Intra Cluster And Inter Cluster Communication Computer Science Essay

Published:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Cognitive radio is deemed to be a breakthrough technology because it has the capabilities to sense and learn from the operating environment, as well as to be applied in real time for network performance enhancement. Due to the dynamicity of spectrum availability in cognitive radio networks, designing of protocols and schemes at different layers of network stack has been challenging. Traditional clustering algorithms are not suitable to be applied in cognitive radio networks because it could not adapt to the channel dynamics in cognitive radio networks. Therefore, new clustering algorithms are designed to address the new challenge. Reinforcement learning has gained attention by solving issues associated with the dynamicity of distributed wireless networks. It allows nodes to take actions based on its observed operating environment. In clustering, as gateway nodes relay packets from one cluster to another, they may become bottlenecks. In this research project, we present how reinforcement learning can overcome this problem. In order to ensure the uniqueness of the problem and novelty of the proposed solution, wide range of latest research papers are obtained from digital. Then, we design a new scheduling approach and develop it to compare with traditional scheduling approach

Table of Contents

1.0 Introduction

As licensed spectrum has been significantly underutilized, cognitive radio is seen as a promising technology to improve the efficiency of spectrum utilization. Cognitive radio has the capabilities to sense and learn from the operating environment. Capturing real-time information and adapting its own behavior allow cognitive radio to enhance the overall spectrum utilization.

Routing schemes have also been applied to enhance network performance in cognitive radio. Routing schemes apply various mechanisms to enhance network performance, such as reducing end-to-end delay, reducing the amount of control overhead and increasing throughput performance.

Clustering is a process of organizing nodes into clusters in order to provide network-wide performance enhancement. The few main objectives are the enhancement of cluster stability, energy efficiency, cooperative tasks and the reduction of the number of clusters in the network.

In cognitive radio, reinforcement learning has been found to be more suitable compared to other machine learning algorithms due to its online learning approach, which allows decision maker to learn while carrying out its tasks. Additionally, it has the ability to perform individual (or independent) learning in a trial-and-error manner and through direct interaction with the operating environment.

The rest of this section presents a summary of important sections of report, including research background, problem statement, objectives, scope of work, research methodology, proposed outcome, and project timeline.

1.1 Research Background

The few research areas are reinforcement learning, cognitive radio, clustering and routing schemes.

1.2 Problem Statement

An optimal scheduling scheme for channel switches is necessary so that gateway nodes listen to the right channel and to ensure successful transmission while minimizing packet collision and bottleneck in two or more adjacent clusters.

1.3 Objectives

As gateway nodes relay packets from one cluster to another, they may become bottlenecks. The main objective is to investigate an efficient scheduler to reduce bottleneck at gateway nodes hence maximizing network wide performance of distributed cognitive radio networks.

1.4 Scope of Work

To design a new scheduling approach, reinforcement learning is applied to learn from the operating environment and make decision. After that, simulation is run to collect results in order to make comparison with the traditional scheduling approach. Both approaches will be investigated on different network scenarios.

1.5 Research Methodology

There are five stages in this research project. The five stages are as follows:

Literature Review

Literature review will be conducted on the relevant area in cognitive radio, including routing, clustering and the application of reinforcement learning.

Designing New Scheduling Approaches

Reinforcement learning will be applied to design new clustering approaches.

Developing Traditional Scheduling Approaches

Traditional clustering approach will be developed to provide baseline results so that comparison can be made with the new clustering approaches.

Developing New Scheduling Approaches

New clustering approaches will be developed on simulation platform such as OMNeT++ or C programming.

Discussing and Evaluating Simulation Results

Comparison between the traditional and new clustering approaches will be made to investigate any improvements.

1.6 Proposed Outcome

By applying reinforcement learning, a node can find a good schedule to reduce bottleneck and subsequently maximize network-wide performance of distributed cognitive radio network.

1.7 Project Time

The whole project will take approximately 28 weeks. A more specific distribution of workload and time planning are displayed in Gantt chart (see Research Methodology).

1.8 Organization of the Research Proposal

The rest of this research proposal report is organized as follows. Chapter 2 presents problem statement. Chapter 3 presents literature review. Chapter 4 presents objective. Chapter 5 presents scope of work. Chapter 6 presents research methodology. Chapter 7 presents project outcome. Chapter 8 presents conclusion.

2.0 Problem Statement

This section presents the problem under investigation in this project. Section 2.1 presents the research problem; Section 2.2 presents the research question; Section 2.3 presents the research challenge; and Section 2.4 presents the significance of the problem.

2.1 Research Problem

Gateway nodes relay packets from one cluster to another. In cognitive radio networks, the requirement of a scheduler to switch its operating channels according to the dynamicity of available channel has added new challenge to clustering schemes. This problem may be aggravated by the fact that adjacent clusters may use different common channel for intra-cluster communications; and so a gateway node must switch its operating channel so that it transmits and listens to different channels at the right time in order to ensure inter-cluster communications are well performed while improving network performance. The purpose is to reduce bottlenecks at gateway nodes in order to improve successful packet transmission in inter-cluster communications while minimizing packet collisions in both clusters.

As an example, there is a gateway node connecting two adjacent clusters, namely clusters A and B. Gateway node GWA is associated with cluster A; and a backbone CHA‒GWA‒CHB is formed. Packets are transferred from clusterhead A to clusterhead B via gateway node A. As a packet traverses from one cluster to another, its priority level becomes higher. This is because, the closer a packet to its destination node can be, the greater the effort of packet forwarding being wasted. Hence, the priority level of GWA‒CHB transmission is higher than CHA‒GWA, specifically, PGWA‒CHB>PGWA‒CHB. Unfortunately, the GWA‒CHB link may have lower level of white spaces compared to CHA‒GWA, causing a bottleneck at node GWA. In this case, the GWA may transmit packets of CHA‒GWA; however, this may cause congestion at GWA since the departure rate to CHB is lower compared to the arrival rate of CHA. Hence, the GWA may propose a channel switch for intra-cluster communication at cluster B.

2.2 Research Question

This research addresses the following research questions:

How to design a good scheduler to reduce the bottleneck at gateway nodes?

How can the scheduler address the dynamicity of channel availability?

2.3 Research Challenge

The main challenge associated with the design of the scheduler is the dynamicity of the available channels. There are two main reasons. Firstly, the unpredictability of white spaces may jeopardize the successful transmission rate of high-priority packet flows. Secondly, the unpredictability of white spaces may cause packets to be dropped at gateway nodes due to its bottleneck. This research performs investigation to address the aforementioned research challenges.

2.4 Significance of the Problem

Gateway nodes serve as a "bridge" that interconnects adjacent clusters. Generally speaking, clusterheads and gateway nodes interconnect among themselves to form backbone throughout the entire network. Hence, it is highly likely that gateway nodes may become bottleneck. This is aggravated by the fact that in cognitive radio networks, there is an additional form of dynamicity, specifically, the changes in the availability of white spaces in most licensed channels. Without an efficient scheduler at gateway nodes, network performance such as successful transmission rate and end-to-end delay may deteriorate.

3.0 Literature Review

This section presents an overview of cognitive radio, routing schemes, clustering and reinforcement learning.

3.1 Cognitive Radio

Radio frequency spectrum is a scarce natural resource, and most frequency bands (or channels) have been allocated to various wireless or broadcasting services such as TV and radio services. Unfortunately, according to occupancy measurements, it was discovered that the spectrum has been significantly underutilized [1]. Due to the scarcity of underutilized frequency bands, the efficiency of spectrum utilization plays a huge role in the development of future wireless systems; and cognitive radio is seen as a promising technology. Cognitive radio is deemed to be a breakthrough technology because it has the capabilities to sense and learn from the operating environment, as well as to be applied in real time for network performance enhancement. Hence, it has been regarded as a novel approach for improving the overall utilization of radio frequency spectrum.

This rest of this section presents various aspects of cognitive radio, including how it works, advantages, challenges and applications.

3.1.1 How it Works?

In order to enhance the overall spectrum utilization, cognitive radios capture real-time information of their respective wireless environment and use it to adapt its own behavior as time goes by. Real-time information of wireless environment covers channel conditions and the availability of resources [2]. Cognitive radios have capabilities of intelligence and reconfigurability. Intelligence is the capability to capture and learn about information from radio environment. Reconfigurability is the capability to be dynamically programmed according to the radio environment as time goes by [2].

There are three elements of cognitive radio networks that should be focused on. The three elements are are network components, network architecture and spectrum use options as shown in Fig. 1.

Fig. 1 The elements of a cognitive radio network [2]

There are two types of network component, namely primary and secondary networks. Primary networks have higher priority because these devices have the license to use the frequency bands exclusively. Secondary networks, who do not have license for spectrum usage, are allowed to access the freuquency bands owned by the primary networks following an important condition in which they must not affect the performance of primary networks [3].

There are two types of network architecture, namely centralized and distributed approaches. In the distributed approach, cognitive radio devices make their own decisions to access frequency bands; whereas in the centralized approach, a central entity (e.g. base station or acces point) is in charge of the allocation of licensed and unlicensed frequency bands among the cognitive radio devices [2].

There are two types of third spectrum use options, which indicate the types of frequency bands to be chosen for channel access, namely licensed and unlicensed frequency bands. While a primary device has license, its channel access is limited to the licensed channel that it owns. On the other hand, secondary device may use licensed or unlicensed frequency bands to interact with other secondary devices.

When it comes to decision making, such as the selection of the underutilized frequency bands (or channels with white spaces), cognitive radio networks perform four functions, namely spectrum sensing, spectrum decision (or channel selection), spectrum sharing and spectrum mobility [3]. This four functions are known as spectrum management framework as shown in Fig. 2.

Fig. 2 Spectrum management framework [2].

The functions of the spectrum management framework are as follows:

Spectrum sensing enables secondary networks to determine the spectrum holes (or white spaces) through coordination with other neighboring cognitive radios.

Spectrum decision (or channel selection) selects and allocates the best possible frequency bands upon the completion of spectrum sensing in order to meet the user requirements. The selected frequency band should not interfere significantly with the primary networks. This requirement is very important as it ensures the 'friendly' coexistence of primary and secondary networks in order to fulfill the spectrum sharing requirements.

Spectrum sharing enables the secondary networks, which covers the secondary device and its neighoring devices, to coordinate access to frequency bands.

Spectrum mobility enables the secondary networks to use any of the frequency bands; however, if primary networks' activities appear at any specific time instant, the secondary device must vacate and switch to another frequency band.

3.1.2 Advantages

Cognitive radio provides a more flexible and efficient usage of radio frequency spectrum [3]. This technology has the ability to solve issues associated with spectrum congestion and underutilization. It also introduces an innovative way for resource management, generates new revenues to network operators, and increases bandwidth availability at each user in the secondary networks.

3.1.3 Challenges

In cognitive radio networks, there are several main challenges as follows:

Communication may be disrupted because of the lack of the availability of frequency bands. Hence, the challenge is how to provide seamless communication among the nodes in secondary networks.

Interference to primary devices must be minimized in order to improve the network performance of both primary and secondary networks.

Quality of service provisioning is a great challenge, particularly in the provision of guaranteed quality of service requirement among nodes in the secondary networks. This is because cognitive radio devices access the licensed frequency bands in an opportunistic manner, which may jeopardize the quality and quantity of white spaces in the selected operating frequency bands as a result of the unpredictable activities from the primary networks.

3.2 Routing Schemes for Cognitive Radio

There exist three routing schemes in cognitive radio networks, namely intra-system-based, inter-system-based and hybrid-systems-based routing schemes [4]. Within each scheme, they are further categorized based of their types, namely proactive, reactive, hybrid and adaptive per-hop. Further discussion about performance enhancement of the existing routing schemes in cognitive radio networks are presented next.

3.2.1 How it Works?

Intra-system-based routing schemes choose a route that provides the best end-to-end network performance for secondary users; while inter-system-based routing schemes choose a route that minimizes interference to primary users caused by secondary users. Hybrid-system-based routing schemes combine both intra-system-based and inter-system-based routing schemes. There are four kinds of schemes that have been shown to achieve different performance enhancement. The four kinds of schemes are as follows:

Proactive routing schemes search for a route at most of the times. Specifically, a node updates its routes to most destination nodes even though it does not have any packets to send.

Hybrid routing schemes combines both reactive and proactive routing schemes. For instance, higher (lower) priority packets adopt proactive (reactive) routing scheme.

Adaptive per-hop routing schemes search for a suitable next-hop node, rather than a route to the destination. This approach may not be suitable for networks with high levels of dynamicity in which the routes may break often, and so the preceding approaches may not be cost effective (e.g. causing high amount of routing overheads).

3.3 Clustering

Clustering, which is a topology management mechanism is a process of organizing objects (i.e. nodes) into groups (or clusters) whose members have something in common. Therefore, a cluster may be comprised of objects that share "similar" properties or characteristics; and so, those who never share the "similar" properties may belong to other clusters. A simple graphical example to explain clustering is shown in Fig. 3. There are different ways of how this process organizes the objects into groups. For instance, the similarity criterion in a distance-based clustering approach is distance in which objects that are close together are grouped.

Fig. 3 Sample of Cluster [9]

3.3.1 How it Works?

In clusters, there are three types of nodes, namely clusterhead, member node and gateway node [5]. Generally speaking, the clusterhead plays the role as the local point of process, such as monitoring spectrum allocation, routing and spectrum sensing. Member nodes communicate with their respective clusterhead and other member nodes, and so they transmit and receive packets from their respective clusterheads and vice-versa. Next, the clusterhead sends the packets to a next-hop clusterhead until the packets reach the destination node. Gateway nodes are physically located within the transmission range of two or more clusters. Communication between clusterhead and members nodes is called intra-cluster communication; whereas communication between gateway nodes and nodes from neighbor clusters are called inter-cluster communication[5].

The aim of clustering is to achieve network scalability and stability [5]. As spectrum sensing and access are important to normal operation of cognitive radio, clustering helps by supporting these cooperative tasks. Traditional clustering algorithms such as lowest ID and maximum node degree may not be suitable to be applied in cognitive radio networks. One of the reasons is because traditional clustering algorithms do not adapt to the channel dynamics in cognitive radio networks. When channel availability of each nodes changes with time and location, each node may observe different level of white spaces. The lack of common channels among nodes in cluster may lead to loss of connectivity among clusterheads [5]. Hence, there are various new clustering algorithms that have been designed and proposed to address new challenges.

Clustering have two types of characteristics, namely clustering metrics and intra-cluster distance [5]. Various clustering metrics have been applied during cluster formation and maintenance.

There are four kinds of clustering metrics as follows:

Channel availability determines the number of common channels in a cluster and ensures that while nodes form clusters and selecting clusterheads, the number of common channels among nodes in the cluster is high. This is because the higher the number of common channels the higher the cluster stability.

Geographical location represents the location of other secondary users. Physically closer nodes may share the same characteristics; therefore forming clusters with physically close neighbor increases the number of common channels, and eventually enhancing the network performance.

Signal strength and channel quality help secondary users to select clusterheads based on signal strength. Stronger signal strength also indicates higher channel quality.

Node degree determines the number of neighbor nodes. Higher level of node degree helps to reduce intra-cluster distance; hence, it reduces the amount of overhead associated with intra-cluster communications.

Intra-cluster distance is the number of hops between member nodes and their respective clusterhead [5]. Nodes can form a single or multiple hops clusters. Single-hop clusters enhance network stability, parallelism and inter-cluster communication delays. Multiple-hop clusters reduce the number of clusters in network causing lower inter-cluster communication overhead.

3.3.2 Advantages

The two advantages of clustering are as follow:

Scalability is improved because communication overhead is reduced and parallelism is increased. Clusterheads and gateway nodes form a backbone to the base station in order to exchange information. Unlike non-clustered networks in which nodes must exchange information with all other nodes, member nodes exchange information with their respective clusterheads only. Hence, communication overhead can be reduced.

Stability is improved because the global effects of any frequent changes in network-wide performance are minimized. Specifically, member nodes and their respective clusterheads are reconfigured to response to changes; So, any local updates will be performed among member nodes and clusterheads if there are any changes on network dynamics.

3.3.4 Applications

Clustering scheme constructs clustered (or hierarchical) networks, which serve as the underlying architecture for various schemes (e.g. routing, channel sensing and so on) in cognitive radio networks. Two examples are as follows:

In routing schemes, the routing messages may be broadcast and traversed among certain nodes, particularly clusterheads and gateway nodes. Clustering helps to establish clusters with similar or near-similar characteristics so that routing schemes can be readily implemented on cognitive radio networks. Clustered networks have been shown to reduce routing overhead.

In spectrum sensing, a nodal representative combines sensing outcomes from its neighboring nodes. Clustering helps to establish clusters comprised of clusterhead and member nodes. The clusterhead combines sensing outcomes from its member nodes in order to determine the final decision on the existence of primary users' activities in a particular network.

3.4 Reinforcement Learning

Machine learning algorithms have gained attention by solving issues associated with the dynamicity of distributed wireless networks. The algorithm allows wireless nodes to achieve context awareness and intelligence for network performance [6]. Context awareness and intelligence help a wireless node to take actions based on its observed operating environment. There are different types of machine learning techniques, and reinforcement learning is one of them [6]. The choice of machine learning algorithms may be based on the characteristic of a distributed wireless network. For instance, in cognitive radio scenarios, reinforcement learning has been found to be more suitable compared to other machine learning algorithms (e.g. Swarm intelligence, genetic algorithms etc.) due to its online learning approach in which it allows a decision maker (or an agent) to learn while carrying out its tasks as time goes by.

3.4.1 How it Works?

Reinforcement learning is one of the many sub-areas of machine learning that uses a mathematical approach to evaluate the rewards of certain actions against the operating environment [7]. Reinforcement learning is suitable for applications in distributed cognitive radio networks because of the following reasons:

Its ability to learn through direct interactions with operating environment without the need of any external supervision.

Its ability to perform individual learning where a learning agent learns based on local observations only.

Its ability to learn in a trial-and-error manner helps to interact with unknown radio environment; this enables a learner to explore the operating environment itself and learns the optimal action [5].

Q-learning is one of the popular reinforcement learning approaches where it is often applied in distributed wireless networks. In Q-learning, an agent is modeled as three-tuple consisting of {S, A, R} as shown in Fig. 4.

Fig. 4 A standard reinforcement learning model [7]

The three tuples are as follows:

State is the decision making factors observed from the local operating environment. State can be internal (e.g. buffer occupancy rate) or external (e.g. destination node). In order to learn more about the operating environment, the agent must observe the states.

Action is taken against the operating environment, and it is normally based on the continuous observation and interaction with the local operating environment. The agent selects an action that maximizes the current and future (or discounted) rewards.

Reward is received by the agent upon taking actions against the operating environment. For example, when the environment makes a transition to a new state, the agent receives a reward at the next time instant. A reward may represent a performance metric (e.g. transmission delay, throughput and channel congestion level).

4.0 Objective

In clustered networks, each cluster uses a common channel for intra-cluster communication; and adjacent clusters may use distinctive common channels; hence, regular channel switches may be necessary at gateway nodes. Since gateway nodes relay packets from one cluster to another, they may become bottlenecks. Therefore, an optimal schedule for channel switches is necessary so that gateway nodes listen to the right channel in order to ensure successful transmission while minimizing packet collisions in both clusters. The main objective is to investigate an efficient scheduler in order to reduce bottleneck at gateway nodes.

Fig. 5 shows a scenario under investigation. For instance, there are two clusters, namely cluster A and cluster B. Gateway A is associated with cluster A. Cluster A uses channel 1; while cluster B uses channel 2. Hence, gateway A must switch from channel 1 to channel 2 in order to communicate with cluster B. The communication between nodes in cluster A and gateway A using channel 1 is called intra-cluster communication; while the communication between cluster B and gateway A using channel 2 is called inter-cluster communication.

Fig. 5 An example of scenario for further investigation.

There are three types of packets as follows:

Type-1 packets are sent from gateway A to cluster A. Since gateway A is a member node, it sends packets to clusterhead first, which is clusterhead A.

Type-2 packets are sent from cluster A to gateway A.

Type-3 packets are sent from gateway A send to cluster B.

Fig. 6 Priority level of each type of packets.

The scheduler determines which packet to transmit at the next time instant. This depends on the priority level of packets. In Fig. 6, the priority levels of each type of packets are as follows:

Type-1 packets from gateway A to cluster A have the lowest priority level because gateway A is a member node of cluster A.

Type-2 packets from cluster A to gateway A have the higher priority level.

Type-3 packets from gateway A to cluster B have the highest priority. Type-3 packets have higher priority than type-2 packets because they are nearer to the base station, in which bottleneck is more likely to occur compared to nodes further away from the base station.

In Fig. 6, queue 3 has the highest priority, and therefore it has the highest weight; followed by Queue 2 and Queue 1. Traditionally, the weight for each queue is a static value. However, this may not be the case in cognitive radio networks in which the availability of white spaces is dynamic in nature. Hence, even though it has the highest weight factor, priority may still be given to the other queues which have lower priority to transmit. In other words, Type-3 packets, albeit its highest priority, may only be sent when there are white spaces in its channel; and so the scheduler may send Type-1 or Type 2 packets while waiting for the channel of Type-3 packets to become better with higher amount and better quality of white spaces. This is a decision making scenario where dynamicity comes into place, in which reinforcement learning can be applied to traditional scheduling schemes such as round robin and weighted fair queuing. Learning is necessary to maximize network-wide performance of clustering schemes in distributed cognitive radio networks.

Using reinforcement learning, the state, action and reward representations must be defined. For instance, the state represents the availability of white spaces, the action represents the selection of operating channel for a gateway node or its upstream clusterhead, and the reward represents successful packet transmission rate at gateway node. Further research will be conducted to investigate the enhancement of reinforcement learning, as well as the scheduler scheme for gateway nodes.

5.0 Scope of Work

This section focuses on implementing reinforcement learning in scheduler in order to reduce bottleneck. Section 5.1 presents the design of a novel scheduling approach. Section 5.2 presents the comparison between novel and traditional scheduling approach. Section 5.3 presents various network scenarios that will be investigated.

5.1 Designing a Novel Scheduling Approaches

In clustered networks, clusters use a common channel to communicate. Hence, regular channel switches may be necessary at gateway nodes. When gateway nodes relay packets from one cluster to another, they may become bottlenecks as a result of the dynamicity of channel availability. When there are many queues, scheduler has to decide which queue to proceed first in order to reduce bottleneck. By implementing reinforcement learning in the gateway's scheduler, it will help to reduce the bottleneck. For instance, the state represents the availability of white spaces. The action represents the selection of operating channel for a gateway node. The reward represents successful packet transmission rate at a gateway node. This scheme may improve throughput performance.

5.2 Performing Comparison between Novel and Traditional Scheduling Approaches

The two traditional scheduling approaches that will be chosen for comparison with the novel scheduling approach are round robin and weighted fair queuing. Round robin is designed especially for time sharing systems. Time slices are assigned to each process in equal portions and the scheduler handles all processes without priority. Whereas weighted fair queuing assign traffic divide classes with different weights in which higher priority classes get higher weights. It transmits queued packets in a cyclic manner by serving one from each class; but normally a class with higher weight is served more. In the novel scheduling approach, for example, there are three queues with different priorities and weights. Reinforcement learning can be applied to improve the scheduler by telling which queue to transmit first in order to reduce packet loss or wasted packet relaying effort. By right, higher priority queues should be process first but if their respective channels have limited white spaces, transmission of lower priority queues may take place.

5.3 Network Scenarios

Different network scenarios will be implemented to test for different outcome. The three scenarios are listed as follows:

Varying the dynamicity level of the availability of white spaces.

Lower level of white spaces may jeopardize the successful transmission rate of packets. High-priority packet flows may be affected causing the gateway to become bottleneck. It may lead to packet drops at gateway nodes due to bottlenecks.

Increasing the number of upstream and downstream clusters.

Observation may be made to investigate how gateway operates when the number of upstream and downstream clusters increases.

Varying the packet arrival rate.

Different packet arrival rates are used to determine whether packets arrive in a way that may not congest the gateway. If a gateway is congested, it may lead to packet loss or longer delay.

6.0 Research Methodology

This chapter presents the process of the research, from choosing the most suitable simulation platform to designing, performing simulation test on the new clustering approach, and gathering the simulation results. Section 6.1 presents choice of simulation platform. Section 6.2 presents the research flowchart and the five stages in this research.

6.1 Choice of Simulation Platform

There are many types of simulation platforms such as Qualnet, NS2 or NS3. Based on the design of the clustering scheme, we choose the most suitable simulation platform. As for our case, we may be working with only a few nodes, therefore the most suitable simulation platform may be OMNeT++, which uses C/C++ programming language. Unlike other simulation platform, OMNeT++ is a freeware and it provides extensive graphic user interface that allows us to perform debugging tasks.

6.2 Research Flowchart

There are five stages in this research, and the subsequent subsections present each of them. Fig. 7 presents the flowchart of this research.

Fig. 7 Flowchart of the five stages in this research.

6.2.1 Literature Review

In order to ensure the uniqueness of the problem and also the novelty of the proposed solution, research materials are obtain from digital libraries such as IEEE Xplore. Digital libraries provide a wide range of latest research papers. In this case, the literature review focuses on cognitive radio, clustering, reinforcement learning and routing. The application of reinforcement learning to clustering allows gateway nodes to learn from the environment, and hence maximizing network-wide performance of distributed cognitive radio network.

6.2.2 Designing New Scheduling Approaches

Reinforcement learning is applied to allow wireless nodes to achieve context awareness and intelligence for network performance enhancement. Context awareness and intelligence help a wireless node to take actions based on its observed operating environment. When gateway nodes relay packets from one cluster to another, they may become bottlenecks as a result of the dynamicity of channel availability. Therefore, reinforcement learning in applied in the gateway's scheduler to reduce the bottleneck. The state, action and reward representations of reinforcement learning are defined in order to fit in well with the scheduler's objectives.

6.2.3 Developing Traditional Scheduling Approaches

Traditional scheduling schemes, such as round robin and weighted fair queueuing, will be developed and have their simulation results gathered. The simulation results will serve as baseline results as comparison with the new scheduling scheme. In round robin scheduling, multiple classes are divided with similar weight. It transmit queued packet in a cyclic manner, serving one from each class. Packets within a priority queue use first in first out (FIFO) scheduling. In weighted fair queueuing, multiple classes will be divided into different weight. Weight is determined by the level of priority, in which higher priority class gets higher weight. As packet is transmitting in a cyclic manner by serving one from each class, class with higher weights are served more.

6.2.4 Developing New Scheduling Approaches

A new scheduling approach will be developed and have their simulation results gathered. The simulation results will be used for further analysis. As mentioned in Section 6.2.2, by using simple C programming language, a new scheduler will be developed by applying reinforcement learning in the gateway's scheduler to reduce bottleneck.

6.2.5 Simulation Results Discussions and Evaluation

With simulation results gathered from traditional and new scheduling approaches, comparison will be made. A network performance graph will be plotted to show the improvement achieved by the new scheduling approach compared with the traditional scheduling approach. Discussions will also be made to explain further about the improvement achieved.

7.0 Project Outcome

This chapter presents research contributions and Gantt chart

7.1 Research Contributions

This research contributes towards a novel approach to reduce bottlenecks at gateway nodes in order to improve successful packet transmission in inter-cluster communications while minimizing packet collisions in both clusters in scheduling schemes for cognitive radio networks. The new scheduling approach will be tested using simulation platform such as OMNeT++ that uses C programming language. Simulation results of the new scheduling approach and traditional scheduling approach will be compared.

7.2 Gantt Chart

This section summarizes the research stages (see Chapter 6, Section 6.2), and presents a Gantt chart in Fig. 8 as part of the effort to ensure the entire research project is put on course. The research stages are as follows:

Literature review. This stage is to perform in-depth research in cognitive radio, clustering, routing and reinforcement learning. Research materials are obtained from digital libraries as they provide a wide range of latest research paper. This is to ensure that this is a pioneer topic. This stage will take up a total of 20 days excluding weekends, which means each topic will take 5 days on average.

Designing new scheduling approaches. This stage is to design a new scheduling approach by applying reinforcement learning in the traditional scheduling approach so that it can achieve the main objective, which is to implement an efficient scheduler in order to reduce bottleneck at gateway nodes. This stage will take up a total of 15 days excluding weekends.

Developing traditional scheduling approaches. This stage is to develop a traditional scheduling approach using OMNeT++ using C programing language. Results gathered from the simulation will be used as a comparison with the new scheduling approach. This stage will take up a total of 30 days excluding weekends.

Developing new scheduling approaches. This stage is to develop a new scheduling approach using OMNeT++ using C programming language. Results gathered from the simulation will be used as a comparison with the traditional scheduling approaches. Results are used to show network performance enhancement achieved by the new scheduling approach. This stage will take up a total of 30 days excluding weekends.

Simulation results discussions and evaluation. This stage is to discuss and compare the difference among the simulation results of both new and traditional scheduling approaches. Evaluation on the efficiency of the new scheduling approach will be carried out. This stage will take up a total of 30 days excluding weekends.

Other stages. Other stages involve preparing a draft proposal, finalizing proposal, drafting final report, finalizing final report, preparing presentations and hard copies of final project report.

Fig. 8 Progress of project in Gantt chart.

8.0 Conclusions

In cognitive radio networks, scheduler is required to switch its operating channels according to the dynamicity of available channels. In this report, we will be focusing on how to design a good scheduler so that the bottleneck at gateway nodes can be reduced. Also, the focus is conduct to research on how scheduler addresses the dynamicity of available channels. Five stages will be conducted in this research to achieve our objectives. Firstly, we will ensure the uniqueness of the problem and the novelty of the proposed solution by reading recent research materials that can be obtained from digital libraries. Secondly, we will design a new scheduling approach by incorporating reinforcement learning into the traditional scheduling approach. Thirdly, we will develop the traditional scheduling approach using OMNeT++ simulator to obtain baseline results. Fourthly, based on the design of the new scheduling approach, we will use OMNeT++ simulator to obtain results for comparison. Fifthly, by comparing the results from the simulation, we will determine whether the proposed solution can achieve our objectives. We will run this simulation on three different network scenarios in order to get accurate results. The project will take approximately 28 weeks.

Writing Services

Essay Writing
Service

Find out how the very best essay writing service can help you accomplish more and achieve higher marks today.

Assignment Writing Service

From complicated assignments to tricky tasks, our experts can tackle virtually any question thrown at them.

Dissertation Writing Service

A dissertation (also known as a thesis or research project) is probably the most important piece of work for any student! From full dissertations to individual chapters, we’re on hand to support you.

Coursework Writing Service

Our expert qualified writers can help you get your coursework right first time, every time.

Dissertation Proposal Service

The first step to completing a dissertation is to create a proposal that talks about what you wish to do. Our experts can design suitable methodologies - perfect to help you get started with a dissertation.

Report Writing
Service

Reports for any audience. Perfectly structured, professionally written, and tailored to suit your exact requirements.

Essay Skeleton Answer Service

If you’re just looking for some help to get started on an essay, our outline service provides you with a perfect essay plan.

Marking & Proofreading Service

Not sure if your work is hitting the mark? Struggling to get feedback from your lecturer? Our premium marking service was created just for you - get the feedback you deserve now.

Exam Revision
Service

Exams can be one of the most stressful experiences you’ll ever have! Revision is key, and we’re here to help. With custom created revision notes and exam answers, you’ll never feel underprepared again.