Power-Aware and Void-Avoidant Routing Protocol for Reliable ...

3 downloads 146 Views 227KB Size Report
Amirkabir University of Technology ... protocol which uses geographic algorithm with a optimizing ... the reasons for using of a wireless technology in industrial.
Power-Aware and Void-Avoidant Routing Protocol for Reliable Industrial Wireless Sensor Networks Mohammad Reza Soltani Department of Electrical Engineering Amirkabir University of Technology Tehran, Iran {aut_mohammad}@aut.ac.ir

Abstract – Industrial Wireless Sensor Networks (IWSN) are wireless sensor networks which have been exerted to industrial environments such as factories, refineries, power plants and etc for measuring, controlling and monitoring applications. One of the most important issues in these wireless networks is how the deployed nodes can communicate to each other so that appropriate requirements for these situations like optimized energy consumption of nodes, reliable and real time delivery of packets to be satisfied. In this paper we focus on optimizing energy consumption of nodes and reliable delivery of generated packets by source nodes in industrial applications. Thus, for achieving to these desired purposes, we introduce a novel routing protocol which uses geographic algorithm with a optimizing function for selecting the next candidate node and handling encountered voids along routing data packets towards sink where control room is located at there and has the responsibility of controlling and monitoring of data has reached. Keywords – Geographic Routing; Reliability; Energy Consumption; Void Avoidance; Industrial Wireless Sensor Networks(IWSN)

I. INTRODUCTION Industrial Wireless Sensor Networks (IWSN) are wireless sensor networks which have been exerted to industrial environments such as factories, refineries, power plants and etc for measuring, controlling and monitoring applications. One of the reasons for using of a wireless technology in industrial environments refers to the easier installing wireless equipments in these places than wired conventional technologies or even new wired methods such as fieldbus technology [1]. The other fact refers to the length diminution of the cables for making communication between different devices that give better performance with lower expenses in an industrial plant. On the other hand, industrial environments have harsh and adverse situations and its reason allude to the existence of high amount of noise which can be a detrimental factor for routing generated data packets by source nodes that is caused to decline the performance of routing protocol. So representing of routing algorithm for decreasing these harmful effects is crucial and necessary. In addition, it is very important to notice that the wireless nodes in wireless sensor networks have constraint resources such as a power supply and memory available. And as it is known the most important point for power consumption in such wireless nodes, is the communication between two nodes for transmitting of their information to each other.

Seyed Ahmad Motamedi, Samad Ahmadi, Mohsen Maadani Department of Electrical Engineering Amirkabir University of Technology Tehran, Iran {motamedi, samad.ahmadi, maadani}@aut.ac.ir Another issue in IWSN is delivery of data packets to sink in real time with maximum reliability as possible as. As a result, the proposing routing protocol for IWSN should satisfy three major goals as following: optimized power consumption of nodes, delivering of data packets with minimum delay to destination and finally considering reliability in delivering data packets to sink with minimum errors. In this paper we focus on representing a routing protocol for IWSN and industrial applications. Especially about optimizing power consumption and reliable delivery of generated packets. We offer a geographic routing algorithm which selects the next candidate node base on maximizing of a defined function [2]. Since we have incorporated factors of residual energy in wireless nodes and progressive distance towards sink to defined cost function, we expect to achieve similar average residual energy per node to greedy protocols but less standard deviation which means increasing lifetime of network and load balancing among different nodes. In many geographical routing protocols, we confront with holes especially in greedy methods [3] where the protocol fails to progress to destination because the lack of any neighbor nodes which are closer than the forwarding node. So, for achieving our aim, other than above discussed method for selecting the next eligible node, we have used other one algorithm similar to GPSR [4] to facilitate these probably voids along routing, and so, heightens the reliability of protocol. The remainder of this paper is organized as follows: Section II reviews the related works about routing protocols. Section III describes the proposed protocol. Section IV presents the simulation results. The paper is concluded in Section V. II. RELATED WORKS Many researches have been done about efficient routing protocol in wireless sensor networks [5]. Totally we can categorize these protocols into two groups: first group involves reactive protocols which act based on discovery route method so they can determine the path toward destination [6]. In other words, protocols which are included in this group do the routing with transmitted requests by their source nodes which demand a new route to the destination. One of these protocols which its target is optimizing energy usage at a node is called

978-1-4244-6252-0/11/$26.00 ©2011 IEEE

EAR [7]. The other routing protocol that is presented for the IWSN is EARQ [8]. In this protocol the next node is selected based on three parameters. Real time, reliability and energy efficiency of the path between two nodes that they want to communicate to each other. In fact the process of determining above mention parameters is originated from sink to source direction in a similar manner to EAR [7]. The second group consists of proactive protocols which make a route from source to destination that are not based on demands of source nodes so, because the lack of discovery route, they have the amount of delay less than the reactive ones. The reference of [9] describes one of these kinds of protocol. GPSR [4] is another of proactive protocol that uses greedy routing for delivering packets to destination. In this protocol, it is assumed that all the nodes know their geographic positions and it will be used Euclidean distance to the sink for determining the next candidate for receiving packets. So, as a result the node which is nearer to destination than the others is selected. Also, GPSR exerts perimeter mode for overcoming the holes which may exist along the path from source to destination. The protocol is illustrated in reference [2] actually acts analogous to GPSR except that in this method, extra factors such as reliability of nodes, energy harvesting rate from environment and residual energy of nodes are considered for choosing the next node in routing process. However, the issue of the hole is not noticed. The other protocol that it has been demonstrated in reference [10] has combined routing algorithm with other network layers concept in a cross-layer manner. For example handling of congestion issue from transport layer or considering of the quantity of SNR for receiving correctly a packet from the viewpoint of physical layer are the major points in this protocol. III. PROPOSED ROUTING PROTOCOL As mentioned in last chapter, the quantity of end to end delay for delivering data packets to destination is very critical issue for our industrial application.. So we should focus on proactive protocols for our routing scheme. In this section first of all the assumption of this protocol is explained. These assumptions are as following: •

All nodes know their geographical position by getting help from GPS or with every localization and positioning method. Because of our application is industrial and nodes are laid at special places for sensing or actuating an event, therefore this assumption is logical.



Since the location of sink in industrial plants is usually the control room where monitoring and processing received information packets is done, so the place of sink is known for source nodes.



The nodes are fixed and do not have any motion.

Now we are ready to describe our method. This protocol is comprised of two phase: first phase is broadcasting of hello massages and the second one is delivering data phase that are elaborated below respectively:

A. Broadcasting Of Hello Massages Phase In this phase every node broadcasts hello massages periodically to all of its neighbors every T seconds. These packets contain different fields. The first field is geographical information of a node and the second one is residual energy of that. Also different nodes broadcast their hello massages jittered in time because of decreasing the collision of hello massages reached to a node. B. Data Delivery Phase In this phase all of nodes know about geographic position and residual energy of their neighbor nodes. So in first step when the source node wants to transmit its data packets, it examines which of its neighbors is closer to sink than itself. If a neighbor node is located in a closer distance to destination, then source node calculate a defined optimizing linear function for that node. In this function has been incorporated the factors of residual energy and progressive distance to destination and with based of maximized value of function for each nodes, can be selected a node as an eligible next node for receiving the data packets of the source node. This process of choosing the next node is continued based on updating of nodes’ energy in a periodic time intervals until data packets are reached by intermediate nodes to destination. This Linear Optimizing Function (L.O.F) is clarified below: L.O.F = α . Norm(PD) + (1−α) . Norm(Eres).

(1)

Where 0 < α < 1 is a weighted factor and in section IV, we have displayed the results for different values of α. Norm(PD) is normalized Progressive Distance and is defined: Norm(PD) = (dmy-s − dnei-s)/Max(dmy-s − dnei-s).

(2)

Where dmy-s and dnei-s are distances of the forwarding node and the neighbor node to the sink respectively. And Max(dmy-s − dnei-s) is the maximum progressive distance of candidate nodes. Finally Norm(Eres) is the normalized residual energy of the neighbor node and is defined: Norm(Eres) = Eres /Max(Eres).

(3)

Where Eres is residual energy of the neighbor node and Max(Eres) is the maximum of this quantity among all possible nodes. Actually the first part of (1) exposes the minimum delay of delivering packet to the sink with forwarding it to the farthest spot from forwarding node toward destination. The second term of (1) controls power consumption in nods and increasing the lifetime of the network. So far we have delineated the operation of the protocol in the situations of we don’t have voids along the path from source to the sink. So when a forwarding node wants to send its data packet and after examining its neighbors, if it couldn’t find any neighbor node which is closer to destination, then it enters a different mode which is called perimeter mode like GPSR protocol that the conception of planarized graphs is used of that [4]. The location of entering to this mode is recorded and routing in this condition is done same as a method of perimeter forwarding in GPSR. Distance between the place of entering to perimeter mode and destination is called Reference Distance (RD). When a node receives a packet from its neighbor in this

mode, it will calculate the distance between itself and destination and if it is less than RD, it will return to the delivery data phase as said before, but If it is not, then it will be looking for closer neighbors to the sink for discovering if a neighbor node exists which its distance to the sink is smaller than RD, then it will return to the deliver data phase otherwise the routing in perimeter mode will be continued until we reach a node which is satisfied one of two above condition. This process is depicted as a flowchart diagram:

Figure 1. Flowchart of the routing operations

IV. SIMULATION AND RESULTS In this section, simulation results of proposed protocol are presented via a large number of running routing algorithm on different scenarios. For achieving this goal, we have used NS2.34 [11] which is a very fast, effective and event-driven network simulator under Linux Fedora 10 platform. First, we define the different metrics which are measured in this paper: • Packet delivery rate: the ratio of the number of received correct packets in destination to all of the sent packets from sources. • Average residual energy per node • Standard deviation of residual energy per node • Lifetime of the network and which is defined the time of first node dies during the simulation. In the first scenario, we have used the parameters which have been illustrated in table 1. The energy value of the sink (destination) is presumed unlimited. There are three sources which are deployed randomly with this condition that their x coordinate is less than 40m. The reason of this selection is that the sources that will have relatively similar distance to destination and so, the average delay of reaching data packets from these sources to sink is fairly.

TABLE I.

SIMULATION PARAMETERS

Parameter Name

Value

MAC layer Propagation Model SNR Bandwidth (Mb/s) Payload Of Packets (Bytes) Terrain ( m × m ) Nodes Number Nodes deployment The Place Of Sink Radio Range (m) Initial Energy Per Node (joule) Simulation Run Time (second)

IEEE 802.11 (CSMA/CA) Two-Ray 13.5 db 2 32 200 × 200 50, 75, 100, 125 Uniform random (200 , 100) 40 6 250

SNR in the above table shows the threshold level of correct reception of data packets and so, when the amount of SNR is less than this threshold, the packet is not received accurately. Based on this value, we have considered a kind of error model which forwarded packets are lost with a constant rate equal to 0.1. Since in this scenario we don’t want to die any node, the initial energy of nodes has been determined 6 joule. Also, all of nodes exception sources and sink nodes relinquish their energy both when receiving and transmitting the data packets. The sources just exhaust their energy while sending a packet and energy of sink node is depleted when it receives a packet. The frequency of broadcasting hello massages can have positive or negative effects on energy consumption so, for we have selected the value 50s for time intervals of broadcasting of these massages after examining results. Figure 2 shows the average residual energy per node after simulation. Every point in these curves is representative of 50 times simulation. As an obvious point the protocol has nearly the same results for different values of α. Actually in (1), α = 1 is related to the condition that the next candidate node is selected only based on Euclidean distance to destination in addition to considering holes along the path is made. This path is maintained to end of simulation. In the other various values of α, the protocol chooses the next eligible neighbor node based on both Euclidean distance and the amount of residual energy of it and also, the issue of holes are taken into account. So in these cases, assorted paths can be made. It is essential to notice that in recent states, although we have prevented from depletion of a bottleneck node by distributing traffic of packets among its neighbors but reduction of the amount of these alternative nodes energy results the average remainder energy per node corresponding to α = 1. In contrast, as it has been illuminated in Figure 3 the standard deviation of residual energy per node become better when the protocol uses α which its value is less than 1 in (1) rather than α = 1.

Average Residual Energy Per Node(J)

5.1

5

200

α=1 α = 0.75 α = 0.5 α = 0.25

150 Sink

4.9

100 4.8

50 4.7

Source

4.6 50

75

100

0 0

125

20

40

60

80

100

120

140

160

180

200

Numbe Of Nodes

α=1 α = 0.75 α = 0.5 α = 0.25

0.45 0.4 0.35 0.3

400

α=1 α = 0.75 α = 0.5 α = 0.25

350 Network Lifetime (s)

0.5

Figure 4. Deployments of nodes in second scenario.

300 250 200 150 100

0.25

50 0.2 50

75

100

0

125

0.25

Number Of Nodes

Figure 3. Comparison standard deviation of residual energy between different values α

Actually the protocol attains better performance when it dispenses the traffic load among various nodes (α < 1). The best result in above figure is related to α = 0.25 which in this case, the second factor in (1) is dominated in maximizing the L.O.F equation and a neighbor node that has the large amount of remainder energy is an eligible candidate for receiving the packet. It is worthwhile to notice that with increasing the number of nodes the standard deviation of every four curves become approximately convergent to a point. The reason of that is the variation of the first term in (1) is trivial. dmy_s is same for all of nodes can be selected and dnei_s becomes equal when number of nodes and density of them is high. So the candidate nodes are placed near to each other and distance of these nodes to destination (dnei_s) are nearly same. As a result, decreasing of the amount of energy for such selected nodes is alike and so deviation of their residual energy from their mean is converged to a point. In second scenario, we fixed the number of nodes to 50 as depicted in figure 4. All specifications of simulation are same with ones as said before in table 1. Except that the initial energy of nodes are supposed 1.5 joules and simulation time is set to 450 seconds. Because, we decide to examine the performance of proposed protocol in a condition that some of nodes are dead during the simulation. Also for this section, we consider a pair of source and sink nodes that their locations have been determined in figure 4. Lifetime of the network which has been showed in figure 4 for different values of α, has been presented in figure 5.

0.5

1 2 Packet Generation Interval (second/pkt)

Figure 5. Effect of different values of α on lifetime of the network.

As you can see in above figure the span of lifetime for α = 1 and α = 0.75 are same and approximately with decreasing of the value of α, duration of the first node lasts to die, is increased. Figure 6 displays packet delivery rate to destination in second scenario. 1

α=1 α = 0.75 α = 0.5 α = 0.25

0.9

Packet Delivery Rate

Standard Deviation of Residual Energy Per Node(J)

Figure 2. Comparison average residual energy between differen values α.

0.8 0.7 0.6 0.5 0.4 0.25

0.5

1

2

Packet Generation Interval (second/pkt)

Figure 6. Packet delivery rate with different values of α.

Because of with decreasing of α, we can prohibit energy depletion of bottleneck nodes, so the relaying nodes can deliver their packets to the sink in a longer time and as a result the percentage of delivering packet is improved. V. CONCLUSION In this paper, we proposed a routing protocol for communication of wireless sensor nodes which are used in industrial plants and it is called Industrial Wireless Sensor Networks (IWSN). These nodes have the responsibility of transferring vital information which is sensed from different

parts of a plant by source nodes to the sink where control room is located at there. But exerting wireless technologies in such environments exposes many challenges in front of us. Because in these environments, the high amount of noise can corrupt the received packets. In other words the value of SNR is low. The other issue is to reach data packets to the destination with a minimum delay as possible as. Furthermore, the nodes in IWSN have constrained resources such as available memory and power consumption. So presenting a routing protocol which can guarantee correct delivering data packets or reliability and also, optimized energy consumption of nodes is crucial. The protocol presented here is based on geographic location of nodes, and when every node wants to send a packet to its neighbors, it uses their geographic location information and residual energy of them in a defined optimizing function. Then every neighbor node has the maximum value of this function is selected as a next candidate for receiving the packet. This process is continued until the packet reaches to the sink. In this protocol the encountered voids are solved by a similar algorithm to which has been used in GPSR. The results shows better performance of proposed protocol in terms of load balancing and increasing lifetime of the network and also, reliability of delivering packets to destination. REFERENCES [1]

Andreas Willig, member, IEEE, Kirsten Matheus, member, IEEE, and Adam Wolisz, senior member, IEEE, "Wireless Technology in Industrial Networks, " 2005. [2] Kai Zeng, Kui Ren, Wenjing Lou and Patrick J. Moran, " EnergyAware Geographic Routing in Lossy Wireless Sensor Networks with Environmental Energy Supply," Department of ECE, Worcester Polytechnic Institute, January 2009. [3] Yan Yu, Ramesh Govindan, Deborah Estrin, " Geographical and Energy Aware Routing: a recursive data dissemination protocol for wireless sensor networks, "2001. [4] B. Karp and H. Kung, "Gpsr: Greedy perimeter stateless routing for wireless networks, " in ACM MOBICOM, Boston, August 2000. [5] Abinash Mahapatra, Kumar Anand, Dharma P. Agrawal, "QoS and energy aware routing for real-time traffic in wireless sensor networks,"OBR Center for Distributed and Mobile Computing, University of Cincinnati, Cincinnati, February 2005. [6] Charles E. Perkins, Elizabeth M. Royer, "Ad-hoc On-Demand Distance Vector Routing ," wmcsa, pp.90, Second IEEE Workshop on Mobile Computer Systems and Applications, 1999. [7] Rahul C. Shah and Jan M. Rabaey, " Energy Aware Routing for Low Energy Ad Hoc Sensor Networks," Berkeley Wireless Research Center, University of California, Berkeley, IEEE 2002. [8] Junyoung Heo, Jiman Hong, and Yookun Cho, Member, " IEEE, EARQ: Energy Aware Routing for Real-Time and Reliable Communication in Wireless Industrial Sensor Networks," IEEE Transactions on Industrial Informatics, vol. 5, no. 1, february 2009. [9] Y. Xu, J. Heidemann, D. Estrin, Geography-informed energy conservation for ad hoc routing, in: Proceedings of the 7th Annual ACM/IEEE International Conference on Mobile Computing and Networking (MobiCom01), Rome, Italy, July 2001. [10] Mehmet C. Vuran, Member, IEEE, and Ian F. Akyildiz, Fellow, IEEE, "XLP: A Cross-Layer Protocol for Efficient Communication in Wireless Sensor Networks,” IEEE Transactions on Mobile Computing, vol. 9, no. 11, november 2010. [11] http://nile.wpi.edu/NS/

Suggest Documents