On-the-Move and in Your Car - Human-Computer Interaction and Innovation in Handheld, Mobile and Wearable Technologies - page 72

CASE STUDY: VEHICLE
NAVIGATION SYSTEMS

ing the visual demand of an in-car user-interface
(Pettitt et al., 2006; Stevens et al., 2004).

Keystroke Level Model (KLM) To ground many of the issues raised above, a

specific system type has been chosen for further
discussion (vehicle navigation systems). Many
of the individual points made for this system can
be generalised and are applicable to other in-car
computing technologies.
Vehicle navigation systems aim to support the
strategic (e.g. route planning) and tactical (e.g.
route following) components of the overall driv-
ing task. They have the greatest potential to assist
drivers who undertake many unfamiliar journeys,
for instance as part of work, or during leisure trips
(e.g. when on holiday) and those who experience
extreme difficulties with existing methods of
navigation (particularly paper maps). When linked
with reliable, real-time traffic information (thus
providing dynamic guidance), the perceived utility
of navigation systems to the everyday motorist is
significantly enhanced (Bishop, 2005).
The market potential for vehicle navigation
systems has already been demonstrated in Japan,
where the technology has been available since the
early 1990s. Approximately 40% of all vehicles
on Japan's roads now have a navigation system
installed (http://www.jetro.go.jp/en/market/trend/
topic/2004_12_carnavi.html, September 2006). In
many other countries, the popularity of navigation
systems is currently reduced in relation to Japan,
but is predicted to rise rapidly over the next few
years (Bishop, 2005).
The majority of human factors issues relevant
to this form of technology relate to overload, al-
though as shall be seen, underload is increasingly
being researched. With respect to overload, clearly,
a key concern is the potential for driver distrac-
tion and there has been considerable research on
this topic since the mid 1980s (see Young, Regan
and Hammer (2003) and Srinivisan (1999) for
reviews). In using a vehicle navigation system,
drivers must interact with controls (e.g. to enter

The KLM method from the GOMs family of
techniques is well known to HCI researchers and
(to a lesser extent) practitioners (Shneiderman,
1998; Preece, Rogers and Sharp, 2002). It is a
form of task analysis in which system tasks with
a given user-interface are broken down into their
underlying physical and mental operators, e.g.
pressing buttons, moving hand between controls,
scanning for information. This is a method that
is extremely cheap to implement, as there is no
need for participants, and the method can be used
with very basic prototypes early in the design
process. Time values are associated with each
operator and summed to give a prediction of task
times. Researchers have developed new operator
values relevant to the in-car situation (e.g. time
to search a visual display, locate a control, move
hand back to steering wheel) and have reported
strong correlations between predicted task times
and times based on user trials (Green, 2003; Pettitt,
Burnett and Karbossioun, 2005). As noted above,
task times can be related to certain measures of
visual demand for in-car user-interfaces.
In an extension of the KLM method, Pettitt,
Burnett and Stevens (2007) recently developed
new rules that enable designers to develop pre-
dictions for a broader range of visual demand
measures. In particular, the extended KLM
considers a time-line view of an interaction in
which a cycle of vision/non-vision occurs with a
user-interface (similar to the occlusion protocol).
The authors have found that their version of KLM
can differentiate between tasks as effectively as
does the occlusion technique, but recommend
that further development is carried out to ensure
that practitioners can utilise the method reliably.