Next Generation Internet 2011 Kaiserslautern Adaptive and - - PDF document

Keynote Talk @ Conference on Next Generation Internet 2011 Kaiserslautern Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Torsten Braun, Universität Bern, Switzerland

Institute of Computer Science and Applied Mathematics Research Group “Computer Networks and Distributed Systems” braun@iam.unibe.ch rvs.unibe.ch joint work with Philipp Hurni, Benjamin Nyffenegger, Sebastian Barthlomé

Overview

>

Introduction

>

Experimentation in Wireless Sensor Networks

— Wireless Sensor Network (WSN) Testbed — Software-Based Energy Measurement

>

Research Experiments

— Maximally Traffic-Adaptive and Energy-Efficient WSN MAC Protocol — Adaptive Forward Error Correction

>

Conclusions and Outlook

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

2

Wireless Sensor Network Applications

>

Monitoring of buildings

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks Markus Wälchli, Torsten Braun: Building Intrusion Detection with a Wireless Sensor Network, 1st International Conference

• n Ad Hoc Networks, Niagara Falls, 2009

Kaiserslautern, June 27, 2011 3

Wireless Sensor Network Applications

>

Environmental monitoring

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks Thomas Staub, Benjamin Nyffenegger, Desislava Dimitrova, Torsten Braun: Operational Support

• f Wireless Mesh Networks

Deployed for Extending Network Connectivity, 1st International Workshop on Opportunistic Sensing and Processing in Mobile Wireless Sensor and Cellular networks (MobiSense), Bilbao, May 9-11, 2011

Kaiserslautern, June 27, 2011 4

Application Requirements

>

Energy-efficient operation

>

Low delays

>

Reliability

>

Adaptivity to varying link characteristics and traffic load

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

5

Design, Implementation, and Evaluation

• f Wireless Sensor Network Protocols

>

Simulations are only meaningful with accurate calibration of parameters, e.g., energy consumption, transmission characteristics, traffic models.

>

Experiments in testbeds give insights about protocol behaviour in more realistic scenarios and system-related issues, but face several problems

— Experiment control — Scalability — Reproducability — Energy measurements — Mobility

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

6

Wireless Sensor Network Testbed

Testbed Testbed Testbed Testbed Testbed Testbed

Wireless Sensor Network Testbed (WISEBED)

>

Recently finished FP7 FIRE STREP

>

Pan-European federation of 9 wireless sensor network testbeds

>

• approx. 1000 deployed nodes

>

Each partner runs own testbed with different hardware.

>

Requirement: uniform testbed infrastructure, i.e., standardization of access (description of experiments and results)

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 8

General Testbed Infrastructure

>

Portal

— Gateway between Internet and WSN — Used for WSN management

>

Wireless Sensor Network

— Sensor nodes communicate with each other. — Backbone used to communicate with portal.

>

Internet

— connects all WISEBED testbeds

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

9

th Internet

Wireless Sensor Network (WSN)

Backbone Portal DB WSN Testbed part of WISEBED WSN Testbed part of WISEBED

Testbed @ Universität Bern

Ethernet Mesh Node Portal (running TARWIS management system) Internet USB LAN wireless Sensor Node

10 Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Testbed @ Universität Bern

>

• Approx. 50 TelosB/MSB nodes connected to portal

via Ethernet with temperature, humidity, light sensors

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

11

TARWIS System Architecture

12 Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Testbed / Sensor Node Reservation

>

Reservation system maintains per-site reservation database.

>

User Interfaces

— Web-based user interface — iPhone application

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 13

TARWIS Experiment Configuration

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

14

TARWIS Experiment Monitoring

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

15

TARIWS-Generated Experiment Trace

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

16

Software-Based Estimation of Energy Consumption Software-Based Estimation of Energy Consumption

>

Problem: Equipment of sensor nodes with measurement hardware is

— very expensive. — difficult in out-door environments / real-world deployments. — not sufficient to support energy-awareness.

>

Solution: Software-based energy measurement (calibration using measurement hardware)

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 18

Hardware-Based Energy Measurements

>

Measurement of current draw and voltage using Sensor Network Management Devices (SNMD) from KIT

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 19

Simple 3-State-Model

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 20

• A. Dunkels, F. Osterlind, N. Tsiftes, Z. He: Software-based On-line Energy Estimation for Sensor Nodes. IEEE EmNets, 2007

Measured vs. Estimated Energy Consumption

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Approach: Measurement of current draw in different states and energy estimation by

Kaiserslautern, June 27, 2011 21

3-State-Model with State Transitions

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 22

Revised estimation:

Estimation Accuracy

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 23 OLS: Ordinary Least Squares Regression Analysis

MaxMAC: Maximally Traffic-Adaptive and Energy-Efficient WSN MAC Protocol

Energy Consumption of Wireless Sensor Nodes

>

Energy consumption as important issue in WSNs

>

Transmission of 1 bit consumes similar energy as processing 100-1000 instructions.

>

Transmitting and receiving consume similar power.

>

Startup times to be considered Medium Access Control Protocols avoid

>

• verhearing and idle listening by switching off the transceiver

>

collisions and retransmissions

>

control packets

>

frequent transmissions, e.g., by aggregation

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 25

MAC Protocols for Wireless Sensor Networks

1.

Scheduled Protocols

— Multiplexing and Allocation of channels, e.g., time multiplexing, requires accurate time synchronization

2.

Contention-based Protocols

— Channel sharing and allocation on-demand, often: periodic wake-ups — Problems: collisions and delays — Sender must ensure that receiver is awake during transmission

–

Transmissions of long preambles/beacons (B-MAC, X-MAC, WiseMAC)

–

Weakly synchronized wakeups (S-MAC, T-MAC)

–

Receiver signals wakeup (RI-MAC)

— Load adaptation

–

Adaptation of wakeup time dependent on activation events or load (T-MAC, X-MAC)

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Wake-up interval

Kaiserslautern, June 27, 2011 26

WiseMAC

>

Very energy-efficient MAC protocol

>

Unsynchronized nodes wakeup for a short time

>

Tpreamble = min {4θL,T}

— θ: clock drift, L: time since last update, T: duration of a cycle — Max. clock drift: 2θL, sender must start preamble transmission 2θL prior to wakeup and transmit it until 2θL after wakeup.

>

„Piggybacking“ of wakeup times

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Enz et al.: WiseNET: An Ultralow-Power Wireless Sensor Network Solution, IEEE Computer, Vol. 37, No. 8; August 2004 Kaiserslautern, June 27, 2011 27

IdealMAC

>

Idealized MAC protocol: only active, when data are transmitted or received.

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 28

MaxMAC: a Maximally Traffic-Adaptive and Energy-Efficient WSN MAC Protocol

>

is based on sampling of preambles, cf. WiseMAC

>

integrates destination address into preamble to reduce overhearing

>

Additional wakeups for higher rates of received packets (measurement by sliding window)

— Periodic reports in acknowledgements from receiver to sender — State transitions if thresholds T1,T2,TCSMA are exceeded.

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Base state S1 2 * duty cycle S2 4 * duty cycle CSMA RECV

packet rate ≥ T1 packet rate ≥ T2 packet rate ≥ TCSMA packet rate < T1 Lease expired packet rate < T2 Lease expired packet rate < TCSMA Lease expired

Kaiserslautern, June 27, 2011 29

MaxMAC

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Traffic estimation at receiver Rate ≥ T1 (2), double wakeup Rate ≥ T2 (6), fourfold wakeup Rate ≥ TCSMA (10), CSMA Rate<T1, Base state

Signal from receiver to sender Base interval

Sen- der Re- ceiver Preamble Data ACK Wakeup times Addtional wakeups CSMA

Philipp Hurni and Torsten Braun. MaxMAC: a maximally traffic-adaptive MAC protocol for wireless sensor networks. 7th European Conference on Wireless Sensor Networks (EWSN), Coimbra, Portugal, February 2010. Kaiserslautern, June 27, 2011 30

Omnet++ Simulations

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 31

Performance with 8 Node Chain (Simulation)

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Time [s] Time s] Time [s] load [packets/s] Received packets [packets/s] Aggregated power [W] IdealMAC, MaxMAC, CSMA T-MAC WiseMAC CSMA MaxMAC IdealMAC

Kaiserslautern, June 27, 2011 32

Throughput vs. Energy Efficiency

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

configuration (duty cycle, wakeup interval) Energy efficiency [kbit/J] Throughput [packets/s]

Kaiserslautern, June 27, 2011 33

Delay vs. Energy Efficiency

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Energy efficiency [kbit/J] Delay [s]

Kaiserslautern, June 27, 2011 34 Philipp Hurni, Torsten Braun: On the Adaptivity of Today's Energy-Efficient MAC Protocols under Varying Traffic Conditions, International Conference on Ultra Modern Telecommunications (ICUMT) 2009 , St. Petersburg, Russia, October 2009

Throughput vs. Delay vs. Energy Efficiency

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Delay [s] Throughput [packets/s] Energy efficiency [kbit/J]

Kaiserslautern, June 27, 2011 35

• 0.20

0.40 0.60 0.80 1.00 1.20 1.40 MaxMAC WiseMAC T-MAC CSMA B-MAC X-MAC S-MAC

Metric for Energy-Efficient and Adaptive Protocols

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

>

Weighted, normalized Euclidean distance to IdealMAC as measure for adaptivity

>

x: energy efficiency, y: throughput, z: delay

Weights (1:1:1) (1:2:2)

d norm(pi, Id, ω x, ωy, ωz) =

Kaiserslautern, June 27, 2011 36

Experimental Evaluation

>

Threshold parameters: T1=1, T2=2, TCSMA=3 packets / s

>

Base duty cycle: 0.6 %

>

Frame size: 30 bytes

>

Base interval: 500 ms

>

Lease times: 3 s

>

Bit rate: 38.4 kbps

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 37

Wisebed Experiments

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

Kaiserslautern, June 27, 2011 38

Adaptive Forward Error Correction Error Control in Wireless Sensor Networks

>

Wireless channels in sensor networks have varying bit error rates, sometimes up to 20 %.

>

Options

— Automatic Repeat Request (ARQ)

–

Retransmission adds delay.

–

Original transmission was useless, but consumed bandwidth and energy.

— Forward Error Correction (FEC)

–

Small delay (due to encoding and decoding), if error can be corrected.

–

En-/decoding can be costly (several 100 ms for decoding).

–

Too strong codes consume computing resources and bandwidth.

–

Too weak codes might not be able to correct errors. >

Proposed Approach: Adaptive FEC

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

40

Implementation of FEC Library

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

41

>

Repetition Code

>

Hamming Code

>

Double Error Correction Triple Error Detection (DECTED)

>

Bose-Chaudhuri-Hocquengham (BCH)

Adaptive FEC

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

42

>

Stateful Adaptive FEC (SA)

— Selection of current code dependent on success of previous transmission (next higher / lower level) — Quick adaptation

>

Stateful History Adaptive (SHA)

— History of last transmissions (here: 5) — For successful/failed transmissions: storage of next lower/higher level — Selection of level with majority in history — Reacts less quickly than SA-FEC

>

Stateful Sender Receiver Adaptive (SSRA)

— Consideration of number of corrected bit errors by receiver (to be reported in acknowledgement)

Energy Consumption by FEC

>

Additional power consumption by FEC

>

In case of no FEC, MSB430 node can enter lower power mode with IDefault

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

43

Error Occurrence Patterns

>

Measurements on outdoor link and indoor link (5 concrete walls)

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

44

Packet Delivery Rates with FEC Schemes

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

45

Wisebed Measurements

>

Different link characteristics → Deployment of a single FEC scheme would not be most efficient.

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

46

Static vs. Adaptive FEC

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

47

>

Similar error correction performance of static and very adaptive FEC schemes

>

Adaptive FEC advantages

— Lower processing and energy costs — Lower bandwidth and lower interference in multi-hop scenarios — Adapt automatically to different link characteristics

Conclusions

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

>

Contributions

— Methodology for accurate software-based energy measurements — Design of energy-efficient and adaptive protocols and evaluation of these in WSN testbeds

>

Experiences: Development and use of WSN testbed resulted in

— More efficient use of hardware resources — Repeatability and larger number of experiments (statistical significance) — Reproducability of experiments and results

>

Outlook

— Several experiences (testbeds, protocols) to be applied in other areas — Mobility support in wireless sensor / mesh network testbeds

Kaiserslautern, June 27, 2011 48

VirtualMesh: Wireless and Mobile Network Emulation

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

49 Thomas Staub, Reto Gantenbein, Torsten Braun: VirtualMesh: an emulation framework for wireless mesh and ad hoc networks in OMNeT++, SIMULATION: Transaction of the Society for Modelling and Simulation International, Vol. 87,

• No. 1-2, January 1, 2011

Mobility Support in Wisebed by VirtualMesh

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

50 Torsten Braun, Geoff Coulson, Thomas Staub: Towards Virtual Mobility Support in a Federated Testbed for Wireless Sensor Networks, 6th Workshop on Wireless and Mobile Ad-Hoc Networks (WMAN 2011) , Kiel, Germany, March 10 - 11, 2011

Thanks for your attention !

>

Contact: braun@iam.unibe.ch

>

More information: rvs.unibe.ch

Kaiserslautern, June 27, 2011

Torsten Braun: Adaptive and Energy-Efficient Protocols for Wireless Sensor Networks

51