1

SAFIRENET: Next-Generation Networks for Situational Awareness

Nalini Venkatasubramanian

2

Situational Awareness for Firefighters

Questions to be answered: Where are the firefighters? Are they doing well? Any danger?

Limited infrastructure access

High network deployment cost

Challenges

The Problem

Deliver contextual data sensed by firefighters to the incident commander

3

Motivation Multitude of technologies

WiFi (infrastructure, ad-hoc), WSN, UWB, mesh networks, DTN, zigbee

SAFIRE Data needs Timeliness

immediate medical triage to a FF with significant CO exposure

Reliability accuracy levels needed for CO

monitoring

Limitations Resource Constraints

Video, imagery Transmission Power, Coverage,

Failures and Unpredictability Goal

Reliable delivery of data over unpredictable infrastructure

Sensors

Dead Reckoning(don’t send Irrelevant data)

Multiple networks

Information need

DA

TA

NE

ED

S

4

Experiences with Existing Network Technologies…

Lessons Learned: Despite multitudes of technologies, rapidly deployable, self-configuring networks that provide end-to-end & continuous

connectivity are hard to create!!!

5

Experiences in deploying WiFi Mesh

Commercial mesh routers not good enough

5X improvement with new antenna technology

Better signal coverage better building penetration

• Some Setup effort required

• Not always feasible

• Vulnerable to hardware failures

6

SAFIRE Mote Sensor Deployment

IEEE 802.15.4 (zigbee)

Crossbow MIB510 Serial Gateway

Polar Heart Rate Module

Polar T31 Heart rate strap transmitter

Proprietary EMF transmission

To SAFIRE Server

IMU (5 degrees

of freedom)

Crossbow MDA 300CA Data Acquisition board on MICAz 2.4Ghz Mote

Heart Rate

Inertial positioning

Carbon monoxide

Temperature, humidity Carboxyhaemoglobin, light

7

Experiences in deploying mote sensors and Zigbee networksCalibration is essential

Rout e updat e per iod

Re

lia

bil

ity

static

mobile

↑Density↑Reliability

↑Mobility↓Reliability Network convergence, gatewayavailability

↑Size↓Reliability

Frequency matters!!

Topology matters!!

8

(Un) Reliability of Wi-Fi Networks

Ad-hoc 1hop > Ad-hoc 2 hops > Private AP >>> Public AP

• No background traffic• Controllable configuration

• Increased bandwidth share• Reduced

contentions/collisions

• Less interferences• Distributed

Beaconing

• Varying traffic load• Varying level of contentions and congestions• Varying inter-device distance

9

Creating Reliable Networks for Onsite Communication…

Goal: Enabling Robust, Timely Data Transfer by combining technologies

10

Approaches

Exploit multiple networks that together provide connectivity (Mobiquitous 2005, WCNC 2007, INFOCOM 2009)

WiFi mesh – direct connectivity to a mesh router MANETS – hop by hop connectivity to gateway nodes Zigbee adhoc – connect to WiFi backbone through gateway node

Exploit mobility when disconnected Store-and-forward networks (Delay Tolerant Networking)

mobile nodes ferry data to gateway node

Combine connected network clouds and disconnected networks

11

Reliable Content Delivery in Connected Networks Two aspects: Data delivery, message awareness

RADCAST: Flash Broadcast in MANETS (Infocom 2009, Percom 2009) Concurrent dissemination of awareness and content

Data diffusion: based on a mix of push/pull (Pryer) Awareness assurance: network traversal using walkers (Peddler)

Problem: fast network traversal (NP-hard) Minimizing cover time, termination time and transmission overhead

AwarenessAssurance

Fragmentation

DataDiffusion

ReliableContent

Dissemination

Metadata

Content Data

{{

{concurrent

Walker

Walker

concurrent

Assures reception

Pull

Push

concurrent

GuidesRetrievesmissing

Spreads

12

Supporting Varying Reliability Needs in Connected Networks

Reliability Level Reliability Needs Awareness

AssuranceData

DiffusionNetwork Size

Knowledge Cost

Max All reachable nodes receive the content √ √ Ignore High

Lower-Bounded

A specified number of nodes are guaranteed to

receive the content√ √ Exploit Medium

Best-EffortAs many nodes as possible

receive the content, no guarantee is required

√ × N/A Low

13

Situational Awareness in a disconnected environment

Aggregatecontextual data

Incident Commander

Board

Forward bundlesupon device encounters

Forward bundlesupon gateway

encounters

Periodically sensing e.g., WiFi AP fingerprints, accelerometer readings, residue battery, snapshots, audio/video recording, etc.

Visualizing the task execution process spatially and temporally

Easy deployment of one or several mesh routers at the edge

of the area

A Store-Move-and-Forward (DTN) based approach

14

The Store-and-Forward Data Transfer Problem

System Model

• Each device maintains a cache storing bundles from itself and others

• Devices exchange certain bundles in cache upon encounters

Goals

• High reliability• Low storage cost• Low transmission cost• Short latency

• How many copies should be generated for each bundle?Replication

• Which bundles should be forwarded upon device encounters, and in what order?Forwarding

• Which bundles should be removed to accommodate incoming bundles upon cache overflow?Purging

Sub-Problems

15

Store-and-Forward Data Transfer: Solution Overview

Replication

ForwardingContext Sens-ing & Collection{

Purging

Components

Fixed Number of Distinct Copies

Location-Closeness Based

Aliveness-Signi-ficance Based

TaskScheduling

0-1Knapsack

Strategies Modeling

16

Implementation on Mobile Devices

Applications

Middleware DTN-Based

Operating System SymbianMaemo

Emergency Situ-ational Awareness

RADcast

Flash Broadcast

17

Note: Reliable networks ≠Reliable Data Collection Sensing Errors Occur

Visibility Readings vary Occlusions etc.

Spikes in SpCO readings due to FF movement

Read errors due to misaligned sensor strip

Reliability at application level is also needed needed

Sensor Calibration (MMCN08) Heart-rate, CO exposure

Exploitation of Semantics, prediction Exploit application tolerance to

errors

-1000 0 1000 2000 3000 40000

200

400

600

800

1000

1200

Light readings (HZ-FF2)

23 24

Tim e [sec]

Lig

ht

lev

el

-500 0 500 1000 1500 2000 2500 3000 35000

500

1000

1500Light readings (HZ-FF1)

3

Tim e [sec]

Light

level