On the development of a Cognitive Radio Network Simulator based on - - PowerPoint PPT Presentation

Rome, October 9-11, 2013

1 4th Workshop of COST Action IC0902 “Cognitive Radio and Networking for Cooperative Coexistence of Heterogeneous Wireless Networks”

October 9–11, 2013 Rome, Italy

Contribution to Working Group 2 – “Definition of cooperation-based cognitive algorithms, that take advantage of information exchange at a local level”

“On the development of a Cognitive Radio Network Simulator based on OMNeT++/MiXiM”

Giuseppe Caso

• Prof. Luca De Nardis

DIET Department Sapienza University of Rome Rome, Italy Email: {caso, lucadn}@newyork.ing.uniroma1.it

Outline

2

• 1. OMNeT++: Objective Modular Network Testbed in C++;
• 2. MiXiM: the MiXed SiMulator;
• 3. From a “simple device” to a Spectrum Sensing Cognitive Radio

(SSCR);

• 4. Simulation Framework & Results;
• 5. Conclusions & Future Works.

Outline Rome, October 9-11, 2013

OMNeT++ 1/2

3

— Objective Modular Network Testbed in C++ (

http://www.omnetpp.org/);

— Open source tool for discrete event simulations; — The simulator can be used for:

¡ Traffic modeling of telecommunication networks; ¡ Protocol modeling; ¡ Modeling queuing networks; ¡ Modeling multiprocessors and other distributed hardware; ¡ Modeling of scenarios when several entities behave independently and the

• utput of every entity must be analyzed.

— OMNeT++ allows for detailed simulation of the reciprocal effect of

independent communications.

OMNeT++ Rome, October 9-11, 2013

4

—

OMNeT++ represents a “framework approach”:

¡ Instead of containing explicit and hardwired support for computer networks

• r other areas, it provides an infrastructure for writing such simulations;

¡ Specific application areas are catered by various simulation models and

frameworks, most of them open source;

¡ These models are developed completely independently of OMNeT++, and

follow their own release cycles.

—

Partial list of OMNeT++-based network simulators and simulation frameworks:

¡

Mobility Framework: for mobile and wireless simulations;

¡

INET Framework: for wired and wireless TCP/IP based simulations;

¡

Castalia: for wireless sensor networks;

¡

MiXiM: for mobile and wireless simulations.

OMNeT++ 2/2

OMNeT++ Rome, October 9-11, 2013

MiXiM Project

5

—

MiXiM is a merger of several OMNeT++ frameworks written to support mobile and wireless simulations;

—

The predecessors of MiXiM are:

¡ ChSim by Universität Paderborn; ¡ Mac Simulator by Technische Universiteit Delft; ¡ Mobility Framework by Technische Universität Berlin, Telecommunication

Networks Group;

¡ Positif Framework by Technische Universiteit Delft.

— It provides detailed models of the wireless channel and connectivity,

mobility and obstacles models, and many communication protocols, especially at the MAC level;

— It provides a user-friendly graphical representation of wireless and

mobile networks, supporting debugging and defining even complex scenarios.

MiXiM: Introduction Rome, October 9-11, 2013

MiXiM: General Structure

6

• “World” utility module provides global parameters of the environment

(size, 2D or 3D).

• “Objects” are used to model the environment;
• They influence radio signals and the mobility of other objects.
• “ObjectManager” decides which objects are interfering.
• “ConnectionManager” module dynamically manages connections

between objects: signal quality based on interference and on mobility.

• “Nodes” modules simulate entities desiring to communicate;
• Different kinds of nodes can be specified with different characteristics.

MiXiM: Introduction Rome, October 9-11, 2013

Node Modules 1/2

7

— Adjacent layers are connected by

two pairs of OMNeT++ “gates”:

¡ 1st pair: to exchange up and down

data messages, including control messages between nodes;

¡ 2nd pair: to exchange control

messages between the layers.

— The PHY layer is the place to take

care of the reception and collision handling;

— Physical and MAC layer design is

usually tightly coupled and very specific for different communication techniques.

Figure (a). Example of a node

• Submods according to the IP

model:

• Application Layer (appl);
• Network Layer (netw);
• MAC Layer (mac);
• Physical Layer (phy).
• Physical + MAC layer = Network

Interface Card (NIC) module (Figure

(b)).

MiXiM: Node Module Rome, October 9-11, 2013

Node Modules 2/2

8

— “Mobility” module: responsible for the movements of an entity.

Different mobility models can be implemented in MiXiM;

— “Battery” module: used for energy related issues; — “Arp” module handles the Address Resolution Protocol (ARP); — “Utility” module: derived from the Mobility Framework “Blackboard”

• module. Two main tasks:

¡ It provides a general interface for collecting statistical data of a simulation

that has minimal impact on the performance;

¡ It maintains parameters that need to be accessed by more than one

module within a node (e.g., the position of a node, calculated and updated by the mobility module, but also needed by the physical layer).

MiXiM: Node Module Rome, October 9-11, 2013

From a MiXiM-based to a Cognitive Radio Scenario

9

Primary User vs. Secondary and Cognitive Users case of study:

• 1. Introduction of a particular Host simulating the Primary User;
• 2. Introduction of the Secondary Network.

MiXiM-based Cognitive Radio Network Rome, October 9-11, 2013

10

Introduction of peculiar cognitive functionalities in the OSI stack of a generic device:

• 1. (Cooperative) Spectrum Sensing;
• 2. Clustering;
• 3. Multi-Channel possibility;
• 4. …

From a MiXiM device to a Cognitive Radio

MiXiM-based Cognitive Radio Node Rome, October 9-11, 2013

11

—

The simulator was already used in network organization algorithms design and performance evaluation, since it is possible to define different scenarios by adjusting key parameters as PUs behavior, cognitive network features (number of SUs, Energy Detector-based Spectrum Sensing modes, mobility models), operating frequencies (also DVB-T), medium access modalities, propagation channel models, network types and topologies, and so on.

—

Comparative analysis and the so-far obtained results regarding, in particular,

¡

Spectrum sensing performance (both local and centralized cooperative with different hard decision fusion rules);

¡

The impact of mobility and spatio-temporal correlation;

¡

The introduction of cluster-based solutions for the network organization and nodes selection;

¡

The achievable data throughput of the cognitive network in particular contention scenarios.

CRN Simulator: Implementation and Use

Simulation Settings, Results and Discussions Rome, October 9-11, 2013

12

—

DVB-T Transmitter (PU):

v

Located in the top left corner of a square area of 10×10 km2;

v

Tx Power = 200 kW, within a DVB-T 8 MHz channel in the UHF band.

—

A set of SUs forming a CRN:

v

Located at the lower right area of the playground, within a 700×700 m2 area, centered

• n the position of the FC;

v

The SUs are equipped with a data interface used to sense the PU channel and eventually transmit data packets, and a control interface working on a common dedicated channel;

v

Tx Power = 110 mW.

—

When mobility is assumed:

v

the SUs are allowed to move within the working area using a Gauss-Markov mobility model with an average speed v = [5 10 15 20] m/s.

—

Simulation Run Settings:

v

1h of simulated time, during which each collaborating SU takes a local decision exploiting a sensing phase of T = 50µs and then transmits its decision to the FC during the subsequent exchange phase of 1s. Finally, a global decision is taken by the FC each 5s.

Simulation Settings: an example

Simulation Settings, Results and Discussions Rome, October 9-11, 2013

13

—

Achievable Qd (CFAR mode) as a function of the requested target Qfa and for different values of γ, in a CSS with Majority rule scenario [2].

Simulation (OMNeT++/MiXiM) Results:

Validity of the proposed approximation for Majority Rule

1. A good match between the analytical and the simulation results is obtained, confirming the reliability of the ongoing OMNeT++ based CRN simulator and, in particular, the validity of the proposed approximations. 2. A perfect knowledge about the number

• f collaborating SUs (N = 25) and the

average SNR (γ = [0 : 5 : 30]dB) is assumed. 3. As expected, for low γ values, the CRN should not set too stringent targets as this actually leads to worse global cooperative performance. For high γ values the performance greatly increase.

Simulation Settings, Results and Discussions Rome, October 9-11, 2013

14

—

Achievable Qd for CSS with Majority fusion rule, as a function of the CFAR target Qfa and the number of SUs, for schemes without and with nodes selection [1].

Simulation (OMNeT++/MiXiM) Results: Nodes selection framework

Simulation Settings, Results and Discussions Rome, October 9-11, 2013

15

1 2 3 4 5 6 7 8 9 Non-Clustered Model MOBIC Model SENSIC Model Offered Traffic [pkt/s] 3.54 6.53 8.17 Throughput [pkt/s] 3.3 5.53 7.6

SUs Data Throughput – Static Case

Simulation (OMNeT++/MiXiM) Results:

Study of CRNs Data Throughput

Simulation Settings, Results and Discussions Rome, October 9-11, 2013

• Achievable Data Throughput for different CRNs with non-cluster/cluster-based

CSS with Majority fusion rule [3].

16

—

Analysis of CSS schemes under CFAR and CDR operative modes, adopting OR, AND and Majority decision rules.

—

Proposal of novel approximations for the network probability of false alarm (CFAR) and detection (CDR) as a function of individual probabilities, allowing to derive individual performance requirements.

—

Study of the impact of spatio-temporal correlation and mobility on CSS under the CFAR

• perating mode.

—

Proposal of a novel framework for nodes selection, based on the Moran’s I statistical index.

—

Simulation results show that the proposed scheme achieves sensing performance comparable to CSS relying on all network nodes while only involving a reduced number of SUs. Results confirm the idea that when correlation is taken in account, efficient CSS schemes can be defined based on the selection of a subset of SUs.

—

A preliminary assessment of the impact of introducing a quasi-realistic mobility model was carried out. Simulation results, show the impact of more realistic mobility models with respect to results presented in previous works relying on simpler mobility and channel modeling, calling for further studies on the role of mobility in cooperative spectrum sensing.

Conclusions & Future Works

Conclusions & Future Works Rome, October 9-11, 2013

17

Thank you for your attention! Q & A

References: [1] Caso, G.; De Nardis, L.; Holland, O.; Di Benedetto, M.-G.; “Impact of Spatio-Temporal Correlation in Cooperative Spectrum Sensing for Mobile Cognitive Radio Networks”, 10th International Symposium on Wireless Communication Systems (ISWCS’13), Workshop CRAFT’13, Ilmenau, Germania, 27-30/08/2013; [2] Caso, G.; De Nardis, L.; Ferrante, G.C.; Di Benedetto, M.-G.; “Cooperative Spectrum Sensing based on Majority decision under CFAR and CDR constraints”, 25th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’13), Workshop MACNET’13, Londra, UK, 8-11/09/2013; [3] Caso, G.; De Nardis, L.; Perez-Guirao, M.D.; “Cooperative Spectrum Sensing and Network Organization in Mobile Cognitive Radio Networks”, 3rd COST Action IC0902 Workshop, Ohrid, Macedonia, 12-14/09/2012. This work has been supported by the ICT-ACROPOLIS Network of Excellence, FP7 project n. 257626 and by COST Action IC0902 ”Cognitive Radio and Networking for Cooperative Coexistence of Heterogeneous Wireless Networks”.

Rome, October 9-11, 2013