Cognitive Radio - A Necessity For Spectrum Pooling: Friedrich K. Jondral Paris, March 28, 2006

Cognitive Radio A Necessity for Spectrum Pooling
Friedrich K. Jondral Paris, March 28, 2006

Universitt Karlsruhe (TH)
Research University founded 1825
Communications Engineering Lab

INT
Spectrum Utilization Measurements

(550-1000MHz)
electric electric field strength field strength dBV/m dBV/m
density of the density of the time between time between arrivals arrivals dBs-1-1 dBs
Lichtenau (Germany), September 2001


INT
Synopsis Spectrum Pooling HIPERLAN/2 System Overview The Licensed User (LU) System Rental User (RU) System: Physical Layer Issues Rental User (RU) System: Detection and Signaling Cognitive Radio


INT
Spectrum Pooling Vision:

Usage of free capacities in licensed frequency bands: Licensed Users (LUs) lease out spectrum to Rental Users (RUs)
fre qu en cy
time
FDMA / TDMA LU system and HIPERLAN/2 RU system: Two different radio systems Coexistence in the same frequency region ?!

INT
HIPERLAN/2 System Overview

European Wireless Local Area Network Standard
2 ms
MAC Medium Access Control

MAC-Frame MAC-Frame
MAC-Frame
BCH Broadcast CHannel FCH Frame CHannel
flexible Lengths Broadcast Phase Downlink Phase Uplink Phase Random Access
ACH Access feedback CHannel PDU Protocol Data Unit RCH Random CHannel
Preamble
BCH
FCH
ACH
Downlink PDU Streams
Uplink PDU Streams
RCH
Slotted Aloha
SCH Short transport CHannel LCH Long transport CHannel
Preamble
Preamble
Preamble
SCHs
LCHs
SCHs
LCHs
SCHs
LCHs
...
LCHs
SCHs
LCHs
SCHs
Preamble
LCHs
SCHs


INT

Physical Layer: OFDM Transmitter structure and spectrum
bit sequence 1 serial-to-parallel conversion ... Cod ... IDFT (FFT) m 1 ... m 1 ... m parallel-to-serial conversion d1(l)
dr(l) Cod ...
D/A xMT(t)
RFMod.
xRF(t)
...
...
dN(l) Cod
f f 3 f 2 f 1 f0 f1 f2 f3


INT

Total Number of Subcarriers Number of Data Subcarriers KD Number of Pilots KP Total Number of OFDM Subcarriers Used K FFT Bandwidth OFDM Subcarriers Distance f Time Distance Between two FFT Input Samples TA Usable Part of an OFDM Symbol TU Guard Interval TG OFDM Symbol
64 48 4 52 (= KD + KP) 20 MHz 312,5 kHz (= 20 MHz / 64) 50 ns (= 1/20 MHz) 64 TA = 3.2 s (= 1/ f) 16 TA = 0.8 s 4 s (= TU + TG)
INT
The Licensed User System

1. Embedding a RU cell into a LU cell (Hot Spot Scenario)
LU system RU cell
2.
Channel pattern and Occupancy Vector (OV)

LU channel of width 4f
LU LU LU LU LU
carrier distance f OV = ( 0 1 0 0 1 0 0 0 1 1 0 0 0 0 1 0 )

INT
The Licensed User System

3. No LU signaling channel LU detection necessary for the RU system hidden / exposed station problem
frequency f1 frequency f2 LU1 LU2 RU detection area of the RU
The detection has to cover the whole sphere of influence of the RU cell Access delay for the LUs Waiting period 4. No Carrier Sense Medium Access (CSMA) in the LU system.

INT
RU Physical Layer Issues

LU detection and signaling Optimum detection? Quality of detection necessary for coexistence? OV transmission calls for a new protocol
RU system synchronization Necessity of a new preambel concept Optimum positioning of pilots?
Interference reducing measures Disturbances to the LU system caused by the sin(x)/x-shaped OFDM spectrum Disturbances to the RU system by FFT leakage caused by the non orthogonality of the LU signals


INT
10
Detection and Signaling

The engaged / idle decision has to be done by the RU system for each LU channel within the pool
The frequency resolution is realized by the anyhow existing FFT:
1. Sampling of the signal s(k) band limited to the pool width 2. FFT for 64 samples at a time. The process is repeated n times. 3. The spectrum values belonging to one LU channel are integrated into a vector z . 4. Decision based on z and on an optimality criterion.

INT
11
Signal and Detector Model

Structure of the detector in the rental users station Structure of the detector in the rental users station
Bandwidth of one licensed user's subband (af )


INT
12
Joint Conditional PDFs

NLOS assumption and central limit theorem yield: NLOS assumption and central limit theorem yield:
Only noise in case of licensed users absence: Only noise in case of licensed users absence:
Likelihood ratio test according to the NEYMAN-PEARSON-criterion: Likelihood ratio test according to the NEYMAN-PEARSON-criterion:


INT
13
Detector for Uncorrelated Samples

Under the condition Under the condition , ,the PDF the PDF simplifies to: simplifies to:
With these conditional PDFs the likelihood ratio results in: With these conditional PDFs the likelihood ratio results in:
Considering the likelihood ratio as aatransformation of random variables: Considering the likelihood ratio as transformation of random variables:


INT
14
Worst Case Consideration
Maximum deviation from the model, if real/imaginary parts are fully correlated: Maximum deviation from the model, if real/imaginary parts are fully correlated:
Hence, all Hence, all
are identical and can be represented by are identical and can be represented by
::
can be interpreted as aaconcatenated random variable: can be interpreted as concatenated random variable:


INT
15
Detection Probability
is the marginal PDF with respect to aarandomized is the marginal PDF with respect to randomized distribution: distribution:
The detection probability The detection probability
can be calculated as: can be calculated as:


INT
16
Receiver Operating Characteristic (ROC)
Worst-case ROC for Worst-case ROC for varying .. varying


INT
17
Distributed Detection: Single Detection

Every rental user station Every rental user station conducts individual detection conducts individual detection


INT
18
Distributed Detection: Collection

Rental user stations transmit Rental user stations transmit their results to the AP their results to the AP


INT
19
Distributed Detection: Broadcast

After an OR-combination, the AP After an OR-combination, the AP broadcasts the new allocation vector broadcasts the new allocation vector


INT
20
Gain by Distributed Detection

Overall Probability of detection ififat least one of the N stations detects the LU access: Overall Probability of detection at least one of the N stations detects the LU access:
Behavior of the overall false alarm probability Behavior of the overall false alarm probability predetermined: predetermined:
if if
is is


INT
21
Spectrum Efficiency
Impact of the false alarm probability PF on the RU system efficiency
relative spectrum utilization by LUs as detected by the RU system
relative spectrum utilization by LUs


INT
22
Regulation
Today spectrum is regulated by governmental agencies, e.g. the American Federal Communications Commission (FCC) or the German Regulierungsbehrde fr Telekommunikation und Post (RegTP) Spectrum is assigned to users or licensed to them on a long term basis normally for huge regions like whole countries Doing so, resources are wasted Vision: Resources are assigned where and as long as they are needed, spectrum access is organized by the network (i.e. by the end users)


INT
23
Self Regulation Wireless LANs (IEEE 802.11x) ISM band: 2400 2483.5 MHz WLAN band: 5150 5350 MHz and 5470 5725 MHz Ultra Wide Band 40
UWB EIRP Emission Level in dBm 45 50 55 60 65 70 10
0
Part 15 Limit
fc greater than 3.1 GHz fc less than 960 MHz

1
Frequency in GHz
10


INT
24
Advanced Spectrum Management

Spectrum reallocation: The reallocation of bandwidth from government or other long-standing users to new services such as mobile communications, broadband internet access, and video distribution. Spectrum leases: The relaxation of the technical and commercial limitations on existing licensees to use their spectrum for new or hybrid (for example, satellite and terrestrial) services and granting most mobile radio licensees the right to lease their spectrum to third parties. Spectrum sharing: The allocation of an unprecedented amount of spectrum that could be used for unlicensed or shared services.
According to G. Staple, K. Werbach: The End of Spectrum Scarcity. IEEE Spectrum, March 2004, pp. 41-44


INT
25
Cognitive Radio A Cognitive Radio (CR) is an SDR that additionally senses its environment, tracks changes and reacts upon its findings. A CR is an autonomous unit in a communications environment. In order to use the spectral resource most efficiently, it has to be aware of its location be interference sensitive comply with some communications etiquette be fair against other users keep its owner informed


INT
26
Cognition Cycle
A necessary condition for highest flexibility in mobile communications is a general rethinking in spectrum allocation: Open access In order to make open access feasible Cognitive Radios are necessary.
ORIENT Establish Priority
Infer on Context Hierarchie Pre-Process Parse
Immediate
Urgent
Normal
Generate Alternatives Evaluate Alternatives
OBSERVE
Receive a Message Read Buttons New States Prior States
LEARN
PLAN
Register to Current Time
Save Global States
DECIDE
Joseph Mitola III: Cognitive Radio An Integrated Agent Architecture for Software Defined Radio. KTH Stockholm, 2000
Outside World
Allocate Resources Send a Message Set Display Initiate Process(es)
ACT


INT
27
CR Properties
Mitola's cognition cycle is very general. The properties of cognitiv radios may be divided into two groups user centric properties (support functions like finding an appropriate restaurant, recommendation of a travel route, supervision of apointments, . . .) technology centric properties - spectrum monotoring - localization - awareness of processing capabilities (partitioning and scheduling of processes) - information and knowledge processing


INT
28
Technologies to be Implemented
A CR carries location (e.g. GPS or Galileo) sensors. It has to monitor its spectral environment, for example by using a real time broadband FFT. In order to track its location or the spectral environments development, it has to implement learning and reasoning algorithms. When complying with a communications etiquette it has to listen before talk as well as to prevent the disturbance of hidden stations. In order to be fair it has to compromise its own demands with the demands of other users, most probably making decisions in a competitive environment using the results of game theory. It has to contact its owner via a highly sophisticated man-machine-interface.


INT
29
Technology Centric CR
Control
Core
Spectrum Monitoring
Localization
Information & Knowledge Processing


INT
30
TeC-CR
Spectral Environment
RF Signals Monitoring Spectrum Holes Noise Statistics Traffic Volume Reception Interference Temperature Channel measurement and modeling
Transmission
Spectral Environment
RF Signals Monitoring
Spectrum Holes Noise Statistics Traffic Volume Reception Transmission Power Transmission Power and Spectrum and Spectrum Management Management
Interference Temperature Channel measurement and modeling
Available Channel Capacity
Available Channel Capacity
Station A
Station B
INT
31
Conclusions 1. Advanced spectrum management will be a hot topic in future research on wireless networks. 2. Unlicensed (ISM, WLAN) as well as secondary (UWB) spectrum usage are already under way. 3. First spectrum sharing strategies (e.g. spectrum pooling) are under investigation. 4. Advanced spectrum management calls for new developments in networking and terminal devices: Intelligent Networks & Cognitive Radios


INT
32
Questions ?


INT
33
SPECTRUM UTILIZATION MEASUREMENTS (50-550 MHz)
electric electric field strength field strength dBV/m dBV/m
density of the density of the time between time between arrivals arrivals dBs-1-1 dBs


INT
34
SPECTRUM UTILIZATION
(50 MHz-1GHz)
dBV/m


INT
35
BOOSTING PROTOCOL (1)

Signaling RUs + BooSs AP ? Boosting Protocol
Presently valid OV:
LU
LU LU
LU
New constellation (after the detection phase) :
LU
LU
LU
LU
newly occupied LU channel

released LU channel
36

INT

Signaling of newly occupied channels
LU LU LU
Boosting
Mapping phase
LU
LU LU
LU
LU 1 2 1
LU LU 3 4 5 6 7 8
LU
9 10 11 12 13 14 15 16 15 2 9 10 15
9 10 15 2 2 3 4 5 6
9 10 7 8
9 10 11 12 13 14 15 16

INT
37
Signaling of further occupied LU channels
2 1 2
9 10 15 2 3 4 5 6
9 10 7 8
15
9 10
15
9 10 11 12 13 14 15 16
not to be considered
New OV (to be signalized by the AP)
LU
LU
LU
LU


INT
38
SUMMARY (1)
Divided detection combined with a boosting protocol and robust Occupancy Vector signaling solves the LU detection problem and leads to a common base of the Physical Layer (PHY).
2 ms MAC frame MAC frame MAC frame
detection phase
boosting phase
broadcast phase
Parametrized PHY (OV), Parametrized PHY (OV), number of usable carriers is known number of usable carriers is known to the RU systems MAC to the RU systems MAC

INT
39
SUMMARY (2)
The RUs systems efficiency is mainly determined by the interference reducing measures!
OFDM channel f
Power spectrum Power spectrum density SS (f) density xx (f) xx of aasingle of single OFDM-carrier OFDM-carrier
LU
interference to the neighboring channel
LU
LU LU
LU
additionaly deactivated carrier additionaly deactivated carrier


INT
40
OUTLOOK
Integration of the results into our OMNeT++ software demonstrator.
Simulation with respect to all effects and with realistic channel models Tuning of the free parameters
Adaptive modulation for optimal use of disturbed channels (FFT leakage)
MAC layer: Investigation of scheduling algorithms particulary resistent against bandwidth variations


INT
41
ACKNOWLEDGEMENT
Our work on spectrum pooling has been supported by the German Federal Ministry of Research and Technology under grant No. IT-KT-ITKT01338200-01BU152. The work was mainly done by the INT staff members - Dipl.-Ing. Timo Wei - Dipl.-Ing. Fatih Capar - Ihan Martoyo, M.Sc. and by our graduate students - Jrg Hillenbrand - Albert Krohn


INT
42
MULTIDIMENSIONAL GAUSSIAN DENSITY

There is No Line of Sight ( NLOS ) from the LU to be detected to the measuring RU


INT
43
TOTAL DENSITY
Noise component density (AWGN)
fN
Fz z I z = expG J bg 2 2 c h H K
1
T 2 n N 2 N
Resulting density Convolution of the single densities

z = z c no LUh f bg z = G2 C c LUh 2 det 1C + E expF1 z c H bg c h fR
no LU N
fR
LU
2n
SS
2 N
SS
+ E
2 N
J hzI K
1
Derivation of an optimal estimator


INT
44
ESTIMATOR
Neyman-Pearson criterion: Maximization of the detection probability PD for a given false alarm probability PF
PF =
R1
PD =
z z
R1
f R no LU z no LU dz
f R LU z LU dz
c h
Likelihood ratio:
f R no LU
f R LU z LU
c h > z cno LUh

LU
2 N 1 LU
Optimal estimator:
F z e + Eh c Ehjz < bg G C -2 ln c G H
T SS 2 N 1 0
det CSS + 2 E N
2 n N
I c h J ch J K
45


INT
UNKNOWN COVARIANCE MATRIX

Problem: For the optimal estimator CSS must be known
optimal detection Receiver optimal detection Receiver Operating Characteristic Operating Characteristic (ROC) (ROC) for /N ==3dB and nn==33 S / for 3dB and
mismatched estimator!
Uncorrelated spectrum values for the LU receiving process:
PF
completely correlated LU spectrum values uncorrelated LU spectrum values
Completely correlated real parts and imaginary parts of the spectrum values:
PD Universitt Karlsruhe (TH)


INT
46
SNRD ESTIMATION
For which average power of a received LU signal must the LU channel be classified as used? depends on the permissible interferences on the LUs
RU LU r AP RU AP Access Point


INT
47
DETERMINATION OF SNRD
Higher SNRD
lower PF
enhanced efficiency
high P is advantageous for detection high P is advantageous for detection


INT
48
RECEIVER OPERATING CHARACTERISTI CS (ROCs) worst case consideration Simulation results:

maximal mismatched estimator
ROC, ROC, worst case, worst case, nn==55
ROC, ROC, worst case, worst case, nn==15 15
PD
PD


INT
49
ROCs
best case consideration: Uncorrelated spectrum values
ROC, ROC, best case, best case, nn==15 15
Reality is somewhere between best und worst case Choose n for the worst case !?

INT
50
DIVERSITY SOLUTION
Given detection probability : PD = 0,999 Problems: multipaths (fading) too high false alarm probability PF PD cannot be realized with only one measurement station!
Solution: distributed detection (diversity) all mobile terminals and three additional Boosting Stations take measurements jointly PD becomes realizable for moderate PF !

INT
51
DISTRIBUTED DETECTION
BooS RU LU BooS AP RU RU cell BooS

AP
Access Point
BooS Boosting Station LU RU
Licensed User Rental User


INT
52
SIGNALING (1)
RUs + BooSs (AP): Boosting protocol
AP computes the elementwise exclusive or of all OVs


INT
53
SIGNALING (2)
AP RUs + BooSs: Robust time-frequency broadcast
AP
Access Point
BooS Boosting Station LU RU
Licensed User Rental User


INT
54
DIVERSITY-GAIN
What is the gain of divided detection ? diversity PD for a specific RU may be reduced no fading PF decreases
The individual detection results are statistically independent. The individual detection results are statistically independent. IfIfthe receiving conditions at the (m) measuring stations (RUs and BooSs) the receiving conditions at the (m) measuring stations (RUs and BooSs) are similar, we get : : are similar, we get
m
1 2 3 4 10 20
PD
0.999 0.968 0.900 0.822 0.499 0.292
PF
0.982 0.662 0.294 0.100 0.001
PZF
0.982 0.886 0.648 0.344 0.010


INT
55
LU TRANSMISSION MODEL
Signal model
s k C f S x, y
b g
n k C f N x, y
r k C f R x, y
b g
b g
2 2n
Transition to n FFT repetitions
b g b g R R f b yg f b, x ,K, x , y , y ,K, y g f bgwhere z = b yg x, = x = z x,

f S x , y f s x, y :
n 1 2 n T S S 1 2 S
n real parts
n imaginary parts

INT
56

Cognitive Radio - A Necessity For Spectrum Pooling: Friedrich K. Jondral Paris, March 28, 2006

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Cognitive Radio - A Necessity For Spectrum Pooling: Friedrich K. Jondral Paris, March 28, 2006

Încărcat de

Drepturi de autor:

Formate disponibile

Cognitive Radio A Necessity for Spectrum Pooling

Friedrich K. Jondral Paris, March 28, 2006

Communications Engineering Lab

Spectrum Utilization Measurements

electric electric field strength field strength dBV/m dBV/m

Lichtenau (Germany), September 2001

Communications Engineering Lab

Universitt Karlsruhe (TH)

Communications Engineering Lab

Spectrum Pooling Vision:

Communications Engineering Lab

HIPERLAN/2 System Overview

MAC Medium Access Control

BCH Broadcast CHannel FCH Frame CHannel

Downlink PDU Streams

Uplink PDU Streams

SCH Short transport CHannel LCH Long transport CHannel

Universitt Karlsruhe (TH)

Communications Engineering Lab

HIPERLAN/2 System Overview

dr(l) Cod ...

Universitt Karlsruhe (TH)

Communications Engineering Lab

HIPERLAN/2 System Overview

The Licensed User System

Channel pattern and Occupancy Vector (OV)

Communications Engineering Lab

The Licensed User System

Communications Engineering Lab

RU Physical Layer Issues

RU system synchronization Necessity of a new preambel concept Optimum positioning of pilots?

Universitt Karlsruhe (TH)

Communications Engineering Lab

Detection and Signaling

The frequency resolution is realized by the anyhow existing FFT:

Communications Engineering Lab

Signal and Detector Model

Bandwidth of one licensed user's subband (af )

Universitt Karlsruhe (TH)

Communications Engineering Lab

Joint Conditional PDFs

Universitt Karlsruhe (TH)

Communications Engineering Lab

Detector for Uncorrelated Samples

Universitt Karlsruhe (TH)

Communications Engineering Lab

Worst Case Consideration

Hence, all Hence, all

Universitt Karlsruhe (TH)

Communications Engineering Lab

The detection probability The detection probability

can be calculated as: can be calculated as:

Universitt Karlsruhe (TH)

Communications Engineering Lab

Receiver Operating Characteristic (ROC)

Worst-case ROC for Worst-case ROC for varying .. varying

Universitt Karlsruhe (TH)

Communications Engineering Lab

Distributed Detection: Single Detection

Universitt Karlsruhe (TH)

Communications Engineering Lab

Distributed Detection: Collection