Sage Journals: Discover world-class research

Abstract

The availability of smart sensors equipped CMOS cameras made it possible to capture and transmit multimedia data ubiquitously. The real-time performance is essential in many applications including surveillance and healthcare monitoring in multimedia sensor networks (MSNs). One of the real-time performance indicators is to transmit the packets within their deadline. Furthermore, in information-rich, battery-powered, and resource-constrained MSNs, it is a challenging problem to extend network lifetime decreasing communication cost. In previous researches, periodic message exchange for neighbor information maintenance leads to reducing network lifetime. This paper presents a power-aware data transmission for real-time communication in MSNs. The proposed scheme not only provides real-time performance, but also conserves energy consumption through efficient transmission and without periodic message exchange. The simulation results show the effectiveness of the proposed scheme in achieving the desired deadline success ratio and prolonging the network lifetime.

1. Introduction

Advances in microelectromechanical systems (MEMS) technology and wireless communications have put forward the development of smart sensors which handle the various kinds of data and consist of sensing, data processing, and communicating components [1]. The camera-equipped sensors, a kind of smart sensor, are deployed in the target area for collaborative multimedia data processing. Multimedia sensor networks (MSNs) are identified as a dominant research area and have contributed to progress in military and environmental applications such as remote and distributed video-based surveillance, environmental and structural monitoring, and industrial process control [2].

It is imperative that data transmission in such applications meets real-time constraints. The image and video data captured by the sensors should be delivered to the control center within the predetermined time, that is, deadline. As shown in Figure 1, a smart sensor detects the enemy attack and then reports the information to control center through multihop communication. Many real-time routing protocols have been proposed for considering deadline satisfaction or the end-to-end delay to report the time critical data [3, 4]. The performance is usually referred to as a quality of service (QoS) requirement [5, 6]. Most real-time routing protocols have taken account of the end-to-end delay by means of summing up the whole links delay in each path. This causes much wasteful processing and much overheads with respect to route maintenance and delay measurement.

Figure 1

An illustration of multimedia sensor networks.

The consideration of the peculiarities of MSNs such as information-rich data, limited power, hardware constraints, dynamic network topology (easy to die due to limited battery or mobile sensors), and large-scale deployment has posed many challenges such as energy-aware routing protocols in the design and management of MSNs. Above all, the extension of network lifetime through improvement of energy efficiency is a critical design issue in most routing protocols for MSNs. In MSNs compared to the traditional sensor networks, image and video data is much bigger than the scalar data such as temperature, pressure, humidity, or location of objects. Due to the inherent constrains of MSNs and the requirement for real-time applications, it is vital to improve energy efficiency for data delivery through minimizing communication overheads. Communication overheads are involved in periodic message exchange for route discovery and maintenance.

Geographic routing which forwards packets only based on the location of itself, its neighbors, and the destination sets out to remove communication overhead in conventional routing brought by route discovery and maintenance or topology updates. Instead of route discovery and maintenance, it is required to collect the location information of all its direct neighbors [7, 8]. However, it is inevitable to exchange messages periodically to obtain the neighbor information. In contrast, flooding, a reactive technique with reliability, seems to be a good candidate for data transmission because it does not involve neighbor information maintenance, complex route discovery, and expensive topology updates. The flooding, a simple data transmission, suffers from the broadcast storm problem which leads to network lifetime depletion [9]. To solve the broadcast storm problem, several schemes have been proposed to reduce the redundancy in flooding mechanisms [10]. However, these algorithms either perform not good enough to reduce redundant transmissions or require each node to maintain one-hop or two-hop neighbor information using additional message exchange such as control message or HELLO message [11–13]. We have rethought of the periodic message exchange with respect to the overall energy consumption influencing on network lifetime.

In [14], author proposed a data transmission scheme enhancing real-time performance by increasing the deadline satisfaction ratio per unit energy consumption of time-sensitive packets. It improved energy efficiency by flooding but effectively constrained energy consumption by controlling the scale of flooding, that is, flooding only when necessary. If unicasting meets the distributed subdeadline at a hop, CFlood aborts further flooding even after flooding has occurred in the current hop. However, author did not address the energy consumption owing to periodic HELLO message exchange for neighbor information maintenance. As mentioned above, it is required to consider both real-time performance and energy consumption in MSNs.

In this paper, we put forward a power-aware data transmission for real-time communication in MSNs in two aspects. (1) Real-time performance: the proposed scheme transmits data within the deadline by estimating end-to-end delay. Delay is proportional to the number of intermediate (forwarding) nodes on the transmission path between a source and SINK. In order to improve real-time performance, the proposed scheme provides the greedy forwarding which maximizes the distance between consecutive nodes on transmission path. Furthermore, the proposed scheme transmits data based on the neighbor information to enhance real-time data delivery. The information is maintained up to date at the nodes by piggy-backing during transmission. (2) Energy conservation: the proposed scheme decreases the communication overhead in building and maintaining neighbor information with simple data transmission such as a constrained flooding. The constrained flooding screens redundant data transmission and is aware of residual energy for deciding next node called the forwarding node. Therefore, the proposed scheme extends network lifetime through balancing energy consumption. The objective of the proposed scheme not only provides real-time data delivery meeting the deadline but also extends the network lifetime through energy conservation.

Packets are transmitted to SINK within their deadline for real-time communication. The end-to-end delay is bounded to sum the one-hop delay. Each intermediate node on a transmission path can independently decide the one-hop delay using node information piggy-backed in the packet.

To build up the neighbor information efficiently, this is, without periodic message exchange, we introduce a simple and efficient data dissemination mechanism in the beginning as follows: the proposed scheme transmits data based on a constrained flooding-based mechanism which brings down the redundant transmission by reducing the number of the forwarding nodes. The proposed scheme provides the greedy forwarding maximizing the distance between consecutive forwarding nodes. It causes improved real-time performance because delay is proportional to the number of nodes on transmission path between source and SINK.

To extend network lifetime, if delay is bounded in the time requirement, it is considered the residual energy of the node which becomes the forwarding node. Consequently, the proposed scheme provides real-time data delivery extending network lifetime. The energy conservation derived from excluding periodic neighbor information exchange, decreasing the number of the forwarding nodes, and balancing the energy consumption caused by the forwarding node decision considering the residual energy of the nodes. The balanced energy consumption leads to distributing data transmission somewhat evenly and prolonging the network lifetime.

The rest of the paper is organized as follows. Section 2 introduces the related work. Section 3 illustrates the network model and definitions. In Section 4, we present a power-aware data transmission for real-time communication in MSNs. In Section 5, we analyze the simulation results of the proposed scheme by using NS-3. Finally, we conclude the work in Section 6.

2. Related Work

Due to the above peculiarities of the networks, various routing protocols are proposed to provide QoS such as deadline success ratio, delay, and conservation of energy.

Researchers have continuously suggested real-time routing protocols in wireless sensor networks (WSNs). SPEED [15] is a real-time routing protocol in WSNs. It maintains the neighbor information and considers the end-to-end delay requirement taken in the method to sum up each links’ delay. This leads to much wasteful calculation and overhead. It has not comprehensively taken the energy consumption into account. MMSPEED [16] provides the packet forwarding based on the probability distribution function and achieves the reliability by transmitting multiple copies of the packet. It does not consider the residual energy of each node. As a result, imbalance of energy distribution happens, which is to reduce the network lifetime.

Considering the properties of network, researchers have proposed various energy efficient routing protocols such as geographic routing and simple data transmission mechanisms, for example, flooding-based data transmission. Geographic routing does not require the route discovery or maintenance, which leads to conserve energy. It forwards the packet using the location [7, 8]. However, the routing protocol exchanges the neighbor information (location) instead of route maintenance. It causes communication overheads.

There are many researches to improve the flooding efficiency. The location-aided flooding to reduce the number of redundant retransmission has proposed in [11]. It improves the energy efficiency by exploiting the location information. As a result, it prolongs the network lifetime. Another efficient flooding scheme has been introduced in [12]. It improves the performance with decreasing the amount of neighbor information. In the suggested algorithm, the number of hops to maintain in each node is limited to one hop. In [13], it provides both guaranteed data delivery and good bound on the number of forwarding nodes based on the partial two-hop neighbor information. In the flooding based data transmission, it cannot avoid the communication overhead caused by neighbor information maintenance.

An energy efficient flooding mechanism providing the real-time performance has been proposed in [14]. It achieves a higher deadline satisfaction ratio; however, the energy consumption due to periodic HELLO message exchange was not considered. Therefore, the periodic HELLO message exchange causes node depletion. Network dies due to the depleted node. The lifetime of a sensor network is most commonly defined as the time to the first node failure. In MSNs, it is crucial to data delivery failure.

Compared to previous researches, we provide a power-aware data transmission scheme for real-time communication. The deadline success ratio provides real-time performance. In addition, we improve network lifetime with power-aware data transmission without periodic control message exchange and with balancing energy consumption considering residual energy. We consider comprehensive communication cost concerning not only actual data transmission but also periodic control message exchange such as HELLO message for neighbor information maintenance. We could not afford to exclude periodic message exchange overheads with respect to the total communication cost because it is more crucial to prolong the network lifetime in MSNs which spends much more energy to deliver the multimedia data relatively much larger than scalar data in the traditional sensor networks. In this paper, we propose a real-time data transmission scheme without periodic control message exchange, which leads to energy conservation and then extending of network lifetime.

3. Network Model and Definition

3.1. Network Model

We assume a homogeneous MSNs architecture. The nodes are densely and randomly deployed in MSNs. We assume that all nodes in the network are equipped with the global positioning system (GPS) or distributed location services [16]. All the camera sensor nodes are aware of their own locations. They are assumed to know the location of SINK. In general, most of imaging processing systems require the location information about the camera sensor nodes in MSNs. All nodes are static and the transmission range of all nodes is initialized by R. We assume multihop communication. The distance between node $N_{i}$ and node $N_{j}$ is given by $d (N_{i}, N_{j})$ , which is the Euclidean distance between $N_{i}$ and $N_{j}$ . We give the following definitions.

3.2. Definition

We define neighbor, forwarding candidate set, and forwarding node in the proposed scheme as follows.

Definition 1 (neighbors (NBR)).

Let $N_{i}$ and NBR denote a node and the set of neighbors of $N_{i}$ , respectively. That is, nodes in $NBR (N_{i})$ are within the transmission range R of node $N_{i}$ and can receive signals transmitted by node $N_{i}$ . The neighbors of node $N_{i}$ is defined as $NBR (N_{i}) = {N_{j} | d (N_{i}, N_{j}) \leq R, N_{i} \neq N_{j}}$ .

Definition 2 (forwarding candidate set (FCS)).

Let $N_{i}$ and FCS represent a node and the forwarding candidate nodes of node $N_{i}$ , respectively. The $FCS (N_{i})$ is a subset of $NBR (N_{i})$ . Given the source s and the SINK (destination) S, node $N_{j}$ is a node in the transmission path from $N_{i}$ and S. The forwarding candidate set of node $N_{i}$ is $FCS (N_{i}) = {N_{j} | d (N_{j}, S) < d (N_{i}, S), N_{j} \in NBR (N_{i})}$ .

Definition 3 (forwarding node (FWD)).

Let $N_{i}$ and FWD represent a node $N_{i}$ and a forwarding node of $N_{i}$ , respectively. The $FWD (N_{i})$ is one of $FCS (N_{i})$ which received the packet from $N_{i}$ and it forwards the packet. In case the residual energy is higher than the high threshold of the energy, $FWD (N_{i})$ is determined by distance to the SINK S in $FCS (N_{i})$ . $FWD (N_{i})$ is decided by the weighting factor of distance to the SINK S and the residual energy in $FCS (N_{i})$ when the residual energy is larger than the low threshold of energy. $FWD (N_{i}) = {N_{j} | d (N_{j}, S)$ is $MIN d (N_{j}, S)$ if $E_{Residual} (N_{j}) > E_{High}$ and $α \times E_{Residual} (N_{j}) + (1 - α) d (N_{j}, S)$ if $E_{Residual} (N_{j}) > E_{Low}$ , $N_{j} \in FCS (N_{i})}$ where MIN $d (N_{j}, S)$ is the minimum $d (N_{j}, S)$ in $FCS (N_{i})$ and $E_{Residual} (N_{j})$ is the residual energy of $N_{j}$ in $FCS (N_{i})$ . α is the weighting factor. $E_{High}$ is the high threshold of energy and $E_{Low}$ is the low threshold of energy. These two thresholds are configurable parameters.

As shown in Figure 2, a node ( $N_{i}$ ) is one node on the transmission path between the source $(s)$ and the SINK $(S)$ . There are five nodes in communication range of $N_{i}$ . $NBR (N_{i}) is {N_{1}, N_{2}, N_{3}, N_{4}, N_{5}}$ and $FCS (N_{i}) is {N_{1}, N_{2}, N_{5}}$ . $FWD (N_{i}) is {N_{1}}$ , where $N_{1}$ is nearest to S in $FCS (N_{i})$ and has the largest residual energy in $FCS (N_{i})$ . On the transmission path, $N_{1}$ is the upstream node for $N_{i}$ and $N_{4}$ is the downstream node for $N_{i}$ if $N_{i}$ is the forwarding node of $N_{4}$ .

Figure 2

FCS and FWD model.

From Figure 2, we introduce how NBR, FCS, and FWD work together. A node ( $N_{i}$ ) transmits a packet. $NBR (N_{i})$ , ${N_{1}, N_{2}, N_{3}, N_{4}, N_{5}}$ , receives the packet. At first, ${N_{3}, N_{4}}$ discards the packet because $d (N_{n}, S)$ is larger than $d (N_{i}, S)$ . $FCS (N_{i})$ , ${N_{1}, N_{2}, N_{5}}$ , computes, respectively, waiting time before packet transmission depending on delay estimation, residual energy, and distance to S. One of $FCS (N_{i})$ , ${N_{1}}$ , becomes $FWD (N_{i})$ which satisfies the conditions as follows: (i) it can transmit the packet within the predetermined in the packet, (ii) the distance between it and S is the shortest in $FCS (N_{i})$ , and (iii) its residual energy is larger than low energy threshold, $E_{Low}$ , in $FCS (N_{i})$ . The $N_{1}$ forwards the packet. Due to flooding-based data transmission, the packet from $FWD (N_{i})$ , ${N_{1}}$ , results in abortion message to the others of $FCS (N_{i})$ , ${N_{2}, N_{5}}$ , which have already received the packet from $N_{i}$ in Figure 2. This means that I forward this packet, so you do not have to forward it. As a result, the others $FCS (N_{i})$ discard the packet. In sender-based flooding schemes, the sender decides the forwarding node, which introduces communication overheads on maintaining the neighbor information [15] and leads to increasing energy consumption. In contrast, it notes that the receiver itself seems to determine whether to forward the packet or not in the proposed scheme.

Furthermore, the forwarding packet from $FWD (N_{i})$ works as acknowledgment for the sender, $N_{i}$ . The sender knows who its forwarding node is. Consequently, the intermediate nodes on the transmission path know their upstream node and downstream node. Without periodic message exchange, one-hop neighbor information can be obtained after data transmission has proceeded for a while. After then the proposed scheme transmits packet using the neighbor information without flooding-based mechanism. Energy conservation is achieved through the receiver-based data transmission and neighbor information without periodic control message.

4. Power-Aware Data Transmission for Real-Time Communication in MSNs

We present a power-aware data transmission for real-time communication in MSNs. The proposed scheme transmits packets meeting their deadline. To provide real-time performance and conserve energy consumption, we design the data transmission scheme based on delay estimation (Algorithm 2) without periodic control message exchange. The proposed scheme decreases the redundancy in flooding-based operation by means of reducing the number of the forwarding nodes and greedy forwarding that transmits packet to the node closest to SINK. In the proposed scheme, the node broadcasts the packet; however, one node $FWD (N_{i})$ transmits the packet under normal condition.

The proposed scheme does not exchange periodic control message to collect the neighbor information. The node caches the estimated delay time of neighbors when it receives a forwarding packet from the neighbors. In the beginning of communication, the proposed scheme (Algorithm 5) builds up the neighbor information with a constrained flooding-based mechanism. After a while, it enhances the real-time performance using the neighbor information as depicted in Figure 3.

Figure 3

Transmission progress of proposed scheme.

Our proposed scheme consists of three components.

(1)

Waiting time computation and delay estimation: computing the waiting time before forwarding; the waiting time of each node is different according to residual energy and distance to SINK and estimating delay to SINK based on one-hop neighbor information in the transmitted packet.

(2)

Forwarding decision: forwarding the packet provides real-time performance.

(3)

Neighbor information update: for communication, neighbor information is updated.

As shown in notations section the notations with a brief description are used in the proposed scheme.

4.1. Waiting Time Computation

The nodes forward the incoming packet after the waiting time. Each node will have different waiting time, $Tim e_{Wait}$ , depending on the residual energy and the distance to SINK. As shown in Algorithm 1, if the residual energy is larger than the high threshold value of energy, then the node nearest to SINK will forward the packet because it will have the shortest waiting time to forward the packet. Algorithm 1 shows how to compute the waiting time ( $Tim e_{Wait}$ ).

Algorithm 1: Waiting time computation.

(1) if $E_{Residual} > E_{High}$ then $⊳ E_{High}$ is high threshold of energy

(2) $Tim e_{Wait} = d (N_{i}, SINK) / D_{Factor}$

(3) else

(4) if $E_{Residual} > E_{Low}$ then $⊳ E_{Low}$ is low threshold of energy

(5) $Tim e_{Wait} = α \times E_{Factor} / E_{Residual} + (1 - α) \times d (N_{i}, SINK) / D_{Factor}$

(6) else

(7) Time_Wait = $W_{Max}$ $⊳ W_{Max}$ is maximum of $Tim e_{Wait}$

(8) end if

(9) end if

Algorithm 2: Delay estimation.

(1) Delay_OneHop = Time_Rcv − Time_Trans + Time_Wait

(2) Hops_Expected = $⌈ d (N_{i}, SINK) / R ⌉$

(3) ${Delay}_{N to End}$ = Delay_OneHop × Hops_Expected

If the residual energy is between the high threshold of energy and the low threshold value of energy, then the waiting time, $Tim e_{Wait}$ , will depend on both of the residual energy and the distance to SINK. This will help balanced energy consumption among the nodes. If the residual energy is smaller than the low threshold value of energy, then it will wait the maximum time because it will not forward the packet unless all nodes which received the packet have the residual energy smaller than the low threshold value of energy. Due to this processing, it prevents a node from dying quickly even though other nodes have sufficient residual energy.

The value of α is a configurable parameter that is set to any value between 0 and 1. α assigns relative weights to distance to SINK and residual energy. The forwarding decision can be distance-based or energy-based depending on the value of α. For example, the value of α close to 0 favors the distance-based forwarding decision. Figure 4 demonstrates delay performance under different values of α. The distance-based forwarding decision shows better performance than the energy-based forwarding decision in Figure 4. Figure 5 shows average energy consumption under different values of α when the depleted nodes happen in the network. Average energy consumption is not quite different according to the value of α. Therefore, the appropriate value of α is less than 0.5 with respect to better delay performance.

Figure 4

End-to-end delay performance under different α.

Figure 5

Average energy consumption under different α.

As a result of the simulation, the delay and energy performance are better when the value of $D_{Factor}$ is between 200 and 300. The performance is better when the value of $E_{Factor}$ is 10. The value of $D_{Factor}$ is influenced by network size.

4.2. Delay Estimation

For the computation of estimated delay ( $Dela y_{N to End}$ ), we use both of estimated number of hops and real transmission delay between two nodes. In the beginning, the node does not have any real value of transmission delay; therefore, it will use the estimated number of hops only to compute the estimated delay time. The estimated number of hops is calculated by dividing the distance to SINK by transmission range. One-hop delay is calculated at the upstream node using the transmission time and reception time. The elapsed time refers to $Dela y_{OneHop}$ :

\begin{matrix} {Delay}_{OneHop} = {Time}_{Rcv} - {Time}_{Trans} + {Time}_{Wait}, \end{matrix}

(1)

where Time_Rcv is the time when upstream node receives the last bit of packet to the physical link and Time_Trans is the time when downstream node transmits the first bit of packet to the physical link. Time_Wait is the duration that the packet is waited on the upstream node before transmission. Therefore, once upstream node successfully receives the packet, the node measures the elapsed time from transmission of downstream node to reception of upstream node. The total estimated delay time between node $N_{i}$ and SINK is calculated as

\begin{array}{l} {Delay}_{N to End} = \sum ‍ {Delay}_{OneHop} \\ = Dela y_{OneHop} \times Hop s_{Expected}, \end{array}

(2)

where $Dela y_{N to End}$ is the estimated end-to-end delay. $Hop s_{Expected}$ is the expected hop count to SINK, which is estimated by $d (N_{i}, SINK) / R$ . In other words, $Dela y_{N to End}$ is $Dela y_{OneHop}$ multiplied by $Hop s_{Expected}$ . $Dela y_{N to End}$ is compared to Deadline, the predetermined time requirement of the packet. After then the node $N_{i}$ updates the remaining time to Deadline. The deadline at node $N_{i}$ is computed as follows:

\begin{matrix} Deadline at N_{i} = Deadline at N_{i - 1} - Dela y_{OneHop} . \end{matrix}

(3)

When the node $N_{i}$ got the packet, it can know the actual transmission delay from $N_{i - 1}$ to $N_{i}$ because node $N_{i}$ knows the arrival time to node $N_{i}$ and the departure time at node $N_{i - 1}$ with the information in the received packet. The total delay time from node $N_{i - 1}$ to SINK equals the sum of the actual transmission delay from node $N_{i - 1}$ to node $N_{i}$ and the estimated delay from node $N_{i}$ to SINK. Because node $N_{i}$ knows the transmission delay from node $N_{i - 1}$ to node $N_{i}$ , the estimated delay from node $N_{i}$ to SINK can be replaced with transmission delay from $N_{i - 1}$ to $N_{i} \times$ estimated number of hops as (3).

Furthermore, the node $N_{i}$ can receive the forwarding packet from node $N_{i + 1}$ as shown in Figure 2, the node $N_{i}$ can know the round traveling time between node $N_{i}$ and node $N_{i + 1}$ . Therefore, the node $N_{i}$ can compute the transmission delay time including the waiting time at node $N_{i + 1}$ . The node $N_{i}$ can compute the new estimated total delay from node $N_{i}$ . Due to the forwarding packet from next node, each node can know the actual transmission delay time to next node; the node $N_{i}$ will reflect those transmission delays more and more. Eventually, the total estimated delay will reflect actual transmission delay only. This reflects actual traffic condition at each node without additional message exchange as shown in Figure 6.

Figure 6

The estimated delay is replaced with the actual delay in each round.

4.3. Forwarding Decision

As we mentioned earlier, the proposed scheme makes the forwarding decision on receiver-side without periodic control message exchange such as neighbor information. One of receivers becomes the forwarding node according to Algorithm 3. As described at line 5 in Algorithm 1, $Tim e_{Wait}$ becomes the smallest if a node is the closest to SINK or has the larger residual energy. An example of forwarding decision is illustrated in Figure 7. When the node receives the packet, then it checks senders’ location, sequence number of the packet, and $Mod e_{Trans}$ in the header of the transmitted packet.

Algorithm 3: Forwarding decision.

(1) if out-of-range then

(2) DiscardPacket

(3) end if

(4) if Packet_Rcv == Packet_Sent then ⊳ ACK from FWD( $N_{i}$ )

(5) ${Delay}_{N to End}$ = (Time_Rcv − Time_Sent)/2 + Delay_Trans

(6) ⊳ Update the estimated delay using the actual delay

(7) else

(8) if Packet_Rcv == Packet_Waiting then

(9) ⊳ Waiting in the packet list before forwarding

(10) if Mode_Trans == 1 then ⊳ Aid from FCS( $N_{i}$ )

(11) Time_Wait = Time_Wait/2

(12) else ⊳ Abortion from FWD( $N_{i}$ )

(13) ${Delay}_{N to End}$ = Delay_Trans

(14) ⊳ Update the estimated delay using the actual delay

(15) AbortPacket

(16) end if

(17) else ⊳ New packet, I becomes one of FCS( $N_{i}$ )

(18) EstimateDelay

(19) SetupForwardingMode

(20) ComputeWaitingTime

(21) end if

(22) end if

(23) FowardPacket

Figure 7

An example of illustrating forwarding decision.

4.3.1. Out of Range

When the receiver locates out between −60 degree and +60 degree from the line of sender to the sink, in this case most forwarding packet from the forwarder could not be received because there are high chance that the forwarder could locate close to the line from forwarder to sink node. Therefore the received packet will be discarded without receiving forwarding packet and comparing the sequence number of that packet.

4.3.2. $N_{i}$ Receives the Packet from the FWD( $N_{i}$ )

If the sequence number of the received packet is same as the packet that the receiver sent before, then the receiver is the original sender of the packet. Therefore, the packet can be regarded as the acknowledgment (ACK) signal from the first receiver, which means the packet was forwarded without hole. The packet can be used to compute the round trip latency time.

4.3.3. FCS( $N_{i}$ ) Receives the Packet from the FWD( $N_{i}$ )

If the sequence number of the received packet is same as the packet the node received before, then other nodes already forwarded the packet. In addition, $Mod e_{Trans}$ is FWD(0), which means that it does not need aid to forward the packet. Therefore, the node aborts the waiting event for the packet. This prevents flooding the packet.

4.3.4. NBR( $N_{i}$ ) Receives the Packet from the $N_{i}$

When the node receives the packet, it computes the transmission delay time, $Dela y_{OneHop}$ , by checking the arrival time, $Tim e_{Rcv}$ . After deducting the transmission delay time, $Dela y_{OneHop}$ , from Deadline, it compares the result with its estimated delay time, ${Delay}_{N to End}$ . If the result is larger than the estimated delay time, $Dela y_{N to End}$ , then the node updates Deadline and transmission time, $Tim e_{Trans}$ , in the packet fields. This will be done while waiting for $Tim e_{Wait}$ . After $Tim e_{Wait}$ , it forwards the packet. If the result is smaller than the estimated delay time, $Dela y_{N to End}$ , that means that the node has a slim chance to deliver the packet within Deadline. Then the node shortens the waiting time a half, ${Time}_{Wait} / 2$ by changing $D_{Factor}$ and $E_{Factor}$ , and set $Mod e_{Trans}$ bit to 1, that is, AID for help from the neighboring nodes. In addition, the node should modify Deadline and transmission time, $Tim e_{Trans}$ , in the packet header. The waiting event will be fired after $Tim e_{Wait}$ .

4.3.5. $N_{i}$ Receives the Packet Whose $Mod e_{Trans}$ Is 1 (AID)

It means one of NBR( $N_{i}$ ) requires the emergent aid. Therefore, the receiver should forward the packet as soon as possible. The receiver checks whether it already received the packet same as this packet with the sequence number. If two packets have same sequence number and $Tim e_{Trans}$ of both is 1 (AID), then the receiver aborts the waiting event and discards those packets.

If there is no packet with same sequence number or $Mod e_{Trans}$ is changed from 0 to 1 even though two packets have same sequence number, then do the following things. The node deducts the transmission delay time, $Dela y_{OneHop}$ , from Deadline to compute new Deadline. Then it compares the value with the estimated delay time, $Dela y_{N to End}$ , at that node. If new value of Deadline is larger than the estimated delay time, $Dela y_{N to End}$ , then reset $Mod e_{Trans}$ to 0 and wait for $Tim e_{Wait}$ before forwarding the packet. If new value of Deadline is smaller than the estimated delay, $Dela y_{N to End}$ , then the node shortens the waiting time a half, ${Time}_{Wait} / 2$ , and fires the event, that is, forwards the packet.

4.4. Hole Management

If the sender cannot receive any forwarding packet which can be used as acknowledgment during the maximum waiting time after sending a packet, there is no node within the transmission range. It is called a hole. There can be two approaches to deal with hole. First, the sender increases the transmission range twice and retransmits the packet. Secondly, the sender, $N_{i}$ , let the previous sender, $N_{i - 1}$ , know that there exists hole around $N_{i}$ and retransmit the packet. In this case, the node $N_{i}$ will not be involved in forwarding the packet. Therefore, the packet will be forwarded by other nodes. The proposed scheme is the latency-sensitive approach. Thus, we choose the first approach.

4.5. Neighbor Information Update

When the node $N_{i - 1}$ receives the forwarding packet from node $N_{i}$ , the node $N_{i - 1}$ can recognize that the packet is same as the packet it sent by comparing the sequence number of the packet. Then before discarding the packet, it reads the arrival time, $Tim e_{Rcv}$ , of that packet and updates the estimated delay time, $Dela y_{N to End}$ , at node $N_{i - 1}$ in the cache with sum of the estimated delay time, $Dela y_{N to End}$ , at node $N_{i}$ and the round trip time/2, that is, RTT/2. It is depicted in the line 5 of Algorithm 3.

When other nodes which received the packet whose $Mod e_{Trans}$ is $AID (1)$ from the node $N_{i - 1}$ receive the packet same as the packet from node $N_{i}$ , then they update their estimated delay, $Dela y_{N to End}$ , with the estimated delay, $Dela y_{Trans}$ , from node $N_{i}$ before discarding the packet. It is illustrated in the line when other nodes which received the packet whose $Mod e_{Trans}$ is FWD(0) from the node $N_{i - 1}$ receive the packet same as the packet from node $N_{i}$ except $Mod e_{Trans}$ is $AID (1)$ ; then they update their estimated delay, $Dela y_{N to End}$ , with the estimated delay, $Dela y_{N to End}$ , from node $N_{i}$ before forwarding the packet. When other nodes which received the packet whose $Mod e_{Trans}$ is $AID (1)$ from the node $N_{i - 1}$ receive the packet same as the packet from node $N_{i}$ , whether its $Mod e_{Trans}$ is $AID (1)$ or is FWD(0), then they discard it.

By doing this step, every node can use the estimated delay time which reflects actual delay time partially from second packet. As more packets flow, the estimated delay time reflect more actual delay time. Eventually every node involved in the transmission will use the actual delay time as its estimated delay time.

In order that the algorithm here operates, the packet header should have some fields including Deadline, $Tim e_{Trans}$ , estimated delay time $Dela y_{Trans}$ , and $Mod e_{Trans}$ as shown in Figure 8.

Figure 8

Packet header format in proposed scheme.

5. Performance Evaluation

5.1. Simulation Environment

In this section, we evaluate the proposed scheme in NS-3 v.3.19. According to [17], the energy consumption model is as follows:

\begin{matrix} E_{R x} (l) = E_{R x - elec} (l) = l E_{elec} \\ E_{T x} (l, d) = E_{T x - elec} (l) + E_{T x - amp} (l, d) \\ E_{T x} (l, d) = {\begin{cases} l E_{elec} + l ε_{fs} d^{2} & d < d_{0}, \\ l E_{elec} + l ε_{mp} d^{4} & d \geq d_{0}, \end{cases} \end{matrix}

(4)

where $E_{elec}$ is the electronics energy, $l ε_{fs} d^{2}$ is the amplifier energy in the free space and $l ε_{mp} d^{4}$ is the amplifier energy in the multipath fading channel models, l is the data size in bits, d is the transmission distance, and $d_{0}$ is a distance threshold.

The simulation parameters are described in Table 1.

Table 1

Network parameters used in the performance evaluation.

Parameter	Value
Network size	400 m × 400 m
Number of the nodes	80 to 160
Packet size	288 bits
Initial energy	10 J
Transmission range	40 m
Propagation model	Two-Ray Ground
Physical layer	wireless-phy
MAC layer	IEEE 802.11
Data rate	11 Mbps

5.2. Performance Metrics

We evaluate the following performance metrics.

(i)

Deadline success ratio is defined as the ratio of the number of successfully delivered packets which reach SINK within the deadline to the total number of packets transmitted.

(ii)

Average energy consumption includes average energy dissipation for communication, computation, and sensing energy dissipation.

We will evaluate the performance by comparing our scheme with the related works, CFlood [14] and MMSPEED [16], using neighbor information exchanged periodically.

5.3. Simulation Results

5.3.1. Average Energy Consumption

The energy consumption is the most important for the sensor networks. As shown in Figure 9, our scheme performs better than the data transmission scheme with exchanging control message periodically. The average energy consumption is closely related to network lifetime. The larger the average energy consumption is, the shorter the network lifetime is. It illustrates that average energy consumption result bigger difference between our scheme and the scheme with periodic control message exchange. α is a configurable parameter that is set to 0.1, 0.2, and 0.3 as explained in Section 4. In our scheme, the distance-based forwarding decision when α is close to 0 performs better than CFlood and MMSPEED. The energy consumption performance can be maintained when the number of nodes increases in Figure 9. Figure 10 shows that our scheme yields good energy conservation performance according to end-to-end deadline. Therefore, our scheme not only provides real-time performance, but also prolongs network lifetime without periodic control message exchange.

Figure 9

Energy consumption impacted by number of nodes.

Figure 10

Energy consumption impacted by end-to-end deadline.

5.3.2. End-to-End Deadline Success Ratio

The end-to-end deadline is critical in the real-time communication. Figure 11 shows end-to-end deadline success ratio for our scheme and the data transmission scheme such as CFlood and MMSPEED. Figure 11 shows the performance for node density. Our proposed scheme meets the tolerated latency independent from node density and good performance without periodic message exchange when it is compared with the CFlood and MMSPEED. Figure 12 illustrates end-to-end deadline success ratio according to the required deadline. The total time of the simulation is 4,000 seconds. α is a configurable parameter that is close to 0 such as 0.1, 0.2, and 0.3. The distance-based forwarding decision enhances the real-time performance. Our scheme yields an acceptable performance without periodic control message exchange (Algorithm 4).

Algorithm 4: Forwarding mode setup.

(1) Deadline = Deadline − Delay_OneHop

(2) if ${Delay}_{N to End}$ ≤ Deadline then ⊳ Meet the deadline

(3) Mode_Trans = 0 ⊳ I can forward

(4) else

(5) Mode_Trans = 1 ⊳ need Aid, I cannot forward

(6) $E_{Factor}$ = $E_{Factor} / 2$

(7) $D_{Factor}$ = $D_{Factor}$ × $2$

(8) end if

Algorithm 5: Forwarding packet.

(1) timer ← Time_Wait

(2) AddpacketInthePacketList

(3) while timer == 0 do

(4) timer ← timer − 1

(5) end while

(6) TransmitPacket

Figure 11

Packet deadline success ratio impacted by number of nodes.

Figure 12

Packet deadline success ratio impacted by end-to-end deadline.

6. Conclusion and Future Work

In this paper, we present and evaluate a power-aware data transmission for real-time communication in MSNs. It is designed to provide real-time performance delivering multimedia data within the predetermined time (deadline) through delay estimation and greedy forwarding. In addition, it improves the energy conservation without periodic control message exchange and the constrained flooding-based data transmission. In contrast to previous schemes, the proposed scheme builds up neighbor information using the transmitted packet instead of periodic message exchange. Consequently, the proposed scheme conserves the energy consumption and, therefore, extends network lifetime, which is one of crucial design issues in MSNs. Simulations show that the proposed scheme performs good with respect to the deadline success ratio compared to the existing schemes.

There are some issues that remain to be studied. We have not considered the mobility of the nodes including camera sensor nodes and SINK. The network mobility should be taken into account in accordance with the recent researches and characteristic of real-time application in MSNs. In addition, it is desirable to adapt data transmission by applying data aggregation to decrease the amount of information-rich data such as image, video, and multimedia data. Finally, the security is a demanding issue with respect to the application such as surveillance and health-care monitoring which deal with the important data.

Footnotes

Notations

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

References

Akyildiz

I. F.

Melodia

Chowdury

K. R.

Wireless multimedia sensor networks: a survey

IEEE Wireless Communications 2007 14 6 32 39

2-s2.0-37249066703

10.1109/MWC.2007.4407225

Soro

Heinzelman

A survey of visual sensor networks

Advances in Multimedia 2009 2009 21

2-s2.0-68949132673

10.1155/2009/640386

640386

Yuan

Cheng

An energy-efficient real-time routing protocol for sensor networks

Computer Communications 2007 30 10 2274 2283

2-s2.0-34447634483

10.1016/j.comcom.2007.06.002

Quang

P. T. A.

Kim

Enhancing real-time delivery of gradient routing for industrial wireless sensor networks

IEEE Transactions on Industrial Informatics 2012 8 1 61 68

2-s2.0-84863054934

10.1109/TII.2011.2174249

Ehsan

Hamdaoui

A survey on energy-efficient routing techniques with QoS assurances for wireless multimedia sensor networks

IEEE Communications Surveys and Tutorials 2012 14 2 265 278

2-s2.0-84860890708

10.1109/SURV.2011.020211.00058

Mahapatra

Anand

Agrawal

D. P.

QoS and energy aware routing for real-time traffic in wireless sensor networks

Computer Communications 2006 29 4 437 445

2-s2.0-32644433771

10.1016/j.comcom.2004.12.028

Karp

Kung

H. T.

GPSR: greedy perimeter stateless routing for wireless networks

Proceedings of the 6th Annual International Conference on Mobile Computing and Networking (MobiCom '00)

August 2000

243 254

2-s2.0-0034547115

Mauve

Widmer

Hartenstein

A survey on position-based routing in mobile ad hoc networks

IEEE Network 2001 15 6 30 39

2-s2.0-0035509489

10.1109/65.967595

Birman

K. P.

Hayden

Ozkasap

Xiao

Tseng

Chen

Sheu

The broadcast storm problem in a mobile ad hoc network

Proceedings of the ACM and IEEE International Conference of Mobile Computing and Networking

August 1999

151 162

10.

Cai

Hua

K. A.

Phillips

Leveraging 1-hop neighborhood knowledge for efficient flooding in wireless ad hoc networks

Proceedings of the 24th IEEE International Performance, Computing, and Communications Conference (IPCCC '05)

April 2005

347 354

2-s2.0-28444469250

11.

Sabbineni

Chakrabarty

Location-aided flooding: an energy-efficient data dissemination protocol for wireless sensor networks

IEEE Transactions on Computers 2005 54 1 36 46

2-s2.0-13244292297

10.1109/TC.2005.8

12.

Liu

Jai

Wan

Liu

Yao

F. F.

A distributed and efficient flooding scheme using 1-hop information in mobile ad hoc networks

IEEE Transactions on Parallel and Distributed Systems 2007 18 5 658 671

2-s2.0-34247555864

10.1109/TPDS.2007.1023

13.

Khabbazian

Bhargava

V. K.

Localized broadcasting with guaranteed delivery and bounded transmission redundancy

IEEE Transactions on Computers 2008 57 8 1072 1086

2-s2.0-47049086236

10.1109/TC.2008.51

14.

Jiang

Ravindran

Cho

CFlood: a constrained flooding protocol for real-time data delivery in wireless sensor networks

Stabilization, Safety, and Security of Distributed Systems: Proceedings of the 11th International Symposium, SSS 2009, Lyon, France, November 3–6, 2009 2009 5873

Berlin, Germany

Springer

413 427 Lecture Notes in Computer Science

2-s2.0-70649107962

10.1007/978-3-642-05118-0_29

15.

Stankovic

J. A.

Abdelzaher

SPEED: a stateless protocol for real-time communication in sensor networks

Proceedings of the 23th IEEE International Conference on Distributed Computing Systems

May 2003

46 55

2-s2.0-0037968893

10.1109/ICDCS.2003.1203451

16.

Felemban

Lee

Ekici

MMSPEED: multipath multi-SPEED protocol for QoS guarantee of reliability and timeliness in wireless sensor networks

IEEE Transactions on Mobile Computing 2006 5 6 738 753

2-s2.0-33646456047

10.1109/TMC.2006.79

17.

Chu

Jan

A GPS-less, outdoor, self-positioning method for wireless sensor networks

Ad Hoc Networks 2007 5 5 547 557

2-s2.0-33947275846

10.1016/j.adhoc.2006.03.004