AWS Community Day CPH - Three problems of Terraform
Dirk Pesch - Networked systems research at NIMBUS (Cork Institute of Technology)
1. Networked Systems Research @ Nimbus
Wireless Sensor and Vehicular Ad-hoc Networks
Dirk Pesch
Head of Centre
NIMBUS Centre for Networked Embedded Systems
Cork Institute of Technology
dirk.pesch@cit.ie http://www.nimbus.cit.ie
2. Overview
1. Overview of CIT and Nimbus Centre
2. Selected research in wireless sensor networks
for indoor applications and localisation
3. Protocol design for vehicular ad-hoc networks
in road safety applications
3. Cork Institute of Technology
• Ireland’s second largest Institute of Technology located
in Cork (south of Ireland)
• CIT offers Bachelor, Masters and PhD degrees in
Science, Engineering, Business, Art and Music
• CIT has ca. 15000 students and approx. 1000 staff
4. The NIMBUS Centre
• Focus on networked systems
research with applications in
• Energy Management
• Vehicular/Traffic
• Infrastructure Security
• Water Management
• Three research groups
– Adaptive Wireless Systems
• Wireless Network Design
GPS
• Algorithms & Protocols
PDR
Map
Filtering
• Real-time Localisation & Tracking
– Smart Systems Integration
• Sensor Device Integration,
• Miniaturisation and Embedding of Electronics
• Integral Sensing networks
– TEC Centre industry R&D group
5. Main Industry & Academic Partners
• Industry (national/international)
– Intel, UTRC, Bord Gais, Benetel, Redmere, Cylon
Controls, Decawave, SocoWave, Alanya, Lincor,
Eurotech, Seftec, IHG, Viva
– Philips, Schneider Electric, Honeywell, ANA,
BijoData, Daimler, HSG Zander, Arup, Gemalto,
Ennovatis, STM
• Academic(national/international)
– UCC/Tyndall, UCD, TCD, NUIG
– Univ. of Bremen, TU Hamburg, CEA LETI,
Fraunhofer IIS, Embedded Systems Institute/TU
Eindhoven, TU Dresden, Univ. College Antwerp, VTT
7. Open Issues for Protocol Design for WSN
• Reliability of wireless channel is a concern in many
applications
• Node life-times currently one to two orders of magnitude
shorter than required for many sensing applications
– Requires careful duty cycle adaptation
• Standards based WSN protocols are non-optimal
compared to proprietary proposals
• Limited understanding of deployment issues for WSN
– No wireless network design for deployment
– Limited understanding of WSN lifetime once deployed
• No integrated network management approach
• No communication protocol framework to deal with
diverse range applications and QoS requirements
– Results in custom designs every time which increases cost
8. Indoor Wireless Network Design
Methodology and Tool
• First tool to support systematic design and deployment of WSAN in
buildings
– Integrates with IFC BIMs
– Reduces equipment costs by > 20%
– Order of magnitude reduction in design time for non-expert
Wireless Network Design Process
PHASE 1 PHASE 2 PHASE 3 PHASE 4
Requirements Automatic Design Deployment Verification
Gathering & Optimisation
• A. Guinard, M. S. Aslam, D. Pusceddu, S. Rea, A. McGibney, D. Pesch, “Design and Deployment Tool for In-Building Wireless Sensor Networks: a
Performance Discussion”, in Proc. 7th IEEE Performance & Management of Wireless and Mobile Networks (P2MNET 2011), Bonn, Germany, Oct. 2011
• A. Mc Gibney, A. Guinard, D. Pesch, “Wi-Design: A Modelling and Optimization Tool for Wireless Embedded Systems in Buildings”, in Proc. 7th IEEE
Performance & Management of Wireless and Mobile Networks (P2MNET 2011), Bonn, Germany, October 2011
• A. Guinard, A. McGibney, D. Pesch, “A Wireless Sensor Network Design Tool to Support Building Energy Management”, in Proc. of 1st ACM BuildSys (in
conjunction with ACM SenSys), Berkeley, CA, USA, November 2009
9. Design Tool Case Study
Novice Designer Experienced Designer WSAN Design Tool
22% Routing
47% Routing
Traffic Novice Designer
53% Sensor
Traffic
Traffic Experienced Designer
78% Sensor 29% Routing
Traffic WSAN Design Tool
71% Sensor
Traffic Traffic
3 Gateways 5 Repeaters 3 hops max 3 Gateways 1 Repeater 3 hops max 2 Gateways 2 Repeaters 2 hops max
Sensing Data Data transmission Design Cost Design
Comments
Delivery Ratio cost (# packets) cost Savings Time
No previous WSN design experience, follows EnOcean Range
Novice Designer 97.0 % 1.85 € 3300
22% Routing € 0 4h
47% Routing 53% Sensor Traffic
Planning Guide
78% Sensor 29% Routing
71% Sensor
Traffic Traffic Traffic Traffic
Experienced Designer 97.6 % 1.21 € 2940 € 360 30 min WSN Design Expert, Sun SPOT developer Traffic
WSAN Design Tool 98.2 % 1.46 € 2620 € 680 40 min WSAN Design Tool
3 Gateways 5 Repeaters 3 hops max 3 Gateways 1 Repeater 3 hops max 2 Gateways 2 Repeaters 2 hops max
Sensing Data Data transmission Design Cost Design
Comments
Delivery Ratio cost (# packets) cost Savings Time
No previous WSN design experience, follows EnOcean Range
Novice Designer 97.0 % 1.85 € 3300 €0 4h
Planning Guide
Experienced Designer 97.6 % 1.21 € 2940 € 360 30 min WSN Design Expert, Sun SPOT developer
WSAN Design Tool 98.2 % 1.46 € 2620 € 680 40 min WSAN Design Tool
10. DCLA protocol
• The DCLA protocol is based
START
on Q-learning Any frames received?
No Increase
learning rate
Select max
inactive period
max(ai)
• DCLA explores and selects Yes
new actions adaptively Update r(ai)
according to the rewards
received Preliminary Yes Select next
action based on
exploration phase
• DCLA adapts duty cycle in
round-robin
Decrease
No
event-based scenarios
exploration rate
No
• Implemented in OPNET and Stable state
(e = 0)
No Select next
action based on
e-greedy
Greedily selected a
different action?
on telosB motes Yes
Yes Increase
exploration rate
Has the reward changed?
No
Select next action based
Increase Increase
on traffic change & last END
learning rate exploration rate
stable
R. de Paz Alberola, D. Pesch, “Duty Cycle Learning Algorithm (DCLA) for
IEEE 802.15.4 Beacon-Enabled Wireless Sensor Networks”, Ad-hoc
Networks, Elsevier, (http://dx.doi.org/10.1016/j.adhoc.2011.06.006)
11. Average Duty Cycle (DC) selection Average end-to-end delay (D)
Probability of Success (PS) Energy Efficiency
12. Event-based traffic
• Nodes generate traffic
following ON/OFF model
– ON/OFF distribution follows
Pareto distribution
– Packet arrivals follow
truncated normal
distribution Event detection
• A number of PIR sensors
detect the event and 30m
report to the sink
• Other nodes generate
CBR 30m
14. Distributed Duty Cycle Management (DDCM)
• Distributed Duty Cycle Management (DDCM) for IEEE 802.15.4
Beacon-Enabled Wireless Mesh Sensor Networks.
– DDCM uses DCLA to adapt node’s duty cycle to the network traffic and
manages the allocation of time slots as well as the prevention and
resolution of possible slot conflicts within a mesh network in a
distributed fashion.
T ransmit t ed T racked Superframe
Ext ended Broadcast
Beacon Beacons durat ion (SD)
SD SD
Coordinat or 1
SD ESD BSD SD ESD BSD SD ESD
(BO= 3)
Beacon Int erval (BI)
Coordinat or 2 SD BSD BSD SD
(BO= 4)
Beacon Int erval (BI)
Coordinat or 3 SD BSD BSD
(BO= 5)
Mult i-superframe durat ion (MD)
R. de Paz Alberola, B. Carballido Villaverde, D. Pesch, “Distributed Duty Cycle Management (DDCM) for IEEE
802.15.4 Beacon-Enabled Wireless Mesh Sensor Networks”, in Proc. of 5th IEEE International Workshop on
Enabling Technologies and Standards for Wireless Mesh Networking, Valencia, Spain, October 2011
16. IEEE802.15.4 TinyOS Implementation
DCLA
Duty Cycle
Adaptation
Clock Drift Radio
Adjustment CAP Sleep
CC2420 Power 16
Consumption
Estimation
17. Localisation and Tracking
MapUme is an opportunistic localisation system which fuses location related
sensor information that is readily available to localise people and objects
Clients MapUme Server
MapUme – OLS Server
Smart
Phone
GPS WiFi Tag
Sensor data
PDR
Camera networks
Map
Filtering
18. Platform for Safety and Security
Enhancement
Key components:
for large critical infrastructures like airports
Sensor and actuators networks – (Loc. + surveillance +environment system)
Context awareness – (moving objects and unexpected events)
Advanced real-time processing – for collision avoidance and navigation services.
Distributed middleware –scalability, predictability, configurability and continuous
commissioning,
(D)GPS
IIS active RFID,
Symeo LPR,
UWB,
CIT Opportunistic
localisation to
cover the rest
WiFi
WSAN
Cellular net.
19. Opportunistic Localisation
Ground Floor First Floor
Mean Location Error of Different Technologies
8
GSM GSM
GSM
7
6
GPS 5
WiFi
WiFi WiFi 4
GPS
3
No PDR
2
All
1
0
Outdoor Ground Floor First Floor
Outdoor
20. Vehicular Communication Network
Terrestrial
Broadcast
Satellite
UMTS
WiMax WiFi Hotspot
Variable
Message Sign
V2I
V2V: 802.11p, IR
21. Vehicular Adhoc NETworks - VANETs
• Vehicular communications has been primarily
motivated by safety
• Advent of Active Safety Applications
24. Vehicular Communications - VC
• Relevant Standards
– WAVE: Wireless Access in Vehicular Environments
• IEEE 1609 set of standards (incl. 802.11p) for VC
– IEEE 802.11p: 802.11a modification for VC
• V2V: Vehicle-to-Vehicle Communication
• V2I: Vehicle-to-Infrastructure Communication
• Our Focus: Development of a Broadcast protocol
for active safety applications
– Reliable Vehicular Geo-broadcast protocol (RVG)
25. Challenges for Broadcasting in VANETs
• Broadcasting is an extremely expensive
technique
– It floods the medium with a high number of redundant
transmissions
– Making an already unreliable medium more
unreliable
• Broadcasting for Safety Applications MUST
satisfy:
– High Packet Delivery
– Low End-to-End delay
– Minimal Overhead
26. Reliable Vehicular Geo-broadcast protocol
(RVG)
• RVG can disseminate any type of application
data but it has been optimised for the
dissemination of safety related messages
• RVG is focused on high packet delivery, low
delay and low overhead
• Compliant with the IEEE 1609 standards
• M. Koubek, S. Rea, D. Pesch, “Reliable Broadcasting for Active Safety Applications in Vehicular Highway Networks”,
in Proc. of IEEE International Symposium on Wireless Vehicular Communications (WiVeC) 2010, Taipeh, Taiwan,
April 2010M. Koubeck, S. Rea, D. Pesch, “Increasing Multi-Hop Broadcasting Reliability in VANETs”, EURASIP
Journal on Advances in Signal Processing, May 2010
• G. Pastor Grau, D. Pusceddu, S. Rea, O. Brickley, M. Koubek, D. Pesch, “Vehicle-2-Vehicle Communication
Channel Evaluation using the CVIS Platform”, In Proc. of IEEE/ IET International Symposium on Communication
Systems, Networks, and Digital Signal Processing, Newcastle, UK, July 2010
36. Summary and Outlook
• Nimbus research focuses on networked
systems with emphasis on wireless sensor and
vehicular ad-hoc networks
• The main application spaces include WSN for
building energy management and VANET for
traffic safety
• Future plans include to combine building energy
management with electric vehicle charging
• Challenges here include the integration of
widely heterogeneous wireless/mobile ad-hoc
networks to manage these applications