cognitive radio and networking research at virginia tech

1

S-88.4221 Postgraduate seminar on signal Processing I

Sami Kallioinen

Cognitive Radio and Networking Research at Virginia Tech

TKK, S-88.4221 Postgraduate seminar on signal Processing I Sami Kallioinen 4.12.20092

Outline

• This presentation is an overview of research work what has been done in Virginia Tech

• Focus

• Introduction and motivation

• Algorithms for reasoning and learning

• Supporting technologies

• Radio Platforms

• Security and Verification

• Cognitive Networking

• Game Theoretic Analysis

• Applications

• Conclusion


CR research work at Virginia Tech -Focus

• More than dozen Wireless @ Virginia Tech faculty are working to address the broad research agenda of cognitive radio (CR) and cognitive networks (CN)

• Core research team spans the protocol stack from radio and reconfigurable hardware to communications theory to the networking layer

• new analysis methods and the development of new software architectures and applications

• Core concepts and architectures underlying CR and CN

• Focusing on cognitive engine• They describe developments that support the cognitive engine and advances in a

cognitive radio technology that provides the flexibility desired in cognitive radio node

• Securing and verifying cognitive systems

• Cognitive paradigm up the protocol stack to optimize end-to-end network performance

• Analysis of cognitive systems using game theory

• Dynamic spectrum sharing and control of multiple-input multiple output


Introduction and Motivation

A cognitive radio (CR) is “a radio that is aware of its surrounding and adapts intelligently” (Mitola)

The need for CRs is motivated by many factors

- Complexity of the radio systems

- The existence of software define radio (SDR)

- Spectrum efficiency

Dynamic Spectrum Access (DSA) as an “killer application”


The Cognitive Core: Algorithms for Reasoning and Learning

• The CE is intelligent system behind a CR and combines sensing, learning, and optimization to control the radio

• This work provides a theoretical foundation for developing the optimization algorithms required to design waveforms to meet particular QoS requirements under a given set of environmental conditions

• Problem set by Rondeau: a multiple-objective optimization process that trades off objectives like bit error rate, data rate, and power consumption that measure radio performance

a foundation for analyzing CR systems

This work provides examples for using feedback, learning, and knowledge representation in the CE


The Cognitive Core: Algorithms for Reasoning and Learning• The CE is intelligent system behind a CR and combines sensing,

learning, and optimization to control the radio• Used concept of cognition cycle proposed by Mitola is shown in

figure as applied to physical layer waveform adaption• Outer loop is reasoning process• Inner loop is learning process


The Cognitive Core: Algorithms for Reasoning and Learning

• Figure presents how the CE fits into radio systems as a whole

Three extrinsic domains that impact the radio:

• The user domain provides performance requirements to the radio

• The policy domain constrains the CE to work within giver regulatory environment

• The RF environment provides the context in which the transceiver will operate

• Figure shows also a traditional communications stack on the right side


The Cognitive Core: Algorithms for Reasoning and Learning• Figure shows a prototypical CE

architecture• Central to this architecture is a cognitive

controller, the kernel and scheduler of the cognitive

• Other components of particular interest include:

Sensors, which collect and preprocess environmental data

The decision maker, which stores past experience and seed the optimization process

Optimizers, which seek to develop optimized solutions to the problem currently posed by the user’s requirements, the constraints of policy, and the radio environment by building on past experience


The Cognitive Core: Algorithms for Reasoning and LearningGenetic Algorithm Optimizer

• Optimization of wireless system parameters is fundamentally a multiple-objective problem

• A key problem with using GAs as an optimizer is that they are slow, requires thousands of generations to converge

Problem is solved by seeding the GA’s initial population with a carefully chosen set of solutions that correspond to promising areas of the search space

Case-Based Decision Maker• Technique operates on a database of cases, in which each case consist of

a problem faced, an action taken, and a result

• Compare new problem, it uses a similarity function in the database


The Cognitive Core: Algorithms for Reasoning and LearningCognitive Engine Experiments

• Rondeau demonstrated the performance of the prototypical CE both through simulations and in over-the-air experiments with real radios. He performed key set of the over-the-air experiments durin IEEE DySPAN in April 2007


Supporting Technologies

1. Spectrum Sensing

• Key application for the development of CR systems is the promise of DSS, which could lead to more efficient spectrum usage

• They have defined Spectrum Sensing as the combination of signal detection and modulation classification and use the general term Automatic Modulation Classification (AMC)

• Distributed approach to cyclic feature-based AMC in which spectrum sensing is performed collaboratively by a network of radios

• Consist of AMC and DM stages


Supporting Technologies

2. Radio Environment Map (REM)

• A distinctive characteristic of CRs and CNs is theirs capability of making decisions and adaptations based on past experience, on current operational conditions, and also possibility on future behavior predictions

1) Link-level simulations• Consider a scenario in which a CR,

following a random waypoint mobility model, moves through a stationary PU network spread over a circular region

2) Network-Level Simulations• 20 CRs are moving along the streets and

20 PU nodes are stationary and clustered at the street crossing


Radio Platform Issues

• Large amount of fractional frequency bands (“frequency-agility required”)

• Have to operate multiple bands simultaneously• Wish to use one ore more RFC chains to search “white space”

• The ability of the radio platform to adapt dynamically and nearly instantaneously

A. Dynamic Assembly of Radio Structures

1) Module Placement and Relocation

2) Dynamic Module Library Preparation

B. Radio-Frequency and Mixed-Signal Integrated Circuit Design

C. Building Multiband Radios



A. Dynamic Assembly of Radio Structures

Need to adapt to new and unforeseen conditions

• FPGAs offer a potential solution for realizing custom signal-processing pipelines• Difficulties in reconfiguration in a

embedded environment

• Alternative is to synthesize the anticipated signal processing structures in advance so that they can be instanced on demand within the radio platform

1) Module Placement and Relocation

2) Dynamic Module Library Preparation



B. Radio-Frequency and Mixed-Signal Integrated Circuit Design

• RF/mixed-signal IC design• Receiver architectures currently under

consideration for CRa) Zero intermediate frequency (IF) (Direct

Conversion) receiversb) RF bandpass sampling receivers

• Trend is to going to systems where ADC structure is “build-in” to the RF front-end• RF structures are also more selective Reduced filtering requirements• Continuous time delta-sigma has an anti-alias filter

feature separate BB filter are not requires less power consumption



RF bandpass sampling receivers

• To reduce the ADC/DSP power consumption requirements of RF bandpass sampling architecture for CRSignal-process functions may be shifted to the analog domainTo reduce the information conversion load on the receiver ADCs, the multiplication-intensive FFT function is relocated ahead of the ADC into the discrete time analog domain

• A prototype has been demonstrated



Direct conversion set challenges for LO in multi band system

• Generation of LO frequencies with acceptable phase noise/spur performance

• Range 500MHz-5GHzPresented synthesizer provides an fast settling time

(<10ns) (Originally designed for UWB)

• Disadvantages: spurious, no multiple outputs

Related work done in TKK/Nokia: • A Digital Frequency Synthesizer for Cognitive Radio

Spectrum Sensing Applications, Tapio Rapinoja et al.

• Digital period synthesizer (DPS)

• No such a spurious problem

• Frequency can be adjust very accuracy (10Hz)

• Trend-setter • Enabler for dynamic spectrum sensing



C. Building Multiband Radios•Soon single-chip CMOS direct conversion transceivers with sufficient performance and bandwidth almost for any wireless application

•They have been collaboration with Motorola (CMOS SDR RFIC), basis for a prototype simultaneously-multiband public safety radio

•Unsolved problem of multiband radios is antennas

•Focus on their current research is to develop multiband mobile radios using the same type of monopole antennas currently in common use, with performance comparable to existing single- and dual-band radios


Security and Verification of Cognitive SystemSecurity

• CRs and CNs requires robust security mechanism to resist misuse

• Several security issues were investigated


Security and Verification of Cognitive SystemVerification

• Lack of security and reliability in the underlying software that serves an the command and control for the radio system

• Poorly tested and verified code• Malicious inputs

• Buffer overflows

• A verification approach using aggressive program slicing and a proof-based abstraction-refinement strategy


Cognitive Networking

• CRs are expected to deliver seamless adaption, opportunistic use of underutilized spectrum, and increased flexibility in modulation and waveform selection to better fit to current wireless environment

• A CN in a network that is capable of intelligently optimizing the end-to-end performance of a network

• Fundamental aspects: learning and reasoning


Cognitive Networking

• An architecture for CNs


Game Theoretic Analysis of Cognitive Radios and Cognitive Networks• Game theory is a set of analytical tools used for analyzing the interactions of

autonomous agents

• They have recently surveyed the application of these techniques to the analysis of wireless ad hoc networks (power control, waveform adaption, medium access control, routing, and packet forwarding)

• Work in applying game theory to problems in wireless networks usually focused on 1. Showing the existence (and, in some cases, uniqueness) of Nash equilibria for a

given game model

2. Demonstrating an algorithm that will converge to a Nash equilibrium

3. Either showing that the Nash equilibrium (or equilibria) is reasonably efficient of providing a mechanism to entice players to move to a more efficient operating point

• They have reviewed two application:• Analysis of Power and Coding Adoptions

• Interference Avoidance


Game Theoretic Analysis of Cognitive Radios and Cognitive Networks• Analysis of Power and Coding Adoptions

• A key aspect of waveform is the selection of a basic adaption criteria


Game Theoretic Analysis of Cognitive Radios and Cognitive Networks• Interference Avoidance

• In many envisioned scenarios, nodes occupy the same geographical area and can cause significant interference to each other


Applications

• Dynamic Spectrum Access

• Cognitive MIMO

• Parallel SISO


Applications

• Cognitive MIMO


Applications

• Parallel SISO is an another means of utilizing antennas in a CR

• Multiple parallel information streams in different frequency band

• Reduced spectral efficient, but however improves performance

• An important consideration in constant rate applications

• An example is shown in figure:• Parallel SISO approach provides 1-3

dB performance improvement

=>Loss is spectral efficiency


Applications

• Dynamic Spectrum Access (DSA)• O the many application of CR, the most recognized is DSA

• Two basic categories: open access and hierarchical access

• They work differs from existing work in that they do not focus on specific algorithms for spectrum sharing, but rather the fundamental difference between hierarchical approaches: namely, spectral overlay based on interference avoidance and spectrum underlay based on interference averaging


Conclusion and future research

• They have demonstrated the breath of approaches required to address the challenges of CRs and CNs

• To archive the promise of CRs and CNs, researchers with a wide range of expertise must work together to create complex, coordinate systems

• Autonomous radio and network nodes • Cognitive system has the application-specific artificial intelligence

challenges

• In addition radio HW, algorithmic, networking and analysis challenges

• Reconfigurable, multiband, multimode radio HW

• Game theory for understand interaction


Conclusion and future research

• Still much to Do• Interface standardization between peaces of cognitive system

• Development of simulation and emulation tools is very hard• Cross layer nature of CRs and CNs

• A lack of progressing analytical methods makes difficult to clearly understand the engineering tradeoffs

• Robust, low cost and flexible SDR and SW adaptable network platforms are desperately needed to assure continued research progress

• They will continue research work for the creation of a CR and CN testbed

• Nodes will be located to the Virginia Tech campus area

• Each node based on universal software radio peripheral• Include an RF front-end built around Motorola chip

• Remotely configured

• GNU radio and the Open source SCA Implementation-Embedded (OSSIE)

cognitive radio and networking research at virginia tech

Documents

cognitive radio technology

radio performance

cognitive enginethey

cognitive radio node

dd initialsthe cognitive

dd initialscr research

cr systemsthis work

optimization algorithms