efficiency for cars by 10-20%

(Anderson et al., 2014) (Fagnant & Kockelman, 2014) is more significant than expected when

considering braking losses only account for 5-7% of total vehicle energy losses (US Govt, 2015)

(Consumer Energy Centre, 2015) (Bandivadekar et al., 2008). This seeming discrepancy can be

explained by the fact there are multiple benefits rather than simply reducing wasteful braking, such

as smoother acceleration/deceleration (Anderson et al., 2014) that can help avoid brake fade due to

excessively high temperatures from sudden braking (Stephens, 2006) and reduced congestion as

cars are able to stop closer to one another (Lari et al., 2014). These savings are significant, with a

one percent improvement in fuel efficiency equating to over $4b USD in savings (Forrest & Konca,

2007). This indicates that a saving of even 10% would equate to ~$40b USD savings – offsetting

~70% of the annual $55b USD cost estimated for autonomous braking technology based on the

moderate upper total sensor cost estimate. In terms of environmental impacts, this 10%

improvement equates to a saving of 2.7% of the US’s total greenhouse gas emissions from 2013,

this being ~180 million metric tonnes (EPA, 2014). Beyond the energy benefits, however,

autonomous vehicles also have the potential to extend brake lifetime.

The materials audit and distribution of embodied energy for a standard braking system is below:

Table 5 Materials Audit for Braking System

Part Material Qty (kg) Embodied Energy

Brake Disc Cast Iron 9.5 9.5kg × 25MJ/kg = 237.5MJ

Brake Pad Ceramic 2 2kg × 20MJ/kg = 20MJ

Brake Piston Stainless Steel 5 5kg × 56.70MJ/kg = 283.5MJ

Total Brake System -- 16.5 541MJ

*Data Source: (NREL, 2012) (Hammond & Jones, 2011)

U5569470

11

From the Sankey diagram we can clearly see that in terms of embodied energy, the brake pad is by

far the least significant. This indicates that in terms of saving energy, it is preferable to maximise

the lifetimes of the piston and disc in comparison to the pad, as the pad has ~10 times less embodied

energy meaning it is far more preferable as the replaceable part, though both the pad and disc

generally require quite frequent changing due to wear though performance can vary between 40-100

thousand kilometres (AA, 2011). However, the most ideal situation would be to increase the

lifetime of the entire system. It is generally believed that lifetime will be prolonged through

improved braking efficiency by avoiding unnecessary acceleration and braking cycles and hence

unnecessary wear of the braking system (Pyper, 2014) (Wang, 2015). However, it is also worth

considering that autonomously controlled brakes may brake more frequently as they will stop at

every legal requirement whereas humans can make judgement calls (though this decreases motor

safety). Due to autonomous driving technology still being very new, very little literature was found

regarding autonomous braking and material lifetime. An experiment is therefore proposed below to

better understand the impact of human and autonomous driving on brake lifetime. This will

hopefully assist in gauging environmental ramifications such as toxic waste and recycling

difficulties linked to ceramics (metals like iron and steel are much easier to recycle) (Reuter et al.,

2013) (EC, 2007) (Fenton, 1998).

We propose a case study of two fleets of 100 cars – one with autonomous brakes and the other with

human drivers, with the fleet size based on a notable previous study into driver behaviour (Dingus

et al., 2006).

Sample: Car users/drivers will be selected from volunteers who declare that they frequently

need to use passenger vehicles as a means of transport, with “frequently” defined as at least

eight times a week (based four return trips to work).

Controlled Variables: Geographic location will need to be controlled to within a city to try

and control for climate and driving conditions/cultural attitudes (more aggressive drivers can

cost 33% fuel efficiency (Karabasoglu & Michalek, 2013)).

Measured Variables: Condition of the brakes, including material lost from brake

components and change in brake effectiveness, will be assessed monthly for a year to

account for all seasonal variations – weather alters road friction and hence brake use.

Figure 9 Sankey Diagram for Embodied Energy in Braking System (Drawn using Sankeymatic.com)

U5569470

12

Comparative Study: The condition of the brakes will be compared between autonomous

and human drivers to develop an understanding of the impact of autonomous control on

brake lifetime compared to the status quo of human drivers

Additional Benefits: Any brake repairs or replacements will be noted to develop an

understanding of which brake components tend to wear the most severely

8. Importance of the Hybrid Control Case (Case 3) Case 3, the hybrid human-machine control case, was outperformed by Cases 1 and 2 in terms of

reaction time and hence safety performance due to the presence of manual override. However, it

should be remembered that the Case 3 time analysis only analysed the manual override situation

unique to the hybrid case as a worst-case scenario reflecting human proneness to panic in urgent life-

threatening situations, such as an imminent driving threat and subsequently overreacting to hazards

or even perceiving non-existent threats (Hartley & Phelps, 2012) (Raghunathan & Pham, 1999).

In fact, it would be expected that the majority of time would be spent in autonomous mode (Google,

2015), meaning a safety performance similar to Case 2 could generally be expected. This suggests

that Case 3 would generally be an improvement on the current real-world situation of Case 1. This is

further exacerbated by people’s (often unfounded) mistrust of machine-based autonomous decision-

making, such as algorithm aversion (Vedantam, 2015) (Dietvorst et al., 2014), which has been

specifically profiled for autonomous vehicles in a comprehensive public opinion survey conducted in

2014 with respondents from the US, UK and Australia (Schoettle & Sivak, 2014). The qualitative

data provided by the survey builds on findings by a number of other surveys that have found up to

88% of adults are uncomfortable with the idea of riding in an autonomous, driverless vehicle (Seapine,

2014) (Howard & Dai, 2013) (KPMG, 2013). Questions of interest in Schoettle & Sivak’s survey

and graphical representations of qualitative data (quantified using simple count and percentage-based

coding) are shown below. Note that although this survey considers fully self-driving cars rather than

just self-braking, the findings below still illustrate highly relevant points about the degree of human

trust in autonomous technology:

How concerned are about the following possible scenarios with completely self-driving

vehicles (Level 4)? (Note: Level 4 refers to fully autonomous vehicles requiring no human

control during the journey)

Figure 10 Graphical Representation of Quantified Coding of Qualitative Data from First Question of Interest (Schoettle & Sivak, 2014)

U5569470

13

This question is framed to generate qualitative data ranging from ‘Very concerned’ to ‘Not at all

concerned’ for respondents regarding how concerned they are about different forms of transport being

converted to self-driven. Notably, riding in a vehicle with no driver control is deemed the highest

concern (average of ~80% either moderately or very concerned). This is interesting because it

indicates people are most concerned about control of their private vehicles being relinquished – in

other words, giving away control of the vehicle they most often drive themselves. It is also interesting

to note that the question option is ‘Riding in a vehicle with no driver controls’. Generally people are

happy to be passengers in a vehicle with a human driver, even though as a passenger they have no

driving control, yet in the case of autonomous vehicles they become highly concerned. This may be

due to respondents interpreting the question to mean riding when they would normally be driving,

but it nonetheless shows that human unwillingness to trust self-driven cars is not only due to an

unwillingness to give up personal driving control, but also due to a general mistrust of the ability of

self-driving technology. The high concerns over commercial vehicles may again be due to phrasing

of the response options, with the relevant option being ‘commercial vehicles such as trucks’. Whereas

public transport is usually considered as something one rides in rather than a road obstacle, and taxis

are generally smaller than trucks, this emphasis of trucks may have influenced respondents to view

commercial vehicles as a hazard due to their larger size and subsequent dangers posed by trucks. This

combined with a lack of faith in self-driving systems may have contributed to commercial vehicles

being the second biggest point of concern (average ~78% moderately or very concerned).

“If you were to ride in a completely self-driving vehicle (Level 4), what do you think you

would use the extra time doing instead of driving?”

The use of the term ‘extra time’ for this question is interesting as it encourages the respondent to view

this time as a form of spare time rather than travel time. This phrasing encourages the respondent to

choose an alternative activity rather than watch the road, as this time is ‘extra’ as opposed to transport-

related. Even more interestingly, despite this phrasing the option to watch the road was still the most

commonly chosen at an average of 41%. This clearly suggests that people at present are still not

willing to fully trust autonomous vehicles, instead opting to continue watching the road – double

checking the machine controller’s every decision.

Figure 11 Graphical Representation of Quantified Coding of Qualitative Data from Second Question of Interest (Schoettle & Sivak, 2014)

U5569470

14

Clearly, people are generally unwilling to trust a fully autonomous vehicle at present it becomes

increasingly apparent that the hybrid model (Case 3) will be required to ease people from the Case 1

model into the Case 2 model. This is because the hybrid vehicle provides that backup layer of manual

control that acts as a perceived security net in the event the technology should malfunction.

The significance of this extra security net is revealed when comparing responses to the third and

fourth questions of interest:

“How concerned would you be about driving or riding in a vehicle with (Level 3) self-driving

technology?”, and “How concerned would you be about driving or riding in a vehicle with

(Level 4) self-driving technology?”

These two questions are identical except for the level of automation – Level 3 entails limited self-

driving technology where the driver is able to hand over control of safety-critical functions requiring

occasional human driver input; Level 4 is completely self-driving. In essence, Level 3 still gives

ultimate control to the driver (like the hybrid case) while in Level 4 the automated system holds

control. From Figure 12 we observe a significant increase in ‘Very Concerned’ respondents from an

average of 20% to 30% once the human loses ultimate vehicle control, with most of this increase

likely coming from previously ‘Moderately’ and ‘Slightly’ concerned respondents (indicated by the

fact that the number of ‘Not at all concerned’ respondents stays relatively constant between the two

levels of automation, with their lack of concern likely due to full trust in autonomous technology).

Note that the overall number of concerned people is ~88%, agreeing with another recent survey

regarding adults being worried about riding in driverless cars (Borcherding, 2014).

This demonstrates a strong correlation between degree of human control and human willingness to

adopt autonomous technology. The likely causation of this relationship being due to lack of human

trust in fully autonomous technology was also illustrated in the first two questions of interest. From

this it is clear that people are currently far more willing to adopt technology where they still hold

ultimate control. Considering autonomous braking is only one function (i.e. doesn’t encompass

controls like steering) it is more likely to be adopted by humans than a full autonomous package but

there will nonetheless be initial resistance as people require time to become accustomed to driving

Figure 12 Graphical Representation of Quantified Coding of Qualitative Data from Third and Fourth Question of Interest regarding Level 3 vehicles (left) and Level 4 vehicles (right) (Source: Schoettle & Sivak, 2014)

U5569470

15

with this new technology. As such, initially it would be preferable to implement a hybrid case where

humans still have the ability to control the brakes in order to enable a smoother and large scale

transition to a system where there is no human-brake interface (Edwards, 2014) – allowing complete

autonomous brake control and eliminating human-brake errors as in Case 2.

Furthermore, at this point in time the ability of autonomous vehicle control systems to navigate non-

urban environments is a very difficult challenge yet to be solved (Kelly et al., 2006), limiting the

vehicle’s range. All these factors indicate that the hybrid control case (Case 3) will have a vital role

to play in encouraging large-scale adoption of fully autonomous technology such as fully autonomous

braking (Case 2) in order to allow for the public to develop greater trust in autonomous control while

also giving researchers time to improve fully autonomous capabilities like the vehicle’s range.

Also, it appears that current trends in technology and human behaviour will necessitate a shift over

to completely autonomous driving. These trends and their consequences are as follows:

Trend Consequence

Increasing confidence in, and

use of, cell phones by drivers

This is representative of a broader overall increase in driver distraction with the advent of

mobile devices (NHTSA & NOPUS, 2011). Increased driver distraction increases

reaction time and subsequently reduces driver performance leading to more accidents

(Fitch et al., 2013). Makes human drivers more dangerous.

Reduction in driver ability as

they rely more and more

heavily on autonomous

control

This issue feeds into itself, with degrading driver ability encouraging heavier dependence

on autonomy, leading to further decrease in human driving ability (Mars, 2015). An

example of what is called the automation paradox, similar to erosion of airliner pilot skills

due to reliance on autopilot (Mars, 2015). Makes human drivers more dangerous.

Improvements in technology Improvements to relevant technology, such as LIDAR sensors that have a sweeping cycle

in the order of nanoseconds will greatly enhance the computer vision that helps control

the autonomous vehicle systems such as the braking control system. Makes autonomous

control safer

Our dependence on the autonomous control system once placed in the environment of a self-driving

car has already been documented with some drivers having taken up to 17 seconds to respond to

requests by the machine telling the human to take over (Lienert & White, 2015). Should these trends

continue in their current directions, it will become ever more necessary to adopt fully autonomous

control systems in cars to maintain an acceptable level of safety.

9. Conclusion & Recommendations

Based on the above analysis of braking control systems, it can be concluded that fully autonomous

braking systems offer the greatest safety performance out of the three cases considered. There were

clear benefits in terms of reaction time and reducing motor accidents, as well as noticeable energy

savings and potential to improve brake lifetime. There was uncertainty over the economic benefits

of implementing fully autonomous braking technology, with two scenarios (the extreme upper and

moderate upper sensor cost estimates) leading to significant additional costs rather than savings.

U5569470

16

However, it is recommended that these additional costs should not inhibit implementation of fully

autonomous brakes. This is because the use of economic estimates to value human life, though

financially valid, do not properly account for emotional factors and the reality of a poor quality of

life at a personal level. As such, the implementation of this technology should be viewed as an

ethical obligation rather than a savvy economic manoeuvre, potentially signalling a permanent

change in national investments in auto-safety. The one exception to implementing this type of

technology would be if the resources would save/improve more lives if diverted to another problem,

such as to combat an epidemic. However, considering the severe magnitude of the issue of motor

accidents in terms of human life and property damage, auto-safety would be expected to be a very

high priority, at least at present.

Furthermore, the hybrid brake control case was identified as playing a key role in assisting the mass

transition from human-controlled to the fully autonomous case. This is due to its importance in

allowing humans to develop trust in autonomous technology while giving them the feeling of

security associated with wielding ultimate control over the vehicle.

Consequently, this study recommends that hybrid braking be gradually phased into motor vehicle

industries until the public comes to widely accept and trust autonomous technology. At this point, it

is recommended that fully autonomously braking vehicles, without a human braking interface, be

phased into the market. Note that this point between hybrid and fully autonomous will likely occur

before market saturation, meaning that not all drivers need to transition to hybrid before fully

autonomous can begin penetrating the market. More likely, this critical point will correspond to the

point when the ‘early majority’ (as defined in Robinson’s Diffusion of Innovations) begin to adopt

the technology (~17th-41st percentile of the population) (Robinson, 2009).

10. Future Considerations

In terms of future work with, there are a number of aspects that will need to be considered with the

rise of autonomous control in vehicles – in this case, autonomous control of the braking system.

One of the most serious issues that will become apparent in the future is the evolution of new types

of accidents, including system failures (‘death by computer’) and cyberterrorism to maliciously

manipulate everything from targeted vehicles to large-scale traffic flows. Not only will these

necessitate innovations in safety and security, but they will also raise legal and ethical questions:

Who is accountable if someone is involved in an accident due to system failure?

Is it acceptable for someone to die or be injured in a situation where they had no control (e.g.

system failure) if there are less accidents overall?

There will also be a need to account for system dynamics such as human behaviour. For instance, if

people think fuel efficiency is better due to the autonomous technology there seems to be an effect

whereby they choose to travel more, offsetting any environmental benefits. This is known as the

rebound effect (Litman, 2013) and has been observed in the case of improved efficiency from

lightweighting vehicles (Stasinopoulos, 2013).

U5569470

17

Finally, it is important to note that there will always be a need to continue improving the performance

and reliability of autonomous control systems in an effort to reach the point of complete elimination

of accidents. The future holds potential developments like fully autonomous vehicles that could

further increase safety by removing the need for any sort of human driver interface at all. Without

this need for an interface, the space originally occupied by the interface could be redeveloped to

maximise safety, such as removing glass windshields as humans no longer have to always look out

the front window, or replacing the steering wheel and various meters with softer materials that help

reduce impact forces in the event of a collision.

With this in mind, it can be seen that we have already managed to achieve great improvements with

technology thus far – as in the case of autonomous braking. However, there is still room for

improvement, such as eliminating all the human error-related accidents altogether – that would be a

reduction of over 90%.

U5569470

18

Appendix A – Back-of-the-Envelope Calculations:

Estimating distance travelled in 1 additional second by car with velocity 100km/h:

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 × 𝑡𝑖𝑚𝑒; 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 =100𝑘𝑚

ℎ=

100 × 103𝑚

3600𝑠𝑒𝑐= 27.8𝑚𝑠−1

This means that in 1 second i.e. time = 1s

𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑡𝑟𝑎𝑣𝑒𝑙𝑙𝑒𝑑 = 27.8 × 1 = 27.8𝑚

U5569470

19

Bibliography:

(BCG), B. C. G. (2015). Revolution in the Driver’s Seat: The Road to Autonomous Vehicles. bcg

perspectives. Boston, Boston Consulting Group (BCG).

(CDC), C. f. D. C. a. P. (2015). "HIERARCHY OF CONTROLS." 2015, from

http://www.cdc.gov/niosh/topics/noise/images/HOC.png.

(EEA), E. E. A. (2013). Specific CO2 emissions per passenger-km and per mode of transport in Europe,

1995-2011. Copenhagen, European Environment Agency.

(Eno), E. C. f. T. (2013). Preparing a Nation for Autonomous Vehicles

Opportunities, Barriers and Policy Recommendations. D. J. Fagnant and K. M. Kockelman, Eno Center for

Transportation (Eno).

(EPA), E. P. A. (2014). "Sources of Greenhouse Gas Emissions." 2015, from

www3.epa.gov/climatechange/ghgemissions/sources/transportation.html.

(WHO), W. H. O. (2015). "Injuries, Traffic." Health topics. 2015, from

http://www.who.int/topics/injuries_traffic/en/.

ADAC (2012). Comparative test of advanced emergency braking systems. Germany, ADAC.

Anderson, J. M., et al. (2014). Autonomous Vehicle Technology:

A Guide for Policymakers. Washington, D.C., RAND Corporation.

Association, A. (2011). "Brake problems

Wear, corrosion, distortion and other common causes of failure ". 2015, from

http://www.theaa.com/motoring_advice/general-advice/brakes-discs-drums-pads.html.

Australia, S. W. (2011). HOW TO MANAGE WORK HEALTH AND SAFETY RISKS

Code of Practice. Australia, Work Safe Australia.

U5569470

20

AutoAnything (2015). "What are The Best Brake Pads: Ceramic or Semi-Metallic?". from

http://www.autoanything.com/brakes/the-best-brake-pads-ceramic-or-metallic.aspx.

Bandivadekar, A., et al. (2008). On the Road in 2035:

Reducing Transportation’s Petroleum Consumption and GHG Emissions Massachusetts, Massachusetts

Institute of Technology.

Blincoe, L. J., et al. (2015). The Economic and Societal Impact Of Motor Vehicle Crashes, 2010 (Revised).

N. H. T. S. A. NHTSA. Washington, DC, NHTSA, National Highway Traffic Safety Administration.

Borcherding, D. (2014). Study Finds 88 Percent of Adults Would Be Worried about Riding in a Driverless

Car Online poll survey from Seapine Software shows potential roadblocks for the future of driverless cars.

Cincinnati, Seapine Software. 2015.

Bureau of Infrastructure, T. a. R. E. B. (2009). Cost of road crashes in Australia 2006

Report 118. Canberra, Australia, Bureau of Infrastructure, Transport and Regional Economics (BITRE).

Bureau, U. C. (2012). U.S. Census Bureau, Statistical Abstract of the United States: 2012. US Census 2012.

U. C. Bureau. USA, US Census Bureau

US Government.

Bureau, U. S. C. (2012). Motor Vehicle Accidents - Number and Deaths (Table 1103). U. S. C. Bureau.

Maryland, United States of America, U.S. Census Bureau.

CDC (2013). "Deaths: Final Data for 2013." 2015, from http://www.cdc.gov/nchs/fastats/accidental-

injury.htm.

Centre, C. E. (2015). "Energy Losses in a Vehicle." 2015, from

http://www.consumerenergycenter.org/transportation/consumer_tips/vehicle_energy_losses.html.

Collaborators, G. M. a. C. o. D. (2014). "Global, regional, and natinoal age-sex specific all-cause and cause-

specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global Burden of Disease

Study 2013." The Lancet.

U5569470

21

Commission, E. (2007). Ceramic Manufacturing Industry. Europe, European Commission.

Council, A. C. and B. C. Council (2011). Slow Down...Keep Your Distance. New South Wales, Auburn City

Council

Blacktown City Council.

Council, N. S. (2014). Estimating the Costs of Unintentional Injuries, 2012. Illinois, National Safety Council.

Dietvorst, B. J., et al. (2014). "Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing

Them Err." Journal of Experimental Psychology: General.

Dingus, T. A., et al. (2006). The 100-Car Naturalistic Driving Study,

Phase II – Results of the 100-Car Field Experiment. Springfield, Virginia, U.S. Department of

Transportation, National Highway Traffic Safety Administration: 856.

Draw.io (2015). "Draw.io Software." 2015, from https://www.draw.io/.

DriverSide (2011). "Car Brakes: How Do You Know When to Change Them?". 2015, from

www.carcare.org/2011/10/car-brakes-how-do-you-know-when-to-change-them/.

Edwards, J. (2014). Here's The Most Obvious, Terrifying Flaw In Google's Self-Driving Car Prototype: The

'Panic Button'. Business Insider Australia. Australia, Allure Media.

Eldada, L. (2014). Today's LiDARs and GPUs Enable Ultra-accurate

GPS-free Navigation with Affordable SLAM. GPU Technolog Conference, San Jose, USA, Quanergy

Systems, Inc.

Fagnant, D. J. and K. Kockelman (2014). Preparing a Nation for Autonomous Vehicles: Opportunities,

Barriers and Policy Recommendations for Capitalising on Self-Driven Vehicles. 93rd Annual Meeting of the

Transportation Research Board. Washington D.C., Transport Research Board.

Fenton, M. D. (2001). Iron and Steel Recycling in the United States in 1998. Reston, VA, USA, US

Department of the Interior

U5569470

22

US Geological Survey.

Fitch, G. A., et al. (2013). The Impact of Hand-Held And Hands-Free Cell Phone Use on Driving

Performance and Safety-Critical Event Risk. Washington, D.C., National Highway Traffic Safety

Administration.

Forrest, A. and M. Konca (2007). Autonomous Cars and Society. O. Pavlov. Worcester, MA, Worcester

Polytechnic Institute.

Google (2015). Google Self-Driving Car Project

Monthly Report

August 2015. Google Self-Driving Car Project

Monthly Report. Google. USA, Google.

Government, U. (2014). "General rules, techniques and advice for all drivers and riders (103 to 158) " The

Highway Code. 2015.

Government, U. (2015). "Where the Energy Goes: Gasoline Vehicles." 2015, from

http://www.fueleconomy.gov/feg/atv.shtml.

Greco, C. R. (2008). Real-Time Forward Urban Environment Perception for an Autonomous Ground Vehicle

Using Computer Vision and LIDAR. Department of Electrical and Computer Engineering. Provo, Utah,

Brigham Young University. Master of Science.

Guizzo, E. (2011). "How Google's Self-Driving Car Works." 2015, from

http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/how-google-self-driving-car-works.

Hammond, G. and C. Jones (2011). Inventory of Carbon & Energy (ICE)

Version 2.0 Bath, University of Bath.

Hars, A. (2010). Autonomous cars: The next revolution looms. Inventivio Innovation Briefs. Nuremberg,

Inventio.

U5569470

23

Hartley, C. A. and E. A. Phelps (2012). "Anxiety and Decision-Making." Biological Psychiatry 72(2).

Hatamoto, M. (2015). "Australian politician speaks positively of autonomous vehicle tech." Technology in

Vehicles - News. 2015, from http://www.tweaktown.com/news/45530/australian-politician-speaks-

positively-autonomous-vehicle-tech/index.html.

Hoople, D. (2013). The Budgetary Impact of the Federal Government’s Response to Disasters. USA,

Congressional Budget Office.

Howard, D. and D. Dai (2013). Public Perceptions of Self-driving Cars: The Case of Berkeley, California.

93rd Annual Meeting of the Transportation Research Board. Online, Transportation Research Board.

Institute, T. R. and O. S. University (1997). STOPPING SIGHT DISTANCE AND DECISION SIGHT

DISTANCE. Discussion Paper No. 8.A. Oregon, Oregon Department of Transportation.

Isa, K. B. and A. B. Jantan (2005). "An Autonomous Vehicle Driving

Control System." International Journal of Engineering Education 21(5): 12.

Karabasoglu, O. and J. Michalek (2013). "Influence of driving patterns on life cycle cost and emissions of

hybrid and plug-in electric vehicle powertrains." Energy Policy.

Keep, M. and T. Rutherford (2013). Reported Road Accident Statistics. S. a. G. S. Section. London, House

of Commons.

Kelly, A., et al. (2006). "Toward Reliable Off Road Autonomous Vehicles Operating in Challenging

Environments." The International Journal of Robotics Research 25(5-6).

Knight, W. (2013). "Driverless Cars Are Further Way Than You Think." 2015, from

http://www.technologyreview.com/featuredstory/520431/driverless-cars-are-further-away-than-you-think/.

KPMG (2013). Self-Driving Cars: Are We Ready? USA, KPMG LLP.

Laboratory, N. R. E. (2012). U.S. Life Cycle Inventory Database, National Renewable Energy Laboratory.

U5569470

24

Lari, A., et al. (2014). Self-Driving Vehicles: Current Status of Autonomous Vehicle Development and

Minnesota Poilicy Implications. Minnesota, University of Minnesota

Transportation Policy and Economic Competitiveness.

Lee, S. E., et al. (2007). Analyses of Rear-End Crashes and Near-Crashes in the 100-Car Naturalistic Driving

Study to Support Rear-Signaling Countermeasure Development. Washington, D.C., National Highway

Traffic Safety Administration (NHTSA).

Lienert, P. and J. White (2015). Automakers, Google take different roads to automated cars. Reuters:

Technology. Online, Reuters.

Litman, T. (2013). Autonomous Vehicle Implementation Predictions

Implications for Transport Planning. Victoria, Australia, Victoria Transport Policy Institute (VTPI).

Markus Reuter, O. O., Finland and Aalto University, Finland; Christian Hudson, DIW, Germany; Antoinette

van Schaik,, et al. (2013). Metal Recycling

Opportunities, Limits, Infrastructure. United Nations Environment Programme. Online, United Nations

Environment Programme.

Mars, R. (2015). Episode 170: Children of the Magenta (Automation Paradox, pt. 1). 99% Invisible. R. Mars.

USA, 99% Invisible.

Mars, R. (2015). Episode 171: Johnnycab (Automation Paradox, pt. 2) 99% Invisible. R. Mars. USA, 99%

Invisible.

McGehee, D. V., et al. (2000). DRIVER REACTION TIME IN CRASH AVOIDANCE RESEARCH:

VALIDATION OF A DRIVING SIMULATOR STUDY ON A TEST TRACK. Human Factors and

Ergonomics Society Annual Meeting Proceedings.

NHTSA, N. H. T. S. A. (2008). National Motor Vehicle Crash Causation Survey: Report to Congress. N. H.

T. S. Administration. Washington, DC, National Highway Traffic Safety Administration.

U5569470

25

NHTSA, N. H. T. S. A. (2015). Critical Reasons for Crashes Investigated in the National Motor Vehicle

Crash Causation Survey. N. H. T. S. Administration. Washington, DC, National Highway Traffic Safety

Administration.

NHTSA, N. H. T. S. A. (2015). Distracted Driving 2013. Washington, D.C., NHTSA, National Highway

Traffic Safety Administration.

NHTSA, N. H. T. S. A. and N. O. P. U. S. NOPUS (2013). Driver Electronic Device Use in 2011 (NOPUS).

USA, NHTSA.

NRMA (2012). Cost of Road Crashes. New South Wales, NRMA.

Özgüner, Ü., et al. (2007). "Systems for Safety and Autonomous Behavior in Cars: The DARPA Grand

Challenge Experience." Proceedings of the IEEE 95(2).

Project, S. (2009). Vehicle Safety. Europe, European Commission, Directorate-General Transport and

Energy.

Pyper, J. (2014). "Self-Driving Cars Could Cut Greenhouse Gas Pollution ". 2015.

Raghunathan, R. and M. T. Pham (1999). "All Negative Moods Are Not Equal: Motivational Influences of

Anxiety and Sadness on Decision Making." Organizational Behavior and Human Decision Processes 79(1):

22.

Research, T. (2013). Autonomous Emergency Braking (AEB). United Kingdom, Thatcham Research.

Risbey, T., et al. (2010). Social Cost of Road Crashes. Australasian Transport Research Forum 2010,

Canberra, Australia, Australasian Transport Research Forum 2010 Proceedings.

RMIIA (2015). "Cost of Auto Crashes & Statistics." 2015, from

http://www.rmiia.org/auto/traffic_safety/Cost_of_crashes.asp.

Roads, Q. G. D. o. T. a. M. (2015). "Stopping distances." 2015, from

http://www.tmr.qld.gov.au/Safety/Driver-guide/Speeding/Stopping-distances.aspx.

U5569470

26

Sabey, B. E. and H. Taylor (1980). The Known Risks We Run: The Highway. Societal Risk Assessment. R.

Schwing and W. Albers, Jr., Springer US: 43-70.

Safety, N. C. f. R. (2011). How does speeding increase the chances and severity of a crash? New South

Wales, Australia, NSW Centre for Road Safety

NSW Government.

Sanders, B. (2014). "Repairing Our Infrastructure." 2015.

Schoettle, B. and M. Sivak (2014). SURVEY OF PUBLIC OPINION ABOUT AUTONOMOUS AND

SELF-DRIVING VEHICLES IN THE U.S., THE U.K., AND AUSTRALIA. Michigan, The University of

Michigan

Transportation Research Institute.

Shanker, R., et al. (2013). Autonomous Cars

Self-Driving the New Auto Industry Paradigm, Morgan Stanley Research.

Sierhuis, M. (2015). CIO Network: Humans Drivers vs. Driverless Cars. Wall Street Journal. G. Stern. WSJ

Website, Wall Street Journal.

Sivak, M., et al. (1982). "Radar-Measured Reaction Times of Unalerted Drivers to Brake Signals."

Perceptual and Motor Skills 55.

Stasinopoulos, P. (2013). A System Dynamics approach to Life Cycle Assessment. College of Engineering

and Computer Science. Canberra, Australia, The Australian National University. Doctor of Philosophy:

221.

Statistics, B. o. T. (2013). Table 1-11: Number of U.S. Aircraft, Vehicles, Vessels, and Other Conveyances.

USA, Bureau of Transport Statistics

US Department of Transport.

Stephens, A. (2006). Aerodynamic Cooling of Automotive Disc Brakes. School of Aerospace, Mechanical &

Manufacturing Engineering. Melbourne, RMIT. Master of Engineering.

U5569470

27

SY, W., et al. (2008). "Injury-related fatalities in China: an under-recognised public-health problem." The

Lancet.

Szybalski, A. T., et al. (2015). User Interface for Displaying Internal State of Autonomous Driving System.

USA, Google Inc.

Taoka, G. T. (1989). "Brake Reaction Times of Unalerted Drivers." ITE Journal 21.

Transport, D. f. (2011). Reported Road Casualties in Great Britain: 2011

Annual Report United Kingdom, Department for Transport.

Transport, D. f. (2015). Statistical data set:

Contributory factors for reported road accidents (RAS50). D. f. Transport. United Kingdom, UK

Government.

Triggs, T. J. and W. G. Harris (1982). REACTION TIME OF DRIVERS TO ROAD STIMULI. Human

Factors Report, Monash University Human Factors Group.

Undercoffler, D. (2014). Look, Ma, no hands: Google to test 200 self-driving cars. Los Angeles Times. Los

Angeles, Los Angeles Times.

Vedantam, S. and B. Dietvorst (2015). Why We Judge Algorithmic Mistakes More Harshly Than Human

Mistakes. Hidden Brain. D. Greene and S. Inskeep. Online, National Public Radio.

Wang, U. (2015). "Can autonomous cars be a boon for the environment?" Are self-driving vehicles good for

the environment? 2015.

Yang, S. (2015). Self-sweeping laser could dramatically shrink 3D mapping systems. Berkeley News.

California, UC Berkeley.

Yang, W., et al. (2015). "Laser optomechanics." Nature: Scientific Reports 5.

U5569470

28

ZHANG, X., et al. (2012). "Basic Characteristics of Road Traffic Deaths in China." Iranian Journal of Public

Health 42(1).

(Sabey and Taylor 1980, Sivak, Olson et al. 1982, Triggs and Harris 1982, Taoka 1989, Institute and University 1997, Raghunathan and Pham 1999, McGehee, Mazzae et al. 2000, Fenton 2001, Isa and Jantan 2005, Dingus, Klauer et al. 2006,

Kelly, Stentz et al. 2006, Stephens 2006, Commission 2007, Forrest and Konca 2007, Lee, Llaneras et al. 2007, Özgüner, Stiller et al. 2007, Bandivadekar, Bodek et al. 2008, Greco 2008, NHTSA 2008, SY, YH et al. 2008, Bureau of

Infrastructure 2009, Project 2009, Hars 2010, Risbey, Cregan et al. 2010, Association 2011, Australia 2011, Council and Council 2011, DriverSide 2011, Guizzo 2011, Hammond and Jones 2011, Safety 2011, Transport 2011, ADAC 2012,

Bureau 2012, Bureau 2012, Hartley and Phelps 2012, Laboratory 2012, NRMA 2012, ZHANG, YAO et al. 2012, (EEA) 2013, (Eno) 2013, CDC 2013, Fitch, Soccolich et al. 2013, Hoople 2013, Howard and Dai 2013, Karabasoglu and Michalek

2013, Keep and Rutherford 2013, Knight 2013, KPMG 2013, Litman 2013, Markus Reuter, MARAS et al. 2013, NHTSA and NOPUS 2013, Research 2013, Shanker, Jonas et al. 2013, Stasinopoulos 2013, Statistics 2013, (EPA) 2014, Anderson,

Kalra et al. 2014, Borcherding 2014, Collaborators 2014, Council 2014, Dietvorst, Simmons et al. 2014, Edwards 2014, Eldada 2014, Fagnant and Kockelman 2014, Government 2014, Lari, Douma et al. 2014, Pyper 2014, Sanders 2014,

Schoettle and Sivak 2014, Undercoffler 2014, (BCG) 2015, (CDC) 2015, (WHO) 2015, AutoAnything 2015, Blincoe, Miller et al. 2015, Centre 2015, Draw.io 2015, Google 2015, Government 2015, Hatamoto 2015, Lienert and White 2015,

Mars 2015, Mars 2015, NHTSA 2015, NHTSA 2015, RMIIA 2015, Roads 2015, Sierhuis 2015, Szybalski, Gomez et al. 2015, Transport 2015, Vedantam and Dietvorst 2015, Wang 2015, Yang 2015, Yang, Gerke et al. 2015)

{Bogart, 2015 #138;Robinson, 2009 #139}