Connectivity, Energy and Mobility Driven Clustering Algorithm for Mobile Ad Hoc Networks

Transcription

1 Connectivity, Energy and Mobility Driven Clustering Algorithm for Mobile Ad Hoc Networks Fatiha Djemili Tolba University of Haute Alsace GRTC Colmar, France Damien Magoni University Louis Pasteur LSIIT Strasbourg, France Pascal Lorenz University of Haute Alsace GRTC Colmar, France Abstract In the context of mobile ad hoc networks (MANETs) routing, we propose a clustering algorithm called Connectivity, Energy and Mobility driven Clustering Algorithm (CEMCA). The aim of CEMCA consists in appropriately choosing the cluster head to reduce routing overhead. In order to reduce traffic and energy consumption, the control messages are sent only when needed, according to the speed of the node. Each node has a quality that indicates its suitability as a cluster head. This quality takes into account the node connectivity, battery energy and mobility. These parameters are very important for the stability of the cluster. Simulation experiments are carried out to validate our algorithm in terms of stability of the clusters and their members and the quality of the connectivity. The results are compared to a previous approach called Weight Clustering Algorithm (WCA) and they show that CEMCA is performing better. I. INTRODUCTION Routing algorithms for ad hoc networks are responsible to establish low-cost and high quality route in rather dynamic environments. Indeed, the clustering algorithm can cooperate with routing algorithm in order to find a better route. This algorithm consists in dividing the network into a set of interconnected nodes covering the total node population, this set of nodes is called cluster. The clustering algorithm is focused on the reduction of the route discovery overhead and the resources optimization such as battery power, network capacity and bandwidth. Some nodes in the cluster are named cluster heads and are responsible for several tasks among others: formation of cluster and its members and maintenance of the topology of the network. Due to the movement of mobile nodes, the network topology changes frequently. Consequently, the cluster stability will be perturbed and the reconfiguration of cluster head will be necessary. For these reasons, a good clustering algorithm should be stable to the radio motions, that is, should not change cluster configuration drastically and often when a few nodes are moving [1]. The remainder of this paper is organized as follows. Section 2 outlines the previous work proposed for cluster formation and its limitations. Section 3 presents our contribution and details the proposed algorithm. The performance and evaluation of our algorithm are presented in section 4 with simulation results. We conclude the article and we present some future research directions. II. RELATED WORK The concept of clustering in ad hoc networks was treated in several research works. These works can be divided in three major trends of development. The First area of research focused on high connectivity degree [2] [3]. The second focused on Low node identifier (ID) heuristic [4] [5], the possibility to combine these two algorithms was also developed. The last and recent area is the weight of node that takes into account the characteristics of node in the election of cluster head [6] [7] [8] [9] and [10]. In this paper, we propose a new clustering algorithm CEMCA that adapts to the needs of the network and the dynamic nature of nodes. Indeed, CEMCA has the aim to implement a platform usable by the routing algorithms of MANETs in order to facilitate the dynamic setup of the routes. The dissimilarities between CEMCA and other existing work consist in the importance and the equality between the parameters of choice (energy, connectivity and mobility) and in the conservation of both the energy and the connectivity by exploiting the SEMC algorithm [11]. Otherwise, the node with a high mobility or a low reserve of energy are not qualified to contribute to the election of cluster head. In this way, we avoid more cluster head re-elections. III. OUR CONTRIBUTION A. Principle of CEMCA Our aim is the design of an algorithm that selects appropriate nodes as cluster heads and creates clusters in a large ad hoc network. The idea of CEMCA is to find a number of cluster heads that ensures the stability of the network topology. The election of the cluster head is based on the combination of several significant metrics such as: the lowest node mobility, the highest node degree, the highest battery energy and the best transmission range. This algorithm is completely distributed and all nodes have the same chance to act as a cluster head. In the following, we explain the main objectives for the choice of each parameter. Mobility: as we mentioned above, the main characteristic of the ad hoc network is the dynamic topology, so we must adapt the algorithm to support this topology. It is necessary that the cluster head changes as less as possible when it moves. So, we choose as a cluster head the

2 node that moves slowly otherwise the cluster may be broken. For example, when a node moves quickly, it can be detached from their neighbors and that causes a reduction in the node degree. Moreover, it is possible that this node moves into another transmission range, i.e. another cluster. Connectivity: in order to ensure that the cluster head properly functions, it is suitable that the cluster head has some neighbors in its cluster. To avoid producing more cluster than necessary is another advantage of this parameter. For this, we favor in CEMCA the node that has the maximum number of neighbors. Battery energy: each mobile node has a defined amount of energy. If this amount is low the node can neither send nor receive any data. In order to increase the lifespan of the cluster head and to avoid frequent topology changes, it is desirable to take into account the battery energy. Therefore, it is important to choose the node with a high amount of energy. Transmission range: is a main factor in the choice of the cluster head because it can keep connectivity and save on the consumption of energy at the same time. A larger transmission range causes a increase in consumption of energy, while a smaller value saves energy. On the other hand, a larger transmission range maintains connectivity. For these reasons, we have used our SEMC algorithm [11]. In this algorithm the node adapts its transmission range according to the distance between itself and its farthest neighbor. In this case, the node keeps always the same number of neighbors (same connectivity) and it preserves the energy by reducing its transmission range. B. Description of CEMCA In the following we describe CEMCA in mobile ad hoc networks. Before proceeding with the presentation of the various steps of the algorithm we describe the system model. We consider a network topology which is represented by a graph G = (V,E) where V is the set of mobile nodes and E is the set of edges e = (u, v) where e models a wireless link that interconnect two nodes. A pair of nodes u and v are in a direct communication if and only if they are within their wireless transmission range. Specifically, the nodes change the information between them at two hops for a better knowledge of their neighbors. Note here that each node computes locally its own value of each parameter cited above. CEMCA is composed of two main stages. The first stage consists in the election of the cluster head and the second stage consists in the grouping of members in a cluster. The different steps are described below: Stage 1: Election of the cluster head In this stage we need the position of nodes, thus it is necessary that each node broadcasts its ID to all its neighbors in the same transmission range. Each neighbor that received the broadcasted message can estimate its distance from the strength of the signal received. Global Position System (GPS) can be another solution, however it has the disadvantage of more energy consumption. Step 1: Calculate the mobility of the node The mobility is evaluated periodically in order to expect the future state of the network, it is defined as follows: S i = 1 t D i(t) D i (t + t) (1) D i = 1 n dist (i,j) (t) n (2) j=1 dist (i,j) = (x i x j ) 2 +(y i y j ) 2 (3) S i : the relative speed of the node vs the other nodes. D i (t): the average distance between the node i and all its neighbors at time t. n: the number of neighbors of node i. dist (i,j) (t): the distance between node i and node j. x and y: the coordinates of the node. Step 2: Compute the node degree C i (t) = e i,j (t) (4) where e i,j is a direct wireless link between node i and j at time t. Step 3: Calculate the energy level of the node The energy is evaluated periodically in order to allow the adaptation to the future state of the network and is defined as: E i ( t) =Power i t (5) So, the total energy left can thus be expressed as: E i (t + t) =E i (t) E i ( t) (6) Step 4: Calculate the quality of the node Compute the ratio normalized to 1 for each parameter using three defined constants S max, C min and E max. Thus: R s = (1 S max ) S i (7) R c = C i C min N 1 C min (8) R e = E i E max (9) After this, we calculate the quality normalized to 1 for each node by pondering each parameter by a coefficient. We assume for the moment that each coefficient is equal to 1/3. The quality of the node is a measure of its suitability as a cluster head. Q i = 1 3 (R s + R c + R e ) (10) The node broadcasts its quality to their neighbors in order to compare the better among them. After this, the node that has the best quality is chosen as a cluster head.

3 Stage 2: Formation of the cluster and its members This stage is the final step of the algorithm. Here, we present the construction of the cluster members set. Each cluster head defines its neighbors at two hops maximum. These nodes form the members of the cluster. Next, each cluster head stores all information about its members, and all nodes record the cluster head identifier. This exchange of information allows the routing protocol to function in the cluster and between the clusters. Due to the dynamic topology of ad hoc networks, the nodes tend to move in different directions and at different speeds, thus the configuration of clusters is necessary. Consequently, the system must be updated periodically. This update concerns the position of the nodes and their speed. The speed of a node is responsible for the change in its position. For this reason, the choice of the update time-slot is be based on the speed of the node. If the mobility of the node is low, we suppose that its position does not change much and consequently we choose a longer time-slot, thus we avoid updates when it is not necessary. A periodical update with a higher frequency results in a great consumption of battery power and consequently, to more configuration changes. IV. SIMULATION STUDY A. Simulation parameters and metrics The performance evaluation of CEMCA is carried out by simulations by using the ns-2 network simulator. We consider networks with 20 and 40 nodes. At first, we use a commonly used mobility model such as the Random Way Point model [12] that represents a particular case of the Random Trip model [13]. In the Random Way Point model, each node chooses its direction and its speed according to a fixed time interval [14]. The move of the node is done during a time t with some pauses. However, the possibility of no pause is also used and this represents the Random Walk model. Other basic parameters used in our experiments are summarized in the table below: TABLE I SIMULATION PARAMETERS Parameters Values Number of nodes 20, 40 Area 100 x 100 m Maximum transmission range 70 m Maximum speed of the nodes 2-10m/s In order to assess the performances of our algorithm, we choose to study two metrics that are: 1) The average number of clusters as a function of the maximum speed of the nodes in the network. 2) The connectivity as a function of the average transmission range and the maximum node speed. We also study the relationship between the average number of clusters, the average transmission range and the connectivity. B. Simulation results All the results in this section have been calculated with a confidence interval of 0.95 and a relative error threshold of 5%. The efficiency of the clustering algorithm is measured by the number of clusters that it produces [1]. So, we compare the number of clusters obtained by CEMCA with those obtained by the Weighted Clustering Algorithm (WCA) [8]. Fig. 1 shows the average number of clusters as a function of maximum speed of node in the network for WCA. We observe that the average number of cluster varies between 6,50 and 7,50 for 20 nodes where this number varies between 8 and 8,50 for 40 nodes according to the speed of node. Fig. 1. Average number of clusters as a function of the maximum speed for the WCA. In fig. 2, we observe that the average number of clusters remains stable for 20 and 40 nodes whatever the speed of nodes. This stability is proved quantitatively by the number of cluster that is limited in a small interval [6,50-6,60] for 20 nodes and [7,40-7,60] for 40 nodes. Further, we observe that the number of clusters produced by CEMCA is lower than these produced by WCA. These results can be explained by the capacity of node to be responsible, for a long time, to its members. This capacity that is expressed by greatest reservation of energy and lowest mobility. For these reasons, the clusters remain stable and the topology less reconstructed in spite of the change of speed. In this section, we take the connectivity as second parameter of evaluation for CEMCA. This is because, in the one hand, the connectivity is primordial for any routing algorithm. In the other hand, the connectivity is essential in order to ensure the responsibilities of cluster head. We define the connectivity by the probability that a node is reachable at any other node in the network. For a single component graph, any node is reachable at other nodes, thus, the connectivity is equal to 1 [8]. So, largest connected component connectivity = N where: N is the total number of nodes in the network. (11)

4 Fig. 2. Average nb of clusters as a function of the max speed for CEMCA. Fig. 4. Connectivity as a function of the transmission range for the CEMCA. Fig. 3 shows the evolution of the connectivity according to the transmission range for WCA algorithm. We observe that the connectivity is very low when the value of transmission range is small (Tr = 10). After this, the connectivity increases slowly until the transmission range is equal to 30 and it will be good when the transmission range is larger. Consequently, a well connected graph can be obtained at the cost of a higher transmission range [8]. Consequently, by using CEMCA, all the nodes can communicate between themselves. The two previous figures show that CEMCA outperforms WCA as it is able to have the same connectivity by using much more reduced transmission ranges thus saving energy and reducing radio interferences. Fig. 5 illustrates the connectivity according to the node speed for CEMCA. We observe also that the connectivity is high whatever the average speed of the nodes. Although the value of the transmission range adapts itself according to the number of neighbors, the value of speed does not influence the connectivity. Fig. 3. Connectivity as a function of the transmission range for the WCA. Fig. 4 shows the connectivity according to the average transmission range for CEMCA. We talk about the average transmission range because we use our SEMC algorithm that allows each node to adapt the value of its transmission range according to its needs. For this reason, we observe that the connectivity remains good in spite of the small value of the transmission range that does not exceed 10. Moreover, quantitatively the connectivity is comprised between 0,948 and 0,965. In this way, we can show that the connectivity is maintained by using our SEMC algorithm at the data link layer [11]. Fig. 5. Connectivity as a function of the maximum speed for the CEMCA. Fig. 6 depicts the relationship between the average transmission range, the average number of clusters and the connectivity in a 3D plot. We see that the average transmission range remains short but the connectivity is high. Moreover, the average number of clusters is low and almost the same. Thus, we confirm that CEMCA makes a good compromise between these essential parameters that play an important role in the communication between the nodes in terms of performances.

5 Fig. 6. Transmission range as a function of both the number of clusters and the connectivity. V. CONCLUSION In this paper we proposed a clustering algorithm called CEMCA that can be applied in wireless mobile ad hoc networks in order to dynamically optimize the cluster head choices. CEMCA is based on an ideal combination of the characteristics of a node namely: battery power, connectivity and mobility. These characteristics play an important role in the stability of the network by reducing the number of cluster head changes. Moreover, CEMCA makes use of our algorithm [11] in order to choose the appropriate transmission range of each node that maintains connectivity and saves energy. We conducted simulation experiments with ns-2 in order to measure the performance of CEMCA. Our results are compared with the Weighted Clustering Algorithm (WCA) [8] and they show a significant advantage of CEMCA over WCA in terms of the number of produced clusters and their stability. Quantitatively, the average number of clusters varies in a small interval from 0.10 to Also, the connectivity between the nodes is typically above 0.95 and better than with WCA in spite of the small value of the transmission range. Consequently, the communication in the cluster or between the clusters will be good. In the future, we intend to test CEMCA with other models of individual mobility such as Weighted Waypoint model [15] or group mobility such as Community Mobility model [16] or Reference Region Group Mobility model [17]. We also plan to implement CEMCA in a MANET routing protocol. REFERENCES [1] G. Chen and I. Stojmenovic, Clustering and routing in mobile wireless networks, in Technical Report TR-99-05, [2] D. Baker and A. Ephremides, The architectural organization of a mobile radio network via a distributed algorithm, IEEE Transactions on Communications, vol. 29, no. 11, pp , [3] A. Ephremides, J. Wieselthier, and D. J. Baker, A design concept for reliable mobile radio networks with frequency hopping signaling, in Proc. of the IEEE, 1987, pp [4] A. Parekh, Selecting routers in ad hoc wireless networks, in Proc. of the IEEE International Telecommunications Symposium, [5] M. Gerla and J. Tasi, Multicluster, mobile, multimedia radio network, ACM Journal of Wireless Networks, vol. 1-3, pp , [6] S. Basagni, Distributed clustering for ad hoc networks, in Proc. of the IEEE International Symposium on Parallel Algorithms, Architectures and Networks, 1999, pp [7], Distributed and mobility-adaptative clustering for multimedia support in multi-hop wireless networks, in Proc. of the IEEE Vehicular Technology Conference, 1999, pp [8] M. Chatterjee, S. Das, and D. Turgut, Wca: A weighted clustering algorithm for mobile ad hoc networks, Cluster Computing Journal, vol. 5, no. 2, pp , [9] F. Nocetti, J. Gonzalez, and I. Stojmenovic, Connectivity based k-hop clustering in wireless networks, in Telecommunication Systems, vol. 22, 2003, pp [10] S. Jang, E. Kwon, J. H. Kim, and J. Lee, A weighted stable clustering scheme based on the mobility prediction in mobile ad hoc networks, in Proc. of the IEEE International Conference on Wireless and Mobile Communication, Bucharest, Romania, [11] F. D. Tolba, D. Magoni, and P. Lorenz, Energy saving and connectivity tradeoff by adaptive transmission range in g manets, in Proc. of the IEEE International Conference on Wireless and Mobile Communication, Bucharest, Romania, [12] W. Navidi and T. Camp, Stationary distributions for the random waypoint mobility model, IEEE Transactions on Mobile Computing, vol. 3, no. 1, [13] J. Y. L. Boudec and M. vojnovic, Perfect simulation and stationarity of a class of mobility, in Proc. of the IEEE Infocom, [14] G. Lin, G. Noubir, and R. Rajaraman, Mobility model for ad hoc network simulation, in Proc. of IEEE Infocom, [15] W. J. Hsu, K. Merchant, H. W. Shu, C. Hsu, and A. Helmy, Weighted waypoint mobility model and its impact on ad hoc networks, in Proc. of the Mobile Computing and Communications Review, [16] M. Musolesi and C. Mascolo, A community based mobility model for ad hoc network research, in Second ACM International Workshop on Multi-hop Ad Hoc Networks, [17] J. Ng and Z. Yan, Reference region group mobility model for ad hoc networks, in Second IFIP International Conference on Wireless and Optical Communications Networks, 2005, pp