Project
This website describes the project: Intelligent agents for Crisis Management.
The project has been performed as part of the ICIS
project. The project started 1 January 2007 and ended 1 February 2009.
The people involved are:
- Jurriaan van Diggelen (Principle
investigator)
- Peter Werkhoven
- Robbert-Jan Beun
- Rogier van Eijk
Feel free to contact any of them with questions or remarks.
Summary
Crisis management involves identifying an incident or a disaster, such as
fire or a traffic accident, and subsequently confronting and resolving it
in order to minimize the damage. Each of these tasks demand the utmost in
human performance as limited time is available, the situation may change unexpectedly
and human failure is potentially hazardous. In this research we investigate
how modern information technology can support collaboration among human crisis
workers and assist them in performing their tasks.
Our approach uses intelligent agents that act as electronic partners of humans,
forming a hybrid human-agent team. Prior investigations have identified three
primary roles played by agents interacting with humans: agents supporting
individual team members; agents supporting the team as a whole; agents assuming
the role of a team member. In the following, we will describe three possible
applications corresponding to each of these roles, together with some technological
groundwork.
Individual Support: A Gateway to Ubiquitous Information
An application where agents support individual team members, could be a system
that provides crisis workers via their PDA’s with relevant information.
In such a system, the PDA is modeled as an agent which acts in a ubiquitous
computing environment. This means that hundreds of networked sensors and devices
are embedded in the environment. The PDA automatically discovers these sensors
and extracts relevant information from them to present it to the user, i.e.
the PDA serves as a gateway to ubiquitous information. As an example, think
of a firefighter obtaining information about a fire by consulting a PDA that
extracts and interprets information from thermometers and smoke sensors in
the environment.
Designing the information infrastructure for these kinds of systems is far
from straightforward. Because different devices may be developed by different
manufacturers, they are likely to represent their information heterogeneously,
which comlicates the sharing of information. Furthermore, the system is open,
i.e. it is not known beforehand which sensors, PDA's and other devices will
constitute the system. This makes it impossible to fix the communication infrastructure
in advance.
To investigate different techniques to represent, share, and discover information,
we developed a prototyping tool, called Ubismart. This allows us to easily
add sensors, PDA’s and other computing devices after which we can run
simulations of the system to study the amount and quality of information exchange.
Team Support: Automatic Notification in Human Agent Teams
An application of agents supporting a team as a whole could be a system which
facilitates coordination in a team by automatically notifying the relevant
people of certain events. For example, after someone has finished performing
an action which has been requested by another agent, that person must notify
the requester after it has finished performing the action.
We have implemented this functionality using policies, i.e. authorizations
or obligations which govern the behavior of agents. For example, we could
state that an agent is obliged to send a notification to the requester after
it has finished performing a requested action. We use the KAoS policy language
to define such rules. Because the rule is written in a machine readable language,
the computer understands what needs to be done, and could take over this task
which would otherwise have to be performed by humans.
We tested the policies using a Mars mission scenario where a hypothermic astronaut
needs to be brought to the habitat for recovery.
Agents as Team Members: Treating Agents as Equals
By allowing agents to assume the role of an equal team member, we could fully
exploit the capabilities of agents. For example, in the Mars mission scenario,
we could allow a robot to fulfill part of the teamwork by bringing the hypothermic
astronaut to the habitat. To allow agents in a team to be treated as equal
to humans, agents should behave a little like humans. We have studied this
type of human-like behavior in agents by identifying patterns of human teamwork
and implementing these using KAoS policies. For example, we have defined a
policy which states that a leader is responsible for its team, and a policy
which states that another team member should take responsibility if the leader
is absent or temporarily inaccessible. These policies directly affect the
behavior of the agents in the team and might make them more predictable and
understandable to humans. Of course this does not fully achieve the ambitious
goal of human and agent teamwork, but we believe we have captured some important
aspects in a rather concrete way.
Software
Please contact any of the project participants to obtain a copy of the source code of these applications.
This application shows a policy based approach to teamwork, which simulates
a use case from the MECA
project. The project builds on KAoS.
Ubismart, is a simulation environment for ubiquitous computing. It builds
on the Protege ontology
editor, developed at stanford.
Presentations
- J. van Diggelen, Human-Agent Teamwork,
Workshop
on Rich Cognitive Models for Policy Design and Simulation , 16 January
2009.
- J. van Diggelen, Fulfilling Collective Obligations
in Human-Agent Teams using KAoS Policies, IHMC
Afternoon Lecture Series, 10 September 2008.
- J. van Diggelen, Information Supply Mechanisms in
Ubiquitous Computing, Crisis Management and Workflow Modelling, 7th
International Workshop on Task Modelling and Diagrams (TAMODIA), 25 September
2008.
- J. van Diggelen, Agent Communication in Ubiquitous
Computing: the Ubismart Approach, Seventh
International Conference on Autonomous Agents and Multi-agent Systems (AAMAS),
14 May 2008.
- J. van Diggelen, Modeling Decentralized Information
Flow in Ambient Environments, Ambient
Intelligence Developments, 17 September 2007.
Publications
J. van Diggelen, R.J. Beun, R.M. van Eijk, P.J. Werkhoven, Efficient
Semantic Information Exchange for Ambient Intelligence, the Computer Journal,
Special Issue on Artificial Societies for Ambient Intelligence, 2009
J. van Diggelen, R.J. Beun, R.M. van Eijk, P.J. Werkhoven, Information
Supply Mechanisms in Ubiquitous Computing, Crisis Management and Workflow
Modelling, Proceedings of the 7th International Workshop on Task Models
and Diagrams (TAMODIA), LNCS 5247, pp. 72-83, Springer Verlag, 2008
Arlette van Wissen, Jurriaan van Diggelen, Virginia Dignum, The Effects of
Cooperative Agent Behavior on Human Cooperativeness, proceedings of BNAIC
2008
J. van Diggelen, R.J. Beun, R.M. van Eijk, P.J. Werkhoven, Agent
Communication in Ubiquitous Computing: the Ubismart Approach, Proceedings
of the Seventh International Conference on Autonomous Agents and Multi-agent
Systems (AAMAS08), ACM Press, pp. 813-820, 2008
J. van Diggelen, R.J. Beun, R.M. van Eijk, P.J. Werkhoven. Modeling
Decentralized Information Flow in Ambient Environments, in Developing Ambient
Intelligence, Proceedings of the second international conference on Ambient
Intelligence Developments (AmI.d'07), pp. 23-33. Springer Verlag, 2008
To appear
Jurriaan van Diggelen, Robbert-Jan Beun, Peter J. Werkhoven, Intelligent Agents in Crisis Management: from PDA to TDA, submitted to ICIS Scientific Book 2009. Please contact first author for more details.
Jurriaan van Diggelen, Tim Grant, Mark Neerincx, Policy-Based Design of Human-Machine Collaboration in Manned Space Missions, submitted to SMC-IT 2009. Please contact first author for more details.
Jurriaan van Diggelen, Jeffrey M. Bradshaw, Matthew Johnson, Andrzej Uszok,
Paul J. Feltovich, Implementing Collective Obligations in Human-Agent Teams
using
KAoS Policies, Submitted to COIN workshop in January 2009. Please contact
first author for more details.
Arlette van Wissen, Jurriaan van Diggelen, Virginia Dignum, The Effects of Cooperative Agent Behavior on Human Cooperativeness, Extended Abstract, AAMAS 2009, to appear.