Supélec Seminar - Cognitive Radio in HF Communications

Transcription

Supélec Seminar - Cognitive Radio in HF Communications
Cogni7ve(Radio(in(HF(Communica7ons(
Laura(Melián(Gu7érrez(
lmelian@ide7c.eu(
IDeTIC(
Universidad(de(Las(Palmas(de(Gran(Canaria,(Spain(
Rennes,(6th(November(2014(
Outline(
•  IDeTIC:(A(li/le(bit(about(us.((
•  Communica7ons(in(the(HF(band(
•  Cogni7ve(Radio(in(HF(communica7ons(
–  Using(wideband(HF(receivers(for(spectrum(sensing(
–  Learning(with(Hidden(Markov(Models(
–  Decision(making(for(dynamic(spectrum(access(
•  Conclusion(
1(
IDeTIC(
•  Research(Ins7tute(
within(ULPGC(
Research(ins7tute(within(
Universidad(de(Las(Palmas(
de(Gran(Canaria(
Around 65 people
! 
! 
! 
! 
! 
Staff((PhD):(((( (
(
(
Staff((no(PhD): (((((( (
(
Hired((no(permanent(staff):(
MSc(students: (
(
(
BSc(students:( (
(
(
(29(
(((8(
(16(
(((6(
(((3(
… and 12 external collaborators
www.ide7c.eu(
2(
IDeTIC:(Academic(ac7vi7es(
PhD'program'on''
Communica.ons'Systems'
Master'on'home'
automa.on,'
architecture'and'
sustainable'for'
tourism'
Expert'course'on'
IT'and'tourism'
Master'on'IT'solu.ons'
for'the'environment'
and'welfare'
Expert'course'on'
aeronau.cal'and'airport'
management'
Short'course'on'
“the'port'city”'
3(
IDeTIC:(Research(fields(
Biometrics'&'
Signal'Processing'
Wireless'Photonics'
&'InKhome'
Services'
IT'applied'to'Social'
Sciences'
Mari.me'
communica.ons'&'
surveillance'
RF'Systems'&'Radar'
Sensor'networking,'
SmartSpaces'&'IoT'
Automa.c'geoloca.on'
of'wildfires'
Space'&'AircraL'
Electronics'
4(
HF(Communica7ons(
•  Research(line(between(IDeTIC((ULPGC)(
and(GAPS((UPM).(
77%
•  LongWdistance(HF(communica7ons(
•  HFDVL(System((HF(Data+Voice(Link):(
–  Based( on( SoZware( Radio( with(
mul7carrier(modula7ons.(
–  Digital( interac7ve( voice( and( highW
data(rate(transmissions.((
–  O p e r a t e s( w i t h( c o m m e r c i a l(
transceivers.(
–  SISO( &( SIMO( (up( to( 4( Rx)(
configura7ons.(
•  Three(configura7ons(in(data(
transmission:(
–  File(transfer(
–  Short(message((SMS)(
–  HFMail(
50%
87%
67%
57%
91%
66%
[EXAMPLE
,
]
5(
HFDVL:(Real(Tests(
Tests done by the Spanish Department of Defense
Arnomendi ship (60 days)
Flight from Manas to Zaragoza (14 h)
Hercules C-130 airplane
6021 Km
3864'Km
7950'Km
Canary(
Islands
6(
HFSA_IDeTIC_F1_V01(Database(
Spectrum(ac7vity(in(the(14(MHz(amateur(band.(
Yagi
antenna
Broad-band HF transceiver
Agilent
Vector Signal Analyzer
PC with:
System Vue and
VSA software
Spectrum power
measurement
•  Power(measurements(in(the(frequency(domain.(
•  600(kHz(bandwidth((200(channels(simultaneously):(
amateur(band(and(other(sta7ons.(
•  Dura7on:(10(minutes.(
•  Weekdays(&(Weekends.(
•  Each(sample(represents(a(3kHz(channel(in(2(seconds.(
7(
HFSA_IDeTIC_F1_V01(Database(
Spectrum(ac7vity(in(the(14(MHz(amateur(band.(
8(
Cogni7ve(Radio(in(the(HF(band(
•  Challenges(to(face(in(this(environment(
–  There(is(no(coordina7on(among(users.(
•  Frequency(alloca7on(per(country.(
•  TransWhorizon(behaviour.(
–  Use(of(wideband(receivers:(
•  The(dynamic(range(of(the(received(power(is(wider(than(
cellular(environments.(
•  Strongly(affected(by(narrowband(interference.(
9(
Outline(
Learning/(
Knowledge(extrac7on(
Decision(
making(
Spectrum(
sensing(
Wireless(
transmi/er(
Wireless(
receiver(
Transmit)
Observe)
Frequency'
spectrum'
10(
NBI(in(wideband(HF(receivers(
11(
NBI(in(wideband(HF(receivers(
12(
NBI(in(wideband(HF(receivers(
Wideband(HF(transceiver(–(Receiver(diagram(block(
ADC
Band Pass
Filters 3-30MHz
RF Amp
Image-rejection
Filter 3-30MHz
BW = 1.0 MHz
112 MHz
≤ 20
dB
BW 1 MHz
10.7 MHz
≤ 75 dB
Low Pass
10.7 MHz
82-109 MHz
DDS
122.7 MHz
µController
CARDS12
AGC
RS232
We(must(detect(and(mi7gate(in(the(analog(domain(
13(
NBI(detec7on(in(wideband(HF(receivers(
Our(proposal(based(on(Compressive(Sensing(for(NBI(detec7on(
(
ADC(with(fs(<<(fNyq(
(
(
((A(parallel(system(to(the(wideband(receiver(with:(
(W(Detec7on(phase(
(W(Mi7ga7on(block(
14(
NBI(detec7on(in(wideband(HF(receivers(
Basics(of(Compressive(Sensing:(
y=Φx"
x =(Original(Signal(
(N(samples)(
Linear(Measurement(
Process(
Φ(
y =(Compressive(Measurements(
(M(measurements)(
M x N transforma7on(
(M < N)(
15(
NBI(detec7on(in(wideband(HF(receivers(
Basics(of(Compressive(Sensing:(
y=Φx"
Compressible(Signal(
Sparse(Signal(
x=Ψs"
Fourier(basis(
16(
NBI(detec7on(in(wideband(HF(receivers(
y =(Compressive(Measurements (
(
(K =(Sparsity(level)
(
Φ =(Measurement(Matrix
"
"
"
"s =(Sparse(Signal(
(
(
(
(Ψ =(Sparsity(Basis(
"
x =(Original(Signal(
"
(
(
17(
NBI(detec7on(in(wideband(HF(receivers(
Our(proposal(based(on(Compressive(Sensing(for(NBI(detec7on(
(
(
(
(
Unpublished(Results(
18(
Outline(
Learning/(
Knowledge(extrac7on(
Decision(
making(
Spectrum(
sensing(
Wireless(
transmi/er(
Wireless(
receiver(
Transmit)
Observe)
Frequency'
spectrum'
19(
Hidden(Markov(Model(
Hidden'Markov'Model'
Doubly( embedded( stochas7c( process( with( an( underlying(
stochas7c(process(that(is(not(observable((it(is(hidden),(but(it(can(
only( be( observed( through( another( set( of( stochas7c( processes(
that(produce(the(sequence(of(observa7ons.(
...
Box 1
Box 2
Box N
20(
Hidden(Markov(Model(
•  The(three(basic(problems(of(HMM(
–  The(evalua7on(problem:( ((
(Forward(step(of(ForwardWBackward(procedure(
–  The(decoding(problem:((
(Viterbi(algorithm(
–  The(learning(problem:((
(BaumWWelch(method(
21(
Data(segmenta7on(
1 minute
2 s.
1 0 0
...
...
1 0 0 1 1 1
3
...
1 1 1 0 1 0
2
3
1
1 1 0 0 0 0
...
...
0 0 0
1
10 minutes
22(
HF(Spectrum(Ac7vity(Predic7on(
•  Main(model:(
(Ergodic(HMM(
(10(minutes(sequences(
HMM1
•  3(submodels:(
(LeZWright(HMM(
HMM3
HMM2
(Observa7on(symbols(for(1(
(minute(
1
2
3
4
...
38
39
40
23(
HF(Spectrum(Ac7vity(Predic7on(
Observations
OT
T-1
t (min)
T
Submodel 1
λ1 = (A1,B1,π1)
Submodel 2
λ2 = (A2,B2,π2)
Submodel 3
λ3 = (A3,B3,π3)
P(OT|λ1)
P(OT|λ2)
P(OT|λ3)
B’11
B’22
B’33
Submodels
High-level model
B’11 0 0
B’ = 0 B’22 0
0 0 B’33
High-level model
λ = (A,B’,π)
New definition of
the high-level model:
max(P(OT+1|λ))
Predicted state
...
Previous
detected states
ST
T-2
T-1
T
t (min)
24(
HF(Spectrum(Ac7vity(Predic7on(
16
Average error rate (%)
14
12
10
8
6
Global performance
Normal activity
High activity
4
1
2
3
4
5
Acquired knowledge (min.)
6
7
8
25(
Outline(
Joint(work(at(Supélec(
Learning/(
Knowledge(extrac7on(
Decision(
making(
Spectrum(
sensing(
Wireless(
transmi/er(
Wireless(
receiver(
Transmit)
Observe)
Frequency'
spectrum'
26(
Decision(making(with(UCB(
Upper(Confidence(Bound((UCB)(algorithm(to(provide(HF(
secondary(users(with(dynamic(access(to(the(spectrum.(
Exploita7on(
Joint(work(at(Supélec(
Explora7on(
27(
Decision(making(with(UCB(
Unpublished(Results(
Joint(work(at(Supélec(
28(
Conclusion(
The(applica7on(of(Cogni7ve(Radio(to(HF(
communica7ons(might(be(feasible(
(
•  Both( Learning( with( HMM( and( decision( making( with(
UCB(can(help(secondary(HF(users(to(avoid(collisions(
with(other(users.(
•  A(compressive(detector(can(be(used(in(wideband(HF(
receivers(to(detect(NBI.(
29(
Thank(you(for(your(a/en7on!(
Cogni7ve(Radio(in(HF(Communica7ons(
Laura(Melián(Gu7érrez(
lmelian@ide7c.eu(
IDeTIC(
Universidad(de(Las(Palmas(de(Gran(Canaria,(Spain(
Rennes,(6th(November(2014(