UCAmI 2013 Costa Rica · December 2-6
E-Flow: A communication system for user notification in dynamic evacuation scenarios
Augusto Morales Technical University of Madrid
Picture by David Gordillo
The Evacuation Problem
• The evacuation analysis is a long-studied problem.
• There are Route modeling systems, decision support systems, middlewares, WSN and so on.
• There are several gaps of current evacuation plans regarding the way users perceive the information and are aware of the situation
2
Building and evacuation plan including the user perspective
• Nowadays, there many solutions that indeed permit WSN environments to provide information in evacuation situations.
• Nevertheless, there are other challenges , more related to the human factor, such as: – reducing the reaction time of
personnel, – reducing the interpretation time, – providing easy interaction mechanisms
and – Providing dynamic signaling.
3
How to improve the chances of survival from the user
perpective?
To offer better information awareness!
Communication System for Dynamic evacuation scenarios
Users Safety staff
Emergency Response Teams
M2M Sensors/Actuators/
Signalling
H2M Communication Capabilities
Communication Infraestructure
Our solution is the Eflow system
Local Network
Infraestructure
Pub/Sub
Client
MQTT
Client
XMPP
Servers
Mobile
Safety Staff
Emergency
Response Teams
Faculty and
Students
subscribe()
publish()
802.15
Sensors
Proxy
MQTT-to-Web
Remote Networks
Mobile
Broker
Concentrators
Wi-fisubscribe()
publish()
WSN
Wi-fi
MQTT
Brokers
The BIG Picture of the E-Flow system
Messaging Capabilities
User Awareness
E-Flow Implementation
Concentrator
MQTT Client
Wireless Sensor Network (WSN)
SensorsSensors
TemperatureTemperature
HumidityHumidity
Evacuation
signalEvacuation
signal
Node 2Node 4 Node 3 Node 5Node 1
(Central Node)
Usb
ActuatorActuator
Integration of Sensor/Actuators
Design from scratch - low-power OMAPL127 (DSP+ARM) of TI - 802.15.4 - PSOC3 of Cypress Semiconductor - USB battery re-chargable - Connected with Linux (Montavista) - Sensors: presence, humidity, Temperature, luminance
Web-based application Mobile Application (Android)
- Nodejs. Websockets-MQTT - Openfire. XMPP - Mosquitto MQTT broker - Android (asmack, Alljoyn)
E-Flow Deployment
First Deployment at the Telecommunication School (UPM)
Second Deployment at Hospital de San José (Carabanchel, Madrid)
Second Deployment at Hospital de San Pastor (Carabanchel, Madrid)
results
End-to-End MQTT-to-Web using Websockets
Pub/Sub Network MQTT only
Mean Median SD Mean Median SD
a) Fixed devices
30,977 29,770 1,056 4,740 4,515 0,904
b) Mobile devices
246,983 250,989 11,893 65,904 66,938 4,111
c) Proxy Delay
4.402 4.520 0.830 - - -
Does the E-flow system provides acceptable user awareness levels?
We have checked the normality of the three samples with the Shapiro-Wilk normality test [23] and conclude that the three samples came from a Normal distribution (p-value <= 0.05).
User experience and awareness are still acceptable for mobile cases. This affirmation is based on the work of Jacob et al., which states that delays less than 1.0 seconds need no special feedback.
summary
E-flow system provides user awareness
Local Network
Infraestructure
Pub/Sub
Client
MQTT
Client
XMPP
Servers
Mobile
Safety Staff
Emergency
Response Teams
Faculty and
Students
subscribe()
publish()
802.15
Sensors
Proxy
MQTT-to-Web
Remote Networks
Mobile
Broker
Concentrators
Wi-fisubscribe()
publish()
WSN
Wi-fi
MQTT
Brokers
future work
stronger experimentation
Better WSN Integration
Integrate decision support system
UCAmI 2013 Costa Rica · December 2-6
E-Flow: A communication system for user notification in dynamic evacuation scenarios
Augusto Morales Technical University of Madrid
Picture by David Gordillo