Reliable Graph Routing in Industrial Wireless Sensor Networks

Hindawi Publishing Corporation International Journal of Distributed Sensor Networks Volume 2013, Article ID 758217, 15 pages http://dx.doi.org/10.1155/2013/758217

Research Article Reliable Graph Routing in Industrial Wireless Sensor Networks Jing Zhao, Yajuan Qin, Dong Yang, and Junqi Duan National Engineering Laboratory for Next Generation Internet Interconnection Devices, School of Electronic and Information Engineering, Beijing Jiaotong University, Beijing 100044, China Correspondence should be addressed to Dong Yang; [email protected] Received 5 August 2013; Revised 7 October 2013; Accepted 7 October 2013 Academic Editor: Yu Gu Copyright © 2013 Jing Zhao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Research studies on smart cities have been conducted, which will enable a better management of the available resources. Industrial wireless sensor networks (IWSNs) are important part of smart city. IWSNs are used for process measurement and control applications in harsh and noisy industrial environments. As substitutes for traditional wired industrial networks, IWSNs are more flexible, scalable, and efficient. However, resource limitation of the sensor nodes and unreliability of low-power wireless links, in combination with stringent quality of the service (QoS) requirements of industrial applications, imposes many challenges in designing efficient routing for IWSNs. Existing standards propose a simple and reliable routing mechanism named graph routing. In this paper, we propose novel routing algorithms to discover reliable paths and construct reliable routing graphs. In our approaches, the centralized manager selects parent nodes for each node in the network to satisfy reliability requirements of the intended application. We try to maximize the number of reliable nodes and change the parent nodes selection strategy along with the link quality and the link correlation. Our design is evaluated using simulation where we show that our algorithm could achieve a balance between routing reliability and overhead.

1. Introduction The ubiquitous computing research studies and the deployment of urban wireless infrastructures make smart cities take shape [1–5]. The concept of smart cities is a natural evolution of research studies about smaller-scale smart spaces [6, 7]. Research studies on smart cities have been conducted by not only some organizations but also several governments. Smart cities form advanced collaborations to optimize resource allocation by monitoring the urban environment, collecting and conveying information, and establishing automated control. Wireless sensor networks (WSNs) can be used to integrate with the surrounding environment and interact with the management system. In smart cities, IWSNs as a smaller smart space can intercommunicate and interoperate with other networks of smart cities providing a unified and personalized interface for their managers. IWSNs are new industrial networks appearing after field bus in industrial control field [8–13]. IWSNs should be designed specifically due to the extreme industrial environments and the stringent industrial application requirements.

WirelessHART [14] approved by IEC is the first global industrial wireless communication standard. It is the extension of wired HART for wireless process monitoring and control. On September 6, 2011, ISA-100.11a [15] is recognized by IEC as a new standard for process automation (PA) and factory automation (FA). Nowadays, more and more research studies about IWSNs have emerged. In industrial systems, the existing standards advocate centralized control. Unlike the distributed control adopted by the traditional WSNs, IWSNs push the complexity of ensuring reliable and accurate data transfers to the centralized manager. The construction of the global reliable routing is performed by the centralized manager to meet stringent requirements [16–18]. The centralized control could achieve relatively complicated algorithms and makes control decisions based on the whole network’s information. Therefore, the centralized control could improve network performance by global information and globally optimal control laws. In typical IWSNs, each field device collects data and publishes its data to the manager, and then the manager dispatches the control messages to field devices periodically. The

2 broadcast messages are propagated through the network to configure the node information. Uplink has a sample rate of each device to transmit its process message to the centralized manager. The broadcast and uplink traffic in industrial applications is mainly a cyclical communication with a predicable interval. In some situations, the manager needs to generate downlink path for each device to forward specific control messages. For industrial control communication, lost packets may lead to industrial application malfunction. But there are many challenges for industrial field devices to achieve reliable transmission. On the one hand, too many wireless devices in the 2.4 GHz frequency band likely cause mutual interference. On the other hand, the industrial environment is extremely severe which may lead to signal reflection and scattering. How to satisfy the reliability requirements is a challenging problem in routing algorithm. A simple and reliable routing mechanism named graph routing is presented by ISA 100.11a [15] and WirelessHART [14]. However, these current standards do not specify a routing algorithm. How to construct a reliable graph has not received enough attention and how to satisfy the strict reliability requirements in the graph routing is still a challenge. Now several research works have been devoted on the reliable graph routing [19–21]. But existing approaches educe the routing graphs based on the fixed routing strategy, which may decrease flexibility to balance communication performances. We hope that the routing strategy could decrease overhead while the reliability can be guaranteed. Therefore, we present flexible route selection algorithms which consider the reliability requirement together with network overhead. The route selection of each node depends on the link quality with neighbor node. In a word, we propose the efficient algorithms to construct reliable graphs: determining the routing that (1) achieves the maximum proportion of nodes which successfully transmit packets to the destination node (2) and a tradeoff between the reliable routing and the network overhead. In this paper, our algorithms choose reliable parent nodes to construct reliable graphs. We maintain maximum number of reliable nodes and present an evaluation method of node reliability. The remainder of this paper is organized as follows. Section 2 presents related works of reliable routing. Section 3 introduces the architecture of IWSNs. Section 4 describes the proposed centralized reliable graph routing. This section details how to construct reliable graphs. Section 5 presents the performance of our approach compared with other graphs. Finally, we conclude our approach and summarize the future work in Section 6.

2. Related Works In this section, we first present previous works about reliable routing in wireless sensor networks (WSNs). Then two routing approaches defined by existing industrial standards are described. 2.1. Reliable Routing in Wireless Sensor Networks. IWSNs are meant for industrial measure and control with high reliable

International Journal of Distributed Sensor Networks requirements. Link error is usually due to network congestion, node failure, signal interference, and so on. How to select a reliable path is an important research for routing protocols. We also focus on how to repair path if route entries become invalid. Some routing protocols regard link error rate as a routing metric for reliability [22, 23]. In [22], energy-aware routing taking into account the retransmission can choose a reliable path by energy cost. Because link error rate can cause the potential energy cost. In [23], the source node sends a redundant packet via alternate path to achieve reliability sometimes. If the reliability of the selected path does not meet the requirements, the source node will randomly select one additional successor. Some routing protocols achieve fault tolerance by providing multiple path [24–29]. Multipath can improve communication reliability, realize load-balancing, and relieve congestion. Ad hoc on-demand multipath distance vector (AOMDV) [26] is the multipath extension of a single path routing protocol known as ad hoc on-demand distance vector (AODV). AODMV maintains disjoint multipath which fails independently. The disjoint paths include link-disjoint path and node-disjoint path. Link-disjoint is effective in keeping links available while node-disjoint is appropriate for relieving congestion. Most of these works focus on identifying reliable path and constructing multiple path. IWSNs should adapt to harsh environment and meet stringent requirements. ISA100.11a [15] and WirelessHART [14] as main industrial standardization groups present graph routing which also supports multipath. 2.2. Source Routing and Graph Routing. These industrial standards define two primary routing approaches: source routing and graph routing. Source routing is a general routing supported by simple protocol and a single direct route from a source to a destination. The next hop of a packet will be directed to a specific neighbor. The ordered list of routing is included in the network layer header. Source routing only establishes a single path between the source and the destination. In graph routing, the packet is disseminated to its destination according to routing graph. The routing graph is a set of directed links and identified by unique graph ID. All nodes will be preconfigured with relevant graph information about which neighbor the packets can be forwarded. The network layer header of packets just need to include the graph ID. In a graph, the neighbors which can be forwarded are not single. So graph routing makes it possible to provide redundant communication path. The source routing and graph routing information defined in separate network header fields could coexist in industrial standards. Reference [19] presents efficient algorithms to construct three types of reliable routing graphs for different communication purposes. This paper ensures reliability by redundant parent nodes. To construct reliable graph, this paper choose two different parent nodes so that there are two communication paths at least. The experiment results show that the algorithms can achieve highly reliable routing at the cost of

International Journal of Distributed Sensor Networks

3

modest configuration overhead compared with other baseline methods.

3. Industrial Wireless Sensor Networks Architecture The architecture of IWSNs mainly consists of centralized manager, backbone routers (BBRs), and field devices (see Figure 1). The centralized manager not only takes charge of the management of the entire network but also represents infrastructure device to communicate with external networks. In IWSNs, each field device, sensor or actuator, has a designed sample rate to publish its data to manager and receive control commands. These field devices attached to the plant devices collect data, in this paper, and also route message [16]. The BBRs unite multiple wireless subnets and make industrial wireless systems interoperate with control networks. Multiple BBRs make the system robust with redundancy and optimize the communication path. Multiple BBRs provide multiple access points in IWSNs in order to facilitate node arrangement [30]. So the ISA100.11a architecture employs multiple BBRs and they are all connected by wire [15]. Both ISA100.11a and WirelessHART adopt IEEE 802.15.42006 [31] as the bottom of their communication stack. These standards define TDMA, channel hopping, and industrystandard AES-128 ciphers, and keys. In network layer, the backbone network conforms to Internet Protocol (IP), IPv4 and newer IPv6. In IWSNs, these standards present two primary routing approaches: graph routing and source routing. The centralized manager is responsible for maintaining upto-date routing and guaranteeing the communication reliability.

4. Centralized Reliable Graph Routing In this section, we present the content concerning how to achieve the centralized reliable graph routing. The mechanism of graph routing is described in detail by industrial standards. We proposed the algorithms to construct reliable routing graph. We describe three reliable routing graphs for different communication purposes. For three communication types, the reliable routing graph should meet different requirements. 4.1. Communication Types. There are three communication types for industrial process monitoring and control (see Figure 2). The BBRs link up the field network and the control system. The types of communication are broadcast, uplink, and downlink. 4.1.1. Broadcast. Broadcast refers to the direction from the BBRs towards entire network. The centralized manager propagates the common configuration and control data to field devices periodically. 4.1.2. Uplink. Uplink refers to the direction from field devices towards the BBRs, in the reverse direction of broadcast. Every

Manager

Backbone router

Field device

Figure 1: The architecture of IWSNs.

field device has a specific sample rate to propagate its process data to the manager. 4.1.3. Downlink. Downlink refers to the direction from the BBRs towards any field device. The centralized manager propagates a specific control message to an individual device by the BBRs. The routing graphs have different characteristics for different communication types. The entire network can share the same broadcast graph and uplink graph. However, the centralized manager should generate the corresponding downlink graph for each field device. The reliability requirements are also different when applied to different communication types. 4.2. Notations. This section presents the notations used in this paper. In wireless network 𝐺 = (𝑉, 𝐸), 𝑉BBR denotes the set of BBRs, and 𝑉𝑁 denotes the set of field devices, and 𝑀 denotes the centralized manager. In this paper, we assume that centralized manager and 𝑉BBR are connected by wire, and the BBRs are wireless access points. The edge 𝑒𝑖,𝑗 ∈ 𝐸 represents V𝑖 and V𝑗 can communicate directly for all V𝑖 , V𝑗 ∈ 𝑉. Then uplink graph and broadcast graph are represented by 𝐺𝑈(𝑉𝑈, 𝐸𝑈) and 𝐺𝐵 (𝑉𝐵 , 𝐸𝐵 ), respectively. 𝐺𝑁(𝑉𝑁, 𝐸𝑁) represents the downlink graph of the node V𝑁. In this paper, we define 𝑅𝑁 as the expected reliability of 󸀠 as the publishing a packet from the 𝑉BBR to V𝑁 while 𝑅𝑁 expected reliability of sending a packet from V𝑁 to the 𝑉BBR . 𝑅𝑖,𝑗 denotes the expected reliability of communication from 󸀠 denotes the expected reliability of 𝑉BBR to V𝑖 via V𝑗 and 𝑅𝑖,𝑗 communication from V𝑖 to 𝑉BBR via V𝑗 . The expected reliability means the probability of sending a packet from source device to destination device successfully. For an edge 𝑒𝑖,𝑗 , we use 𝐿 𝑖,𝑗 to denote link reliability of single hop communication 󸀠 between V𝑖 and V𝑗 . Therefore, 𝑅𝑖,𝑗 and 𝑅𝑖,𝑗 are given by 𝑅𝑖,𝑗 = 𝑅𝑗 𝐿 𝑖,𝑗 , (1) 󸀠 = 𝑅𝑗󸀠 𝐿 𝑖,𝑗 . 𝑅𝑖,𝑗

4


Manager

Field device

Manager

Field device

BBR

BBR (a) The network topology

Manager

Field device

BBR

(b) Uplink

Manager

Field device

BBR (c) Broadcast

(d) Downlink

Figure 2: The Types of Communication.

In order to construct reliable graph, individual node should be positioned relative to the BBRs. So we use rank 𝑟𝑁 to denote relative position between the BBRs and a field device. 𝑟BBR = 0 at the BBRs and the rank of field devices connected with the BBRs directly is 1. The rank of a node is equal to the rank of its parent nodes plus one. If a node has more than one parent node, we choose the first selected parent node for rank calculation. The rank of a node is a scalar representation of location; therefore, the rank monotonically decreases towards the BBRs, increasing in the down direction and decreasing in the up direction. We choose reliability rather than hop count as criteria to choose parent node to improve reliability and reduce energy consumption. Equation (1) shows that the path with fewer hops has higher reliability in general, but more hops may be better when additional forwarding nodes are inserted according to certain principles to shorten the length of individual hops [32, 33]. In our approach, we define a reliability parameter 𝑅𝑑 to measure the reliability of the industrial application. The nodes with low rank are more critical and

need to be more reliable because more data are transited by 𝑟 these nodes. Therefore, our approaches set up a threshold 𝑅𝑑𝑖 to measure the reliability of V𝑖 . If the reliability of a node 𝑅𝑖 is 𝑟 greater than 𝑅𝑑𝑖 , the node is a reliable node in the graph and included in 𝑉𝑖𝑛 which is the reliable node set. Otherwise it will be included in the unreliable node set 𝑉𝑒𝑛 . 4.3. Link Correlation. A feature of wireless communication is that link errors are dependent between neighbor nodes. Recent works [34–36] have shown that correlated interference might lead to correlated packet errors because the interference signal can corrupt neighbor links simultaneously. Link correlation has significant effects on reliability attributes for our route selection. To illustrate the link correlation, we give an example to demonstrate the reliability attributes with correlation, as Figure 3. Link independence means that 𝐿(𝑖, 𝑗2 | 𝑖, 𝑗1 ) equals 𝐿 𝑖,𝑗2 , where 𝐿(𝑖, 𝑗2 | 𝑖, 𝑗1 ) is the link success probability between V𝑖 and V𝑗2 when the link between V𝑖 and V𝑗1 fails. When node V𝑖 forwards a packet, if V𝑖 forwards the packet


j1 Correlated L(i, j2 |i, j1 ) L(i, j1 |i, j2 )

L i,j1

L i,j2

i

5 retransmission process continues until either the packet is forwarded successfully or all parent nodes have been tried, as follows: 𝑅𝑖 = 1 − 𝑃 (𝑅𝑖,𝑗1 , 𝑅𝑖,𝑗2 , . . . , 𝑅𝑖,𝑗𝑛 ) , 𝑅𝑖󸀠 = 𝐿 𝑖,𝑗1 𝑅𝑗󸀠1 + 𝐿 (𝑖, 𝑗2 , 𝑖, 𝑗1 ) 𝑅𝑗󸀠2

j2

L i,j3

j3

Figure 3: The link correlation.

to V𝑗1 , the packet delivery probability is 𝐿 𝑖,𝑗1 . If V𝑗1 does not receive the packet and the packet is forwarded to V𝑗2 , the packet delivery probability is 𝐿(𝑖, 𝑗2 | 𝑖, 𝑗1 ). Therefore, when the packet is forwarded to V𝑗3 , the packet delivery probability is 𝐿(𝑖, 𝑗3 | 𝑖, 𝑗1 , 𝑖, 𝑗2 ). The node V𝑖 receives the broadcast packet successfully as long as one of its parent nodes transfers successfully. Therefore, the packet transmission probability is 1 − 𝑃(𝐿 𝑖,𝑗1 , 𝐿 𝑖,𝑗2 , 𝐿 𝑖,𝑗3 ), where 𝑃(𝐿 𝑖,𝑗1 , 𝐿 𝑖,𝑗2 , 𝐿 𝑖,𝑗3 ) is the probability that none of packets are received from V𝑗1 , V𝑗2 , V𝑗3 , which equals (1−𝐿 𝑖,𝑗1 )(1−𝐿 𝑖,𝑗2 )(1−𝐿 𝑖,𝑗3 ) if link independence. 4.4. Fault Tolerant and Reliability Attributes. The errors that occur on the communication links can be classified into two categories: transient errors and permanent errors. For permanent errors, the routing should be recovered by the centralized manager. But recovering from transient errors, hop-to-hop (data link layer) and end-to-end (transport layer) retransmission can be performed. The link layer retransmission cannot work sometimes in case of signal interference. And end-to-end retransmission has significant latency that would be unacceptable to some real-time industrial applications. The fault tolerant of our routing can be categorized into replication and retransmission mechanism for different communication types [37]. In 𝐺𝐵 and 𝐺𝑁, the replication of packet is a mechanism used to provide fault tolerant routing. The nodes in the network broadcast it to all their next hop when they receive a new broadcast packet. Therefore, a node could receive the packet from all of its parent nodes. When a node receives multiple copies of a packet, it only rebroadcasts the packet once. Anyone of parent nodes transmits the packet to the node successfully, and then the node receives the packet successfully, as (2). In 𝐺𝑈, every node tries to forward packets to the manager and every node has its paths to the manager. The hop-to-hop transmission is successful, as long as the packet is forwarded to any next hop successfully. When a node has a packet to transmit, it firstly forwards the packet to one of its parent nodes. If the packet is lost, the node will forward the packet to an alternate parent node. This

+ ⋅ ⋅ ⋅ + 𝐿 (𝑖, 𝑗𝑛 , 𝑖, 𝑗1 , 𝑖, 𝑗2 , . . . , 𝑖, 𝑗𝑛−1 ) 𝑅𝑗󸀠𝑛 ,

(2) (3)

where V𝑗1 , V𝑗2 , . . . , V𝑗𝑛 are selected as parent nodes. 𝑅𝑖 and 𝑅𝑖󸀠 are the expected reliability of the communication between the 𝑉BBR and a field device V𝑖 . 𝐿 𝑖,𝑗 is link reliability for single hop communication between V𝑖 and V𝑗 . 𝑃(𝑅𝑖,𝑗1 , 𝑅𝑖,𝑗2 , . . . , 𝑅𝑖,𝑗𝑛 ) is the probability that that none of packets are received by V𝑖 from the centralized manager via V𝑗1 , V𝑗2 , . . . , V𝑗𝑛 . In IWSNs, the centralized manager will determine the importance of the industrial massage for the desired reliability 𝑅𝑑 . In our approach, field devices have two attributes 󸀠 . These reliability attributes are used of reliability, 𝑅𝑁 and 𝑅𝑁 for constructing reliable graphs. When a new device V𝑖 joins network, the centralized manager chooses parent nodes for the device based on reliability attributes. In this paper, there are redundant BBRs as access points for reliability and load balancing. They are connected by 󸀠 , are 1. wire and the reliability attributes, 𝑅BBR and 𝑅BBR The new node can calculate the reliability attributes by its parent nodes’ reliability attributes, link reliability, and link correlation. Link reliability can be estimated by observing packet success and loss events as follows: 𝐿 𝑖,𝑗 =

1 ∑ ACKpac . Pac pac∈Pac

(4)

Pac is a set of packets forwarded from V𝑖 to V𝑗 , and pac is one packet of Pac. When packet pac is forwarded successfully, V𝑖 can receive ACKpac . Every node measures the reliability of links and the centralized manager maintains link reliability and calculates reliability attributes for routing. The link correlation also can be estimated by correlated packet reception [35]. 4.5. Process to Construct Graph. In our approaches, we divide nodes into two categories: in the graph and out of the graph. The nodes out of the graph are incrementally selected to construct graph. The new node has neighbor nodes which can communicate with each other and only neighbor nodes in the graph can be selected as parent nodes; then Figure 4 illustrates an example. The communication area may not be a circle because each node may have a different communication range. We choose eligible nodes as parent nodes, and then the node becomes one of the sets in the graph. When all nodes in the network are in the graph, the process of constructing graph is over. 4.6. Reliable Broadcast Graph. Reliable broadcast graph needs 𝑅𝑖 of the node V𝑖 to meet desired reliability for the broadcast message being propagated to the entire network.

6


i

BBR

Out of the graph In the graph

Figure 4: Process of constructing graph.

Initially, all nodes are out of the graph and not included in 𝑉𝐵 except for 𝑉BBR . The manager chooses parent nodes in the graph for the new node joining graph. In the broadcast graph 𝐺𝐵 , for all V𝑖 , V𝑖 is a new node joining the 𝐺𝐵 . If the selected parent node V𝑗 is 𝑉BBR , the broadcast path is 𝑉BBR → V𝑖 . If not, V𝑗 is a field device, and V𝑗 can also act as a children node and find its parent node. This process terminates until it finds that its parent node is 𝑉BBR . Then the broadcast path is 𝑉BBR → ⋅ ⋅ ⋅ → V𝑗 → V𝑖 . Therefore, when V𝑖 find its parent nodes from the set of nodes in the graph, the new node also has its broadcast paths. The number of selected parent nodes depends on the reliability attribute and the link correlation. We maintain 𝑉𝐵 , a set of nodes which have joined 𝐺𝐵 and 𝑉𝐵 is initialized as 𝑉BBR . The edges 𝐸𝐵 are also initialized to include all wired links among 𝑉BBR . The reliability attribute of 𝑉BBR is 1 because of wired connection. Then a node V𝑖 is incrementally selected from 𝑉 − 𝑉𝐵 . The node V𝑖 has one or more neighbor nodes included in 𝑉𝐵 . For a neighbor node V𝑗 as parent node, the reliability attribute is 𝑅𝑖,𝑗 , where 𝑅𝑖,𝑗 means the expected reliability that V𝑖 receives the broadcast message from V𝑗 . Sort its neighbor nodes from 𝑉𝐵 according to their reliability attributes 𝑅𝑖,𝑗 and select V𝑗1 with maximum reliability attribute. The rank 𝑟𝑖 is also determined by V𝑗1 . 𝑟 If 𝑅𝑖,𝑗1 is greater than 𝑅𝑑𝑖 , we only choose one parent node and 𝑅𝑖 equals 𝑅𝑖,𝑗1 (Line 17). If not, we will find the node V𝑗2 with maximum 𝑅𝑖 among its neighbor nodes correlated with V𝑗1 , where 𝑅𝑖 equals 1 − 𝑃(𝑅𝑖,𝑗1 , 𝑅𝑖,𝑗2 ). If 𝑅𝑖 is greater than 𝑟 𝑅𝑑𝑖 , we choose V𝑗1 and V𝑗2 as parent nodes of V𝑖 . If not, this 𝑟 process continues until 𝑅𝑖 is greater than 𝑅𝑑𝑖 . Otherwise all its neighbor nodes become its parent nodes (Line 16–22). At the end of this process, V𝑖 is a reliable node if 𝑅𝑖 is greater than 𝑟 𝑅𝑑𝑖 ; otherwise it is a unreliable node. For each node from the reliable node set 𝑉𝑖𝑛 , we sort their rank and choose the node with minimum rank and add it to 𝑉𝐵 (Line 26–28). If for all V𝑖 , V𝑖 ∈ 𝑉 − 𝑉𝐵 and V𝑖 ∉ 𝑉𝑖𝑛 , there is no reliable node available, and we sort their reliability attribute, choose the node with maximum 𝑅𝑖 , and

(1) Initially 𝑉𝐵 = 𝑉BBR and 𝐸𝐵 contains all wired links among 𝑉BBR . (2) 𝐺(𝑉, 𝐸) is the original graph, construct 𝐺𝐵 (𝑉𝐵 , 𝐸𝐵 ) (3) while 𝑉𝐵 ≠ 𝑉 do (4) Find 𝑆, 𝑆 ⊆ 𝑉𝐵 , ∀V𝑗 ∈ 𝑆, V𝑗 is a neighbor node of V𝑖 , and Pt is the parent set of V𝑖 (5) if 𝑆 = 𝜙, the neighbor set 𝑆 of V𝑖 , ∀V𝑖 ∈ 𝑉 − 𝑉𝐵 then (6) The network topology is disconnected (7) return FAIL; (8) else (9) for all node V𝑖 ∈ 𝑉 − 𝑉𝐵 do (10) for all node V𝑗 ∈ 𝑆 do (11) Sort the reliability 𝑅𝑖 = 1 − 𝑃 (𝑅𝑖,𝑗 , 𝑃𝑡) (12) if Pt = 𝜙 then (13) 𝑟𝑖 = 𝑟𝑗 + 1 (14) end if (15) Choose the node V𝑗 with max 𝑅𝑖 as a parent node, V𝑗 ∈ Pt 𝑟 (16) if 𝑅𝑖 ≥ 𝑅𝑑𝑖 then (17) V𝑖 ∈ 𝑉𝑖𝑛 break; (18) else if 𝑆 − V𝑗 = 𝜙 then (19) V𝑖 ∈ 𝑉𝑒𝑛 break; (20) else (21) 𝑆 = 𝑆 − V𝑗 (22) end if (23) end for (24) end for (25) end if (26) if 𝑉𝑖𝑛 ≠ 𝜙 then (27) Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑖𝑛 , with min 𝑟𝑖 (28) Add V𝑖 to 𝑉𝐵 , add 𝑒𝑗,𝑖 to 𝐸𝐵 (29) else (30) Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑒𝑛 , with max 𝑅𝑖 (31) Add V𝑖 to 𝑉𝐵 , add 𝑒𝑗,𝑖 to 𝐸𝐵 (32) end if (33) end while (34) return SUCCESS; Algorithm 1: Constructing reliable broadcast graph.

add it to 𝑉𝐵 (Line 29–32). This process can be repeated and terminates when all nodes in 𝑉 are contained in 𝑉𝐵 or an error is reported, shown in Algorithm 1. The worst-case complexity of the algorithm is |𝑉|

∑ (|𝐸| + (|𝑉| − 𝑘) ∗ 𝑘) = 𝑂 (|𝑉|3 ) . 𝑘=|𝑉BBR |

(5)

4.7. Reliable Uplink Graph. The algorithm to construct reliable uplink graph 𝐺𝑈(𝑉𝑈, 𝐸𝑈) is similar to 𝐺𝐵 (𝑉𝐵 , 𝐸𝐵 ), and all nodes in IWSNs share the same uplink graph. 𝑅𝑖󸀠 is different from 𝑅𝑖 because the fault tolerant mechanism of 𝐺𝑈 is different from 𝐺𝐵 . Therefore, the parent nodes of a node in 𝐺𝑈 may not be the same as 𝐺𝐵 . The construction of 𝐺𝑈(𝑉𝑈, 𝐸𝑈) is summarized by Algorithm 2.


7

Initially 𝑉𝑈 = 𝑉BBR and 𝐸𝑈 contains all wired links among 𝑉BBR . (2) 𝐺(𝑉, 𝐸) is the original graph Construct 𝐺𝑈 (𝑉𝑈 , 𝐸𝑈 ) by applying Algorithm 1, where 𝑅𝑖󸀠 = 𝑅𝑖 . (4) while 𝑉𝐵 ≠ 𝑉 do Find 𝑆, 𝑆 ⊆ 𝑉𝑈 , ∀V𝑗 ∈ 𝑆, V𝑗 is a neighbor node of V𝑖 , and Pt is the parent set of V𝑖 (6) if 𝑆 = 𝜙, the parent set 𝑆 of V𝑖 , ∀V𝑖 ∈ 𝑉 − 𝑉𝑈 then The network topology is disconnected (8) return FAIL; else (10) for all node V𝑖 ∈ 𝑉 − 𝑉𝑈 do for all node V𝑗 ∈ 𝑆 do (12) Sort the reliability 𝑅𝑖󸀠 Choose the node V𝑗 with max 𝑅𝑖󸀠 as a parent node, V𝑗 ∈ Pt 𝑟 (14) if 𝑅𝑖󸀠 ≥ 𝑅𝑑𝑖 then V𝑖 ∈ 𝑉𝑖𝑛 break; (16) else if 𝑆 − V𝑗 = 𝜙 then V𝑖 ∈ V𝑒𝑛 break (18) else 𝑆 = 𝑃 − V𝑗 (20) end if end for (22) end for end if (24) If 𝑉𝑖𝑛 ≠ 𝜙 then Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑖𝑛 , with min 𝑟𝑖 (26) Add V𝑖 to 𝑉𝑈 , add 𝑒𝑖,𝑗 to 𝐸𝑈 else (28) Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑒𝑛 , with max 𝑅𝑖󸀠 Add V𝑖 to 𝑉𝑈 , add 𝑒𝑖,𝑗 to 𝐸𝑈 (30) end if end while (32) return SUCCESS; Algorithm 2: Constructing reliable uplink graph.

4.8. Reliable Downlink Graph. The downlink graph 𝐺𝑁(𝑉𝑁, 𝐸𝑁) is different from 𝐺𝐵 (𝑉𝐵 , 𝐸𝐵 ) and 𝐺𝑈(𝑉𝑈, 𝐸𝑈) because nodes in IWSNs have their own downlink graph and these graphs do not involve all nodes of 𝐺(𝑉, 𝐸). We should gradually construct reliable downlink graph 𝐺𝑖 (𝑉𝑖 , 𝐸𝑖 ) for each node V𝑖 . If 𝑅𝑖 satisfies the reliability criterion in 𝐺𝑖 (𝑉𝑖 , 𝐸𝑖 ), the node V𝑖 is a reliable node for downlink. We maintain 𝐷, a set of nodes which have their reliable downlink graphs. Initially, 𝐷 only contains 𝑉BBR . To construct 𝐺𝑖 , V𝑖 should find reliable parent nodes from 𝐷. The manager prefers to add the reliable nodes to 𝐷. If the reliable nodes exist, this algorithm will construct downlink graph for the node with minimum 𝑟𝑖 . If not, the node with maximum 𝑅𝑖 will be selected. If the parent node V𝑗 is 𝑉BBR , the downlink path is 𝑉BBR → V𝑖 . If not, the downlink path can be 𝑉BBR → ⋅ ⋅ ⋅ → V𝑗 in 𝐺𝑗 firstly, and then V𝑗 → V𝑖 . Therefore, 𝐺𝑖 should contain the downlink graph of its parent nodes. As shown in Algorithm 3, a node V𝑖 is incrementally selected from 𝑉 − 𝐷. We select parent nodes by calculating the

reliability 𝑅𝑖 and add the optimal node to 𝐷. Then 𝐺𝑖 contains 𝐺𝑗1 , 𝐺𝑗2 , . . . , 𝑒𝑗1 ,𝑖 , 𝑒𝑗2 ,𝑖 , . . ., and V𝑖 . 4.9. The Percentage of Reliable Nodes and the Link Quality. In harsh and noisy industrial environment, the link quality is the major barrier to maintain reliable nodes. The bad link quality will make the network unreliable. We evaluate the relationship between the percentage of reliable nodes and the link quality. We conducted some experiments to evaluate the percentage of reliable nodes by changing the link success probability. In Figure 5, we observe that with 50, 100, and 150 nodes in the experiments, when the link success probability is above 0.7, most nodes are reliable nodes by multiparent nodes. The percentage of reliable nodes drops quickly along with the decreasing link quality below 0.7. Figure 5(a) shows that there are few reliable nodes in the network when the link success probability drops to 0.4 but the value is nearly 0.6 in Figure 5(b) because the fault tolerant mechanism of

8


Initially 𝐷 = {𝑉BBR }, D is a set of nodes with 𝐺𝑁 . while 𝐷 ≠ 𝑉 do (3) Find S, 𝑆 ⊆ 𝐷, ∀V𝑗 ∈ 𝑆, V𝑗 is a neighbor node of V𝑖 , and Pt is the parent set of V𝑖 if 𝑆 = 𝜙, the neighbor set S of V𝑖 , ∀V𝑖 ∈ 𝑉 − 𝐷 then The network topology is disconnected (6) return FAIL; else for all node V𝑖 ∈ 𝑉 − 𝐷 do (9) Choose V𝑗 and find 𝑉𝑖𝑛 by Lines (10–25) of Algorithm 1 end for end if (12) if 𝑉𝑖𝑛 ≠ 𝜙 then Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑖𝑛 , with min 𝑟𝑖 Add V𝑖 to S, 𝐺𝑖 = V𝑖 ∪ 𝑒𝑗1 ,𝑖 ∪ 𝐺𝑗1 ∪ 𝑒𝑗2 ,𝑖 ∪ 𝐺𝑗2 ∪ ⋅ ⋅ ⋅ (15) else Choose the node V𝑖 , V𝑖 ∈ 𝑉𝑒𝑛 , with max 𝑅𝑖 Add V𝑖 to S, 𝐺𝑖 = V𝑖 ∪ 𝑒𝑗1 ,𝑖 ∪ 𝐺𝑗1 ∪ 𝑒𝑗2 ,𝑖 ∪ 𝐺𝑗2 ∪ ⋅ ⋅ ⋅ (18) end if end while return SUCCESS;

1

1

0.9

0.9

0.8

0.8

0.7

0.7

Reliable nodes (%)

Reliable nodes (%)

Algorithm 3: Constructing reliable downlink graph.

0.6 0.5 0.4 0.3

0.6 0.5 0.4 0.3

0.2

0.2

0.1

0.1

0

0.4

0.5 0.6 0.7 Link success probability

50 nodes 100 nodes

0.8

0.9

150 nodes

(a) Reliable nodes of broadcast graph

0 0.55

0.6

0.65


50 nodes 100 nodes

0.85

0.9

150 nodes

(b) Reliable nodes of uplink graph

Figure 5: Percentage of reliable nodes and link quality.

𝐺𝑈 is different from 𝐺𝐵 . The performance of fault tolerant mechanism is compared in Section 5. Based on these results, we conclude that it is difficult to achieve reliable nodes with bad link quality. 4.10. Loop Avoidance. In IWSNs, the node mobility and failure would incur topology change and routing failure. The graph routing with our algorithm would generate routing failure, take time to converge, and introduce much overhead. However, our approaches will not cause the loop problem. The nodes in the network are added to the graph in order.

We can sort these nodes based on the time when they join the graph. In 𝐺𝐵 and 𝐺𝑁, a packet is always transmitted from an older node to a newer node while in 𝐺𝑈 it is from a newer node to an older node. The packets move following the order of the nodes to avoid a loop.

5. Performance Evaluation This section shows some simulation results illustrating features of reliable graph routing. The reliable performance of graph routing is largely independent of MAC layer protocols.

9

Number of links

International Journal of Distributed Sensor Networks 400 350 300 250 200 150 100 50 0 0.3

0.4


0.8

0.9

150 nodes

50 nodes 100 nodes

Figure 6: The topology of simulation.

The simulation topology consists of 100 nodes. The link success probability of all links is the mean percentage of successful links assuming link independence. The link success probability is varied from 0.0 to 0.9. In our simulations, there are three BBRs interconnecting with the centralized manager and the topology is shown in Figure 6.

Number of links

(a) Number of links in broadcast graph

400 350 300 250 200 150 100 50 0 0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

Link success probability 50 nodes 100 nodes

150 nodes

(b) Number of links in uplink graph

5.1. Simulation Parameters. Our simulation for graph routing performance is based on 100 nodes. We evaluate the impact for graph routing performance when the size of the network varies. Figure 7 shows the number of links along with the size of the network. In Figure 7(a), the number of links decreases when the link success probability is 0.5, while it is 0.65 in Figure 7(b). The curves for three kinds of the network have similar trends. The other important parameter is 𝑅𝑑 . 𝑅𝑑 as a reliability criterion depends on the reliability requirement of industrial application. The graph routing performance is affected by 𝑅𝑑 , as shown in Figures 8 and 9. We vary 𝑅𝑑 from 0.86 to 0.98 for high reliability requirements. When the link success probability is low, there is no difference in the performance with different 𝑅𝑑 because our approaches select all parent nodes in routing graph. The difference appears when 𝑅𝑑 is greater than 0.4 in Figure 8 and 0.55 in Figure 9. The result shows that higher 𝑅𝑑 introduces more links and gets more successful communication. The curves with different 𝑅𝑑 have similar trends, so we choose 0.9 as our simulation parameter. 5.2. Performance of Graph Routing. In the experiment, we conducted a series of experiments to evaluate the performance of routing reliability, number of links, and expected number of transmissions in different routing graphs. Then the link success probability in the network is gradually increasing from 0% to 90%. The fault tolerant mechanism of 𝐺𝐵 and 𝐺𝑁 is the same so that some experiment results of 𝐺𝑁 are similar to 𝐺𝐵 in Figures 10 and 12. We compare our approaches for the reliable graphs with three other approaches. The first approach constructs a single tree using breadth-first search, the second approach chooses two parents to construct graphs [19], and the third approach

Figure 7: Number of links with node number.

contains all parents in graphs for max reliable. When a node is added to the graph in these approaches, one, two, or all parent nodes from the current graph are chosen by the average hop between the new node and BBRs. In our approach, reliability attributes are used to support a balance between the reliability and overhead. In our experiment, the number of links is measured by the number of links that the periodic messages used to maintain these links available. More parents need to exchange more information to maintain these links. The expected number of transmissions is defined as the expected number of links for the communication between every nodes and BBRs. In 𝐺𝐵 , the communication means the centralized manager broadcasts a message to the entire network. In 𝐺𝑈 and 𝐺𝑁, the communication means each node sends a packet to the manager and receives a specific packet from the manager. The first experiment evaluates the performance of reliable broadcast graph. Figure 10 summarizes all results of 𝐺𝐵 . Figure 10(a) shows that the percentage of successful communication varies along with the increasing link success probability. When the link success probability is less than 0.5, our approach performs best as max parent graph. Because the link reliability is low, our approach also includes all parents in graph to improve transmission success rate. But more than 0.5, the performance of our approach is different from max parent graph. In spite of the increasing link success probability, the reachable nodes also drop after 0.7 because of the less links. Figure 10(b) shows that the number of links of our approach decreases along with the increasing link success

10

International Journal of Distributed Sensor Networks 1 240

0.9

220 0.8

Number of links

Successful communication (%)

260

0.7 0.6

200 180 160 140 120

0.5

100 0.4 0.4

0.5

0.6

0.8

0.7

80 0.3

0.9

0.4

0.5

Link success probability Rd = 0.94 Rd = 0.98

Rd = 0.86 Rd = 0.9

0.6

0.7

0.8

0.9


Rd = 0.86 Rd = 0.9

(a) Reliability of broadcast graph with different 𝑅𝑑

(b) Number of links in broadcast graph with different 𝑅𝑑

Figure 8: Performance of Broadcast Graph with Different 𝑅𝑑 .

260 240

0.9

220 0.8 Number of links


1

0.7 0.6

200 180 160 140 120

0.5

100 0.4 0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

80 0.5

0.55

0.6


Rd = 0.94 Rd = 0.98

(a) Reliability of uplink graph with different 𝑅𝑑

0.65 0.7 0.75 0.8 Link success probability

Rd = 0.86 Rd = 0.9

0.85

0.9

Rd = 0.94 Rd = 0.98

(b) Number of links in uplink graph with different 𝑅𝑑

Figure 9: Performance of Uplink Graph with Different 𝑅𝑑 .

probability when the link success probability is more than 0.5. The number of links of other three approaches is constant, and the max parent graph is much higher than the other three approaches. When the link reliability can satisfy the needs of communication reliability, our approach need not maintain many links for communication. The decline of overhead also leads to the decline of reliability. Figure 10(c) shows the expected number of transmission when the manager broadcasts a message to the entire network. The higher percentage of reliable links makes more nodes reachable from the manager, and then these nodes also broadcast the message

by introducing expected number of transmissions. If the percentage of reliable links is 0, any node cannot receive the broadcast message and introduce the expected number of transmissions. And the graph maintaining more links also introduces the expected number of transmissions because every node has more neighbor nodes to communicate with. We observe that the expected number of transmissions increases along with the increasing link success probability in other three approaches due to their constant links. In our approach, the expected number of transmissions increases when the link success probability is less than 0.5. However,


11

1 250

0.8 200

0.7 Number of links


0.9

0.6 0.5 0.4

150

100

0.3 0.2

50

0.1 0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0

0.9

0

0.1

0.2

0.3

Link success probability Max parent graph Reliable graph

One-parent graph Two-parent graph

0.4

0.5

0.6

0.7

0.8

0.9

Link success probability One-parent graph Two-parent graph

(a) Reliability of broadcast graph

Max parent graph Reliable graph

(b) Number of links in broadcast graph

Expected number of transmissions

250

200

150

100

50

0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9



(c) Expected number of transmissions

Figure 10: Performance of broadcast graph.

when the link success probability is more than 0.5, the expected number of transmissions did not vary like max parent graph because of the decreased links. The few links need to be included in our graph along with the increasing link reliability. Therefore, the result is affected by the two factors jointly as Figure 10(a). So the curve of the expected number of transmissions increases slowly around 0.5 to 0.7, and then drops quickly after 0.7. In the second experiment, the performance of 𝐺𝑈 is similar to 𝐺𝐵 , as shown in Figure 11. In other three approaches, the successful communication and the expected number of transmissions all increase along with the increasing link success probability and the number of links is constant. In Figure 11(a), the successful communication of our approach does vary as max parent graph when link success probability

is more than 0.7. That is because the fault tolerant mechanism of 𝐺𝑈 is different from 𝐺𝐵 so that the reliability attributes are different. When the node V𝑖 has the same parent nodes in different graphs, 𝑅𝑖󸀠 is lower than 𝑅𝑖 with the same number of links. Therefore, the nodes in 𝐺𝑈 which meet the reliability requirement need to maintain more links than 𝐺𝐵 , as shown in Figure 11(b). Figure 11(c) shows the expected number of transmissions when every node in the network forwards a packet to the manager. So the expected number of transmissions is greater than 𝐺𝐵 . The third experiment shows the performance of reliable downlink graph. As 𝐺𝑁 and 𝐺𝐵 have the same reliability attributes and fault tolerant mechanism, the performance of reliability and the number of links are the same, as shown in Figures 12(a) and 12(b). Figure 12(c) shows the expected

12

International Journal of Distributed Sensor Networks 1 250

0.8 200

0.7 Number of links


0.9

0.6 0.5 0.4 0.3 0.2

150

100

50

0.1 0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0

0.9

0

0.1

0.2

0.3



0.4

0.5

0.6

0.7

0.8

0.9


(a) Reliability of uplink graph


(b) Number of links in uplink graph

550 Expected number of transmissions

500 450 400 350 300 250 200 150 100 50 0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9




Figure 11: Performance of uplink graph.

number of transmissions when the manager sends a special packet to every node in network. Every packet should be broadcast in the 𝐺𝑁 of its destination, so 100 packets propagated in the network will introduce a large sum of transmission. The communication of 𝐺𝑁 and 𝐺𝑈 is between the manager and every node in the network, but the fault tolerant mechanism of 𝐺𝑈 introduces less expected number of transmissions.

6. Conclusion In this paper, we present the mechanism to support reliable graph routing in IWSNs. We summarize three communication types for industrial application and construct reliable

graphs for different communication purposes. Our approach constructs the routing graph based on the link reliability and link correlation. When the link success probability is low, our approaches provide reliable communication with compromising network overhead. When the communication links are reliable, we pursue not the optimal reliability but less overhead. Indeed, the performance and overhead also rely on the network topology and the node density. Therefore, we should further improve and detect our approach. In future work, our system will be deployed in real largescale manufacturing factory and will face more challenges and problems in industrial environments. We will do the experiment about the packet reception correlation among neighbor nodes. Then we should look for more efficient


13

1 250

0.8 200

0.7 Number of links


0.9

0.6 0.5 0.4 0.3 0.2

150

100

50

0.1 0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0

0.9

0

0.1

0.2

0.3



0.4

0.5

0.6

0.7

0.8

0.9



(a) Reliability of downlink graph

(b) Number of links in downlink graph

Expected number of transmissions

4000 3500 3000 2500 2000 1500 1000 500 0

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9




Figure 12: Performance of downlink graph.

algorithms to construct reliable graph and improve other network performances. As a future extension of this work we will focus on the data link layer communication schedule based on routing graphs for multihop transmission. We want to have a cross-layer approach where the network and MAC layer will work together to achieve end-to-end reliable and real-time communication.

Acknowledgments This work was supported by the National Natural Science Foundation of China (Grant nos. 61201204 and 61271201), the Fundamental Research Funds for the Central Universities (Grant no. 2013YJS014), and the National Major Projects of China (Grant no. 2012ZX03005003).

References [1] S. Karnouskos and T. N. De Holanda, “Simulation of a smart grid city with software agents,” in Proceedings of the UKSim 3rd European Modelling Symposium on Computer Modelling and Simulation (EMS ’09), pp. 424–429, IEEE, November 2009. [2] F. Gil-Castineira, E. Costa-Montenegro, F. Gonzalez-Castano, C. Lpez-Bravo, T. Ojala, and R. Bose, “Experiences inside the ubiquitous Oulu smart city,” Computer, vol. 44, no. 6, Article ID 5765905, pp. 48–55, 2011. [3] H. N. Li, G. B. Song, J. T. Kim, and T. H. Yi, “Sensing methodologies and sensor networks for health monitoring of civil infrastructures,” International Journal of Distributed Sensor Networks, vol. 2012, Article ID 358286, 2 pages, 2012. [4] M. Al-Hader, A. Rodzi, A. R. Sharif, and N. Ahmad, “SOA of smart city geospatial management,” in Proceedings of the UKSim

14

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14] [15]

[16]

[17]

[18]

[19]

[20]

International Journal of Distributed Sensor Networks 3rd European Modelling Symposium on Computer Modelling and Simulation (EMS ’09), pp. 6–10, IEEE, November 2009. Y. Yang and B. Honary, “Smart sensor networks: theory and practice,” International Journal of Distributed Sensor Networks, vol. 2012, Article ID 602893, 2 pages, 2012. S. Commuri and M. K. Watfa, “Coverage strategies in wireless sensor networks,” International Journal of Distributed Sensor Networks, vol. 2, no. 4, pp. 333–353, 2006. T. H. Kim, W. C. Fang, C. Ramos, S. Mohammed, O. Gervasi, and A. Stoica, “Ubiquitous sensor networks and its application,” International Journal of Distributed Sensor Networks, vol. 2012, Article ID 310271, 3 pages, 2012. A. Willig, “Recent and emerging topics in wireless industrial communications: a selection,” IEEE Transactions on Industrial Informatics, vol. 4, no. 2, pp. 102–124, 2008. D. Chen, M. Nixon, and A. Mok, Wirelesshart: Real-Time Mesh Network for Industrial Automation, Springer, New York, NY, USA, 2010. V. C. Gungor and G. P. Hancke, “Industrial wireless sensor networks: challenges, design principles, and technical approaches,” IEEE Transactions on Industrial Electronics, vol. 56, no. 10, pp. 4258–4265, 2009. K. Koumpis, L. Hanna, M. Andersson, and M. Johansson, “A review and roadmap of wireless industrial control,” in The Industrial Wireless Book Email Newsletter, vol. 4 of Available lastactualization, p. 2009, GGH Marketing Communications, 2006. J. Song, S. Han, A. K. Mok et al., “WirelessHART: applying wireless technology in real-time industrial process control,” in Proceedings of the 14th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS ’08), pp. 377– 386, IEEE, April 2008. A. Willig, K. Matheus, and A. Wolisz, “Wireless technology in industrial networks,” Proceedings of the IEEE, vol. 93, no. 6, pp. 1130–1151, 2005. “Wirelesshart,” http://www.hartcomm.org. “Isa-100. 11a-2009 wireless systems for industrial automation: process control and related applications,” http://www.isa.org/ ISA100-11a. J. Zhao, D. Yang, Y. Qin, T. Zheng, J. Duan, and M. Gidlund, “Issues of routing protocol for wireless industrial sensor networks,” in Proceedings of the 38th Annual Conference on IEEE Industrial Electronics Society (IECON ’12), pp. 3225–3230, IEEE, 2012. D. Yang, M. Gidlund, W. Shen, Y. Xu, T. Zhang, and H. Zhang, “CCA-embedded tdma enabling acyclic traffic in industrial wireless sensor networks,” Ad Hoc Networks, vol. 11, no. 3, pp. 1105–1121, 2013. T. Lennvall, S. Svensson, and F. Hekland, “A comparison of wirelesshart and zigbee for industrial applications,” in Proceedings of the 7th IEEE International Workshop on Factory Communication Systems (WFCS ’08), pp. 85–88, IEEE, May 2008. S. Han, X. Zhu, A. K. Mok, D. Chen, and M. Nixon, “Reliable and real-time communication in industrial wireless mesh networks,” in Proceedings of the 17th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS ’11), pp. 3–12, IEEE, April 2011. J. Zhao, Z. Liang, and Y. Zhao, “ELHFR: a graph routing in industrial wireless mesh network,” in Proceedings of the IEEE International Conference on Information and Automation (ICIA ’09), pp. 106–110, IEEE, June 2009.

[21] G. Gao, H. Zhang, and L. Li, “A reliable multipath routing strategy for wirelesshart mesh networks using subgraph routing?” Journal of Computational Information Systems, vol. 9, no. 5, pp. 2001–2008, 2013. [22] S. Banerjee and A. Misra, “Minimum energy paths for reliable communication in multi-hop wireless networks,” in Proceedings of the 3rd ACM International Symposium on Mobile Ad Hoc Networking and Computing, pp. 146–156, ACM, June 2002. [23] J. Heo, J. Hong, and Y. Cho, “EARQ: energy aware routing for real-time and reliable communication in wireless industrial sensor networks,” IEEE Transactions on Industrial Informatics, vol. 5, no. 1, pp. 3–11, 2009. [24] M. Radi, B. Dezfouli, K. A. Bakar, and M. Lee, “Multipath routing in wireless sensor networks: survey and research challenges,” Sensors, vol. 12, no. 1, pp. 650–685, 2012. [25] S.-J. Lee and M. Gerla, “Split multipath routing with maximally disjoint paths in Ad hoc networks,” in Proceedings of the IEEE International Conference on Communications (ICC ’01), vol. 10, pp. 3201–3205, IEEE, June 2000. [26] M. K. Marina and S. R. Das, “On-demand multipath distance vector routing in ad hoc networks,” in Proceedings of the International Conference on Network Protocols (ICNP ’01), pp. 14–23, IEEE, November 2001. [27] D. Ganesan, R. Govindan, S. Shenker, and D. Estrin, “Highlyresilient, energy-efficient multipath routing in wireless sensor networks,” ACM SIGMOBILE Mobile Computing and Communications Review, vol. 5, no. 4, pp. 11–25, 2001. [28] A. Nasipuri and S. R. Das, “On-demand multipath routing for mobile ad hoc networks,” in Proceedings of the IEEE 8th International Conference on Computer Communications and Networks, pp. 64–70, IEEE, 1999. [29] A. Nasipuri, R. Casta˜neda, and S. R. Das, “Performance of multipath routing for on-demand protocols in mobile Ad Hoc networks,” Mobile Networks and Applications, vol. 6, no. 4, pp. 339–349, 2001. [30] Y. Ishii, “Exploiting backbone routing redundancy in industrial wireless systems,” IEEE Transactions on Industrial Electronics, vol. 56, no. 10, pp. 4288–4295, 2009. [31] Ieee 802. 15. 4 wpan task group, http://www.ieee802.org/15/ pub/TG4.html. [32] J. Gomez, A. T. Campbell, M. Naghshineh, and C. Bisdikian, “PARO: supporting dynamic power controlled routing in wireless ad hoc networks,” Wireless Networks, vol. 9, no. 5, pp. 443– 460, 2003. [33] S. Doshi, S. Bhandare, and T. X. Brown, “An on-demand minimum energy routing protocol for a wireless ad hoc network,” ACM SIGMOBILE Mobile Computing and Communications Review, vol. 6, no. 3, pp. 50–66, 2002. [34] A. Basalamah, S. M. Kim, S. Guo, T. He, and Y. Tobe, “Link correlation aware opportunistic routing,” in Proceedings of the 31st Annual IEEE International Conference on Computer Communications (INFOCOM ’12), pp. 3036–3040, IEEE, 2012. [35] T. Zhu, Z. Zhong, T. He, and Z. L. Zhang, “Exploring link correlation for efficient flooding in wireless sensor networks,” in Proceedings of the 7th USENIX Symposium on Networked Systems Design and Implementation (NSDI ’10), pp. 49–64. [36] K. Srinivasan, M. Jain, J. Choi et al., “The 𝜅 factor: inferring protocol performance using inter-link reception correlation,” in Proceedings of the 16th Annual Conference on Mobile Computing and Networking (MobiCom ’10), pp. 317–328, ACM, September 2010.

International Journal of Distributed Sensor Networks [37] H. Alwan and A. Agarwal, “A survey on fault tolerant routing techniques in wireless sensor networks,” in Proceedings of the 3rd International Conference on Sensor Technologies and Applications (SENSORCOMM ’09), pp. 366–371, IEEE, June 2009.

15

International Journal of

Rotating Machinery

Engineering Journal of

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

The Scientific World Journal Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014


Distributed Sensor Networks

Journal of

Sensors Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014


Volume 2014


Volume 2014

Journal of

Control Science and Engineering

Advances in

Civil Engineering Hindawi Publishing Corporation http://www.hindawi.com


Volume 2014

Volume 2014

Submit your manuscripts at http://www.hindawi.com Journal of

Journal of

Electrical and Computer Engineering

Robotics Hindawi Publishing Corporation http://www.hindawi.com


Volume 2014

Volume 2014

VLSI Design Advances in OptoElectronics


Navigation and Observation Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014



Chemical Engineering Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Volume 2014

Active and Passive Electronic Components

Antennas and Propagation Hindawi Publishing Corporation http://www.hindawi.com

Aerospace Engineering


Volume 2014


Volume 2014

Volume 2014




Modelling & Simulation in Engineering

Volume 2014


Volume 2014

Shock and Vibration Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Advances in

Acoustics and Vibration Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014