eHMI Design: Theoretical Foundations and Methodological Process
Y. Shmueli and A. Degani
General Motors R&D, Israel
Keywords: External HMI (eHMI), Multiple Resources Theory, Stimulus-Coding-Response Compatibility Principle.
Abstract: In the last decade, substantial efforts have been dedicated to the problem of pedestrian’s encounter with
driverless autonomous (L-4/5) vehicles. Different communication schemes, involving different design
concepts, modalities, and communication formats have been conceived and developed to communicate and
interact with pedestrians. It is expected that only a limited subset of these options, perhaps only one, will be
selected as an international standard (with some allowance for branding and adaptations to different cultural
norms and expectations). Naturally, the selection of the communication scheme has to rely on a valid
theoretical foundation, not only to satisfy automotive regulatory agencies, but also as a precursor to a similar
communication scheme for robots in the public space. In this paper, we provide an eight-step process which
supports the development of an effective communication design. We use Wickens’ (1984, 2002) Multiple
Resources Theory (MRT), as the theoretical foundation for our work, and the Stimulus Coding Response
(S-C-R) compatibility principle (Wickens et al. 1984) as an organizing principle for eHMI design.
1 INTRODUCTION
The transition from manual driving to fully
autonomous driving requires a shift in our
conceptualization concerning the interaction between
vehicles and other road users. Currently, pedestrians
primarily use implicit motion-inherent cues such as
the vehicle’s speed and distance, deceleration rates,
and braking profile to anticipate the behavior of
vehicles on the road and make the crossing decision
(Cohen et al., 1955; Lehsing et al., 2019; Domeyer et
al., 2020). In addition to these physical cues, non-
verbal explicit cues such as eye contact, head nodes
and gestures help pedestrians to interpret the situation
and support the establishment of trust between the
pedestrian and the driver (Šucha, 2014, Rasouli et al.,
2017; Gueguen et al., 2015; Ren et al., 2016;
Lagström and Malmsten Lundgren, 2015; Dey,
2021).
Naturally, the absence of a driver in autonomous
vehicles precludes the possibility of any
communication. As such, autonomous vehicle
technology requires additional features that will allow
the public to interact with such vehicles and perceive
it as safe and accommodating.
Sophisticated interfaces can be devised to
substitute for the missing pedestrian-driver
interaction and may even achieve higher reliability
than the current signaling methods (e.g., use of
headlights, hazard lights, and the car horn). These
communication methods are naturally idiosyncratic
and at times quite ambiguous. For example, does the
driver’s use of the headlights means that he or she is
taking the right of way or giving it to the pedestrian?
The question is how to use this opportunity of the
forthcoming need for external HMI (eHMI) signaling
communication to not only substitute the driver but to
make such communication better. At its most basic,
the communication between a robotic agent and a
human pedestrian should meet four main
requirements: (i) Effectiveness – establish the
necessary communication between pedestrians and
autonomous vehicles, (ii) Efficiency – be simple,
intuitive, and non-intrusive, (iii) Acceptability – form
public “trust” in this new technology, (iv) Satisfaction
– be elegant, induce comfort, and invoke a rewarding
experience.
We propose a step-by-step process for the
verification and synthesis of eHMI design solutions
to fulfil these requirements: Step 1. Requirements’
derivation based on initial conceptual analysis, Step
2. Requirements’ derivation based on an empirical
needs study, Step 3. Proposal of a generic
communication protocol, Step 4. Content selection
for eHMI displays, Step 5. Allocation of the selected
content to media and modalities, Step 6. Media
Shmueli, Y. and Degani, A.
eHMI Design: Theoretical Foundations and Methodological Process.
DOI: 10.5220/0011686600003417
In Proceedings of the 18th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2023) - Volume 2: HUCAPP, pages
201-212
ISBN: 978-989-758-634-7; ISSN: 2184-4321
Copyright
c
 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
201

realization - Representation solutions within different
media, Step 7. Verification of design proposals, and
Step 8. Validation of design solutions.
2 METHODS
2.1 Initial Conceptual Analysis
We begin by analysing generic interaction patterns.
This analysis lists the potential touchpoints between
a vehicle (with a driver or without) and a pedestrian
in a pedestrian crossing scenario. Framing the
interaction between the two as a dialogue will enable
us to pinpoint generic user needs. We outline
potential actions made by the pedestrian (in
aquamarine) and those that can be made by the
vehicle (in black). Pedestrian’s states are marked by
numbers and the vehicle’s states are marked by
letters:
The vehicle is driving, approaching the scene [a].
The pedestrian is walking (0), facing (or with his back
to) the approaching vehicle. The person may look or
glance at the vehicle before reaching the curb,
communicating the message: “I can see you” (1). If
the pedestrian is not planning to cross, his/her body
movements, posture and facial expressions will
convey the message “I am not crossing” (2). If, on the
other hand, there is an intent to cross, the body
movements, posture and facial expressions will
convey the message “I am about to cross” (3). This is
where people seek a confirmation from the vehicle [“I
intend to stop (for you)” - b] (Rasouli et al, 2017;
Habibovic et al. 2018; Dey 2021). While the vehicle
is slowing down, it conveys the message [“I am
slowing down and stopping” – c], or [“I am not
stopping – d] if it cannot stop on time.
The intent to stop is primarily communicated by
the deceleration profile of the vehicle that should be
sharp enough to be easily recognizable (Lehsing et
al., 2019); a vehicle that stops short of a crosswalk
can be interpreted as yielding for a pedestrian and not
simply responding to traffic signage (Risto et al.,
2017; Domeyer et al., 2020). Leaving the curb, the
pedestrian’s message progresses into “I am on the
pavement, starting to cross” (4). From the stopped
vehicle’s perspective, the message changes to [I have
stopped for you – e]. While crossing the road, the
pedestrian may portray various messages using
implicit and explicit modes of communication. These
will vary from a short glance (or no glance at all) to
explicit hand gestures and head movements, claiming
the space (Rasouli et al, 2017).
Vulnerable pedestrians may have special
requirements. Consider for example, people with
mobility, cognitive or perception impairments (Xiang
et al., 2006), or age-related difficulties expressing
vulnerable body language (5). Using eye contact and
a reassuring facial expression, the driver in the
vehicle may enhance the sense of safety for slow
pedestrians crossing the street. The message for the
normal population: [“I am giving you the right of
way” – f], may add an extra sense of patience and
protection to slow pedestrians such as elderly or
disabled users: [“I am respecting your space and will
not act against you”- g], [“I am also looking around
you by being attentive to the surroundings and other
vehicles that may infringe this space” - h]. Upon the
completion of the crossing act, the transaction ends
from the perspective of the pedestrian - “Bye” (6).
The vehicle in return, can communicate its intent to
leave [“I am leaving”- i]. This is not that important
for the pedestrian who has completed the crossing act
but may be useful for those that are planning to start
crossing the road while the vehicle is in a stop
position. Starting to drive ends the transaction [“Bye,
I am starting to drive” - j].
There are several variations to the above
sequence, however, in all variations the pedestrian,
who is more vulnerable, has priority over the
autonomous vehicle. The sequence will be similar if
the pedestrian is standing at the edge of the curb,
waiting for the vehicle to stop and yield. However,
when the pedestrian is already crossing the street
while the vehicle is approaching the scene, the sense
of vulnerability is the greatest and pedestrians tend to
establish an eye contact with the driver as the vehicle
gets closer (Dey et al., 2019).
2.2 Needs Study
This conceptual analysis is followed by a Wizard of
Oz (WoZ) needs and concerns study, which focuses
on the encounter between pedestrians and a fake
autonomous vehicle. We aim to identify what people
actually expect from autonomous vehicles in the
public space and what would make them feel more at
ease and accepting of this technology.
One of the most interesting questions is how can
grounding be established in the absence of a human
driver, what elements would be missing? Shmueli &
Degani (2019) conducted a naturalistic study to get a
clearer understanding of pedestrians’ needs and
reactions during mundane, non-urgent, crossing
scenarios. This study was conducted at General
Motors campus where employees experienced an
encounter with an autonomous vehicle driven
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
202

manually. As an alternative to the GhostDriver
technique (Rothenbücher et al., 2016), different
methods were used to eliminate eye contact with the
safety driver and minimize pedestrians’ expectations
of potential interventions by the safety driver: the
driver was wearing a helmet and avoided head
movements and hand gestures that could have implied
pedestrians that they had been noticed. In addition,
the driver was instructed to keep his hands low on the
steering wheel and apply a more robotic driving style.
People encountered the vehicle and were then stopped
by the researchers for a short interview to better
understand their experience during the encounter. No
specific communication language was used.
30 encounters of 36 pedestrians (7 females and 29
men) were documented. In-depth interviews that were
conducted following the encounter with the vehicle
provided insights into the participants’ intellectual
and emotional confusion, fears and wishes. Users’
comments were collected and classified into the
following categories: (i) Intent to stop & Yield, (ii)
Wait & Give way, and (iii) Intent to drive. Consistent
with the literature (Rasouli et al, 2017; Habibovic et
al. 2018; Merat et al., 2018; Dey, 2021), the
autonomous vehicle’s “intent to stop & yield” was
identified as the main user need with 32 comments:
the interviews yielded 10 comments regarding the
wish for a general notification, 14 comments
specified the need to be personally acknowledged by
the vehicle: “That the vehicle is planning to stop for
ME”, 6 comments indicated the wish to get guidance
and warnings from the vehicle, and 2 comments
described a sophisticated dialogue that can be formed
with the vehicle. In contrast, only 6 comments
mentioned the need for a continuous indication while
the autonomous vehicle is in a stop position. Some
participants mentioned that other people, especially
slower users (elderly people, people with young
children, or people who suffer from some form of
disability) may require to be acknowledged also while
crossing. Finally, 10 comments cited the need for a
dedicated indication when the vehicle resumes
driving for the reasons outlined in the Analysis
section. Overall, the conceptual analysis and the
empirical study suggest that some reassurance is
needed once driverless autonomous vehicles are
introduced in the public space and that some people
are likely to require more reassurance than others.
2.3 Protocol: Representation
Based on the conceptual analysis (in section 2.1), and
a better understanding of how people respond to the
technology in context (in section 2.2), we derived a
set of general communication needs. The set,
presented in Figure 1, is more comprehensive than the
three main touchpoints discussed in the previous
section.
Figure 1: Communication protocol.
We have expanded it to include other needs such
as acknowledgement from a distance (“I see you”)
and deceleration and acceleration that allow for the
assessment of a more complete design sequence. The
ability to augment the dynamic profile of the vehicle
by eHMI design is intuitively promising and will be
further discussed in Section 2.6. We also included a
“protection” state for pedestrians with special needs
and for those who are more vulnerable.
 These
touchpoints define
 the full  set of information content
that can be portrayed to the pedestrian.
Now comes the question of what kind of format
will each communication touch point employ? The
first phase in the design process is content selection
(“what to convey”); the second is media allocation
(“which medium – and hence, modality - to say it in”)
modality; the third, media realization (“how to say it
in that medium; how to design the content?”) and the
fourth media coordination (``how to coordinate
several media'') (Maybury, 1993).
2.4 Content Selection for eHMI
Displays
The method starts by classifying the spectrum of
contents that can be communicated by the
autonomous vehicle into 7 different information types
that were identified by ISO (ISO/TR 23049): Mode,
State, Perception, Recognition & Acknowledgement,
Belief state, Intent, Guidance. Car manufacturers may
differ in the information types that they wish to
include in their communication language. The
selected solution may depend on advances in sensing
capabilities and on available hardware as these may
eHMI Design: Theoretical Foundations and Methodological Process
203

limit the safety area that one can guarantee. In
addition, various policies and international standards
may serve as a filter with regards to what to present
to guarantee safety. Guidance for example - The
autonomous vehicle can act as a semaphore (e.g., like
a traffic light) providing supervisory control to road
users. Providing full guidance involves great
responsibility (and potential liability issues since, at
the moment, car manufacturers are not equipped with
all traffic information). Therefore, it is recommended
to avoid communicating that it is safe to pass, so long
that we cannot guarantee the safety of the pedestrians
from all sides of the road. Other considerations may
also involve planning constraints, the wish to avoid
visual clutter, or simply cost.
2.5 Media Allocation: Codes and
Modalities
The next consideration is how the information
selected should be represented and presented. We
propose Wickens’ Multiple Resources Theory (MRT)
(1984) as a framework for making representation
decisions. The MRT asserts that people have a limited
set of resources available for mental processes, from
sensing to response execution. It consists of four
dimensions: (1) stages of attentional processing (2)
processing codes, (3) perceptual modalities, and (4)
response execution.
1. Resources used for perceptual and cognitive activities,
are shared, and are functionally separate from those
underlying the selection and execution of responses.
2. Spatial activity uses different resources than
verbal/linguistic activity, as evident by working
memory studies (Baddeley, 1986) and action studies
(e.g., speech vs. manual control; Liu & Wickens, 1992;
Wickens & Liu, 1988).
3. Auditory perception uses different resources than visual
perception.
4. Manual and vocal reactions rely on separate resources.
Figure 2: Multimodal Design Space.
In 2002, Wickens added an important qualification
regarding foveal and peripheral vision that can take
place concurrently: Focal vision primarily, but not
exclusively foveal, supports object recognition and in
particular high acuity perception such as that involved
in reading text and recognizing symbols. Ambient
vision distributes across the entire field and (unlike
focal vision) can preserve its competency in
peripheral vision. Ambient vision is responsible for
perception of orientation and movement, for tasks
such as those supporting walking upright in targeted
directions or lane keeping in the highway (Horrey et
al., 2006). This qualification is of a great importance
to the discussion of the potential augmentation of the
ambient perception of vehicle movement by an
external communication language which may involve
focal vision.
Figure 2 outlines the model which is based on the
theory. The processing stages appear on the X axis,
the modalities appear on the Y axis and the codes
appear on the Z axis. Following the MRT, media
allocation in eHMI design can be framed in terms of
Codes (verbal & spatial) and Modalities (visual &
auditory). In visual displays, communication may
rely on symbolic representations: visual text, icons,
and/or on visual-spatial information. In auditory
displays, it may rely on speech and/or non-verbal
(abstract) earcons and auditory icons (everyday
sounds), but also on spatial cues via the engine sound
in internal combustion engine vehicles, or the
external sound system in electric vehicles.
The method proposes that the assignment of
information into codes and modalities should be
based on the pedestrians’ dynamic needs throughout
the interaction with the autonomous vehicle and the
available physical platform. Clearly, the presentation
of text and icons require a minimal display height, so
that the observer will be able to see it from a distance;
a single LED strip/band would not be suitable for this
purpose. Nevertheless, the compatibility of the design
with the processing requirements of the different
touchpoints during the encounter with the vehicle is a
central aspect and may affect the selected platform.
The protocol presented in section 2.3 can be therefore
fragmentized into sub-tasks that fall into one of the
following processing codes: verbal (conceptual) or
spatial, or both. According to the Stimulus-Coding-
Response (S-C-R) compatibility principle of the
MRT (Wickens et al. 1984), tasks with verbal central-
processing demands will be best served by voice
input and output channels, but verbal-visual contents
such as visual text and icons will also produce an
effective resources usage due to their compatibility
with the processing demands. Similarly, tasks with
spatial demands will be best served by visual-manual
channels but may benefit also from spatial auditory
information. Hence, code representations for each
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
204

sub-task determine the optimal stimulus type and
response code.
2.6 Media Realization: eHMI Design
Requirements
eHMI design should be artistically pleasing but most
importantly, be semantically appropriate. Wickens’
MRT is highly useful when seeking logic, specifically
when considering compatibility of the perceptual
content with the processing requirements of the task.
Next, we translate the communication protocol into
spatial and conceptual sub-tasks.
2.6.1 Driving Mode: Conceptual
Some autonomous vehicles will have a unique
appearance which cannot be mistaken with a manually
driven vehicle, while other vehicles may maintain a
more traditional look. The determination of driving
mode is primarily a conceptual “verbal” sub-task that
should be made from a distance. In terms of media
allocation & realization, this calls for a verbal-
conceptual visual solution. To support visibility from a
distance, large visual text and icons would require large
platforms that may induce visual pollution. A more
minimalistic solution can be used to meet the
Efficiency criterion. The proposal made by SAE
J3134, a marker lamp in cyan that will indicate on a
single LED band, or a larger visual panel the
autonomous driving status of the autonomous vehicle,
could be considered. At the auditory level, a unique
sound language for electric autonomous vehicles may
promote their recognition by visually impaired users.
2.6.2 “I See You”: Conceptual (Potentially
Spatial)
Understanding that the vehicle notices the pedestrian is
a conceptual task with a spatial component. As noted
earlier, pedestrians want to know that the vehicle
detects them. As for media allocation & realization, a
mild cue which reflects the spatial location of the
pedestrian, before an intent to stop is broadcasted, may
be a useful cue if it can be perceived from a distance.
2.6.3 Intent to Stop & Yield: Conceptual
(Potentially Spatial)
Understanding the vehicle’s intent to stop is a
conceptual task that currently relies mainly on its
deceleration profile. This is not always a sufficient
cue (Šucha et al., 2017; Schieben et al., 2018; Klatt et
al., 2016; Ren et al., 2016). At any rate, the intent to
stop & yield should attract the visual attention of the
pedestrians who may be located 40-60 meters away.
The visual effect should be carefully planned, it
should be clear yet non-distractive to avoid visual
capture by pedestrians and other drivers. Another
challenge is communicating the intent of the vehicle
without providing an explicit guidance.
2.6.4 Deceleration: Spatial
Assessing the autonomous vehicle’s deceleration rate
is a crucial element in the pedestrian’s ability to
assess the Time to Arrival (TTA) of the vehicle. The
assessment of TTA is primarily a spatial task; people
rely on their visual (foveal and peripheral) and
auditory perception to perform it. Motion cues and
users’ trajectories can be used by automated driving
systems to promote better integration into the traffic
environment (Domeyer, 2020). However, studies
have shown that estimates based on motion can be
biased by occluding objects and by the vehicle speed.
Vehicle size can also affect these estimates, with
larger vehicles being estimated to arrive earlier (Caird
and Hancock, 1994; Delucia, 1991a, 1991b). In
addition, factors associated with capabilities of other
road users have also been identified as affecting
crossing decisions. For example, older adults and
young children suffer from a reduced ability to
estimate TTA (Andersen & Enriquez, 2006; Dommes
& Cavallo, 2011).
There is a potential to strengthen the
understanding of the pedestrian, especially in low-
speed situations where vehicle kinematics may be
ambiguous (Domeyer et al. 2020) and avoid
perceptual mistakes by using an explicit signal. This
may assist the ambient vision in poor visibility
conditions (poor weather, poor lighting), help
pedestrians to correct perceptual biases, and assist
pedestrians with restricted visual and hearing
capabilities. The analysis of the MRT calls for a
visual-spatial media solution rather than a symbolic
one. A visual-spatial animation that implies on speed
reduction seems advisable.
2.6.5 Actual Stop: Conceptual
This state reflects the situation where the vehicle is
approaching 0 MPH, after which it enters the wait
state and has a strong conceptual component. This
touchpoint can be communicated visually and
auditorily by verbal or symbolic means.
2.6.6 Wait: Conceptual (Potentially Spatial)
In this state, the design aims to replace the driver’s
eye contact with the pedestrian. While some people
eHMI Design: Theoretical Foundations and Methodological Process
205

may trust that the vehicle will stay stopped once it has
reached 0 MPH, others may need extra assurance
regarding its intent to let them cross the road. The
solution must be intuitive and easy to understand
across cultures and age groups. Nevertheless,
communicating that the vehicle will wait patiently is
a conceptual message that should not be confused
with an explicit positive guidance to cross.
Naturally, a conceptual solution that
communicates patience may involve symbols or an
abstract, universal, animated effect. An additional
spatial component can be optionally added; a spatial
tracking of the pedestrian’s reflection may replace the
driver’s eyes that follow the pedestrian while
the vehicle is waiting (See various design
implementations of this principle by Nissan,
Mercedes Benz: F015 & The cooperative vehicle,
AutonoMI, Volvo 360C). The pedestrian’s reflection
tracks the spatial position of the person/s in real-time.
Their perception can be conveyed by an animated
effect on the lighting fixture confirming recognition
before, at the beginning of the crossing, or during the
act of crossing, when they feel the most vulnerable.
2.6.7 Intent to Drive: Conceptual
This message is essential for those who want to start
crossing after the agent decided that it is leaving. The
‘Intent to drive’ precedes the acceleration state
(spatial) but it is nonetheless effectively a conceptual
message, or task. An ‘intent to drive’ can be
communicated visually using an attention-grabbing
visual animation, supplemented by an auditory cue to
accommodate visually impaired pedestrians, or by
verbal or symbolic means.
2.6.8 Acceleration: Spatial
Similar to deceleration, acceleration is a spatial task,
and any visual and auditory spatial representation of
speed increment should refer to the solution provided
for the deceleration state. In addition, the visual
animation could potentially be synchronized with the
overall pace of speed increment of the vehicle.
2.6.9 Failure to Stop Warning: Conceptual
Warnings are conceptual in nature and are directed
toward all road users in the vicinity, perhaps in an
omnidirectional manner. This calls for an auditory
solution. There is a possibility to enhance the auditory
warning by a compatible visual animation that shares
temporal and spectral characteristics with the sound.
2.6.10 Pedestrian-Specific Warning:
Spatial
In situations that warrant warning of specific
pedestrians, where the information is directed toward
a specific area, the auditory warning may have a
directional spatial nature.
2.6.11 “Protection”: Both Conceptual &
Spatial
A sophisticated eHMI can provide special
information, in particular to those in need (e.g.,
children, elderly users, disabled road users) and those
that seem apprehensive or reluctant to cross. The
vehicle can send them a message about its
commitment to yield and protect while they are
crossing. Just as in the case of the pedestrian-specific
acknowledgement, the realization of this requirement
may have a directional, spatial nature and can be
coupled with a visual tracking design solution. The
animation itself should be designed so as to convey
the fact that the vehicle is waiting patiently and will
remain stopped until the user completes the task of
crossing the road. A very subtle pulsation of the entire
display in case of a non-pedestrian-specific design, or
the pedestrian/s’ reflection in case of a pedestrian-
specific design, may boost the pedestrian's
confidence.
Having summarized the resources requirements
and input of each touchpoint, it is important to note
that an elegant design requires some form of
continuity. Subtle and sophisticated animated
transitions may be required while shifting between
different phases.
2.7 Verification
We now wish to verify this proposed method by
looking into several design concepts that have been
published in the literature and assess if they meet the
four criteria listed in section 1: Effectiveness,
Efficiency, Acceptability & Satisfaction. One
interesting example is the pioneering study conducted
by Clamann, Aubert & Cummings (2016) at Duke
University. This study aimed to assess the effect of
display content on the participants’ decision to cross.
Two display types were used: (i) an advisory
(guidance) display that consisted of a ‘Don’t Walk’
symbol while the vehicle was in motion (this
corresponds to all touchpoints except the Wait state
and the Intent to drive in the protocol of section 2.3),
and that switched into a ‘Walk’ symbol when the
vehicle came to a stop, and (ii) an information display
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
206

portraying the vehicle’s speed by means of dynamic
digits. In both cases, the designs relied on visual-
verbal coding that violates the spatial cognitive
processing requirements of the vehicle’s motion cues
in the deceleration assessment task. The two display
formats failed to facilitate pedestrians’ decision to
cross the road. Clamann et al. (2016) study is often
cited as a proof of eHMI failure in producing
improvements in road users’ comprehension of the
autonomous vehicle’s intention and the proposed
account provides an alternative explanation for the
results of this study.
The automated vehicle interaction principle
(AVIP) project which was developed with the
Swedish Victoria ICT (2015) provides a more
comprehensive design (Lagström & Malmsten
Lundgren, 2015, Habibovic et al., 2018). The visual
display consists of a LED band above the windshield.
The design represents the vehicle’s autonomous
driving mode using a centralized light in the centre of
the screen (a conceptual solution). The vehicle’s
Intent to stop is communicated by the expansion of
this light, which continues to expand further while the
vehicle is decelerating (a visual-spatial solution that
merges the conceptual Intent to stop and the spatial
Deceleration state). This visual expansion of light is
compatible with the effect of looming, or the visual
enlargement of an object as it approaches the viewer).
The light reaches its maximal size when the vehicle is
Stopping, and subtle pulsation of the full band,
imitating human breathing rhythm, communicates the
fact the vehicle is Waiting for the pedestrian to cross
the road (a conceptual solution that does not
acknowledge specific pedestrians). When the vehicle
Intends to drive, the light converges back to the centre
in a smooth animation. There is neither a
representation of Acceleration, nor a representation of
Warning information. This eHMI design is partially
consistent with our method and was successful in
inducing a sense of safety and improved confidence
in the vehicle’s automation technology (Lagström &
Malmsten Lundgren, 2015; Habibovic et al., 2018).
In 2018, Habibovic et al. mentioned that several
pedestrians stated that the pulsating light during the
Wait state was not contributing to their experience
and suggested that it could be removed to make the
interface easier to understand.
Interestingly enough, the European project
INTERACT used the same breathing light metaphor
to communicate the vehicle’s Intent to stop and yield
and the Wait state, merging these two conceptual
elements with the spatial Deceleration state that has
no unique representation in this design concept. The
only spatial representation in this implementation is
facilitated by a separate tracking lamp (by HELLA)
to acknowledge specific pedestrians crossing the
street. Finally, the vehicle’s Intent to drive is
communicated by the full light band which pulsates
quickly a few times, coupled with an auditory cue.
Similar to the AVIP concept, this eHMI concept was
found to increase participants’ comprehension of the
vehicle’s intention and elevated their level of trust
toward the automated driving technology. The two
concepts and hence meets the Effectiveness,
Efficiency and Acceptability criteria.
To summarize, these three examples show that as
long as there is no strict violation in the compatibility
between the design and the required processing code,
concepts meet the Effectiveness, Efficiency and
Acceptability criteria. A beneficial effect of the eHMI
design is identified: An increased efficiency, along
with increases in perceived safety, comprehension,
and trust.
2.8 Validation
After we filter out incongruent solutions, still there
remains an abundance of eHMI design solutions that
are theoretically valid. It is hard to evaluate in
advance which solution will prove better than the
others. Therefore, there is a need to contrast them
empirically using controlled testing methods (such as
video analyses and VR) in laboratory conditions and
then, externally, in safe test tracks and natural road
context.
Currently, the literature does not provide a full
answer regarding the best, complete, eHMI solution.
There are quite a few examples of validation attempts
of specific concepts (AVIP - Lagström & Malmsten
Lundgren, 2015; Habibovic et al., 2018; The
Mercedes Benz’s Cooperative vehicle - Faas &
Baumann, 2019; Ford concept vehicle - Hensch et al.,
2019b, to name a few). This line of research sheds
light on the intuitiveness and comprehensiveness of
specific solutions but does not tell us which
representation strategy is superior to others. On the
other hand, comparative assessments (Ackermann et
al., 2019; de Clercq et al., 2019; Fridman et al. 2017;
Dey et al, 2020) have been useful in revealing the
relative intuitiveness of specific designs, specifically
regarding the Intent to stop & yield, and the Wait
state. However, apart from a partial attempt by Dey et
al. (2020), who tested abstract visualizations, no
filters have been applied on the Guidance
information-type mentioned earlier. These studies
contrast abstract lighting solutions with colour-based
and icon-based traffic lights solutions, and with
textual messages such as “After you”, “Go ahead”, or
eHMI Design: Theoretical Foundations and Methodological Process
207

“Safe to cross”. As people seek to reduce ambiguity,
they show preference towards non-ambiguous
solutions that may put them at risk.
A more appropriate contrast should involve states
and intents communication designs that are lacking
any component of Guidance. Different visual-abstract
Intent to stop & yield designs, with and without the
visual-spatial Deceleration component, should be
contrasted. In addition, the understanding of the
optimal communication solution for the Wait state
should be based on contrasting several visual-
conceptual solutions that communicate that the
vehicle will wait patiently, with and without a spatial
tracking component that provides acknowledgement
of specific pedestrians. A state-informing textual
message, such as “Waiting” may be added to the
comparison, in spite of the fact that it necessitates
reading capabilities, and, as noted by Dey (2021), if
the message is partially occluded, for example the
”ing” part, both agents will wait for the other to act
and traffic-flow would suffer.
As a final note, even when reverting to novel
cross-cultural, abstract, visual-conceptual solutions,
the issue of guidance emerges. For example, Shmueli
& Degani (in preparation) conducted a small,
controlled, study to assess the meaning assigned to a
central-pulsating light in a fixed position vs. a
pulsating tracking-light. Participants perceived the
pulsating tracking-light as their reflection: “the
vehicle sees me… it recognizes me… I feel safe”. This
did not make them assume that the vehicle is
responsible for the road space beyond them. On the
other hand, when the light was pulsating in the centre
of the lighting display, one participant commented
that he would quickly learn to interpret it as a green
light that guides to cross the road. We believe that this
desire to receive a semaphore signal is common to
many people; The tendency to find positive guidance
in neutral solutions should be thoroughly assessed.
An additional aspect that our method cannot fully
predict is the optimal representation of deceleration
information by abstract visual means. Concepts that
try to tap the deceleration state of the vehicle vary
significantly: light bands diverge (AVIP; Bumper PB
eHMI in Dey et al., 2020b), converge (Volvo 360C),
descend (Shmueli & Degani in preparation). Each has
its own logic for the achievement of the right
semantics of the direction of motion of the
approaching vehicle:
‐ The movement compatibility principle (Warrick,
1947), determines that the direction of movement
in the display should be consistent with the
direction of movement of the approaching
vehicle, suggesting some superiority for flow to
the centre (converging light band)
‐ Looming (Lee, 1976), or the visual expansion of
an object as it approaches, calls for the expansion
of the light band (diverging light band)
‐ The conceptual compatibility with speed
reduction calls for a descending light band
Further research is needed to determine (i) which
visualization conveys best speed reduction
information and (ii), the effect of synchronicity of the
animated effect with the deceleration profile of the
vehicle and with the electric vehicle’s external sound
output.
3 DISCUSSION
An important element in the process discussed above
is Wickens’ MRT and S-C-R Compatibility principle
that are applied to step 5 – Media Allocation – Codes
and Modalities. The theory can be used to identify the
type of codes that are most appropriate for each type
of communication. Designers can optimize their
design solution if they meet the nature of the codes
that underly various touchpoints between human
users and autonomous vehicles. The approach is also
useful as a verification method to analyse the
potential success of eHMI concepts and can pinpoint
specific questions for further research. Evidence
suggests that design solutions that compromise the
compatibility of the perceptual content with the
processing requirements of the task are bound to fail.
Additional decisions should be taken to reach a
unified design:
1. Reach Agreement Concerning “Content
Selection”. Looking at current concepts, it seems
that there is some agreement by car
manufacturers and research institutes about the
need to represent the vehicle’s intents: a
substantial consensus regarding the
representation of the Intent to stop & yield, and
partial agreement regarding the need to represent
the vehicle’s Intent to drive. With respect to the
state information, there is a great variability
between concepts: a few concepts represent the
Wait state in a unique format and a very small
number of concepts try to enhance the
deceleration of the vehicle using visual
animation. Furthermore, there is some
disagreement about the necessity to represent the
vehicle’s mode, which sets the canvas for
subsequent representations.
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
208

2. Reach Agreement About the Optimal “Media
Allocation”. The MRT dictates some design
requirements from the perspective of effective
resources usage: the conceptual processing
requirements of the task will be best served by
conceptual information, whereas the spatial
processing requirements of the task will be best
served by spatial content. Further research
should be conducted to decide which information
should be presented visually, which should be
delivered auditorily, which should be presented
to both modalities, and in the latter case – how
should the information be integrated between
modalities to enhance the communicative value
of the design.
3. Reach Agreement About the Optimal Media
Realization: Design Decisions. This discussion
concerns the selection of colour, animation
design and the interaction between colour and
animation
Colour: Several constraints regarding colour
currently apply by the Federal Motor Vehicle Safety
Standards (FMVSS) in the US and the United Nations
Economic Commission for Europe (UNECE) to
guarantee that eHMI colours should not interfere with
colours already implemented or reserved for specific
functions. Traditionally, the allowable colours for any
moving vehicle consist of white and amber at the
front and sides of the vehicle and red at the rear
(FMVSS 108). Additional colours are reserved to
traffic devices and emergency vehicles. As specified
by SAE J578 Standard, restricted colours are: Red,
Yellow (Amber), Selective Yellow, Green, Restricted
Blue, Signal Blue and White (Achromatic).
Candidate colours to represent states and intents of
autonomous vehicles are therefore Cyan (green blue),
Selective Yellow, Mint-Green, and Purple/Magenta
(Tiesler-Wittig, 2019; Werner, 2018; Dey 2021).
Cyan seems to be a promising choice. It is not
prevalent in traditional urban and highway contexts
and will therefore stand out in a bright daylight. In
addition, there is some precedence for the selection of
cyan in the industry; European and Japanese car
manufacturers use cyan to mark autonomous vehicles
in some design concepts. In addition, the SAE chose
this colour to indicate the autonomous driving status
of autonomous vehicles (SAE J3134). The suitability
of cyan is also supported by several studies (e.g., Faas
and Baumann, 2019; Beggiato et al., 2019; Hensch et
al., 2019b) which show that pedestrians prefer cyan
over white to indicate automation mode. The reliance
on cyan can be elaborated further to express
additional vehicle’s states and intents.
Animation: Another topic that needs to be
addressed is animation. As noted earlier, an abstract
animated eHMI language is an appealing concept,
due to its cross-cultural universal nature and the fact
that it does not require reading capabilities, that is - if
designed correctly. Clearly, some consensus should
be reached regarding the animated effects employed
by autonomous vehicles and the messages that they
represent. Current regulation allows no animation,
except for emergency vehicles (see SAE J845, SAE
J595 and SAE J2498, all were written before the
advent of autonomous vehicles). Similar to the
selection of colour, a substantial body of research
should be used to reverse these restrictions and allow
the use of animated content by autonomous vehicles.
The interaction between Colour and Animation
Colours are loaded with positive and negative
meanings when static. For example, Bazilinskyy et
al. (2019a) noted the compatibility of green with
positive guidance messages such as “please cross”
and recommend avoiding green if the eHMI is
intended to signal a negative guidance message, such
as “please do not cross”. This finding was generalized
also for aquamarine (sRGB 127, 255, 215) by
Bazilinskyy et al. (2021). A pure cyan (sRGB 0, 255,
255) has no associative loading, although Dey et al.
(2020b) report that cyan is perceived as “close to
green” and is hence suitable for yielding signals.
Interestingly, when coupled with a strong-
dynamic animation, cyan (and to a greater extent,
magenta) can be associated with emergency vehicles,
producing negative valence. This interaction between
colour and animation was demonstrated in two
studies by Beggiato et al. (2019) and Hensch et al.
(2019b) on the eHMI design of Ford which consists
of a blinking light to communicate the Intent to drive
and a vigorous dual-sweep animation format to
communicate the Intent to stop and yield message;
These studies show that vigorous light animations
provoke negative valence and a sense of alert, unless
a neutral colour such as white is being used. This
demonstrates the complexity of coming up with an
acceptable design, due to the difficulty of finding
what is the relative importance of each element in the
design space (colour, animation, and prior loading
such as police signals).
4 CONCLUSIONS
This paper provides an eight-step analytical method
starting from analysing pedestrians’ requirements in
their encounters with manual vehicles and
subsequently deduces a comprehensive
eHMI Design: Theoretical Foundations and Methodological Process
209

communication protocol with future, driverless,
automated vehicles. Maybury’s (1993) principles of
content selection (sampling information), media
allocation (assignment of information into media and
modalities), media realization (design methods) and
coordination of different media that can be used to
materialize this protocol into eHMI design while
obeying the recommendations derived from
Wickens’ (1984) MRT and S-C-R Compatibility
Principle. We envision that this method will serve to
reduce the almost endless design space to a much
more manageable space. It will also allow the
industry to use the media-coding verification
suggested here to reject poor designs and harmonize
toward a more optimal design standard. Finally, the
universality of this methodological approach and
development process can be used by other industries
that will need to develop mechanisms for
communication between automated and autonomous
machines and humans; whether in an enclosed space
(e.g., factories and warehouse) or in the public space.
eHMI communication language may also inspire the
design of robots that are not necessarily in the public
space or in immediate space conflict with humans.
Since the automotive industry will be the first to
deploy such robots and provide communication
systems, we believe that other industries that will
deploy robots in the public space (delivery robots,
health care and assistive robots, information kiosk
robots) will eventually be required to also provide
such communication. Naturally, they may look at
regulated solutions in the automotive industry to
inspire their designs.
REFERENCES
Ackermann, C., Beggiato, M., Schubert, S., Krems, J. F.
(2019). An experimental study to investigate design and
assessment criteria: What is important for
communication between pedestrians and automated
vehicles? Applied Ergonomics, 75, 272–282.
Andersen, G. J., Enriquez, A. (2006). Aging and the
detection of observer and moving object collisions.
Psychology and Aging, 21(1), 74–85.
AutonoMI by Leonardo Graziano. AutonoMI Autonomous
Mobility Interface. Video.
Baddeley, A.D. (1986). Working Memory. OXFORD:
OXFORD UNIVERSITY PRESS.
Bazilinskyy, P., Dodou, D., de Winter, J.C.F., (2019a).
Survey on eHMI concepts: the effect of text, color, and
perspective. Transport. Res. F Traffic Psychol. Behav.
67, 175–194.
Bazilinskyy, P., Kooijman, L., Dodou, D., de Winter, J.C.F.
(2021). How should external human-machine interfaces
behave? Examining the effects of colour, position,
message, activation distance, vehicle yielding, and
visual distraction among 1,434 participants. Applied
Ergonomics, Vol 95.
Beggiato, M., Neumann, I., Hensch, A.C (2019). Light-
based communication between automated vehicles and
other road users. Chemnitz University of Technology,
Department of Psychology. Cognitive and Engineering
Psychology. Unpublished presentation Ford_
Color_AVSR-02-17e
Caird, J. K., Hancock, P. A. (1994). The perception of
arrival time for different oncoming vehicles at an
intersection. Ecological Psychology, 6(2), 83-109.
Clamann, M., Aubert, M., Cummings, ML. (2016).
Evaluation of vehicle-to-pedestrian communication
displays for autonomous vehicles. Technical Report.
Cohen, J., Dearnaley, E. & Hansel, C. (1955). The Risk
Taken in Crossing a Road. Journal of the Operational
Research Society. Vol. 6, 120–128.
de Clercq, K., Dietrich, A., Núñez Velasco, J.P., De Winter,
J., Happee, R., 2019. External human-machine
interfaces on automated vehicles: effects on pedestrian
crossing decisions. Human Factors 61, 1353–1370.
Delucia, P. R. (1991a). Pictorial and motion-based
information for depth perception. Journal of
Experimental Psychology: Human Perception &
Performance,17, 738–748.
Delucia, P. R. (1991b). Small near objects can appear
farther than large far objects during object motion and
self-motion: Judgments of object-self and object-object
collisions. In P. J. Beek, R. J. Bootsma, P. C. W. van
Wieringen (Eds.). Studies in perception and action:
Posters presented at the Vlth International Conference
on Event Perception and Action (pp. 94–100).
AMSTERDAM: RODOPI.
Dey, D., Walker, F., Martens, M., Terken, J. (2019). Gaze
patterns in pedestrian interaction with vehicles:
Towards effective design of external human-machine
interfaces for automated vehicles, Proceedings - 11th
international ACM conference on automotive user
interfaces and interactive vehicular applications,
automotive ui 2019, Utrecht, The Netherlands.
Dey, D., Habibovic, A., Pfleging, B., Martens, M., Terken,
J. (2020a). Color and Animation Preferences for a Light
Band eHMI in Interactions Between Automated
Vehicles and Pedestrians, Chi conference on human
factors in computing systems, Hawai’i, Honolulu,
United States.
Dey, D., Holländer, K., Berger, M., P-eging, B., Martens, M.,
& Terken, J. (2020b). Distance-Dependent eHMIs for the
Interaction Between Automated Vehicles and
Pedestrians, 12th international conference on automotive
user interfaces and interactive vehicular applications
(automotiveui ’20), Virtual Event, DC, USA.
Dey, D. (2021). External Communication for Self-Driving
Cars: Designing for Encounters between Automated
Vehicles and Pedestrians. Doctoral dissertation. Future
Everyday, Industrial Design Eindhoven University of
Technology.
Domeyer, J.E., Lee, J.D., Toyoda, H. (2020). Vehicle
Automation—Other Road User Communication and
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
210

Coordination: Theory and Mechanisms. IEEE Access
8: 19860-19872
Dommes, A., & Cavallo, V. (2011). The role of perceptual,
cognitive, and motor abilities in street-crossing
decisions of young and older pedestrians. Ophthalmic
and Physiological Optics, 31(3), 292–301.
DOT HS 813 148. NHTSA (2021). Automated Driving
Systems’ Communication of Intent With Shared Road
Users https://rosap.ntl.bts.gov/view/dot/58325
Faas, S.M., Baumann, M., (2019). Light-based external
Human Machine Interface: color evaluation for self-
driving vehicle and pedestrian interaction. Proceeding
of the Human Factors and Ergonomics Society Annual
Meeting. 63, 1232–1236.
Federal Motor Vehicle Safety Standard 108 (FMVSS 108)
Ford eHMI design https://www.youtube.com/watch?
v=boqG7Ss7chI&t=182s
Fridman, L., Mehler, B., Xia, L., Yang, Y., Yvonne, L.,
Bryan, F. (2017). To walk or not to walk:
Crowdsourced assessment of external vehicle-to-
pedestrian displays.
Gueguen, N., Meineri, S., Eyssartier, C. (2015). A
pedestrian’s stare and drivers’ stopping behaviour: A
field experiment at the pedestrian crossing, Safety
Science, vol. 75, pp. 87–89.
Habibovic, A., Andersson, J., Malmsten Lundgren, V.,
Klingegård, M., Englund, C., Larsson, S. (2018).
Communicating intent of automated vehicles to
pedestrians, Frontiers in Psychology., vol. 9, p. 1336.
Hella-Light. https://www.hella.com/hella-com/en/press/
Company-Technology-Products-12-12-2017-
16333.html
Hensch, A.C., Neumann, I., Beggiato, M., Halama, J.,
Krems, J.f. (2019b). Effects of a light-based
communication approach as an external HMI for
Automated Vehicles — a Wizard-of-Oz Study.
Transactions on Transport Sciences. Vol. 10 (2), 18-32.
Horrey, W. J., Wickens, C. D., Consalus, K. P. (2006).
Modeling drivers' visual attention allocation while
interacting with in-vehicle technologies. Journal of
Experimental Psychology: Applied, 12(2), 67–78.
International Organization for Standardization (ISO)/TR
23049. 2018. TECHNICAL REPORT ISO: Road
Vehicles — Ergonomic aspects of external visual
communication from automated vehicles to other road
users. Technical Report.
Klatt, W. K., Chesham, A., Lobmaier, J. S. (2016) Putting
up a big front: car design and size affect road-crossing
behaviour. PLoS One 11(7): e0159455.
Lagström, T., Malmsten Lundgren, V. (2015) AVIP—
autonomous vehicles interaction with pedestrians: an
investigation of pedestrian driver communication and
development of a vehicle external interface. Master
Thesis. Chalmers University of Technology, Göteborg
Lee, D.N (1976). A theory of visual control of braking
based on information about time-to-collision.
Perception 5, 437-459
Lehsing, C., Jünger, L. & Bengler, K. (2019). Don’t Drive
Me My Way – Subjective Perception of Autonomous
Braking Trajectories for Pedestrian Crossings. In
SoICT’19: The Tenth International Symposium on
Information and Communication Technology,
December 4–6, 2019, Hanoi – Ha Long Bay, Viet Nam.
ACM, New York, NY, USA, 7 pages
Liu, Y. and Wickens, C. D. (1992). Visual scanning with or
without spatial uncertainty and divided and selective
attention, Acta Psychologica, 79, 131-153.
Maybury, M.T. (1993) Intelligent Multimedia Interfaces.
Cambridge, MA: AAAI Press/MIT Press
Merat, N., Louw, T., Madigan, R., Dziennus, M., Schieben,
A. (2018). Communication between VRUs and fully
automated road transport systems: what’s important?
Accident Anal Prev.
Mercedes Benz F015 - luxury in motion Concept Vehicle.
https://www.youtube.com/watch?v=MaGb3570K1U
Mercedes Benz - The cooperative vehicle.
https://www.youtube.com/watch?v=VFu2dyqUQPw
Nissan’s vision for the future of EVs and autonomous
driving. https://www.youtube.com/watch?time_continue
=6&v=h-TLo86K7Ck
Rasouli, A., Kotseruba, I., Tsotsos, JK. (2017). Agreeing to
cross: How drivers and pedestrians communicate. In:
Proceedings of the 2017 IEEE intelligent vehicles
symposium (IV), pp 264–269.
Ren, Z., Jiang, X. and Wang, W. (2016). Analysis of the
influence of pedestrians’ eye contact on drivers comfort
boundary during the crossing conflict, Procedia
Engineering, vol. 137, pp. 399–406.
Risto, M., Emmenegger, C., Vinkhuyzen, E., Cefkin, M.,
Hollan, J. (2017). Human–vehicle interface: the power
of vehicle movement gestures in human road user
coordination. In: University of Iowa (ed) Proceedings
of the 9th international driving symposium on human
factors in driver assessment, training and vehicle
design
Rothenbücher, D., Li, J., Sirkin, D., Mok, B., Ju, W. (2016).
Ghost driver: A field study investigating the interaction
between pedestrians and driverless vehicles. In Robot
and Human Interactive Communication (RO-MAN),
2016 25th IEEE International Symposium on. IEEE,
795–802.
Schieben, A., Wilbrink, M., Madigan, C. K. R., Louw, T.,
Merat, N. (2018). Designing the interaction of
automated vehicles with other traffic participants: A
design framework based on human needs and
expectations. Cognition, Technology & Work, 0.
Shmueli, Y., Degani, A. In preparation. Usability
evaluation of a bioluminescence-inspired eHMI
communication.
Shmueli, Y. & Degani, A. (2019). Vehicle-Pedestrian
Communication I. Analysis and Needs Study Internal
Report
Shmueli, Y., Degani, A., Zelman, I., Asherov, R., Zande,
D. Weiss, J. & Bernard, A. (2013). Toward a Formal
Approach to Information Integration: Evaluation of an
Automotive Instrument Display. Proceedings of the
57th Human Factors and Ergonomics Society Meeting.
San Diego, CA.
eHMI Design: Theoretical Foundations and Methodological Process
211

Society of Automotive Engineers (SAE) (2019) Automated
Driving System (ADS) Marker Lamp (standard no.
J3134). SAE International, Michigan.
SAE Standard J845 - Society of Automotive Engineers
(SAE) (2021). Optical Warning Devices for Authorized
Emergency, Maintenance, and Service Vehicles.
August 2021
SAE Standard J595 - Society of Automotive Engineers
(SAE) (2021). Directional Flashing Optical Warning
Devices for Authorized Emergency, Maintenance, and
Service Vehicles. August 2021
SAE Standard J2498 - Society of Automotive Engineers
(SAE) (2016). Minimum Performance of the Warning
Light System Used on Emergency Vehicles. November
2016
Šucha, M. (2014), Fit to drive: 8th International Traffic
Expert Congress. 8-9 May, 2014, Warsaw
Šucha, M., Dostal, D., & Risser, R. (2017). Pedestrian-driver
communication and decision strategies at marked
crossings. Accident Analysis and Prevention, 102, 41–50.
Tiesler-Wittig, H. (2019). Functional Application,
Regulatory Requirements and Their Future
Opportunities for Lighting of Automated Driving
Systems (tech. rep.).
UNECE regulations ece r-65,’ UNECE, accessed
December 7, 2011.
Volvo 360C Concept Vehicle https://youtu.be/
H5KNPQT72FA Retrieved on Oct 3, 2019
Warrick, M. J. (1947). Direction of movement in the use of
control knobs to position visual indicators. In PM Fitts
(ed.) Psychological Research on Equipment Design:
Army Air Force Aviation Psychology Program
Research Department. pp.137–46.
Werner, A. (2018). New Colours for Autonomous Driving:
An Evaluation of Chromaticities for the External
Lighting Equipment of Autonomous Vehicles.
ColourTurn, 10(3), 183–193.
Wickens, C. D. (1984), Processing resources in attention, in
R. Parasuraman and R. Davies (Eds), Varieties of
Attention, NEW YORK: ACADEMIC PRESS, 63-101.
Wickens, C. D., Vidulich, M., Sandry-Garza, D. (1984),
Principles of S-C-R compatibility with spatial and
verbal tasks: the role of display-control location and
voice interactive display-control interfacing, Human
Factors, 26, 533-543.
Wickens, C.D., Liu, Y. (1988). Codes and modalities in
multiple resources: a success and a qualification,
Human Factors, 30, 599-616.
Wickens, C., Carswell, C. (1995). The proximity
compatibility principle: Its psychological foundations
and its relevance to display. Human Factors, 37(3), pp.
473-494.
Wickens, C.D. (2002). Multiple resources and performance
prediction. Theoretical Issues in Ergonomics Science,
Vol. 3, NO. 2, 159-177
Xiang, H., Zhou, M., Sinclair, S.A., Stallones, L., Wilkins,
J., Smith, G.A. (2006). Risk of vehicle-pedestrian and
vehicle-bicyclist collisions among children with
disabilities. Accident Analysis and Prevention;
38(6):1064-1070.
HUCAPP 2023 - 7th International Conference on Human Computer Interaction Theory and Applications
212