Routing in Cognitive Wireless Mesh Networks

An Intelligent Framework

Nadhir Ben Halima

and Mohammad Ayoub Khan

1,2

College of Computer Science and Engineering, Yanbu Branch, Taibah University, Madina Al Munawwarah, Saudi Arabia

Department of Computer Science and Engineering, Sharda University, Greater Noida, India

Keywords: CN, SDR, QoS, CR.

Abstract: Wireless Networks are limited in energy and resources, are subject to development constraints. The

difficulties are such as the increasing RF spectrum saturation and efficient path discovery. The Cognitive

Wireless Networks, leaning on a form communication model, develop new strategies to mitigate the

inefficient use of the spectrum. The first application of the concept of cognitivity to communications was

focused on exploiting the dynamics in spectrum utilization (cognitive radio), nevertheless network-wide

deployment of such concepts is foreseen in the framework of the “cognitive networks”, where the cognitive

process will be employed to support end-to-end network-wide goals such as QoS. This paper presents a

state-of-art of cognitive networks and proposes a framework, architecture for cognitive networks. This paper

will also discuss mechanisms for self-adaptation, learning and evolutionary functionalities to support

users/applications end-to-end goals.

1 INTRODUCTION

All the wireless networks use radio waves to connect

devices like laptops, sensors. Nowadays we can see

that these devices have become ubiquitous in nature.

There is a seamless integration with the existing

infrastructure.

The ubiquitous computing is one of the most

important trends in networks with a 40-fold increase

between 2010 and 2015(CISCO Inc, 2011). The

applications include security, surveillance,

monitoring, health, vehicular networks are few to

name. The increased demand for wireless

communication has a big challenge for efficient

spectrum utilization. We also understand that due to

increased demand of devices in 2.4 GHz range, the

unlicensed spectrum bands are becoming

overcrowded (Akyildiz et al., 2006, CISCO, 2011).

To address these challenges in wireless networks a

new technique called ‘Cognitive Radio’ (CR)

(Mitola III, 2000, Akyildiz et al., 2006, Thomas et

al., 2005). CR provides an opportunistic access

mechanism to the spectrum using the concept of

cooperation and context awareness. The spectrum

sensing and cooperation between devices allow for

better utilization of spectrums. The CR has three

main components as shown in figure 1.

Figure 1: Components of Cognitive Radio.

The first component cognitive devices has

capacity in hardware or Software Defined Radio

(SDR).The collaboration component interacts among

terminals while as learning component has the

ability to learn about the pattern and behaviour of

spectrum utilization.

Indeed, many scenarios were proposed, focused

on accessing the spectrum using opportunistic

techniques (Akyildiz et al., 2006, Wang et al., 2011).

However, the possibilities offered by implementing

the cognitive process within a communication

network go far beyond improved spectrum

utilization, as cognitive network (CN) technology

can represent a suitable technology to support user

and application QoS requirements. Thomas et al.,

(Akyildiz et al., 2006, Gunawardena and Zhuang,

2011) defines a cognitive network as a network with

a cognitive process that can perceive current

network conditions, and then plan, decide and act on

the basis of such conditions.

The CN is not a radio network but data network

that makes use of various disciplines of computer

science like knowledge representation, machine

Cognitive

Collaboration Learning

Ben Halima N. and Ayoub Khan M..

Routing in Cognitive Wireless Mesh Networks - An Intelligent Framework.

DOI: 10.5220/0005571000430048

In Proceedings of the 6th International Conference on Data Communication Networking (DCNET-2015), pages 43-48

ISBN: 978-989-758-112-0

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

learning and network management to solve some

problems current networks are faced with. The CN

covers all the layers of the OSI model while CR

covers layer 1 and 2 only (Mitola, 2000).

The network can learn from these adaptations

and use them to make future decisions, while taking

into account end-to-end goals. The end-to-end scope

is extremely important and distinguishes cognitive

communications from cognitive networks. Indeed,

there are two levels of cognitivity: at the node level

and at the network level. The end-to-end scope

refers to the collective decision taken by the whole

network to achieve the stated complex goals. Such

end-to-end goals represent the desirable state and

will be reached after a specific set of actions, which

are far too complex to achieve using current

networking solutions.

A CN should be able to decompose the overall

goals into sub-goals and as necessary identify further

dependencies between these sub-goals as the main

benefit of the cognitive learning process.

Starting from the definition, the problem of

designing cognitive networks can then be considered

as a complex, multi-constraint and multi-criterion

optimization problem where a multi-objective

optimization approach should be implemented.

Guaranteeing quality of service (QoS) in

communication networks, particularly in cognitive

mesh networks, represents a typical multi-objective

problem involving simultaneous optimization of the

cost of the communications in the network and

various performance criteria such as: throughput,

average delay of the network, etc. The optimization

of one or more of such metrics is the main objective

of design in most cases, in order to make the

network efficient and possibly adaptable to different

operating conditions.

In this paper, we define a framework for the

design of Cognitive Wireless Mesh Networks

(CWMN), which implements and instantiates the

concepts of CN in a specific scenario where self-

adaptation is a key feature. The goal of such

networks is to foster self-adaptation and to extend it

beyond “simple” topology or resource management

in order to support end-to-end goals by exploiting

conceptual similarities with biological systems.

The structure of the paper is as such: Section 2

introduces the CWMN reference model and related

work on the topic, while Section 3 discusses the

proposed architecture in terms of main modules and

functionalities required. Section 4 concludes the

paper with final remarks.

2 CWMN REFERENCE MODEL

Wireless mesh networks are attracting a great

amount of attention

from networking researchers as

a promising technology for the next generation

access infrastructure, especially due to their

scalability and self-organizing features (Akyildiz et

al., 2005).

A wireless mesh network (WMN) is a mesh

network that is built using wireless access points

(AP) installed at each network user's location. Every

node in the network also acts as a provider to

forward data to the next node. The networking

infrastructure is decentralized and simplified. Each

node needs in WMN only to transmit the data to the

next node as shown in figure 2.

Figure 2: Wireless Mesh Network.

However, due to their distributed nature and

heterogeneity of access technologies, QoS

provisioning, and in general traffic engineering,

represents a challenging issue, involving several

functionalities at different layers of the protocols

stack: routing at layer 3, resource allocation and

reservation at layers 2&3, etc. In a cognitive

network approach, QoS should be granted by self-

adaptation, flexibility and situation-awareness of

network elements.

As a consequence, the characteristics of CNs

clearly represent desirable features on which to

implement completely autonomous and self-

adaptable wireless mesh networks.

The following paragraphs introduce the concept

of Cognitive Wireless Mesh Network as a Wireless

Mesh Network gaining biologically-inspired features

from the Cognitive Network paradigm, with the aim

of bringing the capability of the former to the next

DCNET2015-InternationalConferenceonDataCommunicationNetworking

level by introducing the possibility of autonomously

addressing end-to-end requirements and constraints.

In such a dynamic environment, two

requirements should be taken into consideration for

network control: scalability and mobility, as the

network should be aware of nodes joining and

leaving the system at any time and of their impact on

the performance and other operating parameters of

the network.

Such a lively scenario has significant similarities

with biological systems or colonies, and as a

consequence biologically-inspired approaches can

be considered as promising references due to the fact

that they are highly capable of self-adaptation and

especially evolution, even though relatively slow at

adapting to the changes in the environment.

Adding cognition to the existing WMN

infrastructure will bring about many benefits. In

order to pursue this idea, the paper proposes a

framework to support the design, implementation of

CWMN based on ideas such as evolutionary

programming in bio-nets and multi-agents and

actually starts to bridge the gap between conceptual

similarity and actual design.

The biologically inspired techniques in

information technology is not a new, many attempts

have been concentrated in the area of optimization

(Yu et al., 2010). However we can learn

considerable lessons from biological systems and we

have mainly to concentrate on the adaptability,

scalability, self-organizational and robustness

properties of such systems. Studying the symbiotic

nature of bio-systems can result in obtaining a

beneficial understanding on the behaviour of

distributed systems. Many key factors may be

observed in bio-systems like self-organizational

behaviour which can be considered as the most

important property describing systems consisting of

autonomous entities (ants, bees, etc.). These systems

tend to group together into certain structures very

similar to biological systems where overall state of

the system depends on individual behaviours and its

collection. In general, four basic principles can be

seen in the self-organizational property of bio-

systems which have to be introduced in our

proposed architecture.

First, positive feedback involves reinforcement

to enable the system to evolve, and to promote the

creation. Also, positive feedback behaves as

amplifier for a desired outcome, whereby negative

feedback is to influence from previous adaptations

which were not successful.

Second, bio-inspired systems, in general, do not

rely on any global control unit, but operate in an

entirely autonomous and distributed way. Whereby

the individual units acquires the information,

processes and stores locally. However, in order to

generate a self-organized structure, entities need to

interchange knowledge with each other; either by

direct or indirect interactions among them. Also,

self-organized structures relies on randomness and

fluctuation to enable the discovery of new solutions

and to boost the resilience and the stability of the

system.

Third, self-organization is also seen in swarm

cognition. In biological systems, the intelligent and

well-organized behaviour of insects can often be

observed. For instance, ants solve complex tasks,

like nest-building or food-collection, by delegating

simple tasks to each other. In such emergent

systems, it is the collective work of all single

activities that determines the outcome and not the

individual work. Such a behaviour is usually

referred to as swarm cognition (Yu et al., 2010).The

collaboration of insect societies is based on a

process in which a group of workers is assigned

specialized tasks in parallel in order to increase the

efficiency of the swarm(Yu et al., 2010). All insect

types have a division of labour which can be seen in

its 'organizational structure' that consists of workers

of different reproductive castes and layers. In Ants

colony, the insect is not aware of the global

conditions but gains the input only by interactions

with other members of the species that are locally

close within the rank in the 'structure'. This process

may occur either by direct contact or indirect

interaction. Direct contact can be chemical or visual

while in the indirect interaction, one entity

influences the environment while another responds

to that change later on. For instance, Ants

communicate with each other using pheromones that

other ants follow, thus, reinforcing changed pattern.

This mechanism of indirect coordination is called

stigmergy.

Summarizing, we propose to build a framework

for self-organizing cognitive wireless mesh networks

inspired by the above observations.

3 PROPOSED ARCHITECTURE

The network consists of mesh routers, mesh clients,

WiFi, Sensors, cellular network, conventional clients

and internet as shown in figure 1. This presents the

typical topology of a hierarchical Wireless Mesh

Network.

The proposed architecture aims at providing

global control of the topology using a distributed

RoutinginCognitiveWirelessMeshNetworks-AnIntelligentFramework

control scheme, where nodes are hierarchically

organized in groups or clusters. In our proposed

framework, the group formation and cluster

interconnection is performed in a distributed and

dynamic way. Assuming that at boot time, each node

receives a unique identifier, the election of the head

(H) of the cluster is performed. This process could

be performed by exploiting group formation

algorithms available in the literature for ad hoc

networks, such as (Lin and Gerla, 1997) or by

identifying specific algorithms for mesh networks

(i.e. to exploit higher capacity of mesh routers). The

choice of a specific clustering algorithm is out of the

scope of the paper.

Figure 3: Hierarchical WSN (Akyildiz et al., 2005).

In this framework, cognitive nodes perform local

adaptation, and such behaviour should be

coordinated and optimized to better maintain a

desired global state and achieve end-to-end QoS

guarantees.

Figure 4: Cognitive Node.

Each node receives local input from the

surrounding nodes and provides information about

its current capabilities. By using such information,

the nodes should organize themselves into a

cooperative system.

The proposed approach involves a reformulation

of the end-to-end QoS provisioning problem into a

distributed artificial intelligence problem, in which

cognitive mesh nodes coordinate parameters settings

to maintain the desired goals.

3.1 Problem Statement

The nodes are cognitive, i.e. able to sense their

environment and act intelligently and cooperatively

to achieve a desired global behaviour. More

specifically, nodes can learn to organize themselves

in clusters to increase overall throughput and reduce

the signalling overhead.

In this framework, the end-to-end QoS

provisioning problem can be translated into a

distributed artificial intelligence problem, where

each cognitive node is an agent in a multi-agent

system. Nodes interact with each other by competing

for resources and also by coordinating the access to

the resources to reach a satisfactory state, which is

continually updated. This behaviour represents the

design objective of cognitive networks, i.e. cognitive

nodes are always coordinating and optimizing

towards an ‘unknown’ global optimal state which

does not exist in reality since it is nothing but an

optimal state in a time t and for a given topology due

to the dynamicity of the topology. Each action taken

by a node is local in nature, and may not produce

any noticeable benefit to the node itself.

Collectively, however, the local actions can improve

the global performance of the system.

For example, yielding a resource to a node which

has to deliver urgent (high priority) traffic does not

produce any local benefit to a node, but it does

increase the overall throughput.

Ensuring global QoS provisioning can be seen

from different points of views: at the node level and

network-wide. The network aims at providing QoS

while maximizing the resource utilization, whereas

nodes’ goal is to maintain high throughput for

packets waiting in its queues.

Indeed, the global performance of the network

that should be maximized can be defined as a

function of several parameters:

lobal=

f (P

, P

, … Pi) (1)

Approaches available in the literature aim mainly at

providing the optimal (static) solution given specific

constraints on network setup and complete “a-priori”

knowledge of traffic characteristics and

requirements, or at decomposing the problem within

uniform independent domains (e.g. DiffServ

domains). On the contrary, the main goal of the

proposed Cognitive Wireless Mesh Network is to

perform distributed optimization of the performance

functional while continuously adapting to changing

characteristics of the network and traffic.

DCNET2015-InternationalConferenceonDataCommunicationNetworking

3.2 The Proposed Approach

On the basis of the above discussion, the CWMN

can be considered as an evolving autonomous

system that aims at reaching the optimal status,

which is represented by maximization of function

global

. The network will be dynamically partitioned

into clusters, whose heads (H) will be able to

interact to exchange and build global knowledge.

As a consequence, awareness of global end-to-

end goals will be provided by cluster head nodes

through inter-cluster signalling, while local

optimization actions will be performed through

intra-cluster interaction.

3.3 Modules and Functionalities

- Input Goals : Parameter to optimize

- Routing Engine : distributed optimization

algorithm

- QoS Engine : To guarantee QoS Parameters

- Reasoning and Learning Engine : For best

actions

- Knowledgebase: For Best decisions

Figure 5: Proposed framework.

The proposed framework has input a goal that is

described using a set of input parameters (P

, P

…P

). This is the global parameters. The

architecture has learning and reasoning engine to

deduce best decisions about the routing.

The knowledge based is the collection of

historical data, states and actions that has been taken

in the past. The historical data, states and action help

in taking right decisions. In order to guarantee the

quality we have a QoS engine that ensures quality

during end-to-end transmission.

The routing decisions and path are collectively

computed based on the input parameters from

different components of the proposed architecture

like knowledge base, learning engine, reasoning

engine etc.

4 CONCLUSIONS

In this paper we have discussed that increased

demand for wireless communication which has a big

challenge for efficient spectrum utilization. To

address these challenges in wireless network a new

technique called Cognitive Radio (CR) (Mitola III,

2000, Akyildiz et al., 2006, Thomas et al., 2005) has

been discussed in many literatures. We also

understand that CR provides an opportunistic access

mechanism to the spectrum using concept of

cooperation and context awareness. Since Wireless

Networks are limited in energy and resources, this

makes them subject to development constraints. We

have presented an exhaustive state-of-art of

cognitive networks and proposed a framework,

architecture for cognitive networks. The proposed

architecture has learning and reasoning engine to

deduce best decisions about the routing. We also

proposed a QoS engine to ensure quality during end-

to-end transmission. As a future work, we will

simulate the proposed architecture to demonstrate

the effectiveness.

REFERENCES

Mitola III, J, 2000. Cognitive Radio: An Integrated Agent

Architecture for Software Defined Radio, PhD thesis,

Royal Institute of Technology, Sweden.

Akyildiz, I., W. Lee, M. Vuran, and S. Mohanty, 2006.

“Next Generation/Dynamic Spectrum

Access/Cognitive Radio Wireless Networks: A

Survey,” Computer Networks, vol. 50, no. 13, pp.

2127–2159.

Thomas, R. W., DaSilva L. A. and MacKenzie, A. B.,

2005. Cognitive networks. Proceedings of the IEEE

Symposium on New Frontiers in Dynamic Spectrum

Access Network(DySPAN), November, Baltimore,

MD, USA.

Lin, C. and Gerla, M., 1997. Adaptive Clustering For

Mobile Wireless Networks,” Selected Areas in

Communications, IEEE Journal on, vol. 15, no. 7, pp.

1265–1275.

Amr El-Mougy and Mohamed Ibnkahla, 2013. A Cross-

Layer Framework for Network Management in

Wireless Sensor Networks Using Weighted Cognitive

Maps Hindawi Publishing Corporation, International

Journal of Distributed Sensor Networks, Article ID

RoutinginCognitiveWirelessMeshNetworks-AnIntelligentFramework

568580, 18 pages, http://dx.doi.org/10.1155/2013/

568580.

CISCO System Inc., 2011. Cisco visual networking

index: global mobile data traffic forecast update,

2010–2015.

Mitola, J., 2000. Cognitive Radio: An Integrated Agent

Architecture for Software Defined Radio,” Ph.D.

dissertation, Royal Inst. of Tech., Sweden.

Akyildiz, Ian F., Xudong Wang, Weilin Wang, 2005.

Wireless mesh networks: a survey, Computer

Networks, Volume 47, Issue 4, 15, Pages 445-487,

ISSN 1389-1286,

http://dx.doi.org/10.1016/j.comnet.2004.12.001.

Barıs¸ Atakan Ozgur B. Akan, 2007. BIOlogically-

inspired Spectrum Sharing in Cognitive Radio

Networks, IEEE WCNC ,pp43-48.

Yu, F. Richard et al., 2010. Biologically Inspired

Consensus-Based Spectrum Sensing in Mobile Ad

Hoc Networks with Cognitive Radios, Network, IEEE,

May/June, pp 26-30.

Akyildiz, Ian F. et al., 2005. A Survey on Wireless Mesh

Networks, IEEE Radio Communications, pp. 23-30.

Gunawardena, S. & Zhuang, W, 2011. Capacity analysis

and call admission control in distributed cognitive

radio networks. Wireless Communications, IEEE

Transactions on, 10(9), 3110–3120.

Wang, H., Ding, L., Pan, Z., Liu, N., & You, X, 2011.

QoS guaranteed call admission control with

opportunistic scheduling (pp. 1–5). In Global

telecommunications conference (GLOBECOM 2011),

IEEE.

DCNET2015-InternationalConferenceonDataCommunicationNetworking