huseyin arslan are we ready for 5g vision? what is next?
TRANSCRIPT
Are we ready for 5G vision? What is next?
Flexible and Cognitive RAT for 5G and Beyond
HUSEYIN ARSLAN
http://cosinc.medipol.edu.tr
http://wcsp.eng.usf.edu
Istanbul Medipol University University of South Florida
Bs
Middle East Technical University, Turkey
1987-1992 1992-1993 1993-1998
1998-2002 2002-present
Ms and PhD
Southern Methodist University, Dallas, Tx
ERICSSONResearch Engineer
MsMiddle East Technical University,
Turkey
University of South Florida, Tampa, Florida
2014: Prof.
2014-present
Medipol University,Istanbul, Turkey
ANRITSU
Consulting Engineer
TubitakConsulting Engineer
OUTLINEp Evolution of wireless p 5G and Beyond (Flexibility and adaptability)p Flexibility framework
n Sensing and radio environment monitoringn Cognitive radio and networks (AI and ML)n PHY security and wireless security
p Flexible RAT (PHY and MAC features)n Intelligent Surfacesn Flexible waveforms and frames (hybrid and joint waveforms)
p Integrated sensing and communication
n Flexibility in modulationn Flexible interference management (co-existence and coordination)n Flexible coverage and base stations (aerial base stations) n Flexible band and spectrum usage
Evolution of wirelessp BW
n Increasing (from 30kHz, 200kHz, 1.25MHz, 5MHz, 20MHz, 100MHz, …)
n Getting flexible (fixed earlier, aggregation is now real, dynamic spectrum access (DSA) further along the line). Fractional BW
p Spectrum (started below 1 GHz, later included above 1GHz, now mm-wave frequencies, possibly terahertz and visible light communication in the future)
p Spectrum Efficiencyn Increased number of bits per Hertz
n Re-use (cell size is getting smaller, re-use factor is getting smaller)
n HetNets and Co-existence
p Adaptation and cognition capabilities (ML and AI)p Awareness capabilities
Wireless Evolution p 1G: Cellularp 2G: Digital
p 3G: Data (first attempt to MBB)p 4G: High Speed Data (true MBB)
p 5G: Applications (verticals)p 6G: Flexibility (AI and sensing)=
CR and CN
VOİCE
VOİCE&
DATA
CAPACITY (re-use)
CAPACITY&
SPEED
Diverse applicationclasses
Diverse requirementsand metrics
Single Layer (cells)Single connection pointHomogeneous Nets.MobilityCoveragePersonal Com.
1G(1980s)
Analog
3 G(2000s)
2 G(1990s)
Digital
4 G(2010s)
5 G(2020s)
1) Extreme high traffic2) Extreme variable traffic
Adaptation and flexibilityp 1G and 2G
n Tx : heavily standard defined, not much flexibilityn Channel and Interference: random n Rx: flexible (channel adaptive) - adaptive equalizer, adaptive
channel trackers, etc.p 3G and 4G
n Tx : heavily standard defined but some flexibility (adaptive scheduling, adaptive modulation and coding, MIMO and BF, etc.)
n Channel : random Interference: random (unsuccessful CoMP)n Rx: more flexible
p 5Gn Tx :flexible, standard defines general and essential featuresn Channel: random Interference: controlled (CRAN, BF, mMIMO) n Rx: even more flexible
p 6Gn Tx : extremely flexiblen Channel: controlled (IRS) Interference: controlled (CRAN, IRS,
BF, mMIMO)n Rx: extremely flexible
5G is aimed to provide service for wide variety of applications
with some Flexibility and Adaptation
Applications (diverse and with a range of requirements)
Smart transportation Remote Health
Remote education
Smart Grid
11
IoT
5G AND BEYOND RADIO ACCESS TECHNOLOGIES
LATENCY RELIABILITY CAPACITY SPECTRUM EFFICIENCY
URLLC mMTC eMBB
RE
QU
IRE
ME
NT
SM
ET
RIC
SA
PP
LIC
AT
ION
SR
ES
OU
RC
ES
C
ON
ST
RA
INT
S
SPECTRUMPROCESSING POWER
COST ENERGY
HARDWARE
ENVİRONMENT AND CHANNEL
MIMO & BEAM-FORMINGMASSIVE MIMO
ADVANCED WAVEFORMSENSİNG SMALL CELLS
MM-WAVE COGNITIVE RADIODSA
CRAN
q High data rate & capacity
q Better coverage
q Improved spectral efficiency
q High reliability
q Low latency
q Strong security
q Very large number of devices
q High energy efficiency
q Low device complexity
A flexible air interface is required to meet these diverse requirements!
Metrics and technical expectationsp Capacityp Spectral efficiencyp Energy Efficiency, Power (PAPR, CCDF)p Complexity (computational, hardware)p Cost (device and reduced bit cost)p Accessibility (call blocking, service blocking)p Improved QoS p Data rate (peak rate, average rate, cell edge rate, min. rate)p Coverage (cell edge)p Latency – (no perceived delay)p Reliability (99.999 percent)p Flexibility (Support of variety of services )p Mobilityp SECURITY (Communication security in PHY, MAC, NET)
6G ProjectionsINTELLIGENCE + COMMUNICATION
AI and machine learning for cognitive radio and cognitive networks
Sensing & Communication
Flexible Adaptive Dynamical
Scalable Cognitive Intelligent
Softwarization Degree of Freedom Tunable
Configurable
Adjustable
Programmable
• Security and Privacy• Co-existence and interference management (and dynamic
spectrum access)• Sensing, radar, joint sensing and communications (precision
sensing and actuation)• Channel control through IRS, backscatter communication, etc• Cell-less networks (edge-less networks), coordination, C-RAN
(multi-connectivity)• Massive MIMO, holographic MIMO • Flexible PHY, MAC, and networks (flexible waveforms,
modulations, radios, channel…) • Edge AI (distributed intelligence), smart environments• THz (beyond mmwave), and VLC• Quantum communication• Energy efficiency and energy harvesting
6G keywords (technical)
Expectation from 6G (applications)
• Society and human centric networks and systems (serving society and well-being of people, including disaster, serving remote area)
• Human-machine and human-environment interaction (interconnection of physical, biological, and digital worlds)
• Expansion of communication environment:High altitude platforms, beyond terrestrial coverage – integrated terrestrial and space
• Communication everywhere with everything (IIoT, education, health, agriculture, grids, traffic, city, Industrial internet … more verticals)
• Fully automated driving, internet of bio-things
• Virtual reality, extended reality, augmented reality (holographic verticals and society)
SENSINGIntegratedsensing and
communications
AI and ML
Applications Requirements Requirement SetsService Types
FLEXIBLE RAT PLATFORM
• Flexible multi-band utilization
• Super flexible PHY and MAC
• Super flexible heterogeneous networks
• Co-existence
• Intelligent transmission
• Green communications
• Channel Control (IRS)
PHY Security&
Securecommunication
Learning and reasoning
Adaptation
Cognitive Radio &
Networks
PHY Security&
Secure communication
SENSINGIntegratedsensing and
communications
Adaptation
FLEXIBLE RAT PLATFORM
• Flexible multi-band utilization
• Super flexible PHY and MAC
• Super flexible heterogeneous networks
• Co-existence
• Intelligent transmission
• Green communications
• Channel Control (IRS)
AI and ML
Learning and reasoning
Cognitive Radio &
Networks
Cognitive and Software Defined RadioCognitive radio (highly intelligent radio systems)
Requiring a flexible platform: Flexibility
COGNITIVE CYCLE
Awareness
Sensing
Learning
Soft. Def. Radio
Cog. Radio
Cog. Net.Response
Adaptation
Input (sensing, awareness)Decision making (reasoning, interpretation, learning) Action (implementation of decision, adaptation, parameter change)
•CR requires a flexible radio device. SDRs are ideal platform for CRs.•SDR might not be necessary for CR, but, highly desirable•SCR might possibly do the job for most interpretation of CR•Even multiband and multimode radio devices can do the job for some CR interpretations
Two aspects of CRp Dynamic spectrum utilizationp Fully adaptive and reconfigurable RAN
p Both are requiring user/channel/interference/context awareness
NETWORK LAYER
Demodulator
RF Front End
RSSI
SNRSIR
SINRChannel
Noise Power
Channel Decoder
BER
FERCRC
Packet LossRouting Table Change Rate
Congestion LevelPositions of Nodes
Power Level of Nodes
TRANSPORT LAYER RTT
PHYS
ICAL
& D
ATA
LIN
K LA
YERS
UPPERLAYERS Perceptual Quality
MARKOV MODELSNEURAL
NETWORKSGENETIC
ALGORITHMS
...TO
OLS
LAYE
R IN
PUT
MEMORY
COGNITIVE INTERFACE
DESCRIPTIVE LANGUAGE
EXTE
RNAL
SE
NSI
NG
COGNITIVE ENGINE
ANTENNA
§ Antenna Powers§ Dynamic Range§ Pre–distortion
Parameter§ Pre–equalization
Parameter
RF Front End
§ Transmit Power§ Digital Modulation
Order§ Carrier Frequency§ Operation Bandwidth§ Processing Gain§ Duty Cycle§ Waveform§ Pulse Shaping Filter
Type§ FFT Size (for OFDM)§ Cyclic Prefix Size (for
OFDM)
PHYSICAL LAYER
§ Channel Coding Rate§ Channel Coding Type§ Packet Size§ Packet Type§ Data Rate§ Interleaving Depth§ Channel/Slot Allocation§ Carrier Allocation (in
multi–carrier systems)§ MAC Scheduling
Algorithm§ Hand-off (Handover)§ Number of Slots
DATA LINK LAYER
§ Routing Algorithm§ Routing Metric§ Clustering Parameters§ Network Scheduling
Algorithm
NETWORK LAYER
§ Congestion Control Parameters
§ Rate Control Parameters
TRANSPORT LAYER
§ Communication Modes (Simplex, Duplex, etc.)
§ Source Coding§ Encryption§ Service Personalization
UPPER LAYERS
LAYE
R O
UTP
UT
ADJU
STABLE PARAMETERS
OBSERVABLE PARAM
ETERS
Con
cept
ual M
odel
of C
hann
el A
war
e Ra
dios
Adjustable parameters - FLEXIBILITYControllable parameters:-Modulation options (types and orders) -Power options, transmitted power-Coding options (types and orders)-Waveform options (lattice, shape, etc)-Multiple accessing options-Antenna usage (MIMO, beamforming, precoding)-Scheduling options (types, criteria)-User-cell association options-Carrier Frequency options (mmwave, microwave, THz …)-Bandwidth options-Frequency use options (dynamic, secondary, primary, unlisenced, shared lisenced, etc…)-Access points (like CoMP, how many and which one)-Processing gain-Data rate
Adjustable parameters define: - what we transmit, how we transmit in different layers of the protocol stack- how many different ways we can transmit and receive- how we use the spectrum, power, hardware- how we generate the digital baseband signal- various ways of multiplexing, multiple accessing- how to use resources, how to share resources- how to build the physical signal- how to control the medium (LIS, RIS)- how to use the hardware (antennas, RF)- how often to adjust
COGNITIVE CYCLE
Awareness
Sensing
Learning
Soft. Def. Radio
Cog. Radio
Cog. Net.Response
Adaptation
Input (sensing, awareness)Decision making (reasoning, interpretation, learning)Action (implementation of decision, adaptation, parameter change)
•CR requires a flexible radio device. SDRs are ideal platform for CRs.•SDR might not be necessary for CR, but, highly desirable•SCR might possibly do the job for most interpretation of CR•Even multiband and multimode radio devices can do the job for some CR interpretations
p 5G is somewhat flexible and adaptivep But, still not Cognitive
p 5G + Machine Learning = Cognitive Radio = maybe 6G
Are We Ready for Machine Learning?
Starting of5G Research Starting of
5G Standardization Starting of6G Research
Data: Google Trends, worldwide interest over time
When to use Model and AI-based Approaches?
Hybrid Approach
Model-Based Approach AI-Based Approach
Model is already well-known
Simple problem to solve
Not enough data
Noisy data
Power and computational complexity is important
To merge and avail advantage of both
approaches
System is completely blind
Problem is too complex to solve
Big data is available
Clean data
Complexity is not critical
Applications of AI and Wireless Communication
Physical Layer
Blind signal identification
Modulation identification
Dynamic spectrum access
RF-impairments detection,
compensation
Channel estimation and prediction
Channel modelling
Channel coding
Symbol detection
End-to-end learning
NOMA techniques
User detection
Signal classification
Blind signal analysis
MAC Layer
Channel access mechanism
Quality of service optimization
Scheduling
Mobility management
Hand-off/handover techniques
Routing
BS Switching
Resource management
User association
Cell-sectorization
Energy optimization
Network Layer
Channel Estimation with AI
M. A. Aygül, M. Nazzal, A. Görcin, H. Arslan, “Sparsifying Dictionary Learning for Beamspace Channel Representation and Estimation in Millimeter-Wave Massive MIMO,” submitted to IEEE Transactions on Vehicular Technology.
M. Nazzal, M. A. Aygül, A. Görcin, H. Arslan, “Estimating the Unknown Sparsity in Multiple Dimensions to Realize Compressive Channel Estimation: AMachine Learning Approach,” submitted to IEEE Transactions on Vehicular Technology.
M. Nazzal, M. A. Aygül, A. Görcin, H. Arslan, “Dictionary Learning-Based Beamspace Channel Estimation in Millimeter-Wave Massive MIMO Systemswith a Lens Antenna Array,” IEEE International Wireless Communications and Mobile Computing Conference (IWCMC), Jun. 24-28, 2019.
M. Nazzal, M. A. Aygül, A. Görcin, H. Arslan, “Sparse Coding for Transform Domain-Based Sparse OFDM Channel Estimation,” IEEE SignalProcessing and Communications Applications Conference (SIU), Apr. 24-26, 2019.
Spectrum Sensing and Prediction with AI
M. A. Aygül, M. Nazzal, M. İ. Sağlam, A. R. Ekti, A. Görçin, D. B. da Costa, H. F. Ateş, H. Arslan “Spectrum Occupancy Prediction Exploiting Multi-Dimensional Correlations Through Composite LSTMs,” submitted to Sensors.
M. A. Aygül, M. Nazzal, A. R. Ekti, A. Görçin, D. B. da Costa, H. F. Ateş, H. Arslan “Spectrum Occupancy Prediction Exploiting Time and FrequencyCorrelations Through 2D-LSTM,” IEEE Vehicular Technology Conference (VTC-Spring), May 25-28, 2020.
M. Nazzal, A. R. Ekti, A. Görçin, H. Arslan, “Exploiting Sparsity Recovery for Compressive Spectrum Sensing: A Machine Learning Approach,” IEEEAccess, 7(1), 126098-126110, Sep. 2019.
M. Nazzal, O. Hasekioglu, A. R. Ekti, A. Gorcin, H. Arslan, “Compressed Spectrum Sensing Using Sparse Recovery Convergence Patterns ThroughMachine Learning Classification,” IEEE International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Sep. 8-11, 2019.
Physical Layer Security with AI
M. A. Aygül, S. B. Edibali, D. B.da Costa, H. F. Ateş, H. Arslan “Signal Relation-Based Physical Layer Authentication,” IEEE International Conferenceon Communications (ICC), June 7-11, 2020.
H. M. Furqan, M. A. Aygül, M. Nazzal, H. Arslan, “Primary User Emulation and Jamming Attack Detection in Cognitive Radio via Sparse Coding,”EURASIP Journal on Wireless Communications and Networking, 2020(141), 1-19, July 2020.
M. A. Aygül, H. M. Furqan, M. Nazzal, H. Arslan, “Deep Learning-Assisted Detection of PUE and Jamming Attacks in Cognitive Radio Systems,” IEEEVehicular Technology Conference (VTC-Fall), Oct. 4-7, 2020.
Adversarial Attack
Eavesdropping Spoofing Jamming and PUEA
Waveform Parameter Selection with AI
A. Yazar, H. Arslan, “A Waveform Parameter Assignment Framework for 6G with the Role of Machine Learning,” IEEE Open Journal of VehicularTechnology, 1(1), 156-172, May 2020.
A. Yazar, H. Arslan, “Selection of Waveform Parameters Using Machine Learning for 5G and Beyond,” IEEE International Symposium on Personal,Indoor, and Mobile Radio Communications (PIMRC), Sep. 8-11, 2019.