Improved Situation Awareness
for Public Safety Workers
while Avoiding Information Overload
Marc de Lignie
aBeiBei Hu
bNiek Wijngaards
ca
vts Politie Nederland, PO Box 238, 3970 AE Driebergen
b
D-CIS Lab / Delft University of Technology, PO Box 90, 2600 AB, Delft
cD-CIS Lab / Thales Research & Technology, PO Box 90, 2600 AB, Delft
Extended Abstract
This research stems from the MOSAIC project, a part of the valorization and knowledge transfer effort of the Interactive Collaborative Information Systems (ICIS) research programme (http://www.icis.decis.nl/), supported by the Dutch Ministry of Economic Affairs, grant no.: BSIK03024. ICIS is hosted by the D-CIS Lab (http://www.decis.nl/), the open research partnership of Thales Nederland, the Delft University of Technology, the University of Amsterdam and the Netherlands Organization for Applied Scientific Research (TNO).
1 Information Requirements in Crisis Management
Public safety services for emergency aid and disaster management face the challenge of dealing with the enormous growth of information: any information may add to situation awareness and be relevant for handling a specific incident. Evaluations of past incidents and disasters [1] already show too often that the information present in some information system was not brought to the attention of public safety workers in the field or in a (crisis) control room, making them state: “If only I had known… They should have told me!” This challenge provides an interesting case for testing techniques for distributed context-aware information retrieval and dissemination. An emergency incident or disaster provides a structured context in terms of location, time and nature of the incident and roles and tasks of end users.
2 Approach
A new combination of techniques is applied in conjunction, namely (a) Information retrieval techniques such as accessing heterogeneous data, data-fusion and spatiotemporal and semantic link analysis, (b) Filtering techniques for context specific dissemination of information towards mobile workers and (c) Information processing in distributed systems. The multi-agent architecture is shown in Figure 1, which fits in a wider effort to investigate actor-agent communities for decision support in complex and chaotic environments. The following functional layers are distinguished:
Data access: an agent acts as a specific adaptor for a data source. The agent can deal with access and authentication protocols, usage policies, data models, indexing, relevancy ranking, confidentiality, etc.
Query distribution and data analysis: expert agents adapt, split and translate complex queries into more specific queries towards the data access agents and aggregate the returned results. Aggregation may encompass various functions such as ranking, filtering, de-duplicating, correlating with rule bases (threat analysis), etc. During an incident multiple query distribution agents can be invoked, each with its own range of data sources covered and specific expertise to analyze the results.
Incident assistance: Per incident an agent manages the incident information space. This includes proactive gathering and maintaining all information relevant to the incident itself (situation awareness), as well as the influence on other events and vice versa (super situation awareness). Team assistance: an agent manages the composition of its incident response team as well as filtering
and dissemination information to/from individual team members.
Personal assistance: agents are responsible for the human-computer interaction. This includes negotiations with the team assistance agent about the end user attention as well as distributing information and notifications over the available interaction channels (text, audio, vibrations). This layering realizes a natural separation of concerns and the multi-agent platform supports ad hoc adaptation and integration of techniques and information sources.
other control room applications other partners & crisis coordination teams data sources
public safety workers data access query distribution & analysis team assistance personal assistance incident assistance other control room applications other control room applications other partners & crisis coordination teams other partners & crisis coordination teams data sources
public safety workers data access query distribution & analysis team assistance personal assistance incident assistance
Figure 1. Layers of agents (circles) in the multi-agent system.
The current and ongoing implementation of this multi-agent system is verified by interviews with representatives of Dutch public safety organizations on the basis of progress demonstrations on our research themes:
Personalized information filtering: what factors from the context and user profiles are important to deliver task-based information to the user while preventing information overload.
Super situation awareness: this depends on the ability to integrate and process task-relevant information in ways that support decision-making as well as keep track of work context of users. Knowledge management: apply domain knowledge to various stages of the information retrieval
and dissemination.
3 Conclusions
Dutch public safety organizations acknowledge the need for better use of information available in various sources and reported during the course of an incident. A multi-agent system is suitable for integrating techniques for context-aware information retrieval and dissemination. The ongoing, actual implementation of these techniques for information management in crisis situations requires further research on personalization, knowledge management and the creation of super situation awareness.
References
[1] Advice committee ICT coordination for disaster management (ACIR) 2005, De vrijblijvendheid
voorbij. Towards effective information provisioning for large scale coordinated actions within our
decentralized state, Dutch Ministry of the Interior and Kingdom Relations.