Supporting adaptive processes in tackling emergency situations, such as disasters, is a key issue for any emergency management system. In such situations, various teams are deployed in many sites, at the front-end or back-end of the situations, to conduct tasks handling the emergency. Context information is of paramount importance in assisting the adaptation of these tasks. However, there is a lack of middleware supporting the management of context information in emergency situations. This paper presents a novel framework that manages and provides various types of context information required for adapting processes in emergency management systems.

1. ESCAPE: An Adaptive Framework for
Managing and Providing Context
Information in Emergency Situations
Hong-Linh Truong, Lukasz Juszczyk, Atif Manzoor,
Schahram Dustdar
Distributed Systems Group, Vienna University of Technology
truong@infosys.tuwien.ac.at
W
http://www.vitalab.tuwien.ac.at/autocompwiki
P
A
K O
D
R
1

2. Outline
Scenario, requirements and motivation
Approach
Levels of context information
ESCAPE framework overview
Implementation status
Experiments and current applications
Conclusion and future work
23rd Oct 2007 2
EuroSSC 07, Kendal, UK

3. Scenario
Emergency situations
E.g., Natural disasters (earthquakes, hurricanes, floods)
Multiple support teams
Established on demand
Deployed at front-end (disaster fields) and back-end
sites
Conduct various processes to respond to the
emergency situation
Collecting information, performing relief tasks, etc.
How to achieve effective response processes?
23rd Oct 2007 3
EuroSSC 07, Kendal, UK

4. Requirements and Motivation
Response processes to emergency situations, such as
natural disasters:
Established on demand and changed rapidly, depending context
of the situations
context information associated with entities inherent in
the situation is critical to effective response processes
Context information in emergency situation is complex
Reflects situation, responses, sites, teams and individuals
Related to people, resources and services, their status and
activities.
23rd Oct 2007 4
EuroSSC 07, Kendal, UK

5. Requirements and motivation(cont.)
Many context management frameworks exist
Mostly for normal environments, domotics, HCI for
emergencies,lack of integration with crisis
management systems
Some scientific and engineering issues: how to
Support multiple teams collaborating in pervasive
Grid environments
Make context information available to many crisis
management supporting tools and services
Ensure that context information in emergency
situations are extensible and interoperable
Runtime interpretation and engineering of context info
23rd Oct 2007 5
EuroSSC 07, Kendal, UK

6. Our Approach
Integration with other tools/services
SOAP-based and REST (Representational State Transfer)-
based models
Different techniques for front-end and back-end integration
Extensible context information representations
Support XML-based context information
Accept any representations as long as the information is in XML
Flexibility in pervasive environments
Customizable software components
Tracing capability
Context provenance
This paper focuses on middleware aspects
23rd Oct 2007 6
EuroSSC 07, Kendal, UK

7. Levels of Context Information in
Emergency Situation
Situation Earthquake
Response
Site
Team Recover railways
systems
Individual Individual
Individual Salzburg
Team OEBB
Response
23rd Oct 2007 7
EuroSSC 07, Kendal, UK

8. ESCAPE Framework Overview
•SOAP-based communication
•Mobile adhoc network
REST-based XML-based context
communication information
23rd Oct 2007 8
EuroSSC 07, Kendal, UK

9. ESCAPE Framework Overview
Plugin internal
software sensors
XPath/XQuery
requests
Team member
discovery
Context information is application-specific
The middleware accepts any XML-based context
information
23rd Oct 2007 9
EuroSSC 07, Kendal, UK

10. Query/Subscription and Storage
Query and Subscription APIs
Core APIs support requests based on XPath/XQuery
Application-specific query and subscription APIs built
atop core APIs
Lightweight Data Storage
High-end devices: eXist database with XQuery,
XUpdate, etc.
PDAs: round-robin model in which XML data is
stored in files
23rd Oct 2007 10
EuroSSC 07, Kendal, UK

11. Management of Context
Information
Level
Data aggregration flows
Individual Stored at the
front-end in
team devices
Aggregration without Team
knowing content of
context information
Site
Pull and push manner Stored at back-
XPath/XQuery-based end systems
Response
filters
Aggregation of
snapshots of context Situation
information in [ti, tj]
23rd Oct 2007 11
EuroSSC 07, Kendal, UK

12. Context Provenance Support
For supporting tracing
capability
Types of context provenance
information
Application-specific information
Middleware-specific information
ESCAPE supports
middleware-specific context
provenance
Storing provenance
information into the back-end
system
23rd Oct 2007 12
EuroSSC 07, Kendal, UK

13. Implementation Status
Web services
kSOAP for handling SOAP-based Web services,
between CIMSs
REST model between CIMS and SCIMS
XPath/XQuery/XUpdate for querying and
updating information
Hosting environments:
At the front-end: normal laptops and PDAs
At the back-end: SMP machines
23rd Oct 2007 13
EuroSSC 07, Kendal, UK

14. Experiments: WORKPAD context
information model
23rd Oct 2007 14
EuroSSC 07, Kendal, UK

15. Experiments – Performance Tests
Devices
3 iPAQ 6915 PDAs (Intel PXA 270 416 Mhz, 64 RAM, Windows CE 5.0, 2GB external
miniSD, IBM J9
Dell XPS M1210 (Intel Centrino Duo Core 1.83 Ghz, 2GB RAM, Windows XP)
Dell D620 (Intel Core 2 Duo 2Ghz, 2GB RAM, Debian Linux
Dell Blade (2 Xeon 3.2 Ghz CPUs with hyperthreading, 2GB RAM, Ubuntu Linux)
Network
Mobile adhoc: setting 220Kbits/s (150Kbits/s, observed)
100 MB/s to back-end
Two modes
Single team and multiple team leaders
23rd Oct 2007 15
EuroSSC 07, Kendal, UK

16. Experiments: Performance Tests
Single data transfer between a Concurrent data transfers
CIMS member (PDA) and between a CIMS (PDA) and
CIMS team leader (Laptop) its client (Laptop)
PDA is not suitable for team leader‘s device due to
performance reasons
23rd Oct 2007 16
EuroSSC 07, Kendal, UK

17. Experiments: Performance Tests
Performance of transferring data to the back-end system
~17KB, every 5 seconds
High variation between different runs in PDAs
190 ms – 5170 ms
Should not update an XML document in the back-end with small
documents
Store small XML documents and use back-end high capability to process
information
23rd Oct 2007 17
EuroSSC 07, Kendal, UK

18. Current applications
Enrich GIS management tools with context information
Combine static and heavyweight GIS data (back-end) with
volatile and in-situ context information
Emergency management tools show GIS information annotated
with context information describing current status of entities
within the emergency situation
Provide context information for adaptive process
management
Context information is used in adapting tasks performed in
disaster responses
improving coordination and decision making of response
activities based on real-time context information
23rd Oct 2007 18
EuroSSC 07, Kendal, UK

19. Example: search for relevant
context information
Context information and GIS: Team A wants to reach to a
place „P“. Let‘s check unusable roads leading to „P“
for $infrastructure in
collection('/db/contextinformation')//includeInfrastructure
where $infrastructure/category="ROAD" and
$infrastructure/condition="UNUSABLE"
return $infrastructure
Context information and adaptive processes: Let‘s
send one worker to place „P“ to take a photo
for $worker in collection('db/contextinformation')//SupportWorker
where $worker//hasCamera and $worker/belongsTo/description="Team 1"
and $worker//Activity/status="LOW"
return $worker
23rd Oct 2007 19
EuroSSC 07, Kendal, UK

20. Conclusion and Future Work
ESCAPE (Emergency Situations, Context Awareness,
Pervasive Environments)
Supports extensible context information model and provenance
Makes context information available to front-end and back-end
teams
Provides a basis framework for further developments of smart
sensing and context in crisis situations
Future work
Applications integration: GIS-based and in situ context
information support and adaptive processes in disaster
responses
Quality aspects of context information
Data aggregation based on rules and XML transformation
techniques
Automatically process context information using Event-
Condition-Action and Complex Event Processing techniques
23rd Oct 2007 20
EuroSSC 07, Kendal, UK

21. Thanks for your attention!
Hong-Linh Truong
Distributed Systems Group
Vienna University of Technology
http://www.vitalab.tuwien.ac.at/autocompwiki
23rd Oct 2007 21
EuroSSC 07, Kendal, UK