Optimizing Energy Consumption and QoS in WMSNs Using Queueing Theory

Abstract

The limited resources and enormous amounts of data generated by multimedia sensors require efficient strategies to extend network lifetime while taking into account quality-of-service requirements such as reliability and delay. In contrast, limited battery resources require new techniques to balance energy consumption and multimedia application requirements in wireless multimedia sensor networks (WMSNs). These requirements are very critical, especially for network stability and performance. In this paper, an energy-efficient mechanism based on the M/D/1/B queueing model is proposed. According to the packets in the queue and the waiting time, the nodes decide their activation time, so the nodes wake up for a while to transmit the data in the queue and then go to sleep mode. The simulation results of the proposed algorithm show that the proposed mechanism achieves optimal values to reduce energy consumption while meeting the quality-of-service requirements under different conditions.

1. Introduction

The resource constraint and limited capabilities of sensor nodes mean efficient resource utilization is very important. Sensor nodes have limited and non-replaceable batteries that should work for a long period of time, perhaps years. Wireless multimedia sensor network (WMSN) applications generate a high volume of traffic, which further complicates the resource constraints of sensors. In addition, the traditional techniques used in WSN cannot be directly applied to WMSN. The proposed solutions should handle the massive traffic and quality-of-service requirements of WMSN while extending the network lifetime by considering energy efficiency. To reduce energy consumption, we should realize the process that consumes more energy in WMSN operation.

Sensor nodes perform sensing, processing, and communication with neighboring nodes. Sending and receiving consume more energy than sensing and processing, which are insignificant [1]. Accordingly, the communication process should be designed to optimize energy consumption by changing the process used at the MAC layer. The MAC layer is responsible for performing and organizing the communication process between nodes, where energy is consumed by various states such as inactive enumeration, packet collisions, and sending and receiving data, which are the main factors affecting the energy consumption of the WMSN [2] Most of MAC layer protocols are based on 802.15.4 standard [3] for low-rate WPANs, which outlines the Physical and MAC layers for low-power wireless communication. It was specifically created for applications that require low data rates, low power consumption, and cost-effectiveness, such as sensor networks, home automation, and industrial automation. In the unslotted IEEE 802.15.4 carrier sense multiple access with collision avoidance (CSMA/CA) mechanism, each node in the network has two variables: NBand BE. NB is the number of times the CSMA/CA algorithm has backed off while attempting the current transmission. NB is initialized to zero before every new transmission. BE is the back-off exponent, which is related to how many back-off periods a node must wait before it attempts to assess the channel. The algorithm is implemented using units of time called back-off periods. The parameters that affect the random back-off are BEmin, BEmax, and NBmax, which correspond to the minimum and maximum of BE and the maximum of NB, respectively.

The antenna of the sensor node remains dormant even in the absence of activity, so it is important to put the antenna sensor into sleep mode. Putting sensor nodes into sleep mode reduces energy consumption. On the other hand, putting sensor nodes to sleep delays the transmission of packets because each node must wake up each time it detects activity on the medium. The duty cycle alternates between sleeping without activity and waking up when activity is detected. In other words, it is the ratio of sleep time to work time. As mentioned earlier, the duty cycle helped to reduce energy consumption but compromised the quality-of-service requirements of sensor nodes. Therefore, duty-cycle techniques should make a trade-off between energy consumption and quality-of-service requirements in multimedia transmission [4].

In general, duty-cycle techniques can be divided into synchronized and distributed techniques. In the synchronized techniques, the nodes know their sleep schedule and exchange it, which is impractical and involves more overhead and packet transmissions. In the distributed techniques, the nodes do not exchange their schedule, and the sensor nodes wake up when they detect packets called preamble sent by neighboring nodes. The activities of the sensor nodes should be scheduled to extend the lifetime of the network, although it is not possible to keep the sensors awake all the time [5].

Queueing models can be used to model and analyze the performance of various subsystems in networks, for instance, to estimate the packet loss and packet delay in network paths. Queueing theory therefore plays an important role in the performance modeling and evaluation of communication systems and networks. In particular, queueing models in discrete time are very appropriate to describe traffic and congestion phenomena in digital communication systems, since these models reflect in a natural way the synchronous nature of modern transmission systems, whereby time is segmented into intervals (“slots”) of fixed length and information packets are transmitted at slot boundaries only.

1.1. Problem Statement

The batteries of sensor devices are very limited resources and should work for a long time. In addition, multimedia transmission consumes more resources. Accordingly, energy consumption should be effectively managed and in a planned manner by controlling the sensor modes. In the absence of activities, the sensor nodes go to the idle state, which still consumes energy, so it is feasible to switch sensor nodes to sleep mode. Maximizing sleep time reduces energy consumption but compromises other quality-of-service requirements, such as delay which is another important factor. Managing sensor node resources without compromising the quality-of-service requirements is an important issue that needs to be addressed using various techniques and algorithms, with duty-cycle mechanisms considered a reasonable solution.

1.2. Summary of Contributions

The main goal of this study is to design an asynchronous duty-cycling method for managing sensor modes based on a queueing model that regulates the state of wake-up and sleep depending on the volume of traffic in the queue and the processing time needed. The waiting time is estimated based on the M/D/1/B queueing system. All packets are processed and sent to the next node using the preambles for the wake-up function. In this case, the sensors adjust the wake-sleep cycle so that it is not necessary to wake up at a specific time. The dynamic duty cycle is checked frequently to measure the suitability of the current duty cycle for the incoming traffic and adjusted accordingly. The main contributions of this study are listed below.

• A dynamic duty-cycle mechanism for WMSNs is designed where each node has its own duty-cycle schedule to avoid the overhead of packet transmission.
• A queueing model is proposed along with the duty cycle, where the traffic is handled according to the length of the waiting time in the queue and the duty cycle changes accordingly.
• This study investigates the performance of the proposed mechanism in different scenarios based on the network simulator NS3. The results show that the proposed scheme has good performance in terms of energy efficiency without compromising relevant QoS requirements.

This study is organized as follows: Literature on the WMSN duty cycle is presented in Section 2. Section 3 contains a description of the system design of the proposed algorithm. Section 4 provides details on how the proposed scheme was implemented and evaluated in WMSN. Section 5 discusses the conclusion and future work.

2. Related Work on Duty Cycle over WMSN

Many networks and communication systems rely on queueing theory for evaluation and modeling. If it is possible to describe the traffic shape and congestion encountered in communication systems, many challenges in communication systems can be optimized and the performance of the system can be improved. Based on queueing models. There are many proposals in the literature that use queueing theory for resource management in WMSN. These works are proposed with different objectives and techniques; some aim to solve the congestion problems of the network, while others aim to reduce energy consumption based on duty-cycle operation. In the following section, recent work based on the queueing model for resource management in a wireless sensor network is discussed.

ELBPQA [6] is a queueing algorithm based on IEEE 802.15.6 [7] proposed for wireless body sensor networks. The incoming traffic is classified based on the original location. The local traffic is classified based on data priority, while the traffic generated from remote locations is classified based on the packet’s deadline. The proposed work introduces four queues, each with a different priority. The proposed work utilizes a hardware scheduler to prioritize and process the traffic.

To handle the heterogeneous traffic produced by various sensor nodes, the MK/HyperK/1/M queueing model with Poisson arrival rate and hyper-exponentially distributed service time is presented in ref. [8]. The suggested model looks into how throughput is impacted by buffer size, packet length, arrival rate, and duty cycle. The model is evaluated based on IEEE 802.15.4 in beacon-enabled mode with a star topology and applies to a wide range of medium access control (MAC) protocols proposed for WMSN.

Several priority systems that come pre-installed in devices are integrated with the priority-based energy-efficient (PrEE) MAC protocol [9]. The superframe structure of IEEE 802.15.4 is the foundation for the proposed work. Data priority is used to determine channel access, and slots are assigned at random to avoid overlap. Accordingly, the superframe slots are reserved. The work in ref. [10] proposes an event-driven Queueing Petri net (QPN) to evaluate and simulate the sensor components that consume energy in the communication systems. The Queueing Petri net (QPN) is adapted to be compatible with event-driven operations in WSN. The energy consumption in different states is estimated and modeled to predict the lifetime of the network. S. Ghosh et al. [11] propose a technique for energy efficiency in WMSN, where the antenna of the sensor node is turned on when the number of received packets reaches a certain limit. The M/M/1 queueing model is applied to sensors under stable conditions along with a systematic queue size monitoring method to ensure the ideal limit of the queue.

M. Kempa [12] proposes a technique to make the incoming traffic rate in the empty nodes reasonable, blocking many empty slots at the beginning during sending and receiving. Hence, until a predefined number of N packets have been collected in the queue, the next empty slots are filled with fixed constant time. After the conclusion of the break, the same procedure is again repeated. It uses the M/G/1 queueing mechanism and an infinite buffer size. The traffic rate is based on the Poisson process and prioritized on a first-in, first-out basis. The energy-saving techniques combine the multiple vacations policy and the fixed threshold policy. In ref. [13], a mobile sink is introduced to collect data, and the proposed model allows the correlation between the data aggregation and energy harvesting process and the consumed energy for each activity during data transmission and aggregation. The Markov model is adapted to flexibly evaluate the performance of energy harvesting. A Markovian environment variable is used to correlate energy harvesting and data aggregation. The results show that the temporal correlation of energy harvesting plays an important role in network performance.

H. Hadadian et al. [14] present an analytic technique based on the Markov model and the IEEE 802.15.4 standard with a non-beacon-enabled mode for frequent packets and packet withdrawal. The proposed method is suitable for acknowledgment and non-acknowledgment modes and works for diverse traffic generated by various sensor nodes for crucial application requests. To ascertain the impacts of packet creation time and withdrawn packets on the performance of sensor networks, the performance for the probability distribution function (PDF) of packet delay (PD) and packet delivery ratio (PDR) is evaluated with a high packet rate. In citation b13, active queue management is recommended as a way to calculate the likelihood of packet loss. The suggested work combines a random early detection method with a fuzzy proportional integral derivative (fuzzy PID) controller methodology. When fuzzy logic and PID are combined, it enables monitoring and managing the buffer queue of the sensor node. The packet transmission of the sensor nodes is also evaluated and controlled by a fuzzy logic controller.

The WSN node performance analysis models based on the M/M/1 queueing model with priority and random early detection (RED) were researched by Yonggang Xu et al., [15]. The performance of WSNs is examined using queueing theory, and it is acknowledged that packet priority affects network QoS and packet discarding affects network congestion. The major performance metrics are first developed using the M/M/1-based WSN node queueing model. The major performance metrics of a priority-based WSN node queueing model are then examined for packets with various priorities. In the last section, the RED (Random Early Detection) congestion control technique is presented as a way to improve the first two WSN node queueing models, and it shows how to construct the expressions for the corresponding theoretical performance metrics.

Artificial intelligence algorithms have unique advantages in dealing with this problem. In the literature, many works use AI algorithms to enhance performance. The prediction mechanism used in ref. [16] can be utilized to predict the duty-cycle mechanism and energy consumption in sensor nodes. Where an adaptive recursive fuzzy neural network with GK clustering and hierarchical LM algorithm (GK-ARFNN) is proposed. With this GK-ARFNN, an appropriate fuzzy rule base is obtained by clustering the existing dataset, and high model accuracy is achieved by optimizing the fuzzy rules deeply. In addition, the introduced algorithm can be utilized to explore more solutions and converge them faster. In ref. [17] an enhanced decomposition-based multi-objective evolutionary algorithm is proposed with a self-organizing collaborative scheme (MOEA/D-SCS). The multi-objective evolutionary algorithm based on decomposition (MOEA/D) decomposes a multi-objective optimization problem (MOP) into multiple single-objective subproblems using an aggregation function and optimizes them together using a collaborative approach. MOEA/D exhibits extraordinary optimization ability in solving MOPs.

3. The Proposed Energy-Efficient Mechanism, Based on the Queueing Model

As mentioned previously, the main goal of this study is to maximize network lifetime and consume energy efficiently. First, we study the sensor modes, where it is impractical to put the sensor nodes in an idle state all the time where a considerable amount of energy is consumed. Furthermore, increasing the sleep duty cycle leads to increasing delay which affects the QoS requirements of multimedia transmission. The proposed mechanism is modeled as an M/D/1/B queueing system to study the time spent by packets in the queue. Accordingly, based on the number of packets in the buffer and the service time for each packet, nodes know how long the node needs to stay awake and what the next sleep period is. Increasing the sleep time and decreasing the listening time saves energy on the receiver side, but the cost increases on the sender side because more preambles need to be sent [18]. In addition, a decrease in listening time results in a missing preamble, which causes a larger delay since the sender node must wait for the receiver node to wake up; this delay at each hop affects the end-to-end delay. Therefore, the trade-off between energy consumption and delay should be carefully weighed. In the following section, approximate calculations for delay and power consumption are performed based on queueing theory [19]. By examining the characteristics of the queueing model, delay and power consumption can be determined.

In queueing theory, the M/D/1/B queueing system, which is depicted in Figure 1, depicts a queue length where packets arrive in a Poisson process at a fixed rate and are handled by a single server, with the service time being predetermined and fixed for all packets. The model has a specified capacity of the buffer. The bounded waiting time controls the queue size in addition to the utilization factor. The nodes are assumed to act as clients competing for access to the server. The node that wins the competition is served first and can transmit the data. Thus, if the number of competing nodes, each node takes a portion of the system service [20,21,22]. The embedded Markov chain controls the distribution of queue length at interval time I and the stationary probability distribution at all time intervals, an average number of packets, and the average waiting time [23]. Based on this model, numerous performance metrics can be determined, including delay and energy consumption. Sleeping and awake time based on communication exchange is shown in Figure 2.

Let $X_B\left(t\right)$ the packets be in the queue at a time t and $t_n$ is the time when the nth packet finishes and leaves. The stochastic process ${X_B\left(t_n\right)}_{B\ge 0}$ is a Markov chain with state space $\{0,1,\dots ,B-1\}$, as shown in Figure 3. Accordingly, the probability of arrival packet K at service time t is calculated as per [24,25] and shown in Equation (1) in the following.

$${\alpha }_k=\frac{{\rho }^K}{K!e^{\rho }}$$

(1)

The scheme of embedded Markov chain $\Pi =\left({\pi }_{i,j}\right)$ with a probability transition matrix, is described in the following Equation (2).

$$\begin{array}{c}\begin{array}{c}I=\end{array}\left[\begin{array}{cccccc}{\alpha }_0& {\alpha }_1& {\alpha }_2& \dots & {\alpha }_{N-2}& 1-\sum _0^{N-2}{\alpha }_K\\ {\alpha }_0& {\alpha }_1& {\alpha }_2& \dots & {\alpha }_{N-2}& 1-\sum _0^{N-2}{\alpha }_K\\ 0& {\alpha }_0& {\alpha }_1& \dots & {\alpha }_{N-3}& 1-\sum _0^{N-3}{\alpha }_K\\ ⋮& ⋮& ⋮& \ddots & ⋮& ⋮\\ 0& 0& 0& \dots & {\alpha }_0& 1-{\alpha }_0\end{array}\right]\end{array}$$

(2)

The steady-state probability distribution Q in the finite buffer in the M/D/1/B queue takes the form shown in Equation (3), as per [22,26].

$$q_j=\frac{P_J\left(B\right)}{1-P_N\left(B\right)}$$

(3)

Accordingly, the probability distribution of the number of packets in the M/D/1/B queue is derived from [22] and shown in Equations (4)–(6).

$$P_0=\frac{1}{1+\rho b_{B-1}}$$

(4)

$$P_B=1-\frac{b_{B-1}}{1+\rho b_{B-1}}$$

(5)

$$P_j=\frac{b_j-b_{j-1}}{1+\rho b_{B-1}}$$

(6)

The coefficients $b_0$ and $b_j$ are shown in the following Equation (7).

$$\begin{array}{c}\hfill b_0=1\\ \hfill b_j=\sum _{K=0}^j\frac{{(-1)}^K}{K!}{(j-K)}^K{e}^{(j-k)\rho }{\rho }^K,\forall j\ge 1\end{array}$$

(7)

The mean number of packets $X_B$ of the M/D/1/B system can be calculated using Equation (8).

$$X_B=B-\frac{{\sum }_{K=0}^{B-1}bK}{1+\rho b_{B-1}}$$

(8)

Waiting time T affects the average service time. Hence, T contributes to the calculation of the delay time in the M/D/1/B queue system. The expected delay D time can be calculated using the following Equation (9).

$$D=(B-\frac{1}{2}-\frac{{\sum }_{K=1}^{B-1}{b}_{K-B}}{\rho b_{B-1}})T$$

(9)

The average energy consumption $E_c$ of the queue system can be calculated using the following Equation (10).

$$E_c\left(T_{s,i}\right)=E_r\left(T_{s,i}\right)+E_t\left(T_{s,i}\right)$$

(10)

where $E_r\left(T_{s,i}\right)$ represents the energy consumption during receiving of data packets and $E_t\left(T_{s,i}\right)$ indicates the energy consumption during the transmission of data packets. $E_r\left(T_{s,i}\right)$ and $E_t\left(T_{s,i}\right)$ can be calculated using the following Equations (11) and (12), respectively.

$$E_r=\frac{E_i{T}_i+E_s{T}_s}{T_i+T_s}$$

(11)

where $E_i$ and $E_S$ represent average energy consumption during listening and sleep cycles, respectively. Poisson with an average rate $\lambda$ is used as a packet generation model. The approximate value of the energy consumption to transmit data packets is shown in Equation (12).

$$E_t=E_w{D}_{\lambda }$$

(12)

where $E_t$ is the average energy consumption of packet transmission and $E_w$ can be calculated using the following Equation (13).

$$E_w=\frac{T_{Data}}{D}{P}_{tx}+\left(\frac{D-T_{Data}}{D}\right)\left(\frac{P_{tx}{T}_P+P_i{T}_{wait}}{T_p+T_{wait}}\right)$$

(13)

where $P_{tx}$ shows the transmit energy consumption, $T_{wait}$ is the waiting time, $T_{Data}$ shows the length of the data packet and $T_p$ reflects the preambles. Moreover, $\frac{T_{Data}}{D}(\frac{D-T_{Data}}{D})$ is the weight factor to send data packets and preambles, correspondingly.

4. Performance Evaluation

The effectiveness of the suggested strategy is assessed through several experiments based on the Network Simulator-3 (NS-3), which replicates the fundamental physical layers and MAC of the TCP/IP standard in WSN utilizing IEEE 802.15.4 MAC and physical layers. Clusters of 100 nodes are formed and arranged in a 50 × 50 m space. To study the impact of background traffic, the source nodes are varied from two to eight nodes. Each node is about 7.5 m away from the cluster head. The longest waiting time is twice that of listening and sleeping. The proposed work, which primarily affects the duty cycle of WMSN performance, is evaluated using metrics for energy consumption, reliability, and latency. The simulation settings used in the simulation experiments are shown in

Table 1. Experimental configuration and simulation parameters.

Propagation Model	Shadowing
Path loss exponent	3
Shadowing deviation (dB)	3
Phy-Type	Wireless Phy/802.15.4
Mac-Type	Mac/802.15.4
Frequency	2.4 GHz
BE min, BE max, NB max	2, 3, 2
Back-off duration max	280 microseconds
Preamble, ACK	20 microseconds
Initial Energy	3.6 Joule
Transmission Power	1 MW
CODEC	H.264

. The confidence level for the results of the following studies is 94% and each scenario is repeated.

Initially, the duty cycle of the sensor nodes is tested at different packet rates, as shown in Figure 4. In contrast, as the packet rate increases, the duty-cycle percentage increases, which is common in such a situation. An increase in packet rate means that there are more packets in the queue, which require more processing time. The node resumes its idle state after completing and forwarding the packets. The number of packets in the network grows together with the number of source nodes. In the proposed mechanism, the queueing model in each node calculates the time it will take for packets to clear the queue before adjusting the duty cycle. When the packet rate in the simulation reaches the heist point, 225 P/S, and the number of source nodes is two, the duty cycle reaches 7.5%, which is reasonable. On the other hand, when the number of source nodes is eight, the duty cycle reaches 15%, which is still acceptable, and the energy source is maintained. The duty cycle is adapted to traffic rate as the number of packets in the queue, then goes to sleep mode. In the next round, the node will wake up when it receives a preamble packet as an indicator of new transmission.

Reliability performance is measured as an indicator of the packet reception rate. Figure 5 and Figure 6 show the packet reception rate determined by simulation based on the generated traffic rate and the average of the source nodes. By altering the number of source nodes from two to eight nodes, the simulation is put to the test. The packet delivery ratio is depicted as a function of listening time in Figure 5. As can be seen in the figure, the reliability increases when the eavesdropping time reaches 0.8 ms, and the improvement in delivery ratio increases slowly, which means that the current eavesdropping time is sufficient to handle the coming traffic rate and the time spent on back-off and sending preamble packets. The packet delivery rate at a listening time of 0.2 ms and a quiet time of 0.4 s reaches 93%. The deliver ratio increases slowly in the range between 0.8 ms and 1.8 ms and reaches 98% at a listening time of 2 ms. When the quiet time is 0.4 s, the delivery rate does not exceed 87% at a listening time of 2 ms. The proposed queueing model allows nodes to estimate the coming traffic and the time needed for transmission so that all packets in the queue are transmitted.

Figure 6 shows that the packet delivery ratio decreases as the idle time increases because the number of preamble packets transmitted increases, leading to increased contention during the listening period. When the listening time is 10 ms and the idle time is 0.2 s, the packet transmission rate is 98% and starts to decrease until it reaches 90%. When the listening time was reduced to 4 ms, the delivery rate dropped to 80%. In duty-cycle operation, a trade-off should be made between sleep time and other factors such as delay and power consumption. As shown in Figure 6, at 4 ms, we still achieve a high delivery rate with a reasonable sleep time. The sleeping time is managed in a way that does not affect the delivery ratio which is an important factor in multimedia transmission.

Figure 7 and Figure 8 show the delay of duty-cycle operation; packet delay is defined as the time from the generation of the packet to the successful completion of the transmission. Figure 7 shows the delay compared to duty-cycle operation, and it is clear that the delay decreases as the listening time increases. At 0.4 and 0.6 ms, the delay is 1.8 ms and 2.1 ms, respectively; as the listening time increases, the delay remains at low values. For a higher quiet time, e.g., 1 s, the delay starts at 1.5 s and then gradually decreases as the listening time increases. If the listening time is clear, the receiver must catch up with the preambles sent by the transmitting nodes. On the other hand, if the listening time is long, most packets will be received in the first round, so increasing the listening time up to a certain limit will not affect the delay.

Figure 8 shows the delay as a function of sleep time; the delay increases as sleep time increases. The receiver nodes take longer to receive the preambles and wake up, but according to the queueing model, the nodes go to sleep immediately after sending all the packets in the queue. The delay reaches 1.9 s when the sleep time is 2 s, while the delay decreases when the listening time is 10 ms and reaches 1.2 s at the same time as the sleeping nodes.

Figure 9 and Figure 10 show the energy consumption of the proposed mechanism, which is the most critical factor. The values for energy consumption are calculated as a function of listening and sleeping time. Because energy consumption grows with listening duration, a trade-off must be made between energy expenditure on the one hand and delay on the other. The primary goal is to reduce power consumption while maintaining latency and reliability requirements. As shown in Figure 5 and Figure 7, there is an insignificant change after the listening time of 4 ms. As shown in Figure 10, there are trade-offs between listening and the cost of preamble transmission. When nodes remain idle longer, the energy consumption at the receiver node decreases, while the cost at the sender increases because more preambles must be sent to wake up the receiver nodes.

In Figure 9, it can be seen that the power consumption increases as the listening duration increases. However, as mentioned in the proposed mechanism, the receiver estimates the queueing delay of the packets in the queue and then goes directly to sleep, reducing the total wake-up time of the sensors while maintaining the delay and reliability requirements. Figure 10 shows the energy consumption as a function of sleep time. The energy consumption gradually decreases until the sleep time reaches 1s; then it increases again because the sending nodes need to send more preambles to wake up the receiving node, in addition to the conflicts with other nodes.

5. Conclusions and Future Work

In this work, we proposed an energy-efficient mechanism based on the queueing model for wireless multimedia sensor networks. The nodes estimate the queueing delay based on the queueing model M/D/1/B. The proposed mechanism is based on a distributed duty cycle where each node has a sleep schedule. The embedded Markov chain regulates the distribution of queue length at any given time. The performance evaluation shows The proposed mechanism achieves a high delivery ratio and reaches 98% responding to incoming packet rate which means achieves high reliability. The other important performance metric is delay, where the proposed scheme keeps the delay within an acceptable range for multimedia applications. Finally, the proposed scheme shows energy consumption which is considered to be the main factor for sensor networks is consumed effectively and prolongs the network lifetime. At the same time, the quality-of-service standards, such as delay and dependability, are unaffected. In future work, the proposed mechanism will be tested with different topologies. In addition, other performance metrics for multimedia transmission, such as PSNR and jitter, will be investigated in duty-cycle operation.