International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
Interi
and interact with other individuals and/or their
environment.
The main characteristic of agent-based technology
is that the structure of the software is represented by
a group of agents who collaborate in achieving the
goal of the task in hand. Among different types of
agents, the personal assistant agents are particularly
interesting to this research. This type of agents
operates at the user interface level and actively
assists users by offering information and advice to
the users (Wasson e/ al., 2001). These agents
usually apply a kind of intelligent learning
algorithm so that they can intercept the user input,
examine it and take actions that are more specific to
that particular users’ needs at that moment. These
agents are also called learning or adaptive agents.
However, the assistant agents are different from the
conventional autonomous agents in that they may
not act truly autonomously and their goals must be
adjusted constantly based on their interception of
the users’ goals. According to Wasson ef al. (2001),
the key questions for the design of any assistant
agent are how to model the users’ intent, how to
determine the appropriate actions given that intent
and how to develop an appropriate interface
between the users and the assistant agents.
The functions required for the tourist route advice
system proposed in this research are far too
complicated for a single agent to accomplish. It
must rely on the co-operation/collaboration among
multi-agents, each of which is specialised in solving
a particular aspect of the complex problem. An
agent platform, such as the one specified by FIPA-
OS(FIPA, 2002), provides an infrastructure in
which agents can be deployed, communicate and
collaborate to achieve the goal of the tasks. Figure 1
shows the agent-based architecture of the proposed
tourist route advice system.
The user interaction agent takes the responsibility
for dialoguing with users, i.e., presenting questions
and information to users and record the users’
responses to the system for further analysis. The
presentation agent determines what information and
map to be presented to the user based on his current
position and other information.
The monitoring agent silently observes the users’
spatio-temporal behaviour during a trip. It uses the
dynamic positioning information of the users’
movement to detect the changes of the users’ spatio-
temporal behaviour. When the change is detected, it
informs the inference agent for further analysis and
actions.
The user model agent is responsible for establishing,
maintaining and updating the user profiles for each
individual user. It updates a user profile when it is
informed by the inference agent that changes in
users' interests have occurred. Since the user
profiles are maintained by the user model agent, the
only way for the other agents to access the
information recorded in user profiles is through the
intermediary of this agent.
The Inference agent is the kernel of the whole
system. Based on the information received from the
user interaction and monitoring agents, the inference
agent determines the appropriate actions to be taken,
such as generating a new route, updating the user |
model, presenting updated information to users
and/or starting a new dialog sessions for further |
information when necessary.
The responsibility of the route agent is to generate
routes based on the request from the inference
agent. To do this, it requires the cooperation of the
spatial agent to perform necessary spatial operations
and of the user model agent to retrieve the
information on user profiles.
The database agent performs all the database
management tasks, such as retrieving and updating
the data stored in a GIS and/or Database
Management System (DBMS).
The user interaction and monitoring agents operate
in an active way: they keep acting from the moment
a tourist starts using the system and reside on the
user's light-weight mobile devices. The other agents
act in a more passive way because they are activated
only by those two agents when necessary and they
reside on the server side for heavy tasks.
The personalized tourist route advice system
developed in this research takes advantages of
agent-based technology. The personal assistant
agents resides on the light-weight mobile devices
and they can proactively communicate with the
agents on the server that perform heavy tasks such
as route generation.
Figure 1 Agent-
based system
architecture
-
&
Monitoring User Presentation
agent [Interaction agent
agent
A
Y
Inference
agent
Route User Model
agent agent
A A
|
Y
Spatial «——», Database )
agent agent )
GIS
database
