A Cluster Based Adhoc Cognitive Radio Networks Architecture With ...

CoAd: A Cluster Based Adhoc Cognitive Radio Networks Architecture With Broadcasting Protocol Nafees Mansoor1, A.K.M. Muzahidul Islam2, Sabariah Baharun3, Shozo Komaki4, Koichi Wada5 1,2,3,4

Malaysia Japan International Institute of Technology (MJIIT) Universiti Teknologi Malaysia 54100, Jalan Semarak, Kuala Lumpur, Malaysia. 5 Department of Applied Informatics, Faculty of Science and Engineering Hosei University 3-7-2 Kajino-cho, Koganei, 184-8584, Japan E-mail: [email protected], [email protected], [email protected], [email protected], [email protected] Corresponding Author’s E-mail: [email protected] Abstract- Considering the mounting demand of radio frequencies, proper utilization of the spectrum is very essential. Cognitive radio, which uses an open spectrum allocation technique, can solve this spectrum congestion problem. Effective deployment of cognitive radio networks highly depends on a robust architecture with appropriate communication protocols. Considering the limitations of recently developed cognitive radio network architectures and associated broadcasting protocols, a dynamic cluster-based simple architecture is proposed for Ad-hoc cognitive radio network. For effective routing and load balancing, proposed architecture is dividing Ad-hoc architecture into clusters. The paper also introduces time slotted based broadcasting protocol for the proposed clustered architecture. Simulation results show the clustering efficiency of the proposed architecture. Keywords- Cognitive radio network, network architecture, cluster networks, broadcasting, multi-hoping.

I.

INTRODUCTION

The requirement for radio spectrum, a finite natural resource, is mounting higher with the fast growth in wireless technologies. Currently, radio spectrum allocation uses a static policy, where only the licensees can use the radio spectrums. Furthermore, various surveys on spectrum usages show that radio spectrum is underutilized [1, 2]. The main idea of cognitive radio networks (CRN), which was introduced by J. Mitola [3], is to use the free radio spectrum for communication. Cognitive radio senses the surrounding radio environment, and based on the sensing results, it orients itself to the situation by changing operating parameters such as power, frequency, and modulation etc. Primary users and secondary users are the two types of users in CRN [4]. Primary User (PU)/ licensed user has the sole access right on the radio spectrum. Instead, Secondary User (SU)/ unlicensed user/ cognitive user (CU) has to vacate the spectrum as soon as a PU appears. In cognitive radio system, the secondary user seeks the opportunity to use spectrum when PU is not active [11]. In other words, SU is allowed to use the licensed spectrum when PU is idle [5]. Based on the observation, the unused radio spectrum is used by cognitive radio in an intelligent way for temporary time period. SU is an important component in CRN system [8]. In CRN, spectrum

sensing is the process to identify the free unused where picking the most appropriate unoccupied channel for communication. Clustering in CRN improves network performances in maintenance, mobility, and load balancing. Clustering also leads to a simple and stable cluster backbone that simplifies control over network protocols. This paper proposes a dynamic cluster-based simple architecture for cognitive radio network with a Dynamic Cluster Base Station. The other components of the proposed architecture are Cluster Heads, Cluster Members, and Cluster Gateways. Dynamic Cluster Base Station acts as the controller and the fusion center for the proposed architecture. Broadcasting is a basic network operation in CRN with extensive applications. Blind forwarding is a candid method for broadcasting where each node is required to re-broadcast the broadcasting packet. Due to the redundant transmission, blind forwarding leads to broadcast storm [13]. Probabilistic, counter-based, distance-based and cluster-based are few popular technique to reduce the broadcasting redundancy. The paper is organized as follows. In section II, a brief review on different recently developed architectures for cognitive radio network is discussed. The proposed dynamic architecture for cognitive radio network is described in section III. Section IV, deals with the proposed broadcasting protocol. In section V, the simulation results of the proposed architecture are presented. Future works and conclusion have been discussed in section VI. II.

REECENTLY DEVELOPED ARCHITECTURES

This part of the paper discusses different architectures for CRN, which are proposed in various papers in recent years. Along with the architectures, we also discuss the communication protocols for cognitive radio networks. In the paper [4], the general structures for cognitive radio networks (Figure 1) have been presented. Infrastructure, Ad hoc, and Mesh, these are the three general types of structures in CRN. Cognitive Terminal (CT)/ Mobile Station, Base Station (BS)/ Access Point (AP) and Backbone Network are the key modules for these structures. In Infrastructure architecture (Figure 1 (gridded Circle)), a CT can only communicate with the BS. Inter-cell communication is done

978-1-4799-0400-6/13/$31.00 ©2013 IEEE

using the backbone network. To meet the necessity of the CTs, BS can use diverge protocols.

Figure 1: General structures for cognitive radio networks [4].

In case of the Ad-hoc architecture (Figure 1 (Inner White Circle)), any sort of infrastructural establishment is absent. CTs can setup the communication links among themselves with applicable communication standards/ protocols and form an Ad-hoc network. Similar to the Hybrid Wireless Mesh Networks [10], the Mesh architecture (Figure 1 (Outer White Circle)) of CRN is the combination of the infrastructural and Ad-hoc architectures. Communication between the BS and CT can be done in single hop or multi hops basis. BSs/ APs can act as the gateway when they are connected to the wired backbone networks. If the probable number of spectrum holes is larger, the BSs can have enough wireless communication links to act as the wireless backbone for the network. In [14], architecture for CRN is presented where communications between nodes are done without the common control channel to reduce co-channel interference. Figuring out the home channel of a neighbor is the main challenge for the architecture. Each node can freely select its home channel. In the architecture [14], every node has the set of accessible channels (SAC) of the neighbors. Whenever any node updates its own SAC, it puts a timestamp and exchanges the SAC database with other nodes when they meet in a channel. The proposed algorithm for the selection of the home channel of the neighbor uses a controlled pseudo-random algorithm. This home channel selection of a neighbor is basically a set of repetitive experiments. Using the SAC list, it is possible to compute the home channel of a node without an up-to-date SAC. One of the main limitation of this architecture is it requires extra memory to store the SAC list. Calculating the home channel, from the SAC list, is also an operational surplus for the network. Self-organized cognitive radio network architecture on multi-agent systems (MAS) is presented in the paper [15]. This architecture divides a big network into smaller groups. Spectrum allocation is done separately in each group. Collaborative-Max-Sum-Bandwidth rule is used to allocate spectrum in each group. Each group consists of a master node, which coordinates the group behavior. These master nodes form the gateway that is used for the communication among

groups. Selecting the cutting edge between two groups is the main challenge for this architecture. Depending on the scenario, an edge can integrate two groups, and forms a new group. If an edge has the maximum balance degree in the group then the group is divided into two new groups by the edge and two new master nodes will be elected for the groups. According to the architecture, when there is a necessity to remove an edge, the edge information is sent to the master node. When a master node removes an edge, it needs to check whether it is fit to be the master node or not. To evaluate the system utility, Max-Sum-Bandwidth (MSB) is used. Offering energy competency and spectrum efficiency, a design for Cognitive Radio Based Wireless Sensor Networks (CR-WSN) at the smart grid utility is proposed in [16]. The paper also proposes some key modifications of RPL (Routing protocol for low power and lossy networks) to ensemble the proposed architectural requirements. Nodes in a network are classified into three types, spectrum sensors, coordinator and ordinary nodes. Here, only spectrum sensors and coordinators detect the primary user and update the channels back up list. The Geo-locator database provides the information about the white spaces in a given geographical location. This communication is done in a non-Zigbee channel. A coordinator is the root of the network. In the formation of the network one of the 16 Zigbee channels is used. Whenever a new node is joining the network coordinator uses the Zigbee channels to supervise the process. It also has a schedule transmission period for spectrum sensing. Coordinator gathers reports from all the nodes of the network regarding the free channel. If the packet error rate goes beyond the acceptable threshold of a node, coordinator sends the warning message to that particular node. On the other hand, spectrum sensor carries out the PU discovery and issues a channel switch notification whenever there is a PU detected. In [17], a logical architecture for 4G heterogeneous wireless communication systems with adaptable user-centric network scheme is proposed. The architecture uses the reconfigurability concept, and enables resources sharing among different radio networks. The architecture introduces two new entities for user side and network side naming Resource Manager and Access Controller. Resource Manager is divided into two parts: Resource Unit and Cooperation Unit. The Resource Unit deals with maintaining the QoS of the current network, allocating the common resources among different networks and handover mechanism. Unused resources can be stored in a common logical server, where any network can access the server and borrow the resources from others for load balancing and greater use of the radio resources. Cooperative Unit deals with the common resources issue. When a network wants to share resources to the common resources server, Cooperative Unit determines which resources to be shared and how long it will be shared. Cooperative Unit monitors the status of the borrowed resources and decides the return time of the resources to the first party. Algorithms like dynamic frequency allocations, spectrum management, and resource management can be installed in the Cooperative Unit. Multiuser access and

signaling interactive are the main two tasks for the Access Controller at the network side. Apart from that, the sub unit named Re-configure Unit in Access Controller monitors the network and ensures the communications among different networks. The Resource Unit at the user end manages the resource allocation and monitors QoS of all traffic in the user terminal. Collecting and detecting message of user’s preference and service type are the task for the sub unit named Detection Unit under the Resource Unit. The Access Unit at the user end helps the network sided Access Unit to select the best network for a particular user. The work in [18] allocates dynamic channel among the requesting applications in cognitive radio networks by limiting the transmitted power in the sub-bands using joint power control and link scheduling strategy. A cross layer interaction between the MAC and PHY layers for dynamic channel allocation is the main feature of this architecture. The key benefit of this architecture is the decentralized management scheme where nodes can be added or deleted without the involvement of the central authority where the CR Managers in all nodes make all the central decisions. The Ultra Wide Band is divided into sub-bands by the Channel Scanner, which periodically scans sub-bands for white spaces based upon interference temperature. The Rake Optimization Block computes the number of Rakes or fingers to produce maximum SNR at minimal BER. The Channel Estimation Block monitors the fading condition and the channel error rates. The joint power control and link scheduling strategy help the CR to support high and low priority traffic based on delay sensitivity. CRN research has received a lot of attentions during the recent years. Recently a good number of architectures have been proposed, however a robust architecture for CRN is still absent. The general architecture for CRN proposed in [4] only discusses about the single hop system where the multi-hop communication is absent. The architecture that doesn’t use any common control channel introduces the concept of SAC (Set of Accessible Channels) [14]. Storing the SAC list necessitates extra memory storage. Calculation the home channel from the SAC list, is also an operational surplus for the network. The self-organized CRN architecture [15] divides the network into groups, though the concept of fusion center is absent there, which limits the architecture not to eligible for centralized cooperative sensing techniques. The energy competence CRN architecture presented in [16] has three types of cognitive radio. The proposed architecture assumes a communication link between the coordinator and the geo locator, where the geo locator database provides the information about the white spaces in a given geographical location. The geo locator limits the architecture to be immobile. Using the geo locator also adds infrastructural overhead. The 4G heterogeneous cognitive radio architecture [17] stores the unused resources in a common logical server, where any network can access the server and borrow the resources from others. The common logical server introduces overhead to the network. Accessing and borrowing resources from the common logical server are also operational surplus for the network. Transmission power

control based CRN architecture [18] also adds high volume overhead to each node caused by the orientation of the proposed architecture. III.

PROPOSED ARCHITECTURE

In the adhoc architecture, the Cognitive Terminals are deployed in such a way that it forms a flat topology in which a link exist between two terminals as long as they are in communication range of each other (Figure 2(a)). A graph representation of the flat architecture is shown in Figure 2(b). There is no established structure to facilitate efficient communication in such a flat topology. Therefore, clustering is used to leverage the underlying flat CRN topology and to provide a hierarchical organization.

Figure 2 (a): Flat Architecture, (b) Graph representation

The proposed architecture reorganizes the Adhoc architecture and divides the network into clusters. In the proposed architecture, clusters are formed with the neighboring nodes in an ad-hoc topology. As shown in Figure 3, node with minimum ID acts as a Dynamic Cluster Base Station (DCBS) in the proposed architecture. Cluster formed by the DCBS consists only the DCBS. Other clusters consist of Cluster Heads (CHs) and Cluster Members (CMs). Within a cluster, single hop communication exists among the CMs and CH. Two Cluster Heads are connected by the Cluster Gateways (CGs). In other words, Cluster Gateways are the communication hub for clusters. There will be no connection between any CM with the CG. Each Cognitive Terminal (CT) senses the free spectrum and selects its own operating frequency. This operating frequency is called the home channel. This home channel will be used to receive information for that particular CT. CTs transmit its home frequency’s information through the control channel within the cluster. CHs will maintain a list for the operating frequencies for all the nodes in the cluster. Every CT will listen and talk subsequently. CTs need to listen on its own frequency and control frequency. Nodes will talk in the home channel of the corresponding node and also in the control channel. As mentioned earlier, except the DCBS, all other clusters form with 1 cluster head and one or several cluster members. Communication within a cluster is done through the cluster head using common channels. A new node needs to detect the available channels before joining the network. After detecting the channels, new node sends request message for joining to the neighboring nodes. After receiving the message, neighboring nodes call the select winner procedure [6] to select the winner node from the neighboring nodes. If the CH

is the winner, it sends a message to the new node and the new node joins the cluster. If the new node is 2 hoops away from the Cluster Head, member node carrying the joinn request becomes the CG and the new node becomes the CH and forms a new cluster.

IV.

BROADCA ASTING PROTOCOL

A. Assumptions and Network Modeling M In this section, we introoduce the proposed broadcast protocol for cluster-based sim mple cognitive radio networks architecture. The assumptions of the unstructured CR network G are as follows, a) common control channel scheme for the network, b) each CT will fall into a time-slotted scheme and the duration of a time slot is adequate to broadcast, c) every CT has the knowledge of its 1-hopp neighbor, d) each transmission cycle is called as round where the t transmission cycle is divided into time slots, e) there are maximum m 5 neighboring cluster head nodes to a gateway and 19 neighboring gateways to a cluster head. t slotting B. Collision Avoidance using time

Figure 3: Proposed Cluster-based Cognitive Nettwork CoAd(G)

CLUSTERING Let an Ad-hoc Cognitive Radio Network is presented using an undirected graph (Figure 2(b)), where V is the set of vertices and E is the set of edges. We are considering an edge between two nodes iff they are withinn the transmission range with each other. Nodes in G are parttitioned into nodedisjoined clusters. The clusters are formedd with one leader node known as cluster head and member nodes where the member nodes must have edges with the leader node. Two neighboring clusters are connected to eacch other using an intermediate node known as the gateway nodde. DEFINITION 1 A cluster based structure, CoAd(G) is represented from a graph G=(V, E) with a specific node d, is a connected graph. There exists a spanning tree in CoAd(G) witth root d, which is a subgraph of G that contains every vertex of o G. In CoAd(G), each node has a status either as Cluster heead or gateway or cluster member. PROOF Let G be a connected graph. If G is cirrcuit free, it is a spanning tree. If G has a circuit, then removve an edge from G to get a new graph G1. If G1 has a circuit thenn again remove an edge from the circuit to obtain a new graph G2. Continue until k Gk is a spanning we reach a circuit free graph, Gk for some k. tree for G. DEFINITION 2 Given a graph CW(G), known as the commuunication highway is a sub graph of CoAd(G) and a connectedd graph. CoAd(G) contains only the cluster heads and the gatew ways.

Initiated from the broadcastting node, the broadcast message is flooded over the network CoAd(G). C We are using a TDM (Time Division Multiplexing) scheme to avoid the collision is while broadcasting. Each trannsmission cycle denoted as considered as round that is diivided into 19 time slots. Each relay node is assigned a unique time slot for a given round. The broadcasting stops when all thee leaves receive the message. BROADCAST PACKET The orientation of the brroadcast packet is as follows, , where is the message, number of the hop count. The initial hop 1-hop neighbor nodes and count will start from 0. After each e relay the hop count will be increased by 1 with the previouus count. For instance, if node v receives a message with the so in the relay message v increments the hop count by 1 (i.e. ( ) and transmits. TIME SLOTTING Our broadcasting protocol is i inspired from the broadcasting protocol presented in [7] wherre all are nodes of the network have unique time slots. Accordding to the protocol [7], when a node wants to broadcast, the broadcast b message first needs to send to the root of the netw work. Then the root floods the message to other nodes of thhe network. It is unlikely that collision will occur during the broadcasting process as all the nodes of the network have diffferent time slots to forward the message. For the broadcasting protocol presented in this paper, we are considering each transmisssion cycle as round where is divided into 19 time slots. Eacch relay node will be assigned a unique time slot for a given round. The time slot will be calculated as follows, (1) In (1), n is any positive integger from 1 to 19. The random number generation with the modulus operation allows 19 c 19 time slots as a different time slots. We are considering cluster head can have maximum m 19 gateways as relay nodes.

calculate the time slots using eqquation 1. Only on the calculated time slot one node can transmitt the message. Theorem 1 n CoAd(G), the proposed In a given cluster based network broadcasting protocol effectivvely delivers the broadcasted message to all of the nodes in a finite time.

Figure 4: Time slotting.

Lets consider the proposed time slottingg scheme given in Figure 4. For instance, node A wants to broaadcast a packet and it passes the message to its neighboring nodees (node a, b, c) at time t1. These 3 nodes relay the message at different d time slots (i.e. t1+α, t1+β, and t1+γ), which is calculatedd from equation 1. C. Broadcasting Algorithm Ad(G) follows the The broadcasting message in the CoA following algorithm 1.

start

5.

else while

Proof According to the proposedd protocol, cluster heads (CHs) are connected with each otherr by the gateways (GWs) in the G). Thus, in a finite number of cluster based network CoAd(G steps, the broadcasted messagee reaches all the CHs. It requires one extra step to forward the message m from the cluster heads to the cluster members. Limitations It has been found that the prroposed time slotting works well in a 2-hop communication. However, H when the hop counts increases to 3, the proposedd time-slotting scheme cannot guarantee fully collision free situation. For instance, in Figure 4, it is possible that, t1+α = t2+β. So if node a and node e has a common node at their neighboor residing in the next hop, it is possible to have a collision. We are currently working to resolve this issue by adjusting our proposed time-slotting scheme. Ensuring fully collisioon free situation in the proposed protocol can effectively deliverrs the broadcasted message to all of the nodes in O(D) rounds where w D is the diameter of the graph. V.

endif endwhile 12. endif 13. end

The broadcasting process starts from the source node, which wants to broadcast. If the source node is a cluster member, it passes the message to the cluster head. Else if the way node, it starts source node is a cluster head or a gatew flooding using the proposed algorithm. If thhe current node is cluster head, it broadcasts the message to itts 1-hop neighbors that are the cluster members and gateway. If the current node is n cluster a gateway, it broadcasts the message to the neighboring heads and gateways. This process ends onnce all the cluster members receive the broadcast message whhere no relay node exists. During the whole broadcast process, relay r nodes always

ATION RESULTS SIMULA

We simulate a communication system using the proposed dynamic cluster based archhitecture for cognitive radio networks. For the simulatioon purpose of the proposed architecture, Omnet++ is used.. We have simulated the cluster formation time for 7 nodes to 500 5 nodes network. We consider a 7-node network as a very sm mall Ad-hoc network. Then we increase the node number to 155, 25, 50, 100, 200, and 500. We consider a 500 nodes networkk to review the behavior of the architecture for a large networkk.

Figure 5: Execution tim me for cluster formation.

Cluster formation time witth different number of nodes is showed in Figure 5. In a 7 noode network, the execution time for the proposed dynamic clustter based architecture takes 0.12 ms. With a growing number off nodes of 500, cluster formation of the network according to the t proposed architecture needs

11.73 ms. The simulation environment shows that execution time for the proposed architecture depends on the nodes of the network.

[3]

[4]

[5]

[6]

[7]

Figure 6: Correlation of Clusters and nodes

As we are considering the dynamic architecture where the position of the cognitive terminals vary, we take the average result of number of clusters for each simulated networks with 10 random set of orientations. From the simulation result Fig 6, cluster numbers for the proposed architecture increases with the nodes expansions. VI.

FUTURE WORKS AND CONCLUSION

The proposed architecture breaks the Ad-hoc architecture of cognitive radio network into clusters. Nodes in the network are considered as dynamic. Development of the other communication protocols such as node joining, node leaving, data gathering for the proposed architecture will be our next research step. Moreover, we will develop a fully collision free time-slotting scheme for the architecture.

[8]

[9] [10]

[11]

[12]

[13] [14]

ACKNOWLEDGMENT [15]

This work is partially supported by MJIIT Research Grant, with Vote No. 4J044, Ministry of Higher Education (MoHE) of Technology (MJIIT) of Universiti Teknologi Malaysia (UTM) Year 2012-2013. REFERENCES [1]

[2]

J. Mitola III and G. Maguire, “Cognitive Radio: Making Software Radios More Personal”, IEEE Personal Communications, vol. 6, no. 4, pp. 13–18, 1999. Federal Communications Commission (FCC), Facilitating Opportunities for Flexible, Efficient, and Reliable spectrum Use Employing Cognitive Radio Technologies, FCC Document ET Docket No. 03-108, Dec. 2003.

[16]

[17]

[18]

I. Akyildiz, W. Lee, M. Vuran, and S. Mohanty, Next Generation/Dynamic Spectrum Access/Cognitive Radio Wireless Networks: A Survey, Computer Networks, vol. 50, no. 13, pp. 2127– 2159, 2006. K. C. Chen, Y. J. Peng, N. Prasad, Y. C. Liang, S. Sun, Cognitive Radio Network Architecture: Part I - General Structure, Proceedings of the 2nd International Conference on Ubiquitous Information Management and Communication, ICUIMC-2008, pp. 114-119 Y. Zhao, Performance evaluation of cognitive radios: Metrics, utility functions, and methodology, Proc. IEEE, vol. 97, no. 4, pp. 642–659, 2009. J. Uchida, A.K.M. M. Islam, Y. Katayama, W. Chen, and K. Wada, “Construction and maintenance of a novel cluster-based architecture for ad hoc sensor networks", Journal of Ad Hoc & Sensor Wireless Networks, Vol. 6 No. 1-2, 2008, pp. 1-31. W. Chen, A.K.M.M. Islam, M. Malkani, A. Shirkhodaie, K. Wada, M. ZeinSabatto, “Novel Broadcast/Multicast Protocols for Dynamic Sensor Networks”, IEEE International Symposium for Parallel and Distributed Processing, March 2007. B. Wang, K. J. R. Liu, Advances in Cognitive Radio Networks: A Survey, IEEE Journal of Selected Topics in Signal Processing, vol. 5, no. 1, February 2011, pp. 5-23. I. F. Akyildiz, X. Wang, W. Wang, Wireless Mesh Networks: A Survey, Computer Network, vol. 47, no. 4 (Mar. 2005), pp. 445-487. S. Krishnamurthy, M. Thoppian, S. Venkatesan, R. Prakash, Control channel based MAC-layer configuration, routing and situation awareness for cognitive radio networks, IEEE Military Communications Conference, 2005. MILCOM 2005, pp. 455-460 T. Chen, Z. Honggang, G.M. Maggio, I. Chlamtac, CogMesh: A ClusterBased Cognitive Radio Network, 2nd IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks, 2007, DySPAN 2007, pp. 168 -178. D. Maldonado, B. Le, A. Hugine, T. W. Rondeau, C. W. Bostian, Cognitive Radio Applications to Dynamic Spectrum Allocation, 1st IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks, 2005, pp. 597-600. S. Ni, Y. Tseng, Y. Chen, and J. Sheu, “The broadcast storm problem in a mobile ad hoc network”, MOBICOM, pp. 151–162, August 1999. C. Xin, A Cognitive Radio Network Architecture without Control Channel, IEEE Global Telecommunications Conference, 2009, Nov. 2009. Q. Zhao, S. Qin, Z. Wu, Self-Organize Network Architecture for MultiAgent Cognitive Radio Systems, International Conference on CyberEnabled Distributed Computing and Knowledge Discovery (CyberC), 2011, pp 515-518. A.A. Sreesha, S. Somal, L.I. Tail, Cognitive Radio Based Wireless Sensor Network architecture for smart grid utility, IEEE Long Island Systems, Applications and Technology Conference (LISAT), May 2011. H. Zheng, L. Sun, T. Hui, A Framework of Access Network Architecture for 4G Systems based on Cognitive Radio, 5th International Conference on Wireless Communications, Networking and Mobile Computing (WiCom '09), 2009. C. Ghosh, D.P. Agrawal, ROPAS- cross layer cognitive architecture for mobile adhoc, 2nd International Conference on Cognitive Radio Oriented Wireless Networks and Communications (CrownCom 2007).