Adding Generic Contextual Capabilities to Wearable Computers
Jason Pascoe
Computer Laboratory, University of Kent at Canterbury,
Canterbury, Kent CT2 7NF, United Kingdom
+44 (0)1227 764000 ext.7754
J.Pascoe@ukc.ac.uk
Abstract
Context-awareness has an increasingly important role to
play in the development of wearable computing systems.
In order to better define this role we have identified four
generic contextual capabilities: sensing, adaptation,
resource discovery, and augmentation. A prototype
application has been constructed to explore how some of
these capabilities could be deployed in a wearable system
designed to aid an ecologist’s observations of giraffe in a
Kenyan game reserve. However, despite the benefits of
context-awareness demonstrated in this prototype,
widespread innovation of these capabilities is currently
stifled by the difficulty in obtaining the contextual data.
To remedy this situation the Contextual Information
Service (CIS) is introduced. Installed on the user’s
wearable computer, the CIS provides a common point of
access for clients to obtain, manipulate and model
contextual information independently of the underlying
plethora of data formats and sensor interface
mechanisms.
1. Introduction
Context-awareness [1] is the ability of a program or
device to sense various states of its environment and itself.
In one form or another, context-awareness is already a
basic requirement of many systems and is likely to have
an increasingly important role to play in future wearable
computers and in the integration of wearable and
ubiquitous computing. The intimate association of user
and computer in a wearable system leads to the computing
resources being accessed in a diverse array of situations,
unlike a static desk-bound computer. It is this multitude of
dynamic contextual factors that allows context-awareness
to be exploited particularly well in wearable computers.
This paper addresses the deployment of context-
awareness in wearable computers through three sections.
The first identifies four application-independent
contextual capabilities that can be employed to enhance
the wearable system. In the second section a case study is
given that demonstrates how even a relatively simple
context-awareness can often greatly assist the user. The
final section introduces the Contextual Information
Service, which, based on our development experiences,
we believe is vital to facilitate the wider use of context-
aware technology.
2. Relationship to other work
The case study shares a general theme of exploiting a
knowledge of time and place that is in common with other
applications such as the touring machine [2]. However,
most applications currently use this knowledge to
influence how information is
viewed
(e.g. displaying some
tourist information) whereas our prototype application
uses context-awareness to influence how data is
recorded
.
The design of a Contextual Information Service (CIS)
is a largely overlooked area of research. Contextual
services tend to be of an application-specific level, such as
Cyberguide’s navigator and cartographer [3]. Other
projects that attempt a more application-independent
approach, such as the situated computing service [4], tend
to offer a service that simply abstracts the client
application from the underlying sensors. Our CIS is
designed to provide a more comprehensive service that
models contexts in terms of the objects they belong to and
the relationships between them. Rather than a conveyance
medium operated by discrete calls to a generic sensor
API, our CIS seeks to provide a continuous service via a
rich contextual model that happens to gain some of its
data automatically from sensors.
3. Generic context-aware capabilities
Schilit et al originally defined a set of general
context-aware application categories [1], and subsequent
research projects have defined application specific
categories. However, what we present here are a set of
core generic capabilities that we use as a vocabulary to
identify and describe context-awareness independently of
application, function, or interface.
3.1 Contextual sensing
Contextual sensing is the most basic level of context-
awareness. The wearable computer simply detects various
environmental states and presents them to the user in a
convenient form, effectively augmenting the user’s own
sensory system. For example, if connected to a location
sensor such as a Global Positioning System (GPS)
receiver [5], then the derived latitude and longitude
location can be conveyed to the user via a map display
annotated with a ‘you-are-here’ marker.
3.2 Contextual adaptation
It is not only the user that may be interested in the
contextual data from such a sensory system. Applications
can leverage this contextual knowledge by adapting their
behavior to integrate more seamlessly with the user’s
environment. Rather than providing a uniform service
regardless of the user’s circumstances, the context-aware
computer can tailor itself to the current situation. For
example, adapting behavior for a particular user (where
the user’s presence is the context), turning the screen
back-light on when it gets dark, switching to a
communications link that is less expensive during the
current time of day, etc.
3.3 Contextual resource discovery
Contextual adaptation applies a knowledge of the
wearable computer’s own context (which, for the most
part, also inherently equates to the user’s context). This is
taken a step further if the contexts of other entities are
made available. Using this information the wearable
computer can discover other resources within the same
context as itself and exploit these resources while they
remain in the same context.
The resources that are most readily useful to the
wearable computer are various forms of electronic or
computer equipment. For example, the user’s wearable
computer with limited display capabilities may discover
an unused display screen that is in close proximity and can
temporarily be used to display information on. Less
conspicuous information resources may also lie within the
user’s environment. For example, a service timetable may
be broadcast by a local infrared transmitter installed at a
bus-stop. If the transportation service itself is context-
aware [6], then real-time estimates of arrival times could
be calculated based upon its present position.
An important aspect of contextual resource discovery
is that it bridges the philosophical gap between wearable
and ubiquitous computing [7]. The wearable computer
provides a set of core services constantly at the user’s
disposal, whilst also contextually binding to devices in the
user’s environment (i.e. ubiquitous devices) that can
complement or augment the wearable system.
3.4 Contextual augmentation
Thus far, we have discussed how contextual
information can be used to sense, react, and interact with
the environment. Contextual augmentation extends these
capabilities further through augmenting the environment
with additional information. This is achieved by
associating digital data with a particular context that it is
related to. Depending on the user’s perspective this
coupling of the real and the virtual can be either viewed as
the digital data augmenting reality [8] or reality
augmenting the digital data.
Tour-guides are the most prevalent applications of
augmenting reality with digital data [2, 3]. The tourist
equipped with such an application is presented with
information about the attractions that they are surrounded
by or are approaching. Considering the alternative
perspective of digital data augmented with reality, an
address book program could augment an entry with the
addressee’s present context, such as their current location
and activity. Given this extra information, the user can
route a phone call to the phone nearest the addressee or
postpone the call if the addressee is currently in a meeting.
In addition to reality-information couplings, there are
also reality-process couplings. These facilitate the
execution of a program or particular process, when the
user happens upon a specific situation. In effect, this
extends the event-driven architecture common in the
majority of computer systems to encompass contextual
events of the surrounding environment.
4. A giraffe observation case study
A prototype fieldwork tool is examined in this
section, produced as part of the ‘Mobile Computing in
Fieldwork Environments’ project [9, 10] at the University
of Kent, funded by JTAP. This project has experimented
with the aforementioned context-aware concepts by
constructing prototype systems to assist users in common
fieldwork tasks that are typical across a broad spectrum of
subject domains (e.g. an archeological excavation, a bio-
diversity survey, etc.). The particular system examined
here has been used extensively by an ecologist during a
two-month observational study of giraffe in Kenya.
4.1 Why fieldwork?
Wearable and context-aware computing concepts are
particularly appropriate in the domain of the fieldworker.
The user may traverse large areas on foot, crawl through
thick undergrowth, or face other inhospitable conditions
that make carrying or operating a laptop computer
completely impractical. Therefore, a small, lightweight,
robust, device that can be used on the move is required,
i.e. the wearable computer. Additionally, the extremely
dynamic user-environment also equates to a diversity of
fast changing contexts that can be exploited by the
context-aware wearable computer. For example, one
aspect of the giraffe observations was to record how many
bites a giraffe ate, where it ate them, and at what time –
information that changed from second to second.
The need for a new approach to the provision of
computing resources is not solely driven by the unusual
environment of the user, but also by the nature of the work
itself. In most field observation work, the user needs to
maximize the time they spend observing whilst
minimizing the time they spend on auxiliary tasks such as
interacting with the recording medium. This applies even
more so when there are large volumes of data to be
recorded within short periods. A context-aware wearable
computer helps through its intimate association with the
user, attempting to seamlessly integrate the computing
resource with the user, their task, and their environment.
Finally, data collection and observational work
(which comprise the larger part of fieldwork) focus on
recording various states of the user’s environment, e.g. the
location of an observation point. A context-aware
recording device can thus greatly assist the user in making
observations because it too can ‘observe’ some states of
the user’s surroundings.
4.2 Field assistant hardware
Our directive for work on the first prototype was to
explore and develop context-aware software rather than
construct new hardware designed for the fieldworker’s
ergonomics. Our solution reflects this attitude, comprising
a selection of readily available hardware running the novel
fieldwork tools that we have developed.
Figure 1. The Ecologist records an observation
of a giraffe (see top right of photograph) using
the 3Com PalmPilot linked to GPS-enabled
trousers.
We restricted our selection of computing hardware to
low-cost devices that were widely available and that best
matched the earlier described attributes of being small,
lightweight, and robust. We chose the 3Com PalmPilot
PDA (Personal Digital Assistant) as its small notepad-like
form factor, its Graffiti handwriting recognition system,
and its design philosophy oriented towards simple, easy-
to-use functionality, singled it out as the most suitable
candidate for replacing the fieldworker’s paper notebook.
Already aware of its time context, we augmented this with
a knowledge of location by attaching a Garmin-45 GPS
receiver. This set-up provided the foundation on which to
build programs that utilized a simple understanding of
time and place.
The ecologist’s army trousers were equipped with the
PalmPilot and GPS receiver in the left and right trouser-
leg pockets respectively, where they remained until the
user removed the PalmPilot in order to note down an
observation, see figure 1. An unobtrusive umbilical cable
kept the two stowed devices constantly connected. To
improve the performance of the GPS receiver an external
aerial, which was discretely concealed in the fabric of a
baseball cap, could also be connected. However, in the
giraffe study, the ecologist preferred to use the GPS
receiver without this optional accessory.
4.3 Field assistant software
The prototype software currently concentrates on
providing a set of tools to assist in the fieldworker’s
observation and data collection activities, i.e. helping the
user record information about their environment. This is
an interesting divergence from other context-aware
studies, such as the oft-quoted location-aware tourist
guide, as here information is being authored in a
particular context rather than presented in a particular
context.
The stick-e note context-aware model [11] was used
as the basis for constructing the fieldwork tools. In this
model, electronic notes that are made by the user are
automatically attached to the current context (typically the
user’s location) thereby creating a stick-e note. For
example, a stick-e note could consist of a ‘Saw giraffe
here’ message and the related location context the
message is attached to. This model was extended and
generalized so that a stick-e note now consists of a set of
elements that describe both the content and context of the
note. This is exactly what an observation is: a set of values
that equally describe the content or context of the
observation depending on one’s perspective. For example,
an observation note could consist of a time, location, and
animal field, each of which describe both content and
context. We developed several programs for the PalmPilot
that allowed the user to create, edit, and view stick-e notes
in the field, the two main ones are detailed here:
The StickePad
. All the creating and editing of stick-e
notes is handled by this program. The user can create a
new stick-e note from a template (a set of name and type
pairs created by another program) and then edit the values
in a form-based user interface, see figure 2.
To assist in the observation process the StickePad
employs the system’s knowledge of time and place to
automatically default any location or time fields to the
current contextual values. The user can still edit these
fields manually or request that they be updated with new
contextual readings (useful if the user changes context
whilst editing the note). We also deployed a number of
HCI techniques in this stick-e note authoring tool to
optimize it for the fieldwork environment, utilizing
principles such as indirect modes of operation [12].
However, only the context-aware features will be
discussed within the scope of this paper.
Figure 2. Manually editing a location field of a
stick-e note in the StickePad. The 'Here' button
retrieves the current GPS value.
The StickeMap
. A knowledge of the location context of
both the user and the recorded observations is exploited in
the StickeMap to visualize the collection of stick-e notes
on a map display, see figure 3. There are no background
details such as tracks, hills, etc. as no mapping data was
available for the area at the time of constructing the
prototype. However, future versions will incorporate
mapping data that is currently being generated.
Figure 3. The StickeMap shows the user's
current position (the ‘+’ icon) relative to the
stick-e notes (the note icons).
Selecting one of the stick-e note icons displayed in
the StickeMap invokes the StickePad with the appropriate
stick-e note automatically loaded for editing. This offers
an alternative stick-e note selection mechanism to the
simple sequential note list available in the StickePad.
4.4 The field trial
The ecologist used the prototype for all of her data
collection requirements throughout the course of the two-
month giraffe study. The work carried out during this
period can be divided into three main observational
activities, each associated with a particular stick-e note
template:
Giraffe Observation
. Studying groups of giraffe and
individual giraffe behavior (for which two stick-e note
templates were constructed by the ecologist) was by far
the most common and intensive activity. It would
normally have required the work of two people, one to
make observations and the other to take notes. However,
the automatic entry of location and time contexts into the
stick-e note eliminated the need to operate a stop-watch or
to estimate position (very difficult to do in a homogenous
environment). Additionally, the computer and telescope
could be operated simultaneously, see figure 4. Thus, the
workload was reduced to such an extent that the single
ecologist alone could complete the task.
Vegetation Surveys
. The aim of these surveys was to
establish the impact of giraffe feeding on the Acacia tree
in different parts of the reserve. Recording the information
in a stick-e note that was automatically tagged to location
proved an effective solution.
Fecal Sampling
. Samples of giraffe feces were
routinely collected for analysis in order to investigate
giraffe diet. Stick-e notes were used to record the origin of
any samples by automatically providing the location
context to accompany the description of the state of find.
Figure 4. Recording giraffe behavior.
In addition to automating elements of the
conventional data collection process in each of these
activities, visualizing the location context in the
StickeMap provided novel functionality applicable across
all of these tasks. The map display served two purposes:
(i) to allow the ecologist to orient and locate herself
relative to the stick-e notes she had recorded, and (ii) to
observe any patterns in the distribution of such notes, e.g.
to see that the majority of feces had been collected in a
particular area of the reserve.
The hardware configuration also proved a good
design for the environment. The lightweight, small, and
robust attributes of the equipment allowed it to be used
effectively in the harsh conditions without physically
impeding the user in any way. When making an
observation the device could be operated with one hand,
and at other times kept in a pocket. This ability to
completely remove the device from the user’s awareness
was essential when walking through areas inhabited by
rhino and lions. In such conditions the user could forget
about the computer system and devote their whole
attention on avoiding an undesirable confrontation with a
rhino, or, failing that, running away from one as fast as
possible! Occluding any of the user’s senses or mobility,
even partially, is simply not acceptable in this
environment (though in less hazardous fieldwork settings
this is not necessarily the case). In summary, it was
desirable to have the computing resource as transparent as
possible, effectively removing it when not required, and
providing it in a convenient, easily accessible, manner
when it was.
At the end of the trial approximately 6000 individual
observations had been recorded using the prototype. The
ecologist stated that the system had allowed her to record
more information, quicker and easier than would
otherwise have been possible, proving itself a successful
solution to the fieldworker’s needs.
4.5 An evaluation of context-awareness
Although the prototype proved to be very useful, how
much has context-awareness contributed to this success?
In this first prototype, contextual-awareness is limited to
two types of context: time and location, and it only
exploits two of the four general context-aware capabilities
described earlier: contextual sensing and contextual
augmentation. However, these limited abilities were
effectively targeted on the areas of automating data
collection, transparently providing contextual information
when required, and visualizing collections of observations
whilst in the field. Deploying context-awareness in these
areas was a key factor in greatly improving the speed and
volumes of data collection, and in providing new
opportunities for the fieldworker, such as assessing the
