You are on page 1of 7

A Framework for Simulation and Evaluation of Cross-Layer Techniques in Wireless Sensor Networks

Danny Alexandro Mnera, Natalia Gaviria


Universidad de Antioquia Grupo de Investigacin en Telecomunicaciones Aplicadas, GITA Medelln, Colombia danny.munera@udea.edu.co, nagaviri@udea.edu.co
Abstract Cross-Layer design is a new technique which can be used to improve the performance in wireless sensor networks under the severe restrictions of the energy consumption. Crosslayer techniques intended to create an optimization of the limited resources, taking into account factors associated with different layers of traditional communication schemes, opening a new way for the development of complex communications mechanism. However, because these techniques are based on a non-standard architecture, its structure has a lot of dissimilar characteristics that obstruct a comprehensive comparison of its performance. In this paper, we propose a framework for simulation and evaluation of cross-layer techniques in wireless sensor networks. Our framework promotes a fairness comparison of different communication schemes, a realistic model simulation, an experiment design and a rigorous analysis of results with ANOVA (Analysis of Variance) technique. As a case study, we test our comparative simulation framework with to two popular cross-layer schemes showing the facility of use and the result of the analysis of the data. Keywords-component; Simulation Framework, Wireless Sensor Network (WSN); ANOVA; Cross-Layer Protocols; Simulation Models

Under the analysis of the communication process, there is a new trend of techniques that can achieve a comprehensive communication solution. The main goal of these techniques is to make decisions in the different layers according to the main features involved in the network performance, outperform the single-layer protocols [7][8]. This new technique is called cross-layer (XL). Currently, there are many cross-layer communications designs [9],[10],[11] and [12], which implement various techniques, allowing better network performance in terms of power consumption and quality of service. The comparison in the performance of these techniques is a complex problem, because, they are implemented with very unique structures and behaviors. This paper presents a framework for simulation and evaluation of cross-layer techniques in WSN. In this framework we promote a fairness comparison, a realistic model and a rigorous result analysis whit the ANOVA technique. We develop the simulations models in the framework over the OMNeT++ simulator [13] and the mobility framework [14]. We present a case study as an example of the use of the framework. The comparison of two popular cross-layer schemes for wireless sensor networks, the ALPL (Adaptive Low Power Listening) proposed by R. Jurdak in [12], and XLM protocol proposed by I.F. Akyildiz in [11] is developed and analyzed whit the guidelines of this framework. A thorough of the state-of-art of cross-layer techniques for WSN, shows that the most interesting and novel approaches are the proposed by Jurdak et al. [12] and Akyildiz et al. in [11]. The rest of this article is organized as follows: section II presents related work, section III presents the framework specification; section IV contains the case study. In section V we show the results analysis with the ANOVA tool, followed by the conclusions in section VI. II. RELATED WORK

I.

INTRODUCTION

The use of new wireless communications technologies in areas such as habitat monitoring, environmental control [1][2], industrial control systems [3], and medical applications [4], among others, has increased the interest in technologies like wireless sensor networks (WSN) in recent times. This type of network provides solutions for monitoring of physical and environmental variables in various applications and is very attractive due to features such as ease of implementation, low cost and wide coverage [5]. The nodes that make up these networks are low power and low cost devices based on embedded systems. In addition to capturing environmental data, convert and process information, the nodes act as communication agents using a radio interface. This set of functions implies a high computational complexity, which can translate into high power consumption and therefore limits its autonomy [6]. One of the main technological challenges is to reduce the power consumption of the devices while maintaining the required functionality. Multiples approaches have been proposed in order to achieve this reduction in power ranging from the design of the hardware platform to the application-level optimization and the analysis of network protocols and communications.

A. Cross-layer comunication techniques Cross-layer protocols do not follow the layered structure of traditional networks in order to access the data from each stage of communication and achieve a more effective optimization of resources. It is precisely this feature of not following the standard for conventional networks, which generates that there are multiple proposals with different approaches, behavior and architectures.

A revision that attempt to formulate a cross-layer architecture taxonomy is [7], which classifies these techniques into three different proposals (Figure 1). The first (Figure 1 A) refers to the creation of many additional communication interfaces enabling the exchange of information between the layers. The second (Figure 1 - B) refers to proposals that maintain the layers structure, but implementing a communication module shared between all the layers. Finally (Figure 1 - C) relates to proposals that abandon the layer concept and implements the functionality of the communication through alternative methods (for more information about cross-layer protocol see [15]).

S h a re d D B
A d d . In te r fa c e

with another, it must have access to its routing table to read the optimal listening mode of the node, sending the preamble with sufficient duration and engage the transmission of information. The local state of a node is defined by three elements. The first one is the estimated number of descendents. The second parameter is the current radio duty cycle (if it is high, it indicates that the node has had a high activity). The last parameter is the node role, which indicates the sensing activity. The global optimization is achieved by changing the topology of the network according to some factors. The network layer implements a modified cost function of the routing algorithm by adding some factors that represent the cost of radio activity and the cost per sensing activity. Thus, the cost of using node M as parent node is estimated by the equation (1).

D ir e c t in te r - la y e r c o m m u n ic a tio n A

S h a re d D a ta B a s e B

N e w a b s tr a c tio n ( N o la y e r ) C

(1)

Figure 1. Cross-layer architecture taxonomy

Next, a brief review of the functionality of the protocols worked in the case study is presented. ALPL Jurdak et al.[12], propose a cross-layer optimization framework, with local and global operations. Initially, the nodes are enabled to adjust their behavior at the MAC and routing layers, according to its local state. In the network layer, extending the cost function for routing decisions, in order to include information about MAC or physical layer (radio duty cycle) and information from the application layer (sensors role). This proposal leaves the chance to include QoS information, which may contain values of latency and reliability. The core of this work is the ALPL (Adaptive Low Power Listening), which controls the listening mode of the MAC protocol B-MAC, according to the locally calculated state information. Due to the fact that this work is a framework description, the definition of the states is up to the designer of the network, according to the objective of the implementation and the specific QoS requirements. The network layer uses a proactive protocol, assuming a static monitoring implementation. Furthermore, through regular exchange of routing update packets, the nodes are allowed to exchanging information of its local node status. In the local decision, the nodes can choose the optimal listening mode. At the beginning, all the nodes choose an initial listening mode, called Linit. Then, they send messages of presence and status. With this information, each node creates data flows to the SINK and calculates its local status based on the number of descendants in the routing tree. The node chooses then its optimal listening mode. In the next routing update period, each node announces its listening mode to all its neighbors. This information is stored in the routing table of each node. Thus, if a node will wish communicate

where, is the original cost function based on the link quality, is the node M radio activity cost, is the sensing cost in the application layer, are constants that weighted this factors, is the hop count to the node M and is the hop count to actual routing parent. XLM A unified optimization scheme is worked in the XLM. This framework is developed to achieve an efficient and reliable communication in WSN with minimal energy expenditure. XLM combines features of the common protocols of each layer in a simple XL module, which can be easily implemented with the limited resources of sensor nodes [11]. The operation of the XLM is built on the basis of a new concept: "initiative determination". The value of the initiative determination is calculated based on four variables, each representing a threshold that must be satisfied. The initiative is one if all four conditions are met. The variables are related to the received signal to noise ratio, the transmitted packets (generated and forwarding), and the remaining capacity of the buffer and the remaining energy of the node. If the initiative is denoted as I, it is determined as shown below in equation (2). (2)

The concept of initiative determination is based on the principle that each node decides to participate in the communication. Thus, if a node receives an RTS packet, it

immediately evaluates (2) determining the link quality, local traffic, the transmission buffer overflow and the remaining energy of the node. If all these values satisfy the threshold and is an initiative, the node starts a competition with neighbors for sending the CTS packet. III. COMPARATIVE SIMULATION FRAMEWORK SPECIFICATION

With the results of the ANOVA can easily conclude whether the differences between the various metrics evaluated are significant or not. Furthermore, it allows analyzing the interaction of several network factors that may be also evaluated from the statistical point of view. In this way, the performance analysis of the results test is no longer an argument addressed by observations and trends, now it becomes in a strong and statistically valid result. IV. CASE STUDY Next, we present the results of the implementation of the simulation and evaluation framework, in the comparison of two cross-layer communication protocols quite popular in the scientific community the ALPL in Jurdak et al. [12] and the XLM in Akyildiz et al. in [11]. A. Defining the problematic features There is a difference specifically in the operating conditions of the application, XLM protocol is designed for operation on high-density networks (larger quantity of nodes in a confined space), while the performance of the ALPL does not take into account this factor. In our case study provides a network with a normal node density, so to reach a fair evaluation the XLM protocol was adapted to work in this environment with the inclusion of a preamble in sending the first message. After a thorough analysis of the specification of the protocols we specify that a network application monitoring sensors would have no effect on the proposal inequitable basic schemes. B. Simulation Models Implementation This section details the simulation models implementation of the XLM and ALPL frameworks, under the simulation engine OMNeT++ and the Mobility Framework. 1) Adaptative Low Power Listening Simulation Model The model consists of a set of modules in OMNeT++. The ALPL technique is implemented in the network and MAC component. This cross-layer framework has two basic schemes: the first is the decision of the local state, which decides the nodes optimal listening mode based on current local data. This information is communicated to the MAC layer B-MAC [16]. The second scheme is the global optimization, which is implemented in the routing algorithm by choosing the parent of each node. The local decision system is developed based on B-MAC protocol. In its definition, this protocol enables the use of a special listening mode selected a priori, which remains consistent with the workload of the network. The idea is to make decisions according to the nodes local state, so that it can adaptively and at runtime choose the optimal listening mode according with its workload. The ALPL framework does not describe in detail how to define the state of the node, because it is part of a design decision. In our case, we propose a model of inference from a state machine in which there are three entries representing the nodes current listening mode, the number of descendents and the role of the node (Figure 2). The output of the inference system will be the current listening mode.

This framework proposes some basics steps that can provide an effective performance evaluation these schemes. Initially, we must ensure basic operating principles that generate equity in terms of application features and the simulation model used. Then you need to implement appropriate and realistic simulation models, which follow the main cross-layer features that you want to evaluate. Finally, after choosing the proper experimental design is necessary to process the results from the ANOVA statistical technique for analyzing the results from the statistical rigor. A. Fairness Comparison It is clear that the development of new proposals for crosslayer communication, are often linked to special features of the applications (eg. protocols specifically designed for monitoring applications or protocols specifically designed for high density networks). When it is evaluating cross-layer techniques it is needed to be clear about these special features that affect the fairness of the comparison. If it wants a fair comparison, should reach a common point, where the proposals involved have an acceptable performance. To achieve this goal, we must remember that many proposals raise design framework that provide a degree of flexibility in the implementation of the basic features. B. Simulation Models With the result of previous equity criteria in this step are implemented simulation models of each scheme to be evaluated. This step requires a thorough analysis of the fundamental principles of the scheme to assess the implementation always trying to be a realistic and comprehensive model of the chosen technique. For our case study, the implementation of these simulation models was developed on the platform OMNeT++ and the Mobility framework, which provide useful tools for modeling patterns of cross-layer communication. C. Experiment Design The experimental design is based on a proper statistical theory that enables the achievement of valid and efficient tests. The idea here is to define factors, metrics, scenarios and methods, so that the development of the tests is carried out successfully from a statistical standpoint. On the metrics of interest should always be considered that although the study is based on a reduction in the network energy consumption, quality of service (QoS) metrics play a key papal when evaluate the performance cost of each implementation. D. Results analysis with ANOVA ANOVA statistical tools do possible to obtain conclusive results from samples obtained from the experimental design.

The local decision-making system is implemented in the network layer, so the module has to use a special simulation mechanism to pass information between layers (In the Mobility Framework this tool is the BLACKBOARD). Two roles were defined at the application layer of the nodes. These are represented by the Boolean variable APPSTATE. If APPSTATE = True, the node is in a busy role which indicates a high rate of generated packets. If APPSTATE = False, the node is in a passive role and have a low rate of generating packets. In the link layer, the B-MAC protocol delivers the state of the current listening mode in the node, which can take one of the following values: 1.6s, 0.8s, 0.4s, 0.2s, 0.1s, 0.05s, 0.02 s, 0.01s. Each of these values is represented by a state in the state machine. The network layer stores the number of descendents that has the current node. This value represents the number of nodes that depend on the current node to forward messages to the SINK. Thus, when the number of nodes descendants raise, the nodes occupancy raise too. There are eight states, each corresponding to a listening mode. When the number of descendants of a node increases, it will gradually enter into a state with a corresponding listening mode with lower SLEEP time. Therefore it will be active more frequently and will also increase its duty cycle. However, whether the number of descendents decreases, it means that the node will be free. Thus, the state machine gradually changes to a state with a longer SLEEP time, so that the node will be active with a lower frequency and the radio duty cycle decreases. The global decision system is implemented at the network layer, more precisely in the cost function that chooses the routing parent. Equation (1), the modified cost function, has two additional terms (C(radio) and C(sensing)), which relate to the costs involved in the use of radio and the role of the node, compared with other neighbors. With information from neighboring nodes received through the routing update packages, each node stores in its routing table the relevant information from its neighbors. When it is time to choose its routing father, this information is processed
A P P S T A T E = T R U E
D e s c > 2 D e s c > 3 D e s c > 1

according to (1). 2) XLM Simulation Model The model is composed by a set of modules in OMNeT++. In this case, all the functionality of the network and MAC layers are joined in a single module called XLM. The main functionality of this module resides in the initiative determination technique and the received-based contention. The initiative determination needs some network information to be implemented. Hence, the module has to integrate the network and MAC functionalities together with the information about the congestion and the quality of the link (local information). The only external information needed is the nodes remaining energy, which is obtained via the BLACKBOARD module. For the Received-based contention functions, it was necessary to adapt the original functionality of the protocol. The problem is that this protocol was developed for highdensity WSN, so the first message to initiate the communication is sent without any preamble, assuming that at least one node can hear the communication. Given that our simulation scenario does not provide a high density of nodes, a mechanism was implemented preamble in the Received based contention only for the first message in the handshake process, making sure that neighboring nodes are listening to the initial communication. C. Experiment Design Development The planning of experimentation is crucial in the development process of an investigation, as this directly impacts the quality of design, and cost whether monetary or time. It is reasonable to use an appropriate experimental design methodology under a statistical model that allows obtaining conclusive results. Our main factor to investigate is the effect of the communication protocols scheme on the sensor network performance. In order to achieve this goal, we developed in previous section the simulation model of the two schemes that have been termed ALPL Jurdak et al. and XLM Akyildiz et al. In addition, we also wish to determine the factor of network

D e s c > 4

D e s c > 5

D e s c > 6

D e s c > 7

L is te n in g M o d e 1 .6 s

L is te n in g M o d e 0 .8 s

L is te n in g M o d e 0 .4 s

L is te n in g M o d e 0 .2 s

L is te n in g M o d e 0 .1 s

L is te n in g M o d e 0 .0 5 s

L is te n in g M o d e 0 .0 2 s

L is te n in g M o d e 0 .0 1 s

D e s c = 2 D e s c < = 1 D e s c < 2

D e s c = 3 D e s c < 3

D e s c = 4 D e s c < 4

D e s c = 5 D e s c < 5

D e s c = 6 D e s c < 6

D e s c = 7 D e s c < 7

D e s c > = 8 D e s c < 8

A P P S T A T E

= F A L S E

D e s c > 2

D e s c > 3

D e s c > 4

D e s c > 5

D e s c > 6

D e s c > 7

D e s c > 8

L is te n in g M o d e 1 .6 s

L is te n in g M o d e 0 .8 s

L is te n ig M o d e 0 .4 s

L is te n in g M o d e 0 .2 s

L is te n in g M o d e 0 .1 s

L is te n in g M o d e 0 .0 5 s

L is te n in g M o d e 0 .0 2 s

L is te n in g M o id e 0 .0 1 s

D e s c = 3 D e s c < = 2 D e s c < 3

D e s c = 4 D e s c < 4

D e s c = 5 D e s c < 5

D e s c = 6 D e s c < 6

D e s c = 7 D e s c < 7

D e s c = 8 D e s c < 8

D e s c > = 9 D e s c < 9

Figure 2. Local state decision in the ALPL communication scheme.

traffic to verify its influence on the response variables. The communications protocols schemes aim to improve network performance in order to decrease power consumption. But this always has a cost in other outcome variables of the network that affect the systems QoS. When choosing a specific communication scheme, the network designer has to understand the limitations in terms of QoS, in order to choose the scheme that provides the lowest power consumption and the most acceptable QoS compromise. It is clear that our analysis must take into account other variables beyond the energy consumption of the network. The QoS factors analyzed in this study are: the average end-to-end latency and the goodput perceived by the SINK node. These three response variables of the experiment are discussed below. Average energy consumption per data packet received: This metric represents the average energy cost that the network spends when performing a transaction of a data packet. In the experiment, we used the following equation to estimate this metric: (3)

ALPL and the XLM. The second factor is the network traffic, which also has two levels; high (one packet per node generated every minute) and low (one packet per node generated every 10 minutes). The treatments used in our case are described in Table 1.
Communication Schemes ALPL - XML Traffic High Low Energy Consumption per Packet End-to-End Latency Goodput

Figure 3. Experimental Design.

The experimental unit consist of a WSN with 100 nodes, random deployed in a simulation grid of 100m x 100m. Each node uses the model of the TI CC2420 radio transceiver and is configured to 0.01 mW in the power of the radio. In the TABLE 1 is showed others simulation parameters neded.
TABLE 1. SIMULATION PARAMETERS Parmeter U Value Radio Payload B 16 delaySetupRx s Traffic Exponential delaySetupTx s type delayRxTx s Deployment Area X m 100 delayTxRx s Y m 100 sleepPower W Nodes 100 rxPower W txPower W rxtxPower W txRxPower W setupTxPower W setupRxPower W

where is the average energy consumption per data packet received in the network, is the energy consumption registered in node , is the total number of nodes in the network, and is the total number of packets received in the sink node. Average end-to-end latency: This metric indicates the average delay that a data packet experiences from de origin to the destination. In our experiment, we evaluate this metric using the following equation: (4)

0.001792 0.001792 0.000192 0.000192 6.93E-08 0.06204 0.0363 0.06204 1.06204 0.002109 0.002259

where is the average end-to-end latency, is the latency perceived by the packet and is the total of packets received in the sink node. Goodput: In general, the goodput is defined as the number of successfully transmitted information bits per unit time. In our case, each data packet has a data load of 16 bytes; hence the goodput in bits per second is calculated as follows: (5) where is the goodput of the network, packets received in the sink node, and interval of the network. is the total of is the active time

In this experiment is should be performed repetitions in the simulation process, in order to increase the reliability of the measurements and to estimate that part of the total data variability is due to random error. We choose to perform four repetitions in the development process. In order to endure that each repetition is different, we use a different seed for the random number generator simulator OMNeT++. This has the effect that all random number processes differ, repetition by repetition, changing values as the initial position of the nodes, the backoff timeouts, among many others. Finally, the independence of errors is ensured by performing experimental runs in random order. We show the results of randomization in the TABLE 2.
TABLE 2. TREATMENT RANDOMIZATION. Treatment R1 R2 R3 XL-ALPL LOW 16 8 12 XL-ALPL HIGH 10 7 5 XLM LOW 3 4 6 XLM HIGH 1 11 9 Random Order R4 15 2 13 14

Experimental Unit So far, we have defined the factors and response variables of the experiment. It is time to formalize the experimental unit and then design the methodology to the experiment. Figure 3 shows the experimental unit of this investigation. The first factor is the communication scheme of the network; the levels of this factor are qualitative and are the

1 2 3 4

D. Results Analysis with ANOVA The summary of the results obtained by running the simulation of different treatments are shown in TABLE 3. For

each treatment, the results show the average energy per packet (J), average latency end-to-end (s), and the goodput (bps).
TABLE 3. POPULATION MEANS FOR EACH OF THE RESPONSE VARIABLES Treatment Results Scheme XL-ALPL XL-ALPL XLM XLM Traffic Low High Low High Energy (J) 5,89 2,02 3,69 1,56 Latency (s) 3,30 3,16 1,77 2,38 Goodput (bps) 11,03 73,22 12,59 95,28

interaction is 0.0015 <0.05, which indicates that we reject the null hypothesis, concluding that there was a significant interaction effect of the factors studied on the response variable energy. Figure 4 shows the confidence intervals for the population marginal means, with a confidence of 95%. We can also say that the XLM scheme gives a better performance, taking less energy on average to send a package. The interaction effect in the traffic factor is evident because under high traffic the difference between a scheme and another is very low.

E. Result Analysis with ANOVA The ANOVA analysis can determine whether there are significant differences between the means of the response variables. With the results of ANOVA we can make estimates and hypothesis testing. This is an intermediate step in the analysis as concluding that there are differences. The next step is to find the different means and estimate their difference using complementary methods. For our case, the ANOVA factorial model evaluates the individual and joint effects of the two factors related (schemes of communication and network traffic) on the response variables (latency end-to-end, energy consumed per packet or goodput). In a two-factor model, the effects of interest are three: the two main effects (one for each factor), and the effect of double interaction. The ANOVA factorial model establishes a null hypothesis for each factor and for each possible combination of factors. For a factor, the null hypothesis says the populations mean defined by the factor levels are equal. For an interaction, the null hypothesis says that such effect is zero. For every effect there is a hypothesis and assumptions for each level of significance that will determine whether the hypothesis should be rejected or accepted. It is important to mention that to properly implement the ANOVA method it is necessary to meet the assumptions of normality in the data and homoscedasticity of variances.
TABLE 4 summarizes the results of ANOVA analysis for each of the response variables of the experiment. For the variable average energy consumed per packet on the network, TABLE 4 the statistical Significance testing p-value is less than the value of confidence (0.05), rejecting the null hypothesis of the analysis, which implies that the scheme communication affects the energy consumed per packet. Following the same argument, as the p-value is 0 for the traffic factor in the energy response variable, we reject the null hypothesis, which implies that the network traffic affects the average energy consumption per packet. TABLE 4. RESULT OF ANOVA FOR EACH RESPONSE VARIABLE ANOVA Energy Latency Goodput Source prob>F 0 0,001 0,0058 Scheme 0 0,3975 0 Traffic 0,0015 0,1835 0,0132 Scheme*Traffic

Figure 4. Confidence Intervals for each response variable

As shown in TABLE 4, the p-value of the scheme factor is less than the confidence value, rejecting the null hypothesis of the analysis, which implies that the scheme of communication does affect the network latency. The opposite happens with the traffic factor, because the p-value is greater than 0.05, the null hypothesis is accepted which means that the traffic handled by the network, under preset conditions, does not affect latency network. Following the same analysis of the results of the ANOVA, we found that there is no interaction among the two factors considered in terms of the latency. Given that the only one significant effect found for this variable response is the effect that the scheme factor has on the latency end-to-end, we can say that the communication scheme XLM obtained a better performance in terms of latency network when compared with the ALPL, as shown in Figure 4. Finally, for the goodput response variable, the ANOVA analysis summarized in TABLE 4 shows that all effects are significant, which implies that the communication scheme, the network traffic and the effect of interaction, are significant. With high traffic, it is evident that the XLM protocol has better goodput than ALPL; however when traffic is low this difference disappears and the behavior of both schemes for this variable is very similar. V. CONCLUSIONS This paper presents a framework of simulation and evaluation for cross-layer protocols in wireless sensor networks. In this framework raises a number of steps for making complete comparison between these new kinds of

On the other hand, the null hypothesis for the interaction effect (scheme and traffic) says that the effect of interaction between the two factors to the average energy per packet variable, is zero. If we check the p-value of the ANOVA for

protocols. The main idea whit the framework is promote a fairness comparison of different communication schemes, a realistic model simulation, an experiment design and a rigorous analysis of results with ANOVA (Analysis of Variance) technique. Also this work presents a case study of the simulation and evaluation of two cross-layer techniques for WSNs using this framework. Based on the application of our framework, a series of modifications in the implementation details of the each protocol was made, ensuring fairness in comparing features. We show also the simulation models implementation for each communication scheme. This model was developed under the platform OMNeT++ and the mobility framework. By using this method of comparison we note that the results obtained in our case are large and statistically inconclusive. Then show the conclusions of these results. The behavior of the communication schemes is statistically distinguishable for each of the response variables were defined (average latency end-toend, average energy consumption per packet and the goodput). The XLM scheme got the best results in each of these variables. Analyzing the average energy consumption per packet for both communications schemes was found that in general the XML uses less average energy per packet than the ALPL, achieving, in the case of low traffic, a reduction of approximately 60%. Due to the interaction effect found between the factors according to ANOVA was significant; we can define substantial the effect of the traffic on the performance of the schemes for this variable, because if traffic is high, power consumption between the two schemes is very similar. For the network average latency end-to-end are also significant differences according to ANOVA. In this we find that the XML get about 30% lower average latency in the delivery messages than the ALPL. In this case there is no significant interaction effect, thus having a high or low traffic, the general behavior of the schemes is not affected. For the response variable goodput the XLM keep being more efficient tan the ALPL, getting a 25% more efficient tan the ALPL. The results indicate that under the circumstances chosen the XML schema has a better performance than the ALPL. However, there are some important facts that are in our interest analysis. Initially XLM protocol implements a random routing protocol based on geographic information. It should be noted that this geographic information requires complex systems, whether at the level of hardware (GPS, triangulation) or software level where distances could be calculated indirectly from control messages. Whatever the solution to this problem would generate extra costs in energy consumption and quality of service, either because they require additional hardware modules, implement complex algorithms or even new messages that would increase the overhead of the scheme. It is important to note that current evidence was not taken into account this additional cost. On the other hand is overwhelming the fact that for a WSN under monitoring applications, a proactive random protocol is very efficient as the XLM, the problem would be to achieve an implementation of geographic information in a simple manner.

Finally, we show that despite the marked differences in terms of architecture, implementation and even application between the protocols to be evaluated, our framework proved useful in obtaining conclusive results about their performance. ACKNOWLEDGMENT This work was supported by a grant of the Aliaza Regional de TIC Aplicadas ARTICA, in Medelln - Colombia. REFERENCES
[1] D. Culler, D. Estrin, and M. Srivastava, "Overview of sensor networks," IEEE Computer, vol. 37, no. 8, pp. 41-49, August 2004. [2] T. Torfs, S. Sanders, C. Winters, S. Brebels, and C. Van Hoof, "Wireless network of autonomous environmental sensors," in Proceedings of the 3rd IEEE Conference on Sensors, 2004, pp. 923-926. [3] A. Sleman and R. Moeller, "Integration of Wireless Sensor Network Services into other Home and Industrial networks; using Device Profile for Web Services (DPWS)," in Proceedings of the 3rd International Conference of Information and Communication Technologies: From Theory to Applications (ICTTA08), 2008, pp. 1-5. [4] D. Malan, T. Fulford-Jones, M. Welsh, and S. Moulton, "CodeBlue: An Ad Hoc Sensor Network Infrastructure for Emergency Medical Care," in Proceedings of the Workshop on Applications of Mobile Embedded Systems (WAMES04), 2004. [5] K. Holger and W. Andreas, Protocols and Architectures for Wireless Sensor Networks. England: Wiley, 2005. [6] I. Akyildiz, W. Su, and Y. Sankarasubramaniam, "A survey on sensor networks," IEEE Communications, vol. 40, no. 8, pp. 102-114, August 2002. [7] V. Srivastava and M. Motani, "Cross-Layer Design: A Survey and the Road Ahead," IEEE Communications Magazine, pp. 112-119, December 2005. [8] T. Melodia, M.C. Vuran, and D. Pompili, "The state-of-the-art in crosslayer design for wireless sensor networks," Springer LNCS, vol. 3883, pp. 78-92, June 2006. [9] P.J. Marrn, A. Lachenmann, and D. Minder, "TinyCubus: A Flexible and Adaptive Framework for Sensor Networks," in 2nd European Workshop on Wireless Sensor Networks, 2005, pp. 278-289. [10] W. Su and T.L. Lim, "Cross-Layer Design and Optimization for Wireless Sensor Networks," in Seventh ACIS International Conference on Software Engineering, Artificial Intelligence, Networking, and Parallel/Distributed Computing, Las Vegas, 2006. [11] I.F. Akyildiz, M.C. Vuran, and O.B. Akan, "A Cross-Layer Protocol for Wireless Sensor Networks," in Proceedings of the Conference on Information Sciences and Systems (CISS '06), Princeton, NJ, 2006. [12] R. Jurdak, P. Baldi, and C. Videira Lopes, "Adaptive Low Power Listening for Wireless Sensor Networks," IEEE Transactions on Mobile Computing, vol. 6, no. 8, pp. 988-1004, August 2007. [13] OMNeT++ Community. [Online]. http://www.omnetpp.org/ [14] Mobility Framework for OMNet++ 4.0 wiki. [Online]. http://wiki.github.com/mobility-fw/mf-opp4 [15] O. Karaca and R. Sokullu, "Comparative Study of Cross Layer Frameworks for Wireless Sensor Networks," in Wireless Communication Society, Vehicular Technology, Information Theory and Aerospace & Electronics Systems Technology (Wireless VITAE'09), Denmark, 2009, pp. 896-900. [16] J. Polastre, J. Hill, and D. Culler, "Versatile Low Power Media Access for Wireless Sensor Networks," in Proceedings of the 2nd ACM Conference on Embedded Networked Sensor Systems, Baltimore, 2004.

You might also like