Date post: | 25-Dec-2015 |
Category: |
Documents |
Upload: | gloria-joseph |
View: | 219 times |
Download: | 0 times |
1
Cognitive Radio Architecture Evolution
Karol SchoberBased on paper by Joseph Mitola, III, Senior Member IEEE
2
Introduction
CR prototype introduced 1998-1999 Exchanging of business cards (‘May, I introduce’) Intelligence involved
Past five years - SDR & CR under research concerning : spectrum allocation, market, bussiness and
open architecture Described is architecture for evolving heterogeneous
networks cellular merged with hot spots wireless technologies merged with human interface
technology
3
Outline
Cognitive Radio ArchitecturesArchitecture and Use Case EvolutionSensory Perception in the Evolving CRAQuality of information
4
SDR architecture
•speech
•internet access
•multimedia content
Radio architecture: framework by which products maybe integrated in to evolving sequence of designs with specific rules - public/proprietary
•GSM 900/1800
•CDMA
•Bluetooth
•PSTN….
Most of functionality can be
synthesized in software-based
chips like FPGA, Single-chip
Arrays and Blade servers
AD/DA
convertors are an issue
5
Complexity of SDR
To manage complexity object and layer oriented programming has been adopted
CR in addition to SDR is capable of Sensing the environment and fit content to user
6
Why are resources constrained?
Regulation is due to historical reasons Small/large bands were dedicated to public interest
with respect to economic aspect TV, 1st and 2nd generation cellular networks
Nowadays the radios are capable of multiband transmission and they are practically everywhere (ubiquity)
Heterogeneity of UE allows new way to go – Dynamic Spectrum Allocation
7
Dynamic Spectrum Access
Process of increasing spectrum efficiency employing real-time adjustment of radio resources A real-time spectrum auction between systems with different
purpose, e.g. cellular network( stolen spectrum during peak hours) public safety (more access points for public safety)
Countries treat CR differently TV spectrum in US, European conservativeness, EU direction
for secondary training, Ireland DySPAN 100MHz The ideal iCR is difficult to implement thus XG defined
Simplified CR, simple rule-engine that controls radios air interface to conform to spectrum used policies (policy language) according to Haykin
Near-Future
9
How is iCR different?
Self-awareness, user awareness and machine learning
CR prototype – Wake epoch
Sensory perceptions:
•RF
•Location
•Motion
•Temperature
•Vision
•Speech
•….
Planing technologies
To identify changes in RF scene (IMEC Belgium)
10
Epochs
Wake epoch see previous slide e.g. detection of new RF networks
Sleep epoch Computationally intensive pattern analyses, self-
organizing, autonomous learningPrayer epoch
interaction with higher authorities such as cognitive networks about restrictions , advice
11
Networking and CRA evolution
Neither DSA , not iCRA provide architecture for cognitive wireless networks (CWN’s)
No real-time spectrum auction (standardization necessary) Supportive distributed network Policy language Method of payment
12
Proposals – Fritzek and Katz
CWN CRA characterized by cooperation among intelligent entities
Cooperation considering : game theory, relays power allocation, diversity, cross-layer optimization, stability and security, distributed antennas, cooperative header compression, coding, distributed spatial channel control
13
Proposals- Osaka University
Biological systems –robustness to catastrophe Molecular processes, immune system, social
insects, prey-predator relations Posses : Membership perception, network awareness,
buffer management, message filtering
Strassner’s key issues with lingua francaPlacing the cognition to layers (e.g. to sense
congestions)
14
Architecture and Use case evolution
Media services and Internet changed the use case for Cellular networks
HOW? Product Differentiation
present Multimedia services competition User specific services
Protocol stack IPv4 should be adopted even if not ideal and later drift towards IPv6
OA&M Network management and administration based on self-awareness Agent-based evolved CRAs may assist to overcome misconfigurations
Location Awareness Not a “killer app” , but e.g. Google Maps, MapQuest will be able to move QoS to QoI
15
Continue
Spectrum Awareness before network ordered the UE what band to listen to now the awareness comes to play, switching between WIFI, 3G,
Bluetooth recognition of Emergency calls in D-band U.S 700Mhz
Spectrum auction recently, cross-licensing of small parts of band Jondral’s group: leasing of BW chunks for 5s for browsing and email
for few cents would increase Spectrum utilization by 15-25% (how about revenue?)
User expectations Users adjust their expectations and usage (low/high mobility, hotspot,
3G) Operators move towards femtocells Femtocells recognized by GPS(not indoors ), computer vision,
speech
16
The use case for Femto-cell handover
SAS hotel in Stockholm
The call are dropped when entering
GPS cannot say whether you are inside (glass foyer)
Camera would be a solution
17
First Responder Situation awareness
the prioritizing of transmission (resource management) should be based on users situation, e.g. specific location surrounding (smoke) Movement (trapped) intent (rescue someone, escape)
Radio should be aware of user’s physical setting when getting resources
18
Commercial Sentient Spaces
Confluence of technologies (3G, Wi-Fi, ….) Interference suppression Cognitive load balancing Cooperative power management
Elder/child care with speech and coputer vision Turning off stove Eating pills
Input of data Moving service from administrator to user (car-rental check-in)
Near future with IPv6 -> Users of wireless become devices rather than people
The smart homes are apart of radio, but can help to autonomously adjust radio resource priorities to changing needs of users.
Radio is as well mean to fast deploy the services Radio may be a control for large number of objects in smart-homes.
19
Sensory Perception in the Evolving CRA
the point is to identify aspect of sensory technologies for future cognitive radio
Computer vision Surveillance (fall in parking lot) Internet retrieval The Video scene API should assist to CR to determine proper
speed and priority (car accident) Human Language Translation
For unburdening the user from contact with HMI Can assist to identify user needs better (e.g. interact with
medical devices)
20
Computer speech
with Windows XP does not appear to be in wide use Precision [raw error rate]
3-10% in home office 25% 14.7% topic spotting
Speaker identification Background noise disturbing the model Thus only soft biometric measure (contribute to overall
Authorization )
21
Text Understanding
Used in Business intelligence markets (source WIKI) refers to skills, technologies, applications and practices used to help a business acquire a better
understanding of its commercial context. Business intelligence may also refer to the collected information itself.
Common functions of business intelligence technologies are reporting, OLAP, analytics, data mining, business performance management, benchmarking, text mining, and predictive analytics.
Causal relationship between unstructured customer contact reports (uCCR) Takes a lot of labor work mix word sense disambiguation, named entity detection,
sentence structure analysis Google depends on the laws of very large numbers
Typically there are small amount of uCCRs Android – statistical machine learning
22
Text Understanding -continue
Query tool based on ALICE – AI better in answering than Google
Unstructured comments in wireless networks service, maintenance record CR communities may analyze themselves, optimize
and save costs Functional description languages
23
Quality of Information
QoI=Quantity*Precision*Recall*Accuracy*Detail*Timeliness*Validity
Paremeters are Real [0..1] Are best at 1 Monotonic and approach 0 to degrade the QoI
Quantity No information for given situation, quantity is 0 Older information is better than no (maps) This was economically stupid, in future needed (congestion of
networks)
24
QoI continue
Precision and Recall Degree to which information correspond to user’s need Recall 1 = all relevant documents retrieved Precision 1 = no irrelevant documents retrieved user may provide feedback by rejecting or ignoring
Accuracy Spelling the president name wrong Numerical accuracy Dependence (quadratic, linear, exponential, fractal ….)
Timeliness User’s time when the information is to be employed If needed now: then 1/ε is a good measure There might be window when Timeliness fall to zero immediately (death ) Wake up call should not be 15minutes earlier
Detail If sufficient detail then information provided is complete (I get the directions to
Restaurant) Promoting User as the 8th layer
25
Challenges and opportunities
Some spectrums as GPS, UWB cannot be detected other than with correlation
Some bands are occupied 0.1% time and are necessary 100% (radar bands, airport) → noway to share
Emergency channels have to be clear to have maximal SNR possible (“Mayday” must be heard)
CRA should posses knowledge of forbidden bands Event though many paper exists on spectrum-
auctions and web-page rental of spectrum exists, there is no architecture deployed for real-time auctions