FutureNav
The Design and Evaluation of a Strategic Navigation Interface
for Semi-Autonomous Driving in Urban Environments

Master of Science Thesis in Interaction Design & Technologies

VELKO VECHEV
SIMON NILSSON

Department of Applied Information Technology
CHALMERS UNIVERSITY OF TECHNOLOGY
Gothenburg, Sweden 2016





Master’s thesis 2016:137

FutureNav

The Design and Evaluation of a Strategic Navigation Interface for
Semi-Autonomous Driving in Urban Environments

VELKO VECHEV, SIMON NILSSON

Department of Applied Information Technology
Division of Interaction Design

Chalmers University of Technology
Gothenburg, Sweden 2016



FutureNav
The Design and Evaluation of a Strategic Navigation Interface for Semi-Autonomous
Driving in Urban Environments
VELKO VECHEV, SIMON NILSSON

© VELKO VECHEV, SIMON NILSSON, 2016.

Supervisor: MinJuan Wang and Jonas Landgren, Department of Applied Informa-
tion Technology
Examiner: Staffan Björk, Department of Applied Information Technology

Master’s Thesis 2016:137
Department of Department of Applied Information Technology
Interaction Design & Technologies
Chalmers University of Technology
SE-412 96 Gothenburg
Telephone +46 31 772 1000

Cover: Photo from a real-world evaluation of the navigation prototype designed for
semi-autonomous vehicles.

Typeset in LATEX
Printed by [Name of printing company]
Gothenburg, Sweden 2016

iii



FutureNav
The Design and Evaluation of a Strategic Navigation Interface for Semi-Autonomous
Driving in Urban Environments
VELKO VECHEV, SIMON NILSSON
Department of Applied Information Technology
Chalmers University of Technology

Abstract
Turn-by-turn GPS systems are designed for todays manual drive cars, focusing pri-
marily on providing short-term instructions on steering the vehicle. In the future,
semi-autonomous vehicles will take over much of the steering duty, and in the pro-
cess, change the type of navigation information which is valuable to the driver. We
identify 2 scenarios where information provided by the navigation interface could be
valuable. The first is a challenge: to enhance the driver awareness of their surround-
ings in the case they have to take over operation. The seconds is an opportunity:
to provide strategic long-term control of the vehicle’s heading. In accordance with
these 2 scenarios, we conducted a 50 person survey to study the information needs
of drivers of semi-autonomous vehicles. The results indicate that these information
needs center primarily around Time, Location, Route Conditions, and Self-Driving
Parts. We also find that these vary between different trip purposes: social trips,
home trips, and shopping trips. To test the validity of these findings, we design and
construct a working prototype of a navigation interface, FutureNav, which provides
this information and gives drivers high-level control over it. Testing the hypothesis
that such an interface may increase driver awareness, we conduct an experiment in
the real-world comparing FutureNav to Google Maps as a baseline, using standard
measures for testing situation awareness on all 3 levels. We identify certain features
of FutureNav, such as the ghost car,map landmarks, and the 3D viewpoint as po-
tential sources of increased situation awareness. Finally, we reflect on our design
decisions, and the learning from conducting in-the-wild experiments that focus on
situation awareness, noting that such information may be useful for the design and
testing of future navigation interfaces designed for semi-autonomous vehicles.

Keywords: navigation, autonomous, vehicle, strategic, situation awareness, preview-
ing, tactical, future.

iv



Acknowledgements
We would like to thank our supervisors MinJuan Wang and Jonas Landgren for
providing invaluable support, keeping us on track, helping us to realize our ideas,
and improving ourselves as designers.
We would also like to thank Jan Nilsson at Semcon for providing inspiration for the
project, while trusting us with a fairly open design path. We also appreciate the
support given by Semcon as a whole.
Many people at Chalmers helped support us, including our ‘secret supervisor’ Dim-
itrios Gkouskos, Alexandru Dancu, and many of our fellow class members who par-
ticipated in our field studies, pilots, and design workshops.
We also want to thank the staff and partners at Volvo and Viktoria Swedish ICT
who provided early input and guidance, as well as excellent feedback on our early
prototypes.
Finally, we thank all the people involved along the way, our fellow designers, our
study participants, and friends who gave their support along the way. We appreciate
your time and thoughts.



Contents

List of Figures 1

List of Tables 1

1 Introduction 2
1.1 Research Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Project Goal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 AIMMIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Stakeholders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.7 Ethical issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Theory and Related Work 6
2.1 Automation in Vehicles and Driver Role . . . . . . . . . . . . . . . . 6
2.2 Situation Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2.1 Measuring Situation Awareness . . . . . . . . . . . . . . . . . 8
2.3 Human Factors in Automation . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Driving Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Navigation Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6 Multi-Purpose Trips . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Methodology 12
3.1 Research Through Design . . . . . . . . . . . . . . . . . . . . . . . . 13

3.1.1 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Invention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.3 Relevance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.4 Extensibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.2 Design process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3 Human Centred Design . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4 Expert interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 Literature Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.6 Field Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.7 Participatory Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.8 Prototyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.9 Strong Prototyping . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.10 Oblique Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

vi



Contents

3.11 Bodystorming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.12 Immersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.13 Insight Combination . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.14 Affinity clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

4 Information Needs & Scenario 21
4.1 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.1.1 Field Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.2 Literature studies . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.3 Information Needs . . . . . . . . . . . . . . . . . . . . . . . . 23

4.2 Driving Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Comparing Information Needs . . . . . . . . . . . . . . . . . . . . . . 25

4.3.1 Survey Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.2 Survey Pilot . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3.3 Survey Results . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3.4 Additional Survey Insights . . . . . . . . . . . . . . . . . . . . 27
4.3.5 Participant Comments . . . . . . . . . . . . . . . . . . . . . . 29

4.4 Discussion and Patterns for Future Use . . . . . . . . . . . . . . . . . 29

5 Prototype & Concept Development 31
5.1 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

5.1.1 Design restrictions . . . . . . . . . . . . . . . . . . . . . . . . 32
5.1.2 How might we? . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.2 Early revisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2.1 Co-Design Workshop . . . . . . . . . . . . . . . . . . . . . . . 34

5.3 Integrated prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.3.1 Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.3.2 Technical details . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.3.3 Remote . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3.4 Screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.3.5 The Ghost Car . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.3.6 Current location . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.3.7 Choosing destination . . . . . . . . . . . . . . . . . . . . . . . 40
5.3.8 Landmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.9 Routes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.10 Placemarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.11 Zoom levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

6 Comparative Study 43
6.1 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.2 Pilot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.3 Participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.4 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6.4.1 Environment and Routes . . . . . . . . . . . . . . . . . . . . . 45
6.5 Design and Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6.5.1 Scenario and Tasks . . . . . . . . . . . . . . . . . . . . . . . . 48
6.5.2 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

vii



Contents

6.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.6.1 Situation Awareness . . . . . . . . . . . . . . . . . . . . . . . 49
6.6.2 Subjective Measures . . . . . . . . . . . . . . . . . . . . . . . 50
6.6.3 Participant Comments . . . . . . . . . . . . . . . . . . . . . . 53

6.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.8 SA Aspects of a Navigation Interfaces . . . . . . . . . . . . . . . . . . 56

7 Designing Navigation Interfaces for Semi-Autonomous Vehicles 58
7.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

7.1.1 Challenges and Opportunities . . . . . . . . . . . . . . . . . . 59
7.1.2 Information Needs for Semi-Autonomous Driving . . . . . . . 59
7.1.3 Navigation Prototype . . . . . . . . . . . . . . . . . . . . . . . 60
7.1.4 Time and Accuracy . . . . . . . . . . . . . . . . . . . . . . . . 62
7.1.5 Connecting to Interface Features . . . . . . . . . . . . . . . . 63
7.1.6 Real-time Probes . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.1.7 Strategic Level driving . . . . . . . . . . . . . . . . . . . . . . 64
7.1.8 Navigation as the Nexus of Control . . . . . . . . . . . . . . . 64

7.2 Concept Iteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.3 Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

7.3.1 On Research-through-Design Methodology . . . . . . . . . . . 66
7.3.2 Designing for Future Interfaces . . . . . . . . . . . . . . . . . 67

8 Conclusion 69

Bibliography 69

A Appendix I
A.1 Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.2 Workshop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
A.3 Questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XIII
A.4 Icon Attributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XX

1



1
Introduction

New forms of mobility and technological advancements are changing the way people
move throughout their environment and particularly, in the city (Arthur D Little
Future Lab, 2013). The self-driving car is forecast to be at the center of this tran-
sition, reaching wide availability by 2025 (Berger, 2015). The space around vehicle
automation is undergoing rapid advancement, with vehicle manufacturers such as
Volvo, Mercedes, and Audi developing and testing initiatives. These companies are
developing what are known as semi-autonomous vehicles, able to drive on their own
in some situations, while still needing the driver to take over in others. Examples
of these include Volvo’s drive.me project which aims to continually test over 100
cars around highway roads (Victor et al., 2017), and the recent Mercedes test drive
which covers it’s founders historic test route from Mannheim to Pforzheim cover-
ing complex urban situations, this time in fully automated fashion (Ziegler et al.,
2014). On the other hand, technology companies are also pushing the state-of-the-
art, with Google’s self-driving car program logging over 1 million miles on public
roadways. These companies are preparing to leapfrog traditional manufacturers by
moving directly to fully autonomous vehicles without traditional control systems
like a steering wheel.

As Norman (2009) points out, successful automation brings about its own set chal-
lenges. Even the most advanced autonomous vehicles in testing today require a
supervisor to be available to take over, raising the issue of driver awareness. This
awareness can be influenced by a number of factors: the environment (Leshed et al.,
2008), the state and behaviour of the driver (Merat et al., 2012), and the interfaces
between the driver and the vehicle. In current vehicles, Advanced driver assistance
systems (ADAS) automate and enhance parts of existing interfaces, while in the
future, semi and fully autonomous vehicles will take this a step further by self-
navigating along given GPS paths. In this case, the navigation interface starts to
carry on more importance as the main nexus of control between the car and the
driver. Navigational interfaces in vehicles encompass both the challenges and op-
portunities of automation. The main challenge we identify is providing information
and control to the driver in such a way that supports their ‘situation awareness’.
By removing the need for constant operation of the vehicle, drivers will have the
opportunity to navigate in accordance with their high level trip goals. Studying the
needs of the driver of semi-autonomous cars, and how the navigation interface can
be designed to address these needs will be the focus of the work in this thesis.

2



1. Introduction

1.1 Research Problem

Two main research problems exist which are linked to the challenges and opportu-
nities of designing navigation interfaces for semi-autonomous vehicles.

Driving a semi-autonomous vehicle will not require the same level of attention as
driving a fully manual vehicle, therefore, when transitioning from autonomous to
manual driving modes, the driver may lack the information necessary to take com-
mand of the vehicle. This causes ‘out-of-the-loop’ problems where the driver’s lack
of situation awareness results in degraded performance (Endsley and Kiris, 1995).
The concept of situation awareness will be covered more comprehensively in the
theory section, but briefly, it refers to a drivers ability to understand, comprehend,
and predict their situation. Navigation interfaces, as one of the central information
hubs of the vehicle, will play a key role in keeping the driver aware of the situation,
both in terms of short term and long term readiness.

While addressing the problem of driver awareness is the foundation of the thesis,
there are additional problems that are related to this space. Drivers have gen-
eral navigation information needs, for example, spatial knowledge, environmental
complexity, and knowledge of traffic conditions. These are the navigation needs
where drivers need the most support (Münter et al., 2012). The context of semi-
autonomous driving will likely change these information needs, necessitating further
study in this area. In future driving scenarios, supporting drivers through technol-
ogy will still be a highly desired function (Gkouskos et al., 2015). Therefore, another
issue to address is how to map these needs to an interactive interface that can best
support drivers.

1.2 Research Questions

The research questions are formulated directly on the basis of the research gap
identified concerning using navigation interfaces as a way to raise situation aware-
ness. We propose one primary research question with sub-questions to investigate
its different aspects.

How can we design a navigation interface to support strategic level situation aware-
ness during semi-autonomous driving?

• What are the associated navigational information needs of drivers in this
mode?

• In which scenarios of the future are drivers likely to use this type of navigation
interface?

• Can we design and build a navigation interface that raises situation awareness
when tested under common usage scenarios?

3



1. Introduction

• If so, can we identify aspects of this interface which raise situation awareness?

1.3 Project Goal

Needs for the future are difficult to elicit by simply asking users (Van Elslande and
Fouquet, 2008). Therefore, one goal is to produce an interactive prototype that
might serve as a platform to evaluate the needs of navigation for semi-autonomous
vehicles. For example, various iterations of the prototype can test the mapping
of information to the different parts of a screen. Furthermore, as an artifact that
embodies the design research performed throughout the thesis, it can be used as a
reference design that can inspire further work in the field (Zimmerman et al., 2007).

A second project goal entails that the developed prototype be tested in the real-
world, as opposed to a simulator environment. This is in line with our vision to
create a real, working prototype, which we argue could only be tested to its full
capabilities in real-world scenarios (Broy et al., 2015).

1.4 Scope

First, the thesis focuses on problems related to drivers being out of the loop, which
inevitably is triggered during handovers to manual driving. There are already ex-
isting systems which handle the details of the handovers themselves, and thus, we
explicitly avoid this area. Instead, we restrict the area of study to the parts of the
semi-autonomous drive which are fully autonomous.

Second, the research will be focused on a particular subset of driving scenarios which
contain 3 of the most common type of destinations that drivers visit, are short in
length, and are set in an urban environment. Scenarios involving daily A to B type
trips will not be considered, and long trips on rural or high speed roads will also
not be considered. These scenarios are based on research showing the most common
usage vehicles in the most common type of setting (Krumm, 2012; Mackett, 2003).
The framework section contains a more thorough explanation behind the reasoning
on the developed scenarios.

1.5 AIMMIT

This thesis is part of the work packages contained within the Automotive Integration
of MultiModal Interaction Technologies Project (AIMMIT). As such, the research
and findings in this thesis are closely related to previous projects, particularly thesis
work that has contributed towards an overall vision.

4



1. Introduction

1.6 Stakeholders

This master thesis is in collaboration with Semcon, Chalmers University of Technol-
ogy, Volvo Cars and Viktoria Swedish ICT as a part of the AIMMIT research project.
As the primary stakeholders, Semcon has established the focus of the project and has
provided the office space, resources, and logistics needed to carry out the research.
Semcon bring considerable expertise and knowledge about the automotive industry.
Volvo Cars is a vehicle manufacturer and leading automotive technology provider.
The Volvo Group’s vision is to become the world leader in sustainable transport
solutions. As an industry partner in this project, Volvo provides feedback on tech-
nological approaches, and informing on high level industry direction. Chalmers and
Viktoria Swedish ICT provide additional academic support.

1.7 Ethical issues

When a car enters semi-autonomous control, the implementation by the car manu-
facturer will affect the safety of the driver and the bystanders. Issues arise when an
accident occurs on how responsibility is distributed between the supervising driver
and the autonomous car. Our research aims to increase the situational awareness of
the driver so that the result may be used to increase the safety of driving a semi-
autonomous car, since the driver will be able to instruct the car to avoid critical
situations and be in-the-loop in case a critical situation occurs.

One goal of the project is to evaluate a high fidelity interactive prototype in a moving
vehicle on road. Therefore, the safety of participants in the study will have to be
ensured. This could depend on the distraction factors related to the prototype itself,
and the type of simulation for autonomy that will be performed.

Certain parts of our research methodology involve the dissemination of user question-
naires. In this case, privacy plays an in important role in protecting confidentiality
of participants as well as considerations related to discrimination.

5



2
Theory and Related Work

This thesis aims to undertake work which is at the cross-roads of a number of
established and developing fields. The related works presented below serve to give a
brief overview in these fields, identify areas in which a research gap may exist, and
propose a focus that will be used to guide the thesis work in later processes.

2.1 Automation in Vehicles and Driver Role

Autonomous systems in vehicles vary in terms of their capabilities. Models for clas-
sifying automation in vehicles have been proposed with the aim of making clear the
precise degree to which a particular vehicle automates its functions. The US Na-
tional Highway Transportation Safety Agency (NHTSA) is a government regulatory
body which has issued a set of recommendations for classifying autonomous vehicle
capabilities. There are 5 levels of automation, from 0 through 4 2.1. This thesis will
focus on level 3 vehicles with limited self-driving automation:

“Vehicles at this level of automation enable the driver to cede full control
of all safety-critical functions under certain traffic or environmental con-
ditions and in those conditions to rely heavily on the vehicle to monitor
for changes in those conditions requiring transition back to driver control.
The driver is expected to be available for occasional control, but with suf-
ficiently comfortable transition time. The Google car is an example of
limited self-driving automation."

In this definition, The vehicle itself is functioning in what is known as semi-autonomous
(SA) mode, which implies the vehicle is fully autonomous only some of the time.
The driver of the car acts only as a high level supervisor, although their assistance
may be required with sufficient time to transition to lower levels of control. Sheridan
and Parasuraman (2005) propose a framework for supervisory control which has four
classes of automation centering around information acquisition, information analy-
sis, decision and action selection, and action implementation. This thesis will focus
on the first two classes aiming to provide timely, necessary and useful information
to the supervisor of the vehicle.

6



2. Theory and Related Work

0 1 2 3 4

No-Automation Function-specific Combined Function Limited Self-Driving Full Self-Driving

Figure 2.1: NHTSA Levels of Automation

2.2 Situation Awareness

Situation awareness (SA) is a way to model and measure a driver’s level of awareness
about their environment. Endsley and Kiris (1995) have put forth the most univer-
sally accepted definition of situation awareness as: “the perception of the elements
in the environment within a volume of time and space, the comprehension of their
meaning, and the projection of their status in the near future". The model as de-
scribed in 2.2 has 3 levels: Perception (SA level 1), Comprehension (SA level 2), and
Projection (SA level 3). In her seminal work, Endsley identifies short term working
memory, and attention resources as important constraints on SA. In addition to this,
design guidelines are proposed that could improve SA, for example: “grouping of
information in terms of spatial proximity" and “key in on the top level picture and
easily access greater specificity". These guidelines can help drive design decisions
when creating a navigation interface that support high situation awareness.

Although the concept of situation awareness was originally developed for military
purposes (Endsley, 1988), it has been adapted by others to many other domains,
including driving (Matthews et al., 2001; Bolstad et al., 2008; Sarter and Woods,
1991). As applied to the driving context, Sarter and Woods (1991) suggest breaking
down awareness into the following components: spatial awareness, identity aware-
ness, temporal awareness, and goal awareness such as a navigation plan. Matthews
et al. (2001) builds on this definition, and combines it with Endsley’s model, suggest-
ing that SA during driving is related to the driving goal 2.3. Finally, Matthews et al.
(2001), proposes a research gap which could be addressed by navigation systems:
“Navigation and route-guidance systems illustrate this approach. These systems
have the potential to enhance SA at all three levels".

7



2. Theory and Related Work

Figure 2.2: A model for situation awareness by Endsley.

2.2.1 Measuring Situation Awareness

2.3 Human Factors in Automation

Previous work in the field of automation has shown the importance of considering
human factors when it comes to system design. (Billings, 1991) proposes general
guidelines for automation in aircraft including a set of principles for human-centered
automation that could be equally applicable in vehicle automation. Endsley and
Kiris (1995) characterize out-of-the-loop performance problems as a major conse-
quence of automation where operators have difficulty in taking back manual con-
trol. In their study, higher automation conditions produced lower situation aware-
ness (SA) and an increase in decision time. This was attributed to the difference in
level of information processing, where more automation resulted in a more passive
understanding. Similar studies have been carried out in the automotive domain
Julia Niemann (2011), showing that an active processing paradigm involving the
driver can decrease reaction times during control decisions. Thus, when designing
for higher situation awareness in vehicle interactions, active information processing
can be employed to keep the driver in-the-loop.

8



2. Theory and Related Work

Figure 2.3: How different driving goals affect SA.

In any system where multiple levels of automation are available, transitions can
introduce significant problems related to mode confusion (Sarter and Woods, 1995)
and context switching penalties (Monsell, 2003). Schieben et al. (2011) propose
a layered scheme regarding transitions in automated systems. As applied to au-
tonomous vehicle operation, the four layers consist of (1) explicit switch by the
driver, (2) vehicle proposes, (3) car takes over (minimum-risk maneuver), and (4)
fluid transitions based on implicit parameters such as grip strength of the wheel.
In their study, during take-over requests by the vehicle (layer 2) drivers had diffi-
culty understanding that they needed to perform an extra step to confirm they’re
in control. This suggests there may be an opportunity to design for layer 4 type
transitions, by providing smooth transitions between their level of readiness before
asking to take over control.

2.4 Driving Levels

When designing for interactions in the car, it is important to understand decision
making process of the driver and the kind of information they need to make those
decisions. Michon (1985) breaks down the driving task into three levels: operational,
tactical, and strategic. The strategic level involves high level goals of route plan-
ning, modal choice, and risk evaluation. Tactical level is about obstacle avoidance,
overtaking, and negotiating the immediate surroundings on the time span of seconds
whereas operational level maneuvers spans milliseconds and provide basic controls
of the vehicles such as steering and braking. Keskinen et al. (2004) suggest that
strategic driving can be motivated by even higher goals such as learning to drive or
lifestyle, which are referred to as ‘Goals for life and skills for living’. Current driver
support technology such as Advanced Driver Assistance Systems (ADAS) are more
focused on the operational level, and thus there is an opportunity to design systems
that offer support on the strategic level (Davidsson, 2009). One example of strategic
level support is Tesla’s route planning for long distances, which takes into account
EV charging station locations (See Fig 2.4).

9



2. Theory and Related Work

Figure 2.4: Tesla Trip Planner for EV trips.

Source: http://www.theverge.com/transportation/2015/4/16/8430037/tesla-range-
assurance-hands-on

2.5 Navigation Theory

On a more fundamental level, navigation in a car can be related to human wayfind-
ing behaviour. Cognitive mapping, as defined by (Golledge and Garling, 2002),
consists of establishing locations, understanding distances and relative directions,
and the ability to convert a mental map into knowledge of the surrounding physical
environment. Essentially, the purpose of a cognitive map is to allow a person to
get from one place to another. The next step in the process involves some decision
making in the form of route choice which can be selected a priori (planned) or en-
route (dynamic) (Golledge and Garling, 2002). Dynamic route selection presents
an interesting case, as it is often assumed that the selection criteria should stay
constant throughout a trip, however, a study by (Golledge, 1995) showed the op-
posite. For example, whereas ‘least time’ was the 2nd most important criteria at
the start of the journey, the itinerary undertaken by the participants showed that
in reality, it was 6th in importance. (Bovy and Stern, 2012) refer to this selection
process as ‘adaptive route choice’, which are decisions taken by the traveller based
on changing circumstances. The process that encapsulates all of the above steps,
and guides the traveller through the environment is called navigation (Golledge and
Garling, 2002). Based on the above definitions, the work in this thesis will focus
on providing information that will specifically assist in making en-route decisions
during navigation.

Golledge and Garling (2002) suggest two methods of acquiring route knowledge,
survey and experiential. This 2 methods roughly correspond to the two types of
decision making. The survey method is used when a viewing the landscape from
above, in a ‘birds-eye-view’, while the experiential method refers to physically walk-
ing or moving through a particular environment. Anchor point theory states that we

10



2. Theory and Related Work

build up cognitive maps based on specific locations we identify as salient Golledge
and Garling (2002). These locations are highly personal, although they can also be
common, for example, a church, or a mall, or a person’s home. Through the sur-
vey method, such as looking at a GPS map, or physically viewing all anchor points
from above, we can link their relative locations. Similarly, by experiencing a route,
either by moving through it, or by previewing it, a person’s route knowledge can
be increased. Very few, if any navigation interfaces stress both of these methods of
route knowledge, for example, providing detailed 3D landmarks which are custom to
a person’s journey, while allowing for a quick way to zoom out and see all locations
together.

2.6 Multi-Purpose Trips

Motivations behind vehicle trips was analyzed by (Krumm, 2012) which looked at
data from over 150 thousand U.S. households. The most common destination is
going home, which accounts for 34 percent of all trips, followed by ‘buying goods’
(i.e. errands) and going to work at 12 and 11 percent respectively. However, in
3 destination sequences in combination with home trips, ‘buying goods’ is followed
closely by ‘pick up someone’ and ‘visit friends/relatives’. This suggests that although
work trips may be very common, in more dynamic, non A-to-B type trips, social
needs play a more significant role. A study in the U.K focusing on short trips
(Mackett, 2003) found similar results, placing shopping and social activities as the
two highest usage scenarios for vehicle use. Multi-purpose trips become interesting
in the context of route selection, as many studies have shown that trip purpose is the
primary affecting factor in route criteria (Mackett, 2003; Wachs, 1967; Mahmassani
and Shen-Te Chen, 1993). It therefore becomes the imperative of the navigation
interface, to be able to adapt and provide information that is contextual to the trip
purpose.

11



3
Methodology

The following chapter will explain the research process of this master thesis.

The research were conducted using a design based research approach. We do so to be
able to iteratively produce something that can easily be conceived and interpreted
by all stakeholders and it can be efficiently evaluated in a way for human needs to
be met. Research through design also gives the possibility to produce artifacts that
produces a level of detail that is impossible to attain in a written account (Gaver,
2012).

The process will be performed in a Human-Centred Design ethos. The interdepen-
dence of human-centred design activities is explained in ISO 9241-210 (ISO, 2010).
This is an iterative design process which is not linear but each stage depends on the
output from the other.

The process will be executed iteratively through three major steps:

1. Eliciting information needs in semi-autonomous navigation. Elicit-
ing the important information required for car drivers to efficiently navigate
around an urban commuting scenario.

2. Ideation and implementation of a multimodal prototype that ad-
dresses the information needs. From the findings of 1. there will be
ideation sessions with the desired outcome of a concept that in an intuitive
way addresses the information needs.

3. Evaluation of the design concept. Evaluating the concept created in the
previous stage to see whether it fulfills the goals and to see what aspects affect
the outcome.

12



3. Methodology

3.1 Research Through Design

Research through design is a way of performing Human-Computer Interaction re-
search through a continuous iteration of ideating and critiquing the solution and
problem formulation in an attempt to make the "right thing" (Zimmerman et al.,
2007). The outcome of research through design is a concrete problem formulation
and a formulation of a "preferred state" together with a set of artifacts that com-
municate the result of the research.

The term was first coined by Frayling (1993) in his paper "Research in Art and
Design", where he proposes a way to differentiate different types of research. It
is often explained as the link between research and industry (Zimmerman et al.,
2007; Koskinen et al., 2011; Gaver, 2012), where the design artifact is used as the
communication tool between these two parties. Research through design allows for
a more holistic research contribution (Zimmerman et al., 2007).

The thesis aims to contribute to research in the field so that the knowledge and
insights gathered from the project can be used and extended into further research.
As part of the research process there is a opportunity to use a design artifact as a
tool to conduct design research by iteratively conceptualizing and evaluating and
critiquing the design concepts.

Gaver (2012) discusses about the opinions and expectations of research through
design. He argues that we should not expect to get traditional results from research
through design bur rather see the artifact as the fundamental achievement. He
suggests that research through design should not only strive for a convergence in
the research but that a divergence of many alternatives and ideas and critiquing is
beneficial for the community.

Zimmerman et al. (2007) lists criteria on good research through design which he for-
mulated in 4 categories; Process, Invention, Relevance, Extensibility. These criteria
was used as a guideline to produce a higher quality thesis.

3.1.1 Process

The first proposed criterion is about the process and documentation where the
author states; "Interaction design researchers must provide enough detail that the
process they employed can be reproduced. In addition, they must provide a rationale
for their selection of the specific methods they employed." (Zimmerman et al., 2007)

The thesis aims at explaining the process of each step of the project in the three
chapters to follow. For each chapter it will be an explanation of the process in enough
detail so it could be somewhat reproduced with a similar outcome. The methods
are ought to be motivated and shown with a clear connection to the outcome.

13



3. Methodology

3.1.2 Invention

The second proposed criterion regards the novelty and invention of the outcome.
"The interaction design research contribution must constitute a significant invention.
Interaction design researchers must demonstrate that they have produced a novel
integration of various subject matters to address a specific situation." (Zimmerman
et al., 2007)

To communicate the argument for invension we present an extensive literature review
in the first theory section of this thesis to situate the work and demonstrate the
research’s novelty in the field.

3.1.3 Relevance

The third criterion, Relevance, considers the importance of the result and how this
is communicated. "In addition to framing the work within the real world, interaction
design researchers must also articulate the preferred state their design attempts to
achieve and provide support for why the community should consider this state to
be preferred." (Zimmerman et al., 2007)

The thesis aims to research what aspects increase situation awareness. By making
the driver more aware of the situation and in-the-loop we hope that it can lead to
safer vehicles in the future. The prototype will be used as a manifestation of the
preferred state.

3.1.4 Extensibility

The fourth criterion ensures that the work is extensible for other researchers by
encouraging valuable documentation. "Extensibility means that the design research
has been described and documented in a way that the community can leverage the
knowledge derived from the work." (Zimmerman et al., 2007)

Each stage of the thesis aims at producing an extensible artifact. A framework is
created in chapter 4 on which a researcher could use as a basis to develop his own
version, explained in section 4.4. The concept and prototype explained in chapter 5
is developed such that it could be easily modified to evaluate different parts of the
design aspect. The comparative study explained in chapter 6 provides with a base-
line for future product evaluations and research contributions, where a researcher
could perform a corresponding study and refer to this baseline as reference.

14



3. Methodology

Figure 3.1: Overall design process diagram

3.2 Design process

The overall process of the thesis work is illustrated in figure 3.1. For each chapter
in the thesis the process is illustrated in more details.

The process and it’s methodology is inspired by Human-Centred Design(HCD) with
the goal of capturing a broad viewpoint of the information needs for future drivers.
HCD aims to include an explicit understanding of users and their environment (ISO,
2010). HCD as an approach follow an iterative process where users are involved
during the whole design development. By choosing a Human-Centred approach we
prevent ’designing the users’, as Redström (2006) puts it, meaning a design will
reflect users mental model and support their behaviour.

The three stages inspiration, implementation and validation & concept was per-
formed consequently, each with it’s own product and presented in it’s chronological
order in the thesis but a product of the first stage could be refined after the second
stage has started. Each product was iterated and critiqued both with the help of
users and internally.

A challenge of the process was that it required the researchers to elicit needs for
future users. True semi-autonomous cars does not exist yet as of the writing of this
report and the task was to find the needs of the time when semi-autonomous cars
are in use on public roads.

Querying users directly about their needs is generally not a preferred method (Pat-
naik and Becker, 1999; Laurel, 2003). Eliciting needs of a non-existing product will
add an extra challenge since a need can be relevant to a feature that already exists
but a need can exist for features that has not been discovered yet (Van Elslande
and Fouquet, 2008). By utilizing ethnographic-inspired methodology and scenario
based co-design sessions we uncover an opportunity to extract needs that are not
obvious and that the users are not able to express explicitly.

Each stage of the process includes user involvement and critiquing in an attempt to
evaluate our research and design outcome in an iterative way.

The process follows a form of design process where divergence and inspiration is
prominent in the early phases of the process and convergence of ideas gets more

15



3. Methodology

focused towards the end of the process until there is a refined design suggestion.
The result from the individual studies themselves will also serve as strong research
value.

3.3 Human Centred Design

Human-Centred Design is a design process that has emerged from the fields of
ergonomics, computer science and artificial intelligence (Giacomin, 2014). The In-
ternational Standards Organization defines Human Centered Design as an approach
to interactive systems development that aims to make systems usable and useful by
focusing on the users, their needs and requirements, and by applying human fac-
tors/ergonomics, and usability knowledge and techniques (ISO, 2010). It’s a process
that encourages to examining users in their natural environments, and it requires
the researchers to involve the users in the research process.

The methodology and process of the thesis is inspired by the Human Centered Design
guidelines in a way to try to capture a holistic picture of the future users’ needs.

3.4 Expert interviews

Expert interviews can serve as a tool to collect industry and academic insights and
a broader perspective of the current state (IDEO, 2003). Experts of the field is
interviewed with a set of questions for a period of time while notes are written
down.

Expert interviews served as a basis for knowledge in the area of research and provided
insights and direction of the research in terms of how to frame the research space and
what has been done in previous research. The expert researches was done primarily
in the early stages of the thesis work.

3.5 Literature Study

In research through design, a literature study is used to understand how the work
integrates within existing research, and how it might advance the current state of
the art (Zimmerman et al., 2007). Our literature study involved searching across a
number of different sources, including web searches, related master thesises, specif-
ically within the AIMMIT project, and articles within related journal publications,
specifically DUXU 2015, CHI 2013 to 2015, and AutomotiveUI 2010 to 2015. Google
Scholar served as the main literarature search engine, in part because of it’s broad
coverage (De Winter et al., 2014), and intuitive citing mechanism. Searches targeted

16



3. Methodology

articles with keywords: "vehicle navigation", "semi-autonomous", "situation aware-
ness", "automation factors", "driver needs", "car usage", "research through design",
and many more articles targeted specifically by their title or authors. Articles were
then sorted within the main categories of "navigation", "multimodality", "methodol-
ogy", "visual modality", "haptic modality", "automation". Each article is summarized
and relevant quotations are extracted and added to the overall description of each
article in Trello. This method allowed us to easily find and filter information by
high level topics and identify areas that need further research.

3.6 Field Studies

Performing field studies is an ethnographically inspired method where user studies
take place in the environment in which a product will be used (Dix, 2009). With
little or no instructions, users are observed and/or interviewed while using a product
in a chosen environment. Field studies was performed to study how users naturally
behave today when assigned to navigate a car.

3.7 Participatory Design

Participatory design is a way to utilize the experience and knowledge of people that
have experience and insights that extends the experience of the designers. In par-
ticipatory design, users are treated as experts. The goal is for designers, researchers
and users to collaborate and generate a shared learning (Spinuzzi, 2005). A partici-
patory design session could be done in the form of a focus group or workshop, where
designers and users collaborate in designing a solution.

Participatory design was used as a means of involving users in the process of devel-
oping a preferred design and also as an opportunity to evaluate an existing design
concept and as a way to analyze users’ mental models.

3.8 Prototyping

The creation of a design artifact involves a number of intermediary products, which
can be supported through the creation of various prototypes. Lim et al. (2008) de-
scribe prototypes as a tool for traversing the design space which leads to meaningful
knowledge used towards the construction of the envisioned design. The design space
can be explored by creating prototypes with different purposes. A simple system is
offered by Houde and Hill (1997) which plots prototype dimensions on a triangu-
lar space with 3 features: role, look & feel, implementation. In a similar manner,
Snyder (2003) suggests a system consisting of 4 dimensions: Breadth, Depth, Look,

17



3. Methodology

and Interaction. Prototypes can also be classified as hi-fidelity and lo-fidelity. Hi-
fidelity prototypes integrate the high level design elements (Houde and Hill, 1997)
and can be used for evaluation (Macaulay et al., 2000). Lo-fidelity prototypes cap-
ture early design ideas (Macaulay et al., 2000) and are typically built with paper
Snyder (2003). The selection of which prototype to build is determined by using
an iterative approach to identify the most important open design questions (Houde
and Hill, 1997).

This thesis took use of the models above to decide what types of prototypes that
was needed to do in order to capture the relevant aspects before creating a high
fidelity field-testable prototype.

3.9 Strong Prototyping

Strong prototyping is a bodystorming methodology in which the designers will build
and test their prototypes in an environment that closely resembles the context of
use (Schleicher et al., 2010). The method allows the designers to get a first-hand
look at how the artifact would fit into the context of use. This method was used to
rapidly prototype various interface concepts for the semi-autonomous car by creat-
ing, modifying and evaluating prototypes of different materials while inside a car.

3.10 Oblique Strategies

Oblique strategies is a method that can be used to induce lateral thinking in a
creative process (Taylor, 2003). The method was originally created to help musicians
in their music making process, but has been adapted in various fields where there is
a need to think in new ways.

Oblique strategies uses a set of sentences on a deck of cards or similar. One card
is randomly drawn and read out loud. A sentence on the card could be "Towards
the insignificant" or "Lowest common denominator". The sentence is meant to be
cryptic and to be interpreted by the team and the discussion will hopefully lead to
a new direction of thinking or new ideas.

Oblique strategies was used in the early phases of the concept generation to rethink
and improve on established ideas and to produce additional ideas that could lead to
other focus areas.

18



3. Methodology

3.11 Bodystorming

Bodystorming is a simple method for the designers to physically and mentally ex-
perience a situation (IDEO, 2003). The designer will, with or without props, act
out roles to be able to generate and test concepts that depend on the context and
behaviour.

Bodystorming was used in the early phases of ideation to generate ideas for the
semi-autonomous vehicle, which was hard to mimic by other means since it did not
exist as the writing of the thesis.

3.12 Immersion

Immersion is a method about putting the designers in the context of use (IDEO,
2015). It let’s the designers experience the context of where the product will be
used. An immersion session could be initiated by letting the designers spending
time in the environment in which the target use scenario will take place.

Immersion as a method was interpreted in combination with Bodystorming (3.11)
as a way for the designers to get insights on the experience of navigating a car and
a future of semi-autonomous cars.

3.13 Insight Combination

The insight combination method is a method to generate ideas from insights gathered
from ethnographic research (Kolko, 2010). Insights are written on one type of paper
note and design patterns are written on another type of paper note. These are
then combined and the designers have a short period to come up with an idea that
incorporates the both.

The insight combination method was used to convert insights from ethnographic
research and constraints of the semi-autonomous setting and technology limitations
of the project into a multitude of different ideas.

19



3. Methodology

3.14 Affinity clustering

Affinity clustering is a way to sort information gathered from research according to
similarity (Institute, 2012). Quotes, insights, thoughts or any gathered information
is put on individual notes and then sorted into clusters by the design team by moving
them around on a board, the outcome is then discussed and rearranged and lastly
the clusters are named after the containing notes. The result is called an affinity
diagram.

An affinity diagram was created and used to find patterns in insights from field
studies and literature studies.

20



4
Information Needs & Scenario

The goal of the first research sub-question is to elicit information needs for nav-
igating in a semi-autonomous vehicle. General information needs of drivers have
been studied extensively in the past (Mackett, 2003; Wachs, 1967; Mahmassani and
Shen-Te Chen, 1993), however, these needs may change in scenarios which involve
driving semi-autonomous vehicles. Not only are the information needs change, the
scenarios themselves will also shift as urban travel patterns and car use evolves.
This directs us to our second research sub-question, which asks which are the most
likely scenarios under which semi-autonomous driving will take place in the future.
We employ a multi-pronged, multi-step process described below in more detail. The
end result is a set of information needs that are specific to the scenarios under which
they will be used. We further suggest a way to standardize this process so future de-
signers can perform similar research in accordance with our research through design
methodology (Zimmerman et al., 2007).

4.1 Process

The process of eliciting information needs and creating a framework is summarized
in figure 4.1. The first step involved using a number of inquiry methods to yield
the main set of factors. These were filtered, organized, and ranked in a survey
by 50 participants. Finally the results of the survey were organized in an easy to
understand framework.

Figure 4.1: Process diagram for framework creation

21



4. Information Needs & Scenario

Figure 4.2: Field Study Methods Comparing Observable Roles

4.1.1 Field Studies

Two types of field studies were conducting early on in the process of the thesis with
the purpose of understanding and observing driver behaviour in different roles (See
Fig 4.2), specifically as operator and supervisor in a semi-autonomous setting.

The co-navigator scenario provided insight on how drivers navigate in a multi-
destination scenario. This study was inspired by a similar study performed by
Perterer et al. (2015). A handful of drivers were selected in pairs to navigate in
the city of Gothenburg, Sweden. Each pair was assigned 3 tasks to complete that
varied between studies. Tasks included finding a coffee shop, picking up a friend,
and driving to a shopping destination. Data from the field studies included field
notes, video recordings, and post-hoc interviews.

The authors also test drove a vehicle equipped with Advanced Driver Assistance
Systems (ADAS) including lane keeping and cruise control. Sections of the route
included handing over the control entirely to the vehicle, for up to several minutes
at a time. This was to get a feeling on how the semi-autonomous setting would
impact the navigation of the vehicle.

4.1.2 Literature studies

The bulk of the factors were gathered by conducting a literature study across a
number of papers related to navigation, route choice, and driver needs. A similar
method was used by (Münter et al., 2012) on gathering driving information and sup-
port needs for navigation. They identified 3 relevant dimensions: driver, vehicle, and
environment. From this list, a survey showed that drivers need the most support
with spatial knowledge, particularly in unfamiliar routes. Gkouskos et al. (2014)
looked into future driving scenarios by asking the question “how are vehicles used

22



4. Information Needs & Scenario

in this future?” and eliciting driver needs, ranking them based on their original-
ity, evaluative ability, descriptive richness, and dominance dimensions. Zhang and
Levinson (2008) conducted a study to understand the value of travel information
with 113 participants. Models derived from the results included factors for accuracy
of information, trip purpose, traffic density, road safety, and others. Finally, articles
related to factors in GPS navigation (Leshed et al., 2008; Brown and Laurier, 2012)
rounded out the sources by providing practical needs experienced by drivers today.

4.1.3 Information Needs

Based on the factors identified in the literature review and field studies, a list of
155 factors were extracted onto a spreadsheet. Shaping these factors into some-
thing more manageable requires returning to the original research questions, as
summarized: What are the information needs for strategic navigation and can this
information support a high level of situation awareness? Using this as a baseline,
factors were subjectively assigned a situation awareness level (1-3), as well as Mi-
chon’s driving levels (Michon, 1985). Only factors which affect situation awareness
comprehension or projection (level 2 and 3 respectively) and strategic level driving
control were selected. This resulted in filtered list of 54 factors that affect situation
awareness in strategic car navigation.

TRIP 
PURPOSE

INFORMATION

INTERACTION

DRIVER

Time

Route 
Condtions

Topography

Context & 
Aesthetics

Trip Impact

Control

Information 
Accuracy

Fluency

Input

(Goal, Experience)

Driver Load

Route
Familiarity

Travel Patterns

Personalization

Figure 4.3: Final result of Affinity Mapping. The main categories created un-
der Information Needs are: Topography (later renamed to Location), Time, Route
Conditions, Context & Aesthetics, and Trip Impact. Factors under Driver and In-
teraction are not investigated any further in this thesis.

In the next step, the factors elicited from the filtered list were sorted into categories
using the affinity clustering method. The affinity diagram resulted in 10 categories:

23



4. Information Needs & Scenario

time, motivation, topography, experience, route conditions (dynamic/static), trust,
driver load, driver profile, interaction, and spatial knowledge. Returning again to
the first research question, some of these categories do not fit as an ‘information
need’. Therefore, the categories and factors were re-sorted into 3 higher level cate-
gories: information, interaction, and driver. Furthermore, all high level categories
are affected by the trip purpose, which has been shown in the theory section to be
the primary affecting factor in driver information needs (Mackett, 2003; Krumm,
2012). The final categorization is represented in Figure 4.3. All further work in the
thesis maintains a focus on these core information needs. While the factors in the
driver and interaction categories were henceforth dropped from further study, they
present an equally valid starting point for future research.

4.2 Driving Scenarios

The information needs gathered thus far represent the contemporary driving en-
vironment for the specific purpose of raising situation awareness through strategic
control. These same information needs however, may not apply, or may differ when
looking at new contexts such as semi-autonomous driving in the future. In order to
verify and rank these needs against the proper context, a scenario can be created
which reflects the desired frame of reference.

In our scenario, we include the setting, the tasks, and the time frame. Each of these
is chosen based on the research in the introduction and theory sections. First, the
setting is set in the city, where travelled urban kilometers are expected to triple
by 2050 (Arthur D Little Future Lab, 2013). The tasks in the scenario are based
on the multi-purpose trip analysis presented earlier, which shows that the 3 most
common trip purposes are home, shopping, and social (Mackett, 2003; Krumm,
2012). The time frame is also carefully chosen to coincide with the introduction
of semi-autonomous vehicles in the city, which is projected to happen in 2021-2025
(Berger, 2015).

We argue that the linear nature of A to B type trips such as highway driving to
and from work would not place stringent demands on navigation interface usage. In
addition, the proposed time frame implies that some manual driving will be required,
particularly in complex city environments. The resulting scenario is a non A-to-B
type trip, situated in the city, with multiple stops, requiring some parts of the route
to be driven manually 4.4. On the way to each destination, we strategically place
events that stress different information needs. For example, the ‘Get Location’ and
‘Park’ events are querying for the topography of the destination, while ‘Route
Selection’ emphasizes at the time and route condition needs. We make use this
scenario as the basis of our survey on information needs and empirical evaluation of
navigation interfaces for semi-autonomous vehicles.

24



4. Information Needs & Scenario

H

Plan Trip

Arrange Pick-Up

Get Location

Call

Drive

F
Park

What/Where/When?

Plan Route

Park

Drive

S

H

Time Estimation

Fastest Route
Risk Assesment

Route Selection
Drive

Park

Autonomous Driving

Manual Driving

Figure 4.4: A dynamic city type scenario constructed to test navigation interfaces
for semi-autonomous vehicles. H = Home, F = Friend (Social Trip Purpose), S =
Shop (Shopping Trip Purpose). Manual driving sections are represented in pink.

4.3 Comparing Information Needs

An online survey was created with the purpose of verifying and weighing the im-
portance of the found information needs for navigating a semi-autonomous vehicle
in the specified scenario. A similar survey approach was used by (Pfleging et al.,
2014), with the difference that they used expert focus groups for factor elicitation.

4.3.1 Survey Design

The hypothesis was that different purposes of a trip would affect the information
needs. Therefore, it was decided that each information need would be individually
ranked for each purpose. As such, the survey consisted of 3 main sections, each
corresponding to a trip purpose: picking up a friend, going shopping, and going
home.

The survey was constructed using the scenario as a basis for a story that was textu-
ally explained with short paragraphs, together with a related picture. Each textual
story included a related task assignment. For each story and task the survey included
an importance rating of each information type.

Since the found information types were broad and could be interpreted in different
ways it was decided that the survey would also include some examples of different
information items that could relate to these. 16 examples were extracted from the
same list of factors and traced back to the source to determine a concrete example
of information that could be provided in a navigation interface.

These examples were provided, in addition to the category rating, as individually
selectable options for if the participant considered the information interesting to

25



4. Information Needs & Scenario

complete the task in the story. Each example was slightly modified to make sense in
each story but one were related to only one information type. To reduce bias between
each story the order of the examples were randomized. To prevent participants to
tick all information as interesting and to force them to prioritize the information
they were required to tick less than 7 items.

4.3.2 Survey Pilot

The survey was piloted with 3 participants. The main feedback concerned having too
many example check-boxes, one for each category, per trip purpose. This resulted
in combining the examples from all the categories in one list, which made them
more ambiguous, however, allowed participants to compare them against each other.
In addition, some of the category names were translated to be more readable for
everyday users. Topography was ranamed to Location, and Context and Aesthetics
was renamed simply to Setting.

After the survey had been revised and improved the link to the online survey was
published via social media channels including reddit/r/SelfDrivingCars, Tesla Club
Sweden forum, related facebook groups, as well as individually messaging contacts
of the project team. The full survey is available in appendix A.1

4.3.3 Survey Results

The survey was completed by 50 persons with a mean age of 30.5 (SD 11.2), con-
sisting of 66 % male and 34 % female respondents. Participants were distributed
globally with 58 % in Europe, 22% in North America, 4 % in Asia. Urban settings
included 21 different cities of varying sizes, with large cities (> 400,000 population)
comprising 56% of respondents and small cities 22 %. Driving frequency of partic-
ipants ranged from 52% for frequent drivers (daily or weekly), 28% for infrequent
drivers (monthly or yearly), and 20% for non-drivers.

Survey results showed a clear delimitation between primary and secondary needs
(see 4.5). The categories of time, location, route conditions, and self-driving parts
were ranked significantly higher than the categories of setting and trip impact. From
this point, we discuss these primary information needs, and link them to the specific
information types in Figure 4.6.

Time: Consistently ranked high irrespective of the trip purpose. Time to
destination is ranked slightly higher than the expected time of arrival, and
time delays become more sensitive as the trip purpose switches to going home.

Trip Conditions: Home trips are more sensitive to condition information
needs, which can be linked to time delays in the time category.

26



4. Information Needs & Scenario

Location: Varies widely across trip purposes. Distance is highly desired in
cases of shopping, while location information is needed in both picking up
a friend and shopping scenarios. A picture of the destination is useful in
shopping and social scenarios.

Self-Driving Parts: Based on the results of the individual examples, the
self-driving example checkboxes did not correlate with the parent category of
Setting. The checkbox example was checked at a much higher rate than the
preference for setting information. Therefore, it was factored out as its own
category. Overall, it was rated as useful at exactly the same rate across all 3
scenarios, hence some degree of self-driving part information is always needed
irrespective of trip purpose.

General Information CategoriesGeneral Information Categories

Friend

Shop

Home

-0.8 0 0.8 1.6 2.4

Location

Conditions

Urban Setting

Time

Trip Impact

Figure 4.5: General Information Categories. For each scenario (Home, Friend,
Shop) it is important that I am given information regarding the particular informa-
tion category. Rated on a 5 point Likert Scale.

4.3.4 Additional Survey Insights

Although the sample size is limited at 50 participants, some insights can be made
across geographic and demographic distributions. The most significant finding is
in comparing big cities to small cities, where the need for information regarding
self-driving parts is 78% in big cities. On the other hand, small cities need more
information on estimated time of arrival which is 69% higher than in big cities.
Delving into region differences, drivers in North America had higher needs for delay
information (73%), visual location information (40%), and self-driving parts (52%)
than their European counterparts.

The driving experience also showed a strong influence in information need prefer-
ences. In our data, frequent drivers showed higher interest in traffic, delays, expected
time of arrival information. This type of information might be more associated with

27



4. Information Needs & Scenario

Specific Information TypesSpecific Information Types

Friend

Shop

Home

0 12.5 25 37.5 50

Level of Congestion

Safety of the Road

My Previous Trips

Self-Driving Parts

Environmental Impact

Monetary Cost

Distance from You

Location on a Map

Picture of Surrounding Area

Transportation Links

Scenery

Road Type

Urban Setting

Time to Reach Destination

Est. Arrival Time

Time Delays

Figure 4.6: Specific Information Types: For each scenario (Home, Friend, Shop)
what kind of information would be most interesting? Each Tick of a checkbox is 1
Point.

28



4. Information Needs & Scenario

manual driving information needs. Infrequent and non-drivers showed interest in
location related information such as a picture of their surroundings and a map of
the destination. They also showed slightly higher interest in knowing which parts
of the route would be autonomous. This type of information could be interpreted
as applying a more strategic level of navigation.

4.3.5 Participant Comments

The qualitative results of the survey include participant comments on their choice
of factors, the ’other factor’ free-form entry field along with the general comments
and ideas about navigation in semi-autonomous cars. For the ’other factor’ field, the
most common request was to show available parking locations. Driving directions
(as in turn-by-turn) and road data (one way streets) were also on the list. The pick
up a friend scenario yielded some interesting discussion regarding factor choice. One
participant noted that information accuracy plays a key role in the time dimension:

“When you’re picking up Bobby, you want to know precisely when you’ll get there,
so you can let him know if you’re going to be delayed.”

Another participant prioritized location information in order to make optimal route
choices:

“If he is closer to Macrosoftś shop, then it makes sense to drive in his direction,
otherwise we could also meet half way.”

In the general comments, the low needs in the ‘Urban Setting & Scenery’ cate-
gory can be summarized by one participant’s comment:

“I guess I don’t care about the scenery, I just want to get to where I’m going”.
Several participants requested that the navigation can incorporate a ride-sharing
system “so you can fill up extra seats in your car if you want”. Finally, one par-
ticipant wanted to have their data displayed in “a bright bold H.U.D. projected on
windshield, or flexible clear OLED display”.

4.4 Discussion and Patterns for Future Use

This section of the thesis undertook the task of answering the first 2 research sub-
questions: What are the likely scenarios of car usage in the future, and what are the
information needs of drivers of semi-autonomous vehicles under these scenarios. The
results showed that there are 4 primary information needs: Time, Trip Conditions,
Self-Driving Parts, and Location. Overall, the results may seem somewhat obvious,
these are needs that are even needed in today’s driving. However, we have showed
that these core needs will remain necessary despite the changes in context, and more

29



4. Information Needs & Scenario

importantly, have showed that they vary according to the trip purpose. Furthermore,
we quantify the particular information types that are most needed for each purpose
in Figure 4.6.

The process we have undertaken to identify the information needs of semi-autonomous
vehicles consisted of a identifying a likely future use cases, sourcing base needs from
literature, constructing a likely scenario out of this, and querying current drivers on
their needs based these scenarios. This process yields an interesting design pattern
which may be used when expanding vehicle interfaces to new contexts (e.g. fully au-
tonomous), or to different scenarios (A-to-B type trips). At the heart of this pattern
is the idea of ‘Trip Purpose’, serving as the driving differentiation in information
needs. Using this as a basis, it is possible to construct a scenario consisting of likely
trip purposes (or destinations), and add motivations for information during the trip.

In our case, we studied the scenario of city driving with en-route planning in a
semi-autonomous vehicle. The found information needs are location, time, route
conditions, and self driving parts. And the most common trip purposes are social,
shopping, and home. An example of using the same pattern could be the scenario
of country driving, with pre-trip planning, in a semi-autonomous vehicle. We can
suppose that the designer has found context & aesthetics, as well as trip impact to
be more significant than in the previous example. The trip purposes would also be
quite different depending on the overall goal of the trip, which could be for example,
a vacation on the coastline.

In the next section of the thesis, the scenario and found information needs will be
used to build up a concept for a navigation interface. Further on in the thesis, we
apply the scenario and context of semi-autonomous driving to evaluate this interface.

30



5
Prototype & Concept

Development

We state as one of our thesis goals the production of a real, working prototype
which may be tested in the field. The following chapter explains the process and
outcome of designing a multimodal interactive prototype for strategic navigation in
semi-autonomous cars. The outcome of the previous chapter is used in ideation and
prototyping. The result is a fully working high fidelity prototype with live routing
and GPS support.

Figure 5.1: Process diagram for development of the concept

5.1 Process

The overarching process for generating a concept and prototype is shown in Fig 5.1.
The insights gained from our literature studie, scenario creation, and field studies
in the previous stage of the thesis was used as a basis when ideating on the concept.
A co-design workshop was also held to capture the mental model of users on how

31



5. Prototype & Concept Development

they would prefer to display the information needs identified in the previous chapter.
Numerous prototypes was created to evaluate different parts of the concept.

5.1.1 Design restrictions

Some restrictions and goals for the design steered the concept generation. These are
listed briefly below. The first two follows by the scope of the project explained in
the introduction of the thesis.

1. The concept should primarily support the situation where the car drives au-
tonomously in a semi-autonomous car.

2. The concept should primarily support strategic navigation and improve level
3 of situation awareness (Projection).

3. The design should work well in a three-destination trip in the city. The same
reason for using the scenario created for the survey.

4. The design should be extendable for future research. A wish from the authors
to contribute to future research by building a concept that can be reused for
future research, in the ethos of Research Through Design.

5. The design should be testable in real-life situations. Creating a concept that
was feasible to build as a working prototype for the length of the master thesis
project, which is of interest since something that has, in the context of the
semi-autonomous vehicle, has not yet been extensively reviewed in published
works.

5.1.2 How might we?

By going through the field studies and information framework we could write down
a number of insight statements in short scentences. These insights were categorized
with the use of affinity clustering. The categories would point out common themes
and challenges for the design. The themes challenges were synthesized into 8 "How
might we?" questions as follows.

1. How might we allow different (detail-)levels of navigation input?

2. How might we tailor the information infrastructure to the trip purpose?

3. How might we assist a multi-stage trip?

4. How might we display relevant information for selecting a route?

5. How might we clearly communicate information about autonomous parts. (dif-
ferentiate from other types of information)?

32



5. Prototype & Concept Development

6. How might we relate information to the surroundings?

7. How might we reduce the effort required in interacting with the system?

8. How might we enable a smooth overlap between a high level and low level
interface?

The how might we questions were used in ideation to brainstorm on concepts for
the design.

5.2 Early revisions

The early stages of the process focused on diverge and produce as many ideas as
possible. Bodystorming, immersion sessions and strong prototyping were used to
get a deeper understanding on what could be the future needs of semi-autonomous
driving. Insight combination method, 3-6-5 and Oblique strategies are some of the
other methods used to generate ideas. The ideas generated were prioritized according
to the project goals and limitations. A favored idea was to visualize a timeline of the
current route that the autonomous car is intending to traverse. Many prototypes
was generated during the ideation phase. Depending of the progress of the concept
generation and what aspects of the idea that is unclear, different levels of prototypes
was created.

Figure 5.2: HUD prototypes

Another prominent idea was to utilize a Head Up Display (HUD), a transparent
display shown on top of the windshield, for displaying information in the system.
The theory was that this screen residing in closer proximity to the field of view of the

33



5. Prototype & Concept Development

environment would support a connection between the interface and the environment.
This idea was evaluated first by drawing a paper prototype placed in a car and
second by building a functional prototype to see benefits and disadvantages of such
a display. See 5.2 for a picture of these prototypes. It was concluded that an HUD
had disadvantages that outweighed the choice of a regular screen compared to the
advantage in situation awareness it gave, thus the idea was dropped. Difficulties
with HUD’s arise when the brightness of the surrounding varies, which makes it
difficult reading the display when the color of the surrounding may interfere with
the color of the interface, which in turn limited the range of colors our interface
could use.

Figure 5.3: Lean gesture functional prototype

On the input side of the equation, we also investigated the idea of using head leaning
gestures for additional control of the interface. This idea was evaluated with a
functional prototype, a web based application that used a web camera to track the
position of the users face (See fig. 5.3). The outcome was promising, the gestures
could have the potential of adding a natural interaction. However, there were some
issues identified, primarily because the head movement required to do gestures would
be too much to be comfortable and at the same time not activate unintentionally. It
would also make the interface take up a lot of attention by always moving with the
users head, which could have the effect of lowering situation awareness. This feature
was taken as far as the pilot studies, but was dropped for the final prototype.

5.2.1 Co-Design Workshop

A simple paper prototype created to serve as a ‘role prototype’ Houde and Hill (1997)
was used in a workshop session to elicit ideas and feedback on the eventual use of
the navigation interface during real scenarios. We carried out a co-design workshop
as a form of participatory design with paired users on three occasions. The majority
of participants were interaction design students as a convenience sample and also

34



5. Prototype & Concept Development

for the reason to have other designers opinions of what would be feasible or not to
do.

The primary goal of the workshop was to generate a handful of plausible ways that
the information needs found in the previous stage could be presented in the semi-
autonomous context, and also when this information should be presented to the
driver. The workshop material was based on the information types found from the
research and prior generated examples for the information representations.

A workshop session was initiated with an introduction of the background of the
project and the aim for the workshop. As a warm up, a bodystorming session was
held for the participants. This was to help them get a sense on how it would be like
in a semi-autonomous vehicle and also to spark a creative mindset. The workshop
continued by handing out to each participant three design wireframes on paper.
Along with this a special sheet with examples and inspiration for visual, auditory
or haptic representations of different information types.

The inspiration sheet was divided into five columns of information types: Time, Des-
tination, Conditions, Self-driving parts and Self-driving point of transition. Each
column contained examples of information to show and examples of ways to rep-
resent the information. The examples were created after proposed designs that
was produced during brainstorm sessions beforehand, both good and bad ideas was
presented. The point of the inspiration sheet was to give instructions on what infor-
mation could be shown in the system but primarily as inspiration for the participants
to spark ideas on how the information could be represented.

The participant’s task was to create an interface design on the wireframe sheet using
the information listed on the inspiration sheet. The design sheets also contained the
same scenario as in the survey. The task for the participants was to create a design
for each scenario. After the bodystorming the participants were given instructions
and 30 minutes to create the three designs. The design session was followed with a
presentation session where participants could present their designs and ideas. These
were recorded for later analysis.

After presenting the designs, participants were assigned to cut out their designs and
put them on a timeline on the trip, to indicate at which time the participants would
like to have the information to be displayed in the interface. See an example of a
completed timeline in fig. 5.4. The workshop was documented with pictures and
some video and audio recordings. The data from the workshop was reviewed and
used as a reference in further idea generation and to support other ideas.

5.3 Integrated prototype

The process to follow consisted of settling on a design concept that would fulfill the
restrictions specified in 5.1.1 and also support the desired research goal. Insights
from the research and iterations of functional prototypes and aesthetic prototypes

35



5. Prototype & Concept Development

Figure 5.4: Example of a design timeline created by one participant on a workshop

made way to a design that we thought had potential to be a good manifestation of
a well functioning navigation system for the context of use with support from the
user research that we had done previously.

It was decided to create an integrated prototype that would be testable in a real-
world driving scenario. The integrated prototype and its components are explained
in the sections below.

The prototype explained below is what was used in the comparative study explained
in the next chapter. Data and insights from the comparative study let us iterate on
the design concept and refine it further. The changes made to the prototype and
concept is explained in section 7.2.

5.3.1 Setup

The prototype consists of two parts, the screen and the remote (See fig. 5.5). The
screen is intended to reside behind the steering wheel, straight in front of the driver.
The remote is a touch surface meant to be placed at a position so that the driver
can comfortably reach it while seated in the car.

36



5. Prototype & Concept Development

Figure 5.5: The prototype setup

The setup is adapted to a future where the driver has more free time to do other
things than navigating. The driver might want to sit in a more comfortable position
and focus on another task, but there is still a need for the driver to be in-the-loop
and keep a track of what is happening. For these two arguments it was decided that
the navigation system should have a screen that is close to the windshield to quickly
be able to relate the interface on the screen to the environment. It was also decided
that this would need to be controlled from an external control of some sorts.

5.3.2 Technical details

The final prototype was developed using Unity 3D and Mac OS X as the main
software developing platform and XCode and iOS as the platform for the remote.

The map data was downloaded offline from openstreetmap.org, a crowdsourced map-
ping platform with the possibility to integrate the data into various systems. The
data was used in Unity for rendering the roads in a three dimensional world.

The communication between the remote and the screen was done via Wi-Fi wireless
technology and UDP connectionless communication using a protocol called OSC

37



5. Prototype & Concept Development

that is specifically adapted for real time applications. The remote sends both touch
sensor data for gesture recognition, but also GPS data.

5.3.3 Remote

The remote is used to control the interface on the screen. It’s controlled using
gestures performed on the touch surface.

There are 7 gestures recognized by the remote: Swipe left, swipe right, scrolling up
and down, pinching in, pinching out, tap and hold and double tap. Swiping left and
right performs various actions. Pinching in and out changes the zoom level of the
map. Scrolling up and down moves the "ghost car" to preview a route. Tap and
hold confirms a destination change. Double tap cancels a destination selection.

The gestures were chosen to be forgiving and distinguishable so that it works with
rough inputs in a bumpy car. The gestures were also chosen in regards to be able to
mimic the motion of the interface on the screen, e.g. scrolling up and down moves
the "ghost car" up and down.

The reason to use a touch interface instead of tactile buttons were that the touch
gestures would allow for a wider variety of inputs on a smaller surface and that the
touch surface could be reused for other systems in the car.

5.3.4 Screen

The screen displays all information about the current trip and provides the feedback
and changes information according o the user input.

The screen is divided into two main areas of information: A top bar with static infor-
mation about the current route and the map which displays contextual information.
See fig. 5.6 for reference of the different parts of the system.

The map is essentially a standard road map but showing only the roads available
for traveling by car, and grassland and water indicated by different colors. The map
is oriented in a 3D based perspective view. On the map there are a certain number
of destinations that are available to the driver to select. The current route of the
car is represented with a turquise line that is drawn out from the start of the trip
to the current destination. The current position of the car is represented as a 3d
model of a car.

The top bar (5) displays the information of the current trip and destination: A
representative picture of the destination, the name of the destination, the address of
the destination, the time to get to the destination and the estimated time of arrival.
The time to get to the destination is prioritized in terms of size and position. This
is a consideration based on the survey results in the previous chapter.

38



5. Prototype & Concept Development

9
arrival time

15:34

minBobby

current destination

Norra Allegatan 21

3
min

next manual
driving section

3 minin

Home
Chalmers

2
min

Nordstan 9
min

3D Perspective Map1

Information Overlay4

Landmarks5

Ghost Car7

Current Position8

Placemarks6

Self-Driving Manual Driving

Destination Bubble2 Current Destination3

Figure 5.6: Components in the interface on the screen

5.3.5 The Ghost Car

By swiping up on the remote, the user is able to preview the future of the trip and
how it will progress. The gesture would spawn a semi-transparent car that spawns
from the current position of the car and follows the path of the intended route (2).
The appearance and function of this "ghost car" as it will be referred to from now on
is so that the user could perceive the function so as it would indicate a future self,
meaning that in a certain amount of time the car would be at the position of the
ghost car. Next to the ghost car there is information showing how many minutes it
will take for the car to get to the current position of the ghost car. The ghost car is
also used as a cursor to select where on the route it would zoom in.

The scrolling gesture was chosen for this action for it to create a one to one mapping
of the motion of the gesture to the motion of the ghost car in the interface, with the
intention of an intuitive interaction.

5.3.6 Current location

The current location of the vehicle is represented by a 3d model of a car that is
visible on the map (3). The intention of using a car as the indication for current
location was to enable the analogy of a future self in the case of previewing with the
ghost car.

39



5. Prototype & Concept Development

Figure 5.7: The zoomed in view of the interface screen

5.3.7 Choosing destination

Alternative destinations (4) that are available for the user is shown on the map as a
"bubble" at the position of the destination. The current destination is indicated with
a turquoise flag on the map and the information is on top of the screen. Swiping
left or right on the remote will allow the user to preview alternate destinations.

The alternate destinations is thought to be pre-entered into the system. Entering
destinations could be done on the users phone before entering the car or the system
could intelligently guess based on your information where you would like to go. It
could also be a separate system in the car that would allow for entering destinations.
Entering destinations is not in the scope of this concept.

When previewing an alternate destination in the interface, the route that the car
would take if it were to go there is indicated with a white line in addition to the
current route (See 5.8).

If the driver would like the car to drive to the currently previewed destination he
would tap and hold on the remote. This would make the turquoise flag appear at
the new destination and the information in the top bar would update accordingly.

40



5. Prototype & Concept Development

Figure 5.8: Previewing a destination.

5.3.8 Landmarks

On the 3d oriented map there are certain 3d modeled buildings placed out on the
map (6). These are landmarks that the driver can use to relate the relative position.
Using landmarks for navigation is supported by the multi-point anchoring theory
described in the theory section.

5.3.9 Routes

The current route that the autonomous driving is currently driving on is indicated
with a turquoise line drawn on the map. If there is a part of the trip where the
driver needs to take over and drive manually this part of the route is marked with
a pink line instead. The turquoise line will always be shown on the map while the
autonomous driving is activated.

5.3.10 Placemarks

Certain route conditions that may be of interest for the driver is presented as place-
marks on the route (7). The placemarks could indicate a traffic jam, construction
or a point which the car would not be able to handle the autonomous driving but
the drive would have to take over and drive manually.

41



5. Prototype & Concept Development

5.3.11 Zoom levels

The map can be viewed in two different modes; a strategic view where the camera
is further away from the road so that more destinations are visible and a tactical
view where the camera is closer to the road and more context information may be
displayed. In the strategic view the user is able to make strategic decisions, meaning
that he will be able to select which destination that the car will go towards. In the
tactical view he will be able to make tactical decisions that regards route changes
and traffic situations.

In the zoomed out view the user can use a swipe left or swipe right gesture on the
remote to preview how the car would get to an alternative destination.

When zoomed in to the tactical zoom level swipe left or right would be used for
tactical decisions. This was not implemented in this prototype since it would not
be used in the comparative study.

To switch between the strategic and tactical views the user needs to perform a
"pinch" gesture on the remote by using two fingers in an outwards motion.

Golledge and Garling (2002) explains that there are two ways to learn an environ-
ment; by having an "overview" and by "experiencing" it. This served as a basis for
adding two zoom levels. These are detail levels that are quick to switch between
where the first zoom level, the "overview", gives a wider field of view and provides
a long-term estimation, while the second zoom level gives a closer look at the im-
mediate surroundings (which could be at a different place). The angle in which the
world is viewed on the screen is also changed so that it reflects the purpose of the
zoom level. The zoomed in view follows the direction of where the car is going while
the zoomed out view focuses on the current destination.

42



6
Comparative Study

The goal of the third research sub-question is to investigate if a navigation inter-
face employing information designed for semi-autonomous driving increases situation
awareness. To answer this, we compare against a current baseline. The baseline of
Google Maps was picked for two reasons: it was the only system which could most
closely match the features provided in the concept interface, and it could serve as a
baseline for future researchers to compare their own systems against. We performed
an in-the-wild study comparing the concept we designed in the previous section to
Google Maps. The results showed significant differences between the two system in
terms of situation awareness and subjective preferences.

6.1 Process

The process for creating and working through the comparative study is summarized
in figure 6.1. The experiment design consisted of several steps, starting with building
the experimental setup in the test vehicle, creating appropriate routes for a balanced
study, and recruiting participants via a website. Once the experiment design was
complete, the first pilot studies were carried out, with intensive cycles of integrating
feedback. Once the experiment started, all variables, prototype version, and exper-
imental setup was locked down. The study lasted nearly 3 weeks and included 22
participants, with an average time per participant of 1.5-2 hours. The data from
the experiment was analyzed with the aid of multiple professionals and researchers.
Results include both statistical tests for quantitative measures and qualitative in-
sights based on participant comments and observations. The most salient findings
are presented in the results section of this chapter.

6.2 Pilot

A pilot study was conducted to inform about the design of the experiment and to
gather feedback on the current prototype design. Four expert participants were
recruited: a PhD studying interaction design within the automotive industry, a in-
dustry researcher with a background in automotive interaction, and two consultants

43



6. Comparative Study

Figure 6.1: Process diagram for comparative study

with experience in designing interfaces for semi-autonomous vehicles. The main
feedback concerned the experiment design, which was originally structured using a
SAGAT task interruption called a freeze (Endsley, 1995), which was placed before
a supposed transition to manual driving. This caused a number of issues for partic-
ipants, and was thus changed to a real-time probe methodology. This change also
worked well in combination with another suggestion to use think aloud as a method
to gather more user experience insights. Due to the high novelty factor of both
the environment and the interface, all participants felt they needed more training.
As a result, a short warm-up round was added at the very start of the experiment
to introduce users to the context and experimental procedures. In terms of pro-
totype feedback, participants found it highly intuitive and fast to learn. However,
some revisions were made to the remote interface gesture set, mainly ‘cancel’ gesture
which was changed to a double tap based on previous research findings (Hammar
and Karlsson, 2015), and the select gesture for which the hold time was reduced to
1 second from 3 seconds. Some participants reported a weak link in the information
presented in the map to the information presented in the static overview in the top
of the interface. To address this, the coloring of the current destination was updated
to match the coloring of the information in the overview area.

6.3 Participants

We recruited 22 (8 females) participants with valid driver’s licenses aged between
21 and 36 (mean 26.8, s.d. 4.7) from local universities and industry. Participants
had background included 8 interaction design master students, 3 automotive engi-
neering master degree students, 4 students from an engineering background, and
2 professional researchers studying interaction in the vehicle automation domain.
Additionally, participants reported their driving frequency (7 weekly, 5 monthly, 12

44



6. Comparative Study

yearly) GPS usage (7 never, 4 usually, 8 in new areas, and 3 for long trips) familiar-
ity with self-driving vehicles (6 not familiar, 11 somewhat familiar, 5 very familiar)
and familiarity with ADAS systems (6 not familiar, 13 somewhat familiar, 3 very
familiar). All participants received a movie ticket as compensation for their time.

6.4 Apparatus

In this wizard-of-oz experiment, a special right-hand drive Volvo V40 provided by
Volvo Cars was used as the make shift semi-autonomous vehicle. Within the vehi-
cle, technical equipment was mounted and distributed as shown in figure 6.2. The
technical components components consisted of:

1. Tablet Display: Samsung Galaxy Tab 10.5 SM-T800 (2560 x 1600 resolution)
running Android 6.0, with the following software: Google Maps version 9,
Twomon version 1.1.

2. Remote Input: iPhone 6 running iOS 9.3, running the FutureNav Remote
application.

3. CPU: Macbook Pro Retina running OSX 10.11.4 with the following software:
FutureNav, Twomon version 2.0.42.

4. Immersion Cloth: A black cloth installed between the driver and passenger in
the shape of a triangle measuring 37cm wide by 28cm tall.

5. Additional Equipment: USB cabling between tablet and cpu, power supply
for the cpu, and usb cables to charge all components during the experiment.
GoPros were installed for only some participants and are not part of the core
setup or analysis.

6.4.1 Environment and Routes

The experiment was conducted in-situ, set in the city of Gothenburg, Sweden in
the area of Chalmers University main campus. An in-situ environment was chosen
based on a number of factors as related to our main study objectives: the ability
to compare to a real GPS system such as Google Maps, higher believability in the
experiment tasks.

The routes were carefully designed to include a similar characteristics in terms num-
ber of turns, time to complete, urban density, landmark visibility, and distances to
the displayed destinations on the map.

45



6. Comparative Study

Tablet Display

Remote Input

Immersion Cloth

cpu

tablet

CPU

1

2

4

3

Figure 6.2: The components of the experimental setup

6.5 Design and Procedure

The goal of the experiment was to evaluate map navigation in terms of user expe-
rience and situation awareness for semi-autonomous vehicles set in the city. The
experiment was a within-subjects comparative study with a single independent vari-
able, the Navigation Interface. This variable had two levels: FutureNav as the new
concept to be evaluated, and Google Maps, which serves as the nearest comparable
baseline. The design uses a full balancing of each condition between the two routes
in order to minimize learning effects.

Each participant was introduced to the experiment with a brief overview of the
tasks they would need to complete, a tour of the vehicle, the expectations of the
study, and were told they would not need to drive at all during any time. After this,
a short questionnaire on their background was administered and the experiment
began. Before any conditions were instructed on or tested, each subject was given
a short warm up ride in the vehicle to familiarize themselves with the environment
and the experiment. This practice round lasted between 1 to two minutes, used
no navigation interface, and took an established, looped route near the starting
location. During this time, a representative question was asked similar in nature to
the questions asked in the full route.

46



6. Comparative Study

Route West

Route East

1

2

Go Home

Go Home

Change Dest.

Change Dest.

3

1

2
3

4

4

55

6

6

7

7

88

8

Figure 6.3: The two routes used, displaying the points at which questions were
asked, and the points where the tasks were given.

Next, each