Information Technology Reference
In-Depth Information
to choose the modality that has the greatest weight and to instantiate it by using the
values of attributes having obtained the strongest ratings. However this situation
corresponds, in reality, to the simplest case: that of a unique semantic unit, a unique
user and a unique presentation device. In the more general case, where several users
are in proximity to the device or, conversely, several devices are in proximity to the
user (or even in the case where several users are in the proximity of the same
devices), more complex algorithms have been implemented so as to have several
devices collaborate with one another [JAC 06a]. This aims to ensure a global
coherence of the distribution of presentations while guaranteeing a minimum level
of satisfaction to any of the users.
10.7. Implementation
We have implemented these models and algorithms in the form of a platform
called PRIAM (
PResentation of Information in AMbient intelligence
). As mentioned
above, this platform is built according to agent architecture. The agents are
implemented in Java, which enables them to run in various hardware and software
environments. Furthermore, they can easily exchange messages through the network
via the
remote method invocation
(RMI) mechanism, which is integrated in Java.
As the implementation of full-scale experimentations is relatively hard and takes
a lot of time, we started by making a simulator that allows the final components of
an application to be tested, without having to deploy them in real conditions as early
as the first trials (see Figure 10.8). This simulator enabled us to verify the correct
behavior of the algorithms. For example, Figures 10.8b and 10.8c put two users in
play (someone who is blind and someone who is not), as well as a screen and a
sound device. When the user who can see presents himself, the screen and the sound
system can give him information (b). However, when the blind person comes close,
the screen does not display anything as its output modality is not compatible with
the profile of the blind person (c).
In this example, we note that the multimodal behavior of the PRIAM platform
conforms to what we have specified before. Indeed, when choosing the presentation
method of a semantic unit, the system chooses a modality adapted to the user,
depending on the presentation devices present in proximity.
All the aspects of the algorithms can be tested in this manner, with the necessary
number of presentation devices, users and information sources. For example, Figure
10.9 presents an example of a simulation of the instantiation process of modalities.
Depending on the distance between a given user and the screen, the attributes of
instantiation are different. Thus, the text is displayed in larger characters when the
user is further away from the screen.
