SmartOffice: An Intelligent and Interactive
Environment
Christophe Le Gal, Jerome Martin, and Guillaume Durand
Projet PRIMA - Lab. GRAVIR-IMAG
INRI A Rhone-Alpes
655, avenue de l'Europe
38330 - Montbonnot Saint Martin, France
e-mail: Christophe.Le-GalOimag.fr
Abstract. This paper presents our Intelligent Environment called
Smart Office. In the Smart Office the user can work as in a normal of-
fice. The office's intelligence observes the user in order to anticipate his
intentions and augments his environment to communicate useful infor-
mation. Computers are involved in user activities in order to help in
everyday tasks. The system interacts with users using voice, gesture or
movement.
The SmartOffice provides a test-bed for collaboration and combinaition
of independant modules integrated into a single coherent application. In-
tegration requires a flexible working environment in which module devel-
oppers should not worry about low-level communication between mod-
ules. This paper presents a flexible resource-oriented integration proto-
col, which we argue is necessary to build such an environment. All mod-
ules need not be aware which resources can be provided by each module.
They communicate with the supervisor which acts as a resource-server.
The supervisor is programmed using a rule-based language, in which the
addition or the suppression of a module requires only the suppression of
the corresponding rule.
Two major modules of the Smart Office are considered to illustrate de
role of the supervisor: The first module is the MagicBoard, an ordinary
white board augmented by a camera and a video-projector.
To guess the user-intentions, the system must constantly be aware of
the location of the users in the Smart Office. Therefore we need a user
localization module. This is the second module that we describe. An
example of the tracking architecture is presented.
1 Motivation and Background
Surprisingly, some people claim that computers have introduced a reduction
in the effectiveness of work. For those people, the arrival of computers simply
means that a huge new box takes up space on their desk, and makes tasks that
were simple more complicated. Adams' axiom humorously claims that computers
increase the working speed by 100 % but increase the work load by 300 %
[Add97]. Computers today are intrusive. Even if they claim to be, they are not
P. Nixon et al. (eds.), Managing Interactions in Smart Environments
© Springer-Verlag London Limited 2000
105
truly user-friendly, since users must engage in an explicit oriented dialogue with
the computer, instead of using more natural ways of interaction.
We believe that a computer should not involve a new way of working but
simply augment the current working modes. Therefore, as pointed out by Weiser
[Wei94], computers should be invisible, not demanding any adaptation from the
user, while at the same time bringing him the benefits of data processing power.
The idea of Coen [Coe98] is to make computer-interfaces for people rather than
people-interfaces for computers.
Intelligent Environments (IE) aim to create environments that facilitate the
use of computers, by being aware of the occupants and responding to their voice
and gesture commands. The Smart Office is an intelligent environment being
developed at the GRAVIR-IMAG Lab. In the SmartOffice the user can work
as in a normal (even computer-free) office. The office's intelligence observes the
user in order to anticipate his intentions and augment the user's environment to
communicate useful information.
As mentioned by Etzioni [Etz93], and applied in Intelligent Environments
by Co en [Coe99], software environments (such as IE) provide a test-bed well-
suited for techniques developed in the context of robotics. It would be a mistake
to view an IE as a simulated robot, even though it is partially artificial. A more
accurate view is to consider it a non simulated robot because it manages real
sensors and effectors and also software applications. Consequently, as all robots,
it needs control modules and a supervisor.
Coen [Coe97] proposed a distributed room control architecture based on com-
munication between agents. Each agent is responsible for a specific room func-
tion. Agents can ask others to warn them when a particular event occurs. For
example an agent can ask the speech recognition agent to be warned each time
a specific utterance is spoken. The architecture proposed in this paper follows
the same spirit. The main difference is that a supervisor is responsible for the
communication between agents, thus acting as a resource-manager. All agents
should not be aware of which resources can be provided by each agent. They
should not have to ask Tracker: give me what you know but be able to pose
questions such as "Resource manager: give me the (x,y) position of the user".
As opposed to Coen's agent-oriented communication, we propose a resource-
oriented communication.
Section 2 presents the material architecture of the SmartOffice. The Smart-
Office is composed of fifty sensors (cameras and microphones) and three actua-
tors, a video-projector and two speakers. A description of the supervisor is given
in section 3. The supervisor controls numerous independent modules and makes
them cooperate via exchange of push or pull messages. Sections 4 and 5 describes
two major modules of the Smart Office, the MagicBoard and the user-tracking
procedure. The MagicBoard is the main "actuator" the SmartOffice can use
to augment the user environment. The user-tracking module is a requisite for
predicting user intentions. These applications demonstrates the communication
between modules and the supervisor. Finally, in section 6 a discussion and a
conclusion are given.
