Communication-aware Service Platform for Collaborative UAV ... UAV-g-CW-10-Web - final.pdf · dortmund university Slide 2 Communication Networks Institute Prof. Dr.-Ing. C. Wietfeld

Faculty of Electrical Engineering & Information TechnologyCommunication Networks InstituteProf. Dr.-Ing. Christian Wietfeld

dortmunduniversity

Communication-aware Service Platform for Collaborative UAV applications

Christian WietfeldTU Dortmund University, Germany

September 2013

Talk at UAV-g

dortmunduniversity

Slide 2

Communication Networks InstituteProf. Dr.-Ing. C. Wietfeld


Content

Applications for Collaborative UAV Systems

System Architecture of Service Platform

Selected Platform Services: Communication-aware Networking Algorithms

Dynamic Role Management and Task Assignment

Secure Wireless Mesh Networks for UAV Swarms

Example Application Services Real-time 3D Virtualization

Aerial Relays for Ad-hoc LTE Network Provisioning

Conclusion & Outlook

dortmunduniversity

Slide 3



Where UAV swarms are useful (compared to ground-based systems and single UAVs)

Exploration Ad-hoc Communication

Advantages of a UAV swarm: • Faster task execution• Increased exploration coverage• More, but less complex sensors• Scalable

Advantages of a UAV swarm: • Faster network setup independent of

ground infrastructure• Increased network coverage

NetworkGateway

Air-to-Air

Air-to-Air

e.g.Plume

Air-to-Ground

Serving Cellular

Base Station Users

Movement

Air-to-User

Sensor Data Sink Ground Network

Coverage

MovementInterferingCellularBaseStation

dortmunduniversity

Slide 4



A UAV swarm can achieve more in less time

BUT: control is more complex, as UAVs must cooperate to spread efficiently acrossthe scenario communication within the swarm is key!

dortmunduniversity

Slide 5



Support of emergency response: AIRSHIELD – Sensing for fire

fighters AVIGLE – Avionic Digital Service

Platform ANCHORS – Ad Hoc Networks for

rescue forces AirBeam – Wide-Area Airborne

information for Emergency Situation Awareness

Cooperative UAV research @ CNI

dortmunduniversity

Slide 6



Content









dortmunduniversity

Slide 7



Key Components of the multi-purpose UAS Service Platform

Service ControlGround(SCG)

Service Control

Air(SCA)

Air-to-Ground

Air-to-Air

dortmunduniversity

Slide 8



Service Control AService

Control AService

Control Air(SCA)

Service Control Ground (SCG)

Service Control Air

(SCA)

3D Sensor Exploration

(3DSE)

Aerial RadioAccess Network(ARAN)

…

Air-to-Ground Comm.

Air-to-Air Comm.

CognitiveMobility Control

Task Assignment &Mission Planning

ApplicationServices

Architecture of the multi-purpose UAS Service Platform

C. Wietfeld, K. Daniel, Cognitive Networking for UAV Swarms, in „Handbook for Unmanned Aerial Vehicles“, Ed. K. P. Valavanis; G. J. Vachtsevanos, Springer Reference, 2014.

dortmunduniversity

Slide 9



Service Control

Air(SCA)

Flight ControlAccess Service (FCAS)

Mobility ControlService (MCS)

Payload Management Service (PMS)

Obstacle Map Service (OMS)

Radio Propagation Map Service (RPMS)

System Monitoring & Logging Service (SMLS)

Dynamic Role Mgmt. Service (DRMS)

Mobile Comm. Mgmt. Service (MCMS)

UAV Information Dispatch Service (UIDS)

802.11a .. s

UMTS / LTE

…

Multicopter

Tiltwing

…

Sensors

Comm. Relays

…

Scout

Intermediate Relay

…

CAPF

IPAR

…

ServiceControlGround

(Air-to-Ground)

Other SCAs of AUS(Air-to-Air)

CAPF- Communication-Aware Potential FieldsIPAR – Interference-Aware Positioning of Aerial Relays

Geo-Positioning Service (GPOS)GPS

GALILEO

…

User Equipment

(Air-to-User)

Service Control Air (SCA)

dortmunduniversity

Slide 10



Service ControlGround(SCG)

Application ServiceRegistry (ASR)

Dynamic Task Assign-ment Service (DTAS)

UAV Payload Mmgt. Service (UPMS)

System Monitoring & Logging Service (SMLS)

Mission Preplanning & Safety Service (MPPS)

Mobile Comm. Mgmt. Service (MCMS)

Swarm Information Dispatch Service (SIDS)

802.11a...s

UMTS / LTE

…

3DVE Service.

ARAN Service

…

Optical Sensors

Comm. Relays

…UAS ApplicationServices

Service Control Airs of Swarm(Air-to-Ground)

UAS Radio Propagation Map Service (UPMS)

UAS Obstacle MapService (UOMS)

3DSE – 3D Sensor ExplorationARAN – Aerial Radio Access Network

Service Control Ground (SCG)

dortmunduniversity

Slide 11



Model-based Development Process

Hardware in the LoopValidation

SimulationValidation

Software in the LoopValidation

FullExperimental

Validation

Optimized Parameterization

Refined Control Algorithms & System ConceptIn

itial

Con

cept

Evol

ved

UAV

Sys

tem

ExperimentalValidation

SCG

SCA1SCA2SCA3

Hardware in the Loop

dortmunduniversity

Slide 12



Content









dortmunduniversity

Slide 13



Scientific objectives of context-aware mobility algorithm design

Design context-aware mobility

algorithms

Context: Unknown environment Changing communication

channel Ground station

temporarily not available

Sensitivity: Continuous RSSI

measurements Memory of visited cells

Mobility algorithm:Evaluation and Usage of

controlled mobility“self-placement / self-

configuration “

Sensordistribution

Sensorcoverage

Receivedpower

Swarmcoherence

„Exploration efficiency“ „Connectivity“

dortmunduniversity

Slide 14



Spatial Exploration Ratio

∑

Exploration Distribution Index, T: Theil-Index

Cluster Separation Ratio

∑: # ,

: #

Exemplary Key Performance Indicators

Separation

SER=5%

EDI0

“unequal“

EDI1

“equal“

Cluster

c = 1

c = 2

dortmunduniversity

Slide 15



Key Performance Indicators

Cluster Separation Rate >= 1

SER 100%

EDI 1

time

dortmunduniversity

Slide 16



Pre-planned schemes cannot cope with dynamically changing environments

Example: Deterministic Tour Planning Centralistic, deterministic planned paths good Exploration Ratio

Maintain static formation

BUT: Mission area has to be known a priori Not suitable in dynamically changing

environments Not robust against UAV loss

dortmunduniversity

Slide 17



Mobility Algorithm

Micro Mobility

DeterministicTour Planning

RepellingWalks

Cooperative Area Exploration

StaticFormation

ClusterBreathing

Comm.-awarePotential Fields

Efficient & AutonomousExploration

„Spatial Exploration“ „Connectivity“

Macro Mobility

Communication-aware Mobility

K. Daniel, S. Rohde, N. Goddemeier, C. Wietfeld. “Cognitive Agent Mobility for Aerial Sensor Networks“, IEEE Sensors Journal, Nov 2011

Self-configuringReliable Links with

Guaranteed QoS

dortmunduniversity

Slide 18



Macro-Mobility for Exploration in unknown environments:Probabilistic movement Random Walk revisits

Self Avoidance Walk deadlocks

Self Repelling Walk Less restrictive than Self

Avoidance Walk Deadlock free

Well distributed UAVs high EDI

BUT: Not communication aware Clusterization

dortmunduniversity

Slide 19



Macro-Mobility for Global Exploration:Cluster Repelling Walk for Cooperative UAVs

The swarm “performs” a grid-based Self Repelling Walk Macroscopic swarm movement

Each UAV get the same overlay movement vector

Swarm centroid

visited cell

3. Choose neighbor cell with highest coefficient

4. Move complete cluster

0.8

0.5

0.2

0.1

-1

1. Assign randomized coefficient to each neighbor of the swarm centroid

2. cell not visited increase the coefficient cell visited set coefficient to -1

1.5

1.8

1.1

dortmunduniversity

Slide 20



Micro-Mobility: adapt to changing communication channels

1000

1500

2000

2500

3000

2500 3000 3500 4000 4500 5000 5500

3500

Y-A

xis

[m]

X-Axis [m]

flee

seek

Goodchannel

Badchannel

Example traces of UAV swarm

dortmunduniversity

Slide 21



Micro Mobility– Cluster Breathing (CB)

Continuous automatic adaptation tochanging channel conditions

dortmunduniversity

Slide 22



Overlay of communication-aware microscopic with macroscopic behaviour

Static formation of swarmcannot adapt to changing channelcharacteristics

Adaptive behaviour within swarmtaking into account channel characteristics

dortmunduniversity

Slide 23



Taking into account obstacles: Communication Aware Potential Fields (CAPF)

The total force on each UAV is defined as:a

total number of repelling forces Repelling force of obstacle j to UAV i

a total number of attracting or repelling UAVs Repelling force of UAV j to UAV i

Attracting force of UAV j to UAV i

Attracting force of moving crowd to UAV i

Communication Awareness: Selection of a subset of all UAVs that impose

attracting/repelling forces by means of the RSSI Calculation of force direction and strength by means of

the RSSI cf. Cluster Breathing

dortmunduniversity

Slide 24



Scenario parameters for performance comparison

1500m

1500

m

Parameter Specification

Number of simulation runs

5-10

Weight of UAV 2 kg

Max. velocity 50 km/h

Number of UAVs 3, 5, 8, 10, 15

Transmit power 20 dBm

Frequency 2,4 GHz

Channel models Free Space / Two-RayAreas of shadowing

Cognitive mobilityalgorithms

Self Repelling WalkDetermin. Tour Plan.Coop. Repelling WalkComm.-AwarePotentialFields

Distortion areas

dortmunduniversity

Slide 25



Spatial Exploration Ratio (SER)

DTP: Best SER, but not communication-aware (theoretical limit) Self-Repelling Walk: fast exploration, but severe separations!

Communication-aware strategies are able to explore an area (high SER)

und keep swarm coherence (low CSR) at the same time!

Reference: Self Repelling Walk

CAPF/CB

Reference: DTP

Clu

ster

Sep

arat

ion

(CS

R)

SRW CAPF/CB0

1

2

3

4

5

DTP

dortmunduniversity

Slide 26



Exploration Distribution Index (EDI)

Potential Fields / Cluster Breathing performs worse than Self Repelling walk but better than Deterministic Tour Planning

Reference: DTP

Reference: SelfRepelling Walk

Cluster Breathing/CAPF

EDI0

“unequal“

EDI1

“equal“

dortmunduniversity

Slide 27



Scalability of sensor swarm

0

10

20

30

40

60

50

70

80

90

100

Spa

tialE

xplo

ratio

n R

atio

(SE

R)

0 30 60 90 120 150 180 210 240 270 300Time [min]

10 nodes

Scenario: Cluster Breathing & Cluster Repelling Walk

8 nodes

5 nodes

3 nodes

15 nodes

12,6% 9,6% 7,0% 6,4% 4,8%

SE

R [%

] / N

odes

3 5 8 10 15Nodes

after 60 minutes

dortmunduniversity

Slide 28



Useful to detect unvisible threats such as radiationDetection after 30 min

Most recent Refinements: Detection of Plume Barrier

Detection after 60 minDetection after 90 minDetection after 120 minDetection after 150 min

D. Behnke, Patrick-Benjamin Bök, C. Wietfeld. ''UAV-based Connectivity Maintenance for Borderline Detection'', IEEE 77th Vehicular Technology Conference (VTC-Spring), Dresden, Germany, Jun 2013

dortmunduniversity

Slide 29



Comparison of novel communication-aware algorithmsAlgorithm Communication Robustness /

Self-OrganizationGlobal Spatial

Exploration Ratio

Plume BorderDetection

RatioDTP

Self RepellingWalk

CooperativeRepelling +ClusterBreathing

CooperativeRepelling +Distributed Dispersion Detection(DDD)

Task-specific adaptations lead to improve service performancecombined with reliable communication behaviour

dortmunduniversity

Slide 30



Content









dortmunduniversity

Slide 31



Role-based Relaying Strategies

Release ofautonomouscluster

Exploration Volume

Ground Stations

T ::

Scenario : m x1500m x 1500m

2A A .11 5 - - .

ask Exploration of VolumeSteering CAPF combined with CRW

Size 1500Number of UAVs 5A G UMTS (2,1 GHz) with realistic A2G channel

2 IEEE802 a, GHz Band,omni direct antenna

01500

Exploration Time [min]300

50

100

Spa

tial

Expl

orat

ion

Rat

io[%

]

No Return & ReleaseExploration volume limiteddue to range of ground network

Communication-AwarePotential Fields withReturn & Release

Air-to-AirCoverage

ArticulationPoint

Air-to-GroundCoverage

Air-to-GroundRelay

Return ofautonomouscluster

N. Goddemeier, K. Daniel and C. Wietfeld, "Role-Based Connectivity Management with Realistic Air-to-Ground Channels for Cooperative UAVs", IEEE Journal on Selected Areas in Communications (JSAC), June 2012.

dortmunduniversity

Slide 32



Exemplary study in realistically modelled scenario

2 UAVs (scout and relay) connected to ground station in multi-hop scenario –position of UAVs adapted due to shadowing of buildings

dortmunduniversity

Slide 33



Content









dortmunduniversity

Slide 34



Different routing protocol options analyzed:

Secure Wireless Mesh Networks for collaborative UAVs:

Reactive: AODV, DYMO Proactive:OLSR, BATMAN, BATMAN-AdvancedHybrid: HWMP

Our new approach:Reactive + Secure: PASER

EncryptedRouting Messages

EncryptedData Packets

Mesh LinkLegende:

UAV 1 Sender

UAV 2

UAV 3

UAV 4

GatewayDestination

Data Centre

Sbeiti, M., Wietfeld, C., "PASER: Position Aware Secure and Efficient Mesh Routing Protocol", IETF Internet Draft, November 2012. [draft-sbeiti-karp-paser-00]

dortmunduniversity

Slide 35



Attacker drops encrypted data frames > 1300 Bytes (e.g., videopackets)

Standard WLAN/Internet security mechanisms not sufficientExample routing attack: Wormhole

EncryptedRouting Messages

EncryptedData Packets

Mesh LinkLegende:

UAV 1 Sender

UAV 2

UAV 3

UAV 4

UAV1: How can Ireach the Gateway? UAV3: How can UAV1

reach the Gateway?

UAV2: How can UAV1reach the Gateway?

UAV4: How can UAV1, UAV2 reach the Gateway?

UAV1: How can Ireach the Gateway?


Gateway ismy directneighbour

Wormhole Tunnel


Gateway: UAV1 reachsme directly

UAV1 is my directneighbour Gateway

Destination

Attacker only forwards encrypteddata frames < 500 Bytes ( thresholdfor routing messages)

Gateway: UAV1 reachsme via UAV2, UAV4Gateway: UAV1 reachsme via UAV3, UAV4

dortmunduniversity

Slide 36




PositionSender


PositionSender, UAV2, UAV4


PositionSender, UAV2, UAV4


PositionSender


PositionSender

How Does PASER Combat the Wormhole Attack?

ProtectedRouting Messages

Mesh LinkLegende:

WMNS: Wireless Mesh Networks; UAVs: Unmanned Area Vehicles

UAV 1 Sender

UAV 2

UAV 3

UAV 4

Wormhole Tunnel

UAV1 is reachablevia UAV4, UAV2 Gateway

DestinationUAV2: How can UAV1reach the Gateway?

PositionSender, UAV2

UAV3: How can UAV1reach the Gateway?

PositionSender, UAV3

Incoming PASER packet…- allowed distance: 250m- acquiring GPS position…- measured distance: 800m

Discarding packet!Deleting neighbor!Updating all routes via neighbor!

800m

dortmunduniversity

Slide 37



Experimental Validation of PASER in Wormhole Attack

Iperf UDP Client 3.5Mb/s Iperf UDP Server

M. Sbeiti, J. Pojda and C. Wietfeld, "Performance Evaluation of PASER - an Efficient Secure Route Discovery Approach for Wireless Mesh Networks", 23rd IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Sydney, Australia, Sep 2012.

dortmunduniversity

Slide 38



M. Sbeiti, J. Pojda and C. Wietfeld, "Performance Evaluation of PASER - an Efficient Secure Route Discovery Approach for Wireless Mesh Networks", 23rd IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Sydney, Australia, Sep 2012.

HWMP+SAE OLSR+WPA2

Impact of the Wormhole Attack in an experimental Testbed

PASER is robust against wormhole attack, due to its security features and its geographical leash mechanisms

Reliable Network in case of PASER

Sabotaged Network Despite the IEEE Security Frameworks

PASER HWMP+SAE OLSR+WPA2

Wor

mho

leA

ctiv

atio

nPe

riod

dortmunduniversity

Slide 39



Content









dortmunduniversity

Slide 40



3D Visual Exploration Service

Efficient UAVs withintegrated embedded PC and integrated camera

High datarate meshcommunication forpayload data

Live Task Assignmentand intelligent UAV remote trajectoryalgorithms for 3D reconstruction

Live image transmissionto quasi-online 3D reconstruction process

Principal ReconstructionPoint

dortmunduniversity

Slide 41



- 3D reconstructionstarts after first3 Images (<1Min)

- 3 minutes and 6 images after Take-Off thereis already a partial 3D model withsignificantdetails

- Shown result is~10min after Take-Off

Live 3D Reconstruction During Flight Time

dortmunduniversity

Slide 42



Content









dortmunduniversity

Slide 43



Example: Cell OverloadCompensation

Methodology: Overload scenarios are being

resolved by moving mobile eNodeBs towards the overload area

Benefits: Spectral Efficiency Gains Coverage Gains Traffic Offloading

eNodeB 2

Cell overloadmobile eNodeB

eNodeB 1

Where is the best place to deploy the eNodeB?

S. Rohde, M. Putzke, C. Wietfeld. ''Ad Hoc Self-Healing ofOFDMA Networks Using UAV-Based Relays'', Journal on Ad Hoc Networks (Elsevier), Jul 2012.

dortmunduniversity

Slide 44



Simulation results: scenario overload situation Simulated using NS-3 „Lena“ Linear Movement of mobile LTE eNodeBs

UE Association (Best-Server)

dortmunduniversity

Slide 45



Conclusion and Outlook

Benefits of UxV-Service Platform: Cooperative UAVs can increase the efficiency of task execution

significantly. Dynamically changing communication channels require for channel

adaptive and cognitive swarm mobility CNI UxV service platform automatically manages the trade-off

between application service (e.g. exploration) and communication reliability

Application developers can focus on their applications tasks!

On-going work: New use cases: CNI UxV service platform to be leveraged for

precision farming, critical infrastructure inspection, etc. Heterogenous UxVs: Additional UxVs from different suppliers will be

integrated via the service platform to collaborate.

dortmunduniversity

Slide 46



See you in December 2013 in Atlanta

www-wi-uav.org

Thank you for your attention!

4th IEEE GLOBECOM Wi-UAV International Workshop on

wireless networking & control for unmanned autonomous systems

dortmunduniversity

Slide 47



Contact Information

Address:TU Dortmund UniversityCommunication Networks InstituteOtto-Hahn-Str. 644227 Dortmund

Germany

Head of InstituteProf. Dr.-Ing. Christian Wietfeld

Tel.: +49 231 755 4515Fax: +49 231 755 6136e-mail: [email protected]: http://www.cni.tu-dortmund.de

Thank you for your attention!

Documents

Communication-aware Service Platform for Collaborative UAV ... UAV-g-CW-10-Web - final.pdf · dortmund university Slide 2 Communication Networks Institute Prof. Dr.-Ing. C. Wietfeld