Prediction based opportunistic routing for maritime search and rescue wireless sensor network

Accepted Manuscript Prediction based opportunistic routing for maritime search and rescue wireless sensor network Huafeng Wu, Jun Wang, Raghavendra Rao Ananta, Vamsee Reddy Kommareddy, Rui Wang, Prasant Mohapatra

PII: DOI: Reference:

S0743-7315(17)30210-1 http://dx.doi.org/10.1016/j.jpdc.2017.06.021 YJPDC 3711

To appear in:

J. Parallel Distrib. Comput.

Received date : 14 January 2017 Revised date : 14 June 2017 Accepted date : 26 June 2017 Please cite this article as: H. Wu, J. Wang, R.R. Ananta, V.R. Kommareddy, R. Wang, P. Mohapatra, Prediction based opportunistic routing for maritime search and rescue wireless sensor network, J. Parallel Distrib. Comput. (2017), http://dx.doi.org/10.1016/j.jpdc.2017.06.021 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Prediction based Opportunistic Routing for Maritime Search and Rescue Wireless Sensor Network Huafeng Wua , Jun Wangb , Raghavendra Rao Anantab , Vamsee Reddy Kommareddyb , Rui Wangb , Prasant Mohapatrac a Merchant

Marine Engineering Shanghai Maritime University Shanghai, 201306, China b EECS, University of Central Florida Orlando, Florida 32816 c University of California, Davis

Abstract In recent years, maritime and air crashes occur frequently. The existing rescue measures rely only on board satellite communications equipment, which makes it difficult to confirm the accurate positioning information and vital signs of drowning people. Recently, wireless sensor networks (WSN) are introduced to Maritime Search and Rescue (MSR). WSNs features quick expansion, selforganization, and self-adaptation to the marine environment. However, the constant changing nodes location and link reliability in marine search and rescue WSN makes the routing metrics between nodes highly dynamic. Traditional routing protocols such as AODV that establish a fixed single route based on static nodes information will provide poor packet delivery rate and take no consideration of the limited energy on the irreplaceable WSN nodes. We propose to employ opportunistic routing which can make best use of the broadcast property of radio propagation. The forwarding decisions in opportunistic routing is only based on its neighbor’s information. No network-wide flooding is required to establish routes. In order to maintain the latest neighbor information and minimize the energy cost of collecting these information, we propose a lightEmail addresses: [email protected] (Huafeng Wu), [email protected] (Jun Wang), [email protected] (Raghavendra Rao Ananta), [email protected] (Vamsee Reddy Kommareddy), [email protected] (Prasant Mohapatra)

Preprint submitted to JDPC

July 18, 2017

weight time series based routing metric prediction method to deal with the high communication cost incurred by collecting the latest routing metrics betweens nodes. Results: Our implementation of opportunistic routing protocol achieved 30% more Packet Delivery Ratio compared to the traditional AODV protocol. Also opportunistic routing protocol with prediction performed slightly better than opportunistic routing protocol without prediction. Our approach generated 90% efficiency where as 60% efficiency was achieved using AODV protocol. In achieving this an additional 3% energy consumed by the nodes. We feel additional 2% energy consumption to improve delivery greatly by 30% is a good tradeoff. Keywords: Wireless Sensor Network, Opportunistic routing protocol, topology control, complex network theory, maritime search and rescue. 1. Introduction In recent years, maritime and air crashes occur frequently. The existing rescue measures rely only on board satellite communications equipment, which makes it difficult to confirm the accurate positioning information and vital signs 5

of drowning people, greatly reducing the rescue efficiency. [1-2] first introduced the wireless sensor networks to Maritime Search and Rescue (MSR). A wireless sensor network (WSN) is a collection of spatially distributed autonomous sensors, connected to a Wireless Sensor Node, that collects data and cooperatively pass it through the network to a main location. Features of WSNs include quick

10

expansion, self-organization, and self-adaptation to the environment. Maritime Search and Rescue (MSR) is dependent on a WSN that tracks and provides real-time information on objects in a specified complex range (Huafeng Wu et al., 2013) [1]. In the marine environment, wireless sensor nodes are widely distributed, whose movement is always affected by winds, waves, surge, etc., Due

15

to this, the network topology frequently changes, the communication link is very unstable. Also, the number of nodes is limited (while a general WSN contains a large number of nodes), the sensor nodes cannot be replaced (because they are 2

arranged on lifejackets), and the nodes will be moving constantly [1]. Furthermore, nodes will not always have sufficient energy to broadcast their messages to 20

other nodes. As a result, maintaining a high communication quality and maximize the life time of WSN in MSR is quite challenging. An efficient maritime WSN routing algorithm is the key to improving the quality of communication. There has been a lot of research, most of which dealing with onshore WSNs. But, the topology and routing algorithms onshore cannot adapt to marine en-

25

vironment due to the reasons stated, so they cannot be applied directly. With a good routing technique, the best path can be determined to send the packet from the origin node to the destination node [2]. Good WSN routing algorithm can effectively improve the reliability of network communication and reduce unnecessary energy consumption. Signal propagation environment

30

on the sea is quite different from that on the land [3]. The nodes of communication network show high mobility due to winds and waves, and the signal attenuation shows a strong randomness on the sea. Hence, it is not applicable to simply adopt the signal propagation model of free space [4] to analyze network routing in maritime search and rescue environment. So it is important to

35

resolve the high mobility and signal attenuation randomness in the maritime search and rescue environment for the packet routing. As a routing technique for wireless application environment, opportunistic routing is proposed, which shows consistency with the wireless links and effectively improve the performance of WSNs. Opportunistic routing is a new kind of dynamic routing based

40

on multiple hops [5]. Unlike traditional routing, the forward nodes in this routing technique are not fixed. Instead, the forwarding nodes meeting some requirements are chosen dynamically having some priority orders for different forwarding nodes. If the forwarding to the current optimal node fails, it will switch to the second optimal node and so on, until the data packet reaches the

45

target node successfully. The opportunistic routing technique can make full use of the broadcast characteristics of radio propagation, and significantly improve the success rate of transmission. Opportunity routing mechanism can improve data transmission in the wireless communication network performance, and the 3

key point in the design of opportunity routing algorithm is determining the for50

warding priorities of all candidate neighbor nodes. The biggest challenges in marine search and rescue WSN is that node locations and link quality between nodes are constantly changing. Keeping latest routing metrics betweens nodes will incur high communication cost. We propose a prediction based opportunistic routing called POR to deal with the high communication cost. The same

55

as [18], it takes a combination of geographic location and link quality to compute forwarding priority. Differently, POR employs a light weight time series based prediction method to minimize the frequency of location and link quality information exchange between nodes. To summarize our contribution, for highly dynamic marine search and res-

60

cue WSN, we propose a new light-weight prediction based routing algorithm (by extending the existing opportunistic routing algorithms) that selects and prioritize the forwarding nodes depending on the situation. 2. Literature review For large scale networks, due to the node energy constraints, it is highly

65

difficult to achieve a global control of the network. Hence, we adopted the hierarchical topology that can not only reduce the communication load on the network but also can achieve a balanced energy network to effectively prolong the survival time of the network. Partitioning a set of nodes into clusters, and assigning a cluster head for each cluster is one way in which hierarchical network

70

is designed. In such a network, the nodes of the cluster would be forwarding the data to the cluster head, and the cluster head would be forwarding it to the next level of the hierarchy. LEACH (Low-Energy Adaptive Clustering Hierarchy) [3] was the first proposed clustering algorithm. But because of the randomness of the cluster head selection, resulting in uneven clustering, the

75

overall performance of the network was affected. HEED (A Hybrid EnergyEfficient Distributed Clustering Approach) [4] has improved the problem of uneven distribution of cluster head in LEACH algorithm by choosing primary

4

and secondary parameters, using which the energy balance is maintained across the entire network. Xu [5] assumed that sensor nodes can determine location 80

information, and the clustering can be carried out based on the location. It predicts a node’s residence time in the current cluster and makes adjustments based on the mobility with a good scalability. However, if the cluster head that is randomly selected has less energy remaining, it would exit the network once its energy is depleted, affecting the overall performance of the network. In or-

85

der to solve this problem, [6,7] have improved the algorithm at the cluster head selection process by also taking the remaining energy of nodes as a key factor, thus effectively balancing the network’s energy consumption and prolonging the lifetime of the network. The wireless sensor networks at sea are analogous to the mobile Ad-Hoc

90

wireless sensor networks. In recent years, researchers have proposed many routing solutions to improve its communication quality by introducing several of the more typical routing protocols including AODV, DSR, DSDV, and TORA [8]. These protocols can adapt to the dynamic characteristics of Ad Hoc network topology by maintaining the link, thus avoiding breakage. But they can not

95

guarantee the stability of the selected link, thereby increasing the cost of link reconstruction. The [9] analogized instability of the network link and uncertainty of information introduced the Shannon information evaluation theory entropy to MANET (Mobile Ad-hoc Networks), Sun B [10] [11] [12], which takes the entropy as a measure of cluster head selection. This maintained the

100

stability of the cluster and ensured the stability of the communication link. Zou et al [13] further proposed a stability underwater QOS routing algorithm based on entropy. Reference [1] made the detailed elaboration on the basic prototype of the MSR-WSN and preliminarily proved the feasibility of the maritime search and

105

rescue system. However, the authors did not make the research on the routing algorithm under special circumstances of the sea. In the context of maritime applications, the routing algorithm should not only ensure a better success rate and energy transfer balance but also be faced with highly dynamic nodes. So 5

far, most of the studies on WSN focused on how to ensure the load balance, 110

minimizing energy consumption, data transmission efficiency, and reducing the end-to-end delay. Thomas Watteyne [11] and others pointed out that the data transmission of WSN should happen in the form of multiple hops, and the energy of the WSN finiteness requires that the routing protocol has high addressing efficiency. Watteyne analyzed the flood routing protocol and hierarchical rout-

115

ing protocol of WSN and summarized the basic concept of routing protocol, its improvements, and application of the latest research results with relevant agreements. The flooding routing algorithms include SPIN (Sensor Protocols for Information via Negotiation) routing algorithm, SAR (Sequential Assignment Routing) routing algorithm, LEACH (Low Energy Adaptive Clustering

120

Hierarchy) routing algorithm, clustering routing algorithm, opportunistic routing algorithm and so on. Djamel Djenouri [12] pointed out that different WSN application scenarios had different quality of service requirements, and a WSN routing protocol was proposed based on local service quality. It is also the first WSN based on different application background for the remainder of the time

125

delay of data transmission nodes, reliability, energy requirements have different characteristics. Most WSN routing protocols discussed so far apply only to a single sink network. However, Luca Mottola [13] proposed a multicast routing protocol in which distributed data fusion path and load balancing technique were used. Under the same conditions, this protocol can reduce the total energy

130

consumption by 10% and also the routing time is shorter. In addition, Ayaz M [14] proposed a kind of multiple hops relay, dynamic positioning, good expandability, and short routing time UWSN routing algorithm. Compared to other routing algorithms, this routing algorithm need not have any information and at the same time, need not have any additional hardware

135

overhead. Mari Carmen Domingo [15] proposed a kind of suitable routing algorithm for deep sea environments, i.e., routing algorithm without using GPS. The routing algorithm minimized active routing information exchange, protected time values to reduce the amount of data loss and ensured the communication quality through continuous adaptive school guard, which showed high efficiency 6

140

in the process of signal transmission. At Massachusetts institute of technology (MIT), Biswas S. proposed opportunistic routing mechanism. Opportunistic routing is different from the traditional routing, as in which it selects multiple potential forwarding nodes for the next hop. The forwarding nodes, which are likely to be the next jump, must satisfy certain successively smooth forward-

145

ing activity. Multiple forwarding node collaboration forms the space diversity greatly improving the throughput and transmission reliability. Therefore, the revolutionary routing idea has been widely accepted. The basic idea of this routing is a very good use of the broadcast feature of wireless communication. Improvement of the success rate of the information transmission in the maritime

150

search and rescue environment can fully embody its superiority. G. Huo [16] proposed a routing algorithm based on the RSSI (Received Signal Strength Indication) routing measures, aiming at solving the sparse Mobile WSNs (MWSN) node problem, i.e., the end-to-end path may often be broken or does not exist. In conclusion, although the WSN routing has been widely studied, research on

155

maritime search and rescue of the WSN routing algorithm is a new attempt. Under different WSN application scenarios, different routing algorithms need to be designed. Under the sea background, information dissemination is poorer, and topology changes dramatically. In order to guarantee that the transfer of node information has a higher success rate and long cycle life, some particular

160

studies still need to be done. 3. System Model The Maritime Search and Rescue System based on WSN. An MSR-WSN system includes sensor nodes, sink nodes, communication base stations and the network center of searching. Ordinary sensor nodes are put on lifejackets, which are acti-

165

vated automatically once it senses the seawater. These nodes can be connected with each other by self-organizing themselves in the form of a wireless multihop network. Ordinary nodes send the monitoring information to the sink nodes and then this data is forwarded to the base station on the mothership or rescue

7

Figure 1: The system structure of MSR-WSN

boat. Finally, the rescue center system will receive the information from the 170

base station via the satellite network. Related search and rescue departments and business users can seek for the information on the targets in this system. Fig.2 describes the network model of marine WSN. Each node broadcast information and receive information from neighbor nodes to create a list of reliable nodes based on the energy and predicted distance factor of the neighbor

175

from the sink nodes. All the nodes maintain a list of reliable nodes based on the information, mainly energy and predicted distance factor, received from the neighbor nodes. The listed neighbor nodes are prioritised based on the remaining energy and the distance factor. If an ordinary node with low residual energy is the source then it would be better for the node to send the packet

180

to a neighboring node which has high residual energy to deliver the packet to the sink node, if the sink node is not in the range of the source node. Nodes with higher residual energy forwards the packet to the most prioritised node in its neighbor list. Additionally packets are sent to all the nodes available in the neighbor list of a specific node with a specific timer value attached for each

185

neighbor node. This is done to ensure that if the most prioritised neigbor node

8

does not forward the packet within specific time, the next prioritised neighbors will forward the packet. This approach is followed to minimize packet loss and increase packet delivery ratio.

Figure 2: Network model of MSR-WSN

System Assumption. Wireless sensor nodes are randomly deployed in the sea. 190

As the marine environment is extremely complex, the movements of wireless sensor nodes are randomly affected by wind, wave and surge. In order to facilitate the description and analysis of the problem, we make the following assumptions: 1. The wireless sensor nodes in the sea can obtain precise GPS location for themselves.

195

2. Nodes within a range communicate via broadcast using a symmetric communication link. 3. The transmission power of the wireless sensor nodes is fixed. Node Motion Model. In the MSR-WSN system, the nodes are driven by the action of waves, in which the overall movement of all the nodes follows the

200

same trend and the state of motion of each node would be affected by the local wave forces. Many models were introduced to replicate node movement in sea. Reference Group based Mobility (RPGM) model characteristics closely reflects node motion in sea but the speed of the nodes in this model is very slow. 9

[14] proposed a model for offshore wireless sensor nodes. It takes the rate of 205

population movement and moving direction into account in order to show the moving state. The moving speed of each node is the rate of the group with the addition of a random rate, whereas the moving direction of each node is the moving direction of the group with the addition of a random moving direction, which overall subjects to uniform distribution. To properly test the protocol we

210

needed a model which allows to provide a bit higher node speeds. So for this current simulation a model called RandomWalk model from BonnMotion [21]] package is used for node motion. BonnMotion is a Java software to create and analyse mobility scenarios and is one of the most commonly used tools for the investigating mobile adhoc network characteristics.

215

4. Opportunistic routing The biggest challenges in marine search and rescue WSN is that node locations and link quality between nodes are constantly changing. Keeping latest routing metrics betweens nodes will incur high communication cost. We propose a prediction based opportunistic routing called POR to deal with the high

220

communication cost. The same as [18], it takes a combination of geographic location and link quality to compute forwarding priority. Differently, POR employs a light weight prediction method to minimize the frequency of location and link quality information exchange between nodes. Wireless Energy Propagation Model based on Wave Shadowing Effect. The link quality between nodes in marine environment is highly affected by waves. When the signal is reflected or absorbed by the waves, energy loss will happen. One of the most common radio propagation model is the log-normal shadowing path loss model [24]. The logarithm of sum of losses is assumed to follow Gaussian distribution. The received signal power (unit’s dB) can be calculated with equation (1): Ωr (d) = Ωt − P L(d0 ) − 10β log10 (

10

d )−χ d0

(1)

where Ωr (d) is the received signal power at a distance of d, Ωt is transmitting 225

power at the transmitter, P L(d0 ) is the average path loss for a reference distance d0 , β is the pass loss exponent, and parameter χ denotes the shadowing effect on the propagation path, which follows Gaussian distribution with expectation being zero and variance being σ 2 . As the waves go ups and downs, the quality of the communication link between nodes is affected by wave randomly. Because χ follows Gaussian distribution with the mean being 0 and the variance being σ, the received signal power Ωr (d) also follows a Gaussian distribution with mean being EΩr (d) = Ωt − P L(d0 ) − 10β log10 ( dd0 ) and the variance being σ 2 . Under the assumption

of constant transmitted power, only when the received signal is greater than a threshold T , stable communication between transmitter and receiver can be ensured. Therefore, in shadowing environment, to ensure successful communication at a distance of d, the signal power at the transmitter need to meet the following conditions: Ωt > P L(d0 ) + 10β log10 (

d )+T +χ d0

(2)

The threshold T can be calculated using a Q-function [25]. We use θ to denote the rate of correct reception at the distance d. They are formally expressed as follows: P (Ωr (d) > T ) = θ = Q(

T − EΩr (d) ) σ

where Q is a function [25] defined as: Z +∞ 1 x2 Q(z) = √ exp(− )dx 2 2π z

(3)

(4)

Routing metric estimation based on time series prediction model. Packet advancement prediction In this section, we deal with the geographic location factor when determining the forwarding node in opportunistic routing. As assumed, each node is able to obtain its own location information. Formally, we use ij to represent one of node i’s neighbors. We use dij to denotes the distance factor derived from nodes location. dij is formally defined 11

as the packet advancement to the destination D when packet sent by node i is forwarded by neighbor ij : dij = Distance(i, D) − Distance(ij , D)

(5)

where Distance(i, D) and Distance(ij , D) are the Euclidean distance between i and D and between ij and D, respectively. Getting dij on node i is easy in traditional static networks. What is required is only one hello messages from each of its neighbors. However, in marine environment, each node’s locations changes frequently. Therefore, dij is also changing. Upon every forwarding request, node i has to collect latest location information from all of its neighbors. In the energy restricted marine rescue environment, collecting real time location information is energy prohibitive. One intuitive approach is collecting neighbor location information at a fixed time interval. When making forwarding decisions, a node just used the location information from last interval. The drawback of this approach is that the locations information may be highly outdated, which may cause a further neighbor as a higher priority forwarding node. We propose to exchange neighbors location periodically with a HELLO message and keep the most recent n packet advancement value for each neighbor. For each future forwarding request, we apply simple time series prediction method to this history data to get an estimated packet advancement value. We choose the simple moving average (SMA) prediction method as the first attempt. Specifically, we predict the packet advancement value at the next future interval as the average of the most recent n history value. We denote the n history packet advancement value for neighbor j as: dtij , dit−1 , ..., dt−n+1 . Then the package ij j advancement dˆt+1 at time t + 1 is estimated as: ij dˆt+1 = ij 230

dtij + dit−1 + ... + dt−n+1 ij j n

(6)

SMA is able to reduce the noise associated with fluctuations which makes it easy to capture the general trend. The reason for applying SMA in the marine environment is that the sea movement at a area within a small interval does

12

not change significantly. A second reason is the low computational overhead of SMA. Link quality prediction As nodes move, the distance between node i and each of its neighbor is constantly changing. According to the propagation model introduced above, the received signal strength is also changing. Additionally, even if the distance between two nodes does not change, the sea wave shadowing effect will also cause the received signal to change frequently. As introduced in [26], the received signal strength directly affect the packet reception ratio (PRR) between the two nodes. We use PRR to represent the link quality between two nodes. Ideally, PRR from node i to neighbor ij should be obtained by sending a list of test packet from i to ij and recording the number of successfully received packet. During this test, the locations of node i and neighbor ij should be fixed. Obviously, this method is not practical in the marine rescue environment where both the distance and wave shadowing effects between the two nodes are constantly changing. Fortunately, [27] gives a theoretical model to calculate based on received signal strength. Their analytical results matches very well with PRR results from real experiment. The PRR p is given by: Ωr (d)−Pn 1 1 2 0.64 )8f p = (1 − exp− 2

235

(7)

where Pn is the noise floor. The received signal strength can be obtained through the HELLO message. The same as the prediction of packet advancement, PRR at time t + 1 can be estimated by the average of most recently n PRR value pt+1 ij history. Priority calculation

When a node sends a packet, all its neighbors within

communication range can hear the packet. In opportunistic routing, the key design is to determine which neighbor should forward the received packet first. We take both the packet advancement and link quality into consideration when calculating the priority of a node. The first reason for this design is that selecting the candidate node that can make the larger advancement to the destination will 13

reduce the number of hops. The second reason is that selecting the candidate node with higher PRR will increase the overall probability of successful delivery per hop. The priority Pij for neighbor j of node i can be calculated as: Pi j = d i j × p

(8)

The meaning of Equation (8) can be interpreted as the expected packet ad240

vancement per transmission. Candidate selection. Opportunistic routing attempts to make best use of its neighbors. However, we should not simply include all neighbors that are within the communication range as the forwarding candidates. Regarding the packet advancement, we should only select a neighbor if it can advance a packet to the destination. That is neighbor ij should be included only when its packet advancement dij > 0. After determining the initial candidate forwarding nodes, we sort them based on their priority as introduced earlier. This initial prioritized candidate list is denoted as F0 . To further optimize the set of forwarding nodes, we define a packet delivery rate r for one hop. For example, selecting the first n node in F0 will result in a packet error rate of one hop as: r =1−

n Y

(1 − pij )

(9)

j=1

The users can set a threshold value r0 as the desired packet error ratio per hop. Therefore, the final candidate list will be the first n nodes in F0 that make r ≥ r0 . Sometime, the resulting r from including all nodes in F0 is still

smaller than r0 , we simply include all nodes in F0 . POR’s main objective is 245

to use opportunistic routing to ensure transmission reliability, while reducing the number of transmissions, and therefore, the selection of the candidate node set needs to ensure that the number of backup links can provide the required delivery rate. Priority scheduling. POR uses a timer-based priority scheduling algorithm to

250

coordinate the forwarding of a packet. The neighbor node with the highest 14

priority will attempt to forward the packet first. For other candidate nodes, if they hear a higher-priority node has successfully forwarded the packet, they would terminate their forwarding process. However, if the higher priority node fails to forward the packet, a lower priority node will start to forward this packet. 255

The biggest challenge to make this schema work is to avoid multiple neighbor nodes forwarding the same packet. We add a timer to each candidate neighbor nodes. Each node will start to forward the packet when the timer expires. Each node will turn off its timer when it hears a node with higher priority has successfully forwarded the packet. The waiting time of each timer is based on

260

the node’s priority. A node with a higher priority has less waiting time. We define the network delay T as the time from a node receiving a packet to send it completely. We index the priorities of n candidate node from the highest to the lowest as 0, 1, 2, 3, ..., n. The node with the highest priority has zero waiting time. So within network delay T , if the node priority 2 does not hear that

265

the node with priority 0 has successfully forwarded the packet, it will start to forward the packet. The waiting time for a node with priority i is generalized as (i − 1) × T .

Specifically, the network delay consists of four parts: the processing delay,

queuing delay, transmission delay, and propagation delay. Since we do not 270

consider the network load, which means not considering the queuing delay, the network delay consists of three parts. Assuming that the total time of these three parts is T , if the node priority is i, the timer should be set to (i − 1)T . In our simulation, the packet size is set to 512 bytes. The protocol in MAC layer is

802.11, in which the channel rate is 2 Mbps. So, the transmission delay is equal 275

to 512 × 8 bits/2M bps = 0.002048s. The radio wave propagation velocity in air is equal to the speed of light, namely, 3×108 m/s. However, the distance between

two vehicles who can communicate with each other directly is less than 250 m. So, the propagation delay is equal to 250m/3 × 108 m/s = 0.83 × 10− 6s, and it can be ignored. Through doing multiple times of simulation and analyzing the

280

trace files, we can get the processing delay which is approximately 0.001 0.002 s. Therefore, based on the above analysis, we can conclude that T is about 15

0.004 s. This approach does not solve hidden node problem completely, that is if low priority forwarding node does not receive packet from high priority forwarding 285

node, both of them forward the packet. But the redundancy of a packet is only for a short time as it has to be understood that as the packet moves towards the destination the neighboring nodes are reachable (hidden node problem is less) and then the duplicate packets are dropped. In the worst case it is the responsibility of the destination to drop off the duplicate packets.

290

5. Simulation results and evaluation Results vary based on the number of nodes in the network, receiver threshold range, HELLO message interval, initial position of source and destination and so forth. Ordinary nodes, which are attached to the lifejackets as mentioned in section 3, send the monitoring information to the sink nodes and then this data

295

is forwarded to the base station on the mothership or rescue boat. Destination nodes, which are sink nodes, are always fixed. In the experiments that follow, the destination node is fixed which is node 0. Network Simulator, ns2, is used to perform simulation analysis. Fig.3 is the animated version of initial topology of the nodes deployed in the network. To properly validate the routing algorithm

300

a node which is far from the destination is chosen as a sink node. The simulations were executed for a simulation time of 150 s, for a grid dimension of 1000 m x 1000 m, consisting of 50 nodes. Each node was configured to send HELLO packets with an interval of 5 s and the receive threshold of 500 m. RandomWalk node motion model was used with a minimum and maximum

305

node movement speed of 15 m/s and 10 m/s respectively. Unless explicitly mentioned in the experiments, this environment was used by default. Overall packets delivery rate. Three routing protocols are tested • AODV • Our Opportunistic routing with prediction disabled. 16

Figure 3: Nodes deployed in the network

310

• Our Opportunistic routing with prediction enabled The following experiments are conducted on the above mentioned routing protocols to find out the efficiency or throughput achieved in transmitting data from source to destination. Varying Number of Nodes Fig.4 depicts scenarios with varying number

315

of nodes. In each scenario destination is stationary and the farthest most node which is moving towards the destination for sometime and latter on moves away from the destination is chosen as the source node. We specifically chose that node as source as AODV works only when the source and destination nodes are nearby or in range.

320

It can be observed from the figure that, with any number of nodes the AODV protocol does not perform better than opportunistic routing protocols, both with and without prediction. This is due to that a node in AODV protocol forms a single link and transmits the data. Only when the formed link is broken a new link is formed. Where as in opportunistic routing protocols nodes broadcast it

325

information regularly to neighbor nodes to communicate with the destination node. So multiple links are formed in opportunistic routing protocol. When it comes to comparing opportunistic routing protocols, both perform 17

Figure 4: Packet delivery rate and Energy consumed comparisons of AODV, opportunistic routing and opportunistic routing with prediction algorithms by varying number of nodes in the network

very well in delivering the packet to the destination. But as the number of nodes deployed increases opportunistic routing protocol with prediction performs bet330

ter compared to opportunistic routing protocol without prediction. Congestion increases as the number of nodes increase. In such a situation forwarding nodes should be chosen wisely using prediction algorithm. Opportunistic routing algorithm maintains the history of the neighbor nodes and broadcasts the packet to all the predicted neighbor nodes along with assign-

335

ing priorities to the neighbor nodes. If the highest node does not transmit the packet for some reason the next priority node tries to transmit the packet. This way multiple links are formed and this increases the possibility of delivering the packet to the destination. Opportunistic routing algorithm with prediction focuses on choosing the neighbor nodes wisely, in a mobile environment,

340

by maintaining the neighbor nodes history and predicting the location of the 18

neighbor’s from the previous history and also by checking the link quality of the neighbor nodes using packet error rate of the neighbor nodes. Energy consumed per node is also shown in Fig. 4. As explained AODV algorithm does not broadcast the data but delivers the packet to the desti345

nation by forming a single path. If the path is or link is lost a new path is formed then using hello messages. But opportunistic routing algorithms broadcast hello information in regular intervals of time, for the neighboring nodes to build neighbor history. So it is obvious that the energy consumed is more in opportunistic routing algorithms. Each node has to broadcast its position and

350

remaining energy, for the neighbor nodes to predict the location and check for link quality, in the form of hello packets. All the nodes get their own location by communicating through GPS. Energy consumed for communicating with GPS is calculated as explained in [22] [23]. Varying Communication Range

355

Simulation environment: Fig.5 de-

picts the packet delivery rate against node communication range results of the above mentioned routing protocols. Results show that, as the communication range of the nodes increase the packet delivery ratio also increases in all the three algorithms. If communication range is more AODV algorithm perform better as there is only single

360

link formed. But in marine environment communication range of the nodes are not high, so performance degrades. Opportunistic routing algorithm performs way better than AODV when both communication range is low and high. Opportunistic routing protocol with prediction performs similar to opportunistic routing protocol without prediction. Reason for this improvement is due to bet-

365

ter and strong links are chosen by the nodes using neighbor nodes distance and packet error rate along with predicting the next immediate forwarding node. When the communication range of the nodes is less, the average energy of the network is more, as the less distance a node can communicate, the more are the number of hops required to deliver a message to the destination. It can

370

be observed from fig.5(b) that opportunistic routing algorithms, consumes more 19

Figure 5: Packet delivery rate and Energy consumed comparison of AODV, opportunistic routing and opportunistic routing with prediction algorithms by varying the communication range of the nodes

energy than AODV. The reason for consuming more energy is explained in the previous experiment, that nodes broadcast information in regular intervals of time unlike AODV. Varying Average Node Moving Speed Simulation environment: Fig.6 375

depicts the packet delivery rate against average node moving speed results of the above mentioned routing protocols. In marine environment nodes are always moving. So it is very crucial to consider node speed as a factor. We have varied the average speed of the nodes from 5 m/s to 30 m/s. As can be observed from fig.6, opportunistic routing

380

protocol with and without prediction performs way better than AODV protocol. In AODV protocol the link that gets formed is not stable when the nodes are continuously moving. A new route has to be formed which increases the delay as

20

Figure 6: Packet delivery rate and Energy consumed comparisons of AODV, opportunistic routing and opportunistic routing with prediction algorithms by varying the speed at which the node move in the network

well as reduces the packet delivery ratio. But in opportunistic routing protocol, nodes broadcast the data and all the time nodes try to find a better routing path 385

which increases the probability of delivering the message to the destination. As mentioned in section 5.1.3, nodes are always moving in marine environment and it is very crucial to consider node speed as a factor. Consumption of energy in opportunistic routing protocol with and without prediction is almost the same by varying average speed of the nodes from 5 m/s to 30 m/s but both

390

of the these routing protocols consume more energy compared to AODV protocol. As mentioned earlier this scenario is due to that the nodes broadcast hello messages regularly to indicate neighbor nodes about its existence which is the main reason for nodes to consume more energy.

21

Varying Hello Message Intervals Only two routing protocols are tested 395

in this experiment • Our Opportunistic routing with prediction disabled. • Our Opportunistic routing with prediction enabled

Figure 7: Packet delivery rate and Energy consumed comparisons of opportunistic routing and opportunistic routing with prediction algorithms by varying the frequency at which the nodes broadcast Hello messages

As explained before, opportunistic routing algorithm with prediction predicts the location of the neighbor nodes unlike opportunistic routing which uses 400

the last received location from the neighbor’s to forward the data. According to the observations the packet delivery ratio is better by predicting the neighbor’s actual position with a negligible amount of energy loss, which can be observed from fig.7. It can also be interpreted form fig.7(b) that as the broadcasting interval is increased, nodes end up consuming less energy. Increasing broadcasting

405

interval makes nodes prioritize its neighbors based on the old history more often 22

rather than the actual updated information from the neighbor nodes. This leads to decrease in performance with respect to packet delivery ratio. Overall, considering the fact that the percentage of performance improvement in-terms of packet delivery rate, the loss in energy consumption with 410

respect to AODV protocol is negligible. 6. Conclusion A topology control algorithm for determining an optimal MSR-WSN was proposed in this paper which is based on the complex network theory.The complex network model of the MSR-WSN was established by taking into account

415

its characteristics. According to this protocol 3% to 4% more energy is consumed but there is around 30% to 40% improvement in delivery of the packet, compared to AODV, which is a very good trade off Future work. As of now we have investigated the routing algorithm using Bonnmotion topology model which has limits to model real marine environment. We

420

will investigate more efficient topology control algorithms in constructing the optimal marine environment and apply the proposed communication protocols to an efficient topology which resembles marine environment. We will also test the proposed routing algorithm under the environments where multiple source nodes exist.

425

7. Acknowledgment This work was supported by National Natural Science Foundation of China (51279099, 51579143); The Innovation Program of Shanghai Municipality Education Commission (13ZZ124); “Shu Guang” Project of Shanghai Municipal Education Commission and Shanghai Education Development Foundation (12SG40);

430

The project of applied basic research of the Ministry of Transportation (2013329810300); U.S National Science Foundation Grant CCF-1337244, 1527249 and 171388. We would like to give our sincere thanks to Dr. FAN Ye from Stony Brooks for helping us generously improve the presentation of this paper. 23

References 435

[1] A. Ahmadi, A. Bigdeli, M. Baktashmotlagh, and B. C. Lovell. A wireless mesh sensor network for hazard and safety monitoring at the port of brisbane. In Local Computer Networks (LCN), 2012 IEEE 37th Conference on, pages 180–183. IEEE, 2012. [2] E. M. Carapezza and A. Bucklin. Intelligent maritime security system with

440

sensor networks for coastal environmental and homeland security applications. In Optics/Photonics in Security and Defence, pages 67360F–67360F. International Society for Optics and Photonics, 2007. [3] B. Chen, K. Jamieson, H. Balakrishnan, and R. Morris. Span: An energyefficient coordination algorithm for topology maintenance in ad hoc wireless

445

networks. Wireless networks, 8(5):481–494, 2002. [4] J. Chen, H. Shen, and H. Tian. Energy balanced data gathering in wsns with grid topologies. In Grid and Cooperative Computing, 2008. GCC’08. Seventh International Conference on, pages 362–368. IEEE, 2008. [5] P. Closas, A. P. Zamora, and J. A. F. Rubio. A game theoretical algorithm

450

for joint power and topology control in distributed wsn. In Acoustics, Speech and Signal Processing, 2009. ICASSP 2009. IEEE International Conference on, pages 2765–2768. IEEE, 2009. [6] H. Denkilkian, A. Koulakezian, R. Ohannessian, M. S. Chalfoun, M. K. W. Joujou, A. Chehab, and I. H. Elhajj. Wireless sensor for continuous real-

455

time oil spill thickness and location measurement. Instrumentation and Measurement, IEEE Transactions on, 58(12):4001–4011, 2009. [7] W. Guo, B. Zhang, G. Chen, X. Wang, and N. Xiong. A pso-optimized minimum spanning tree-based topology control scheme for wireless sensor networks. International Journal of Distributed Sensor Networks, 2013,

460

2013.

24

[8] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan.

An

application-specific protocol architecture for wireless microsensor networks. Wireless Communications, IEEE Transactions on, 1(4):660–670, 2002. [9] V. Kawadia and P. Kumar. Power control and clustering in ad hoc networks. 465

In INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and Communications. IEEE Societies, volume 1, pages 459–469. IEEE, 2003. [10] J.-J. Kim, C.-H. Hong, D.-J. Kim, B.-B. Lee, J.-D. Kim, and K.-N. Ko. The device for generation the distress signal and monitoring system for a sur-

470

vivor based on wsn. In Electronics and Information Engineering (ICEIE), 2010 International Conference On, volume 1, pages V1–81. IEEE, 2010. [11] L. M. Kirousis, E. Kranakis, D. Krizanc, and A. Pelc. Power consumption in packet radio networks. Theoretical Computer Science, 243(1):289–305, 2000.

475

[12] M. Kubisch, H. Karl, A. Wolisz, L. C. Zhong, and J. Rabaey. Distributed algorithms for transmission power control in wireless sensor networks. In Wireless Communications and Networking, 2003. WCNC 2003. 2003 IEEE, volume 1, pages 558–563. IEEE, 2003. [13] N. Li and J. C. Hou. Topology control in heterogeneous wireless networks:

480

problems and solutions. In INFOCOM 2004. Twenty-third AnnualJoint Conference of the IEEE Computer and Communications Societies, volume 1. IEEE, 2004. [14] S. Narayanaswamy, V. Kawadia, R. S. Sreenivas, and P. Kumar. Power control in ad-hoc networks: Theory, architecture, algorithm and implemen-

485

tation of the compow protocol. In European wireless conference, volume 2002, pages 156–162. Florence, Italy, 2002. [15] M. Pugliese, L. Pomante, and F. Santucci. Topology optimization and network deployment algorithm in wsns for mobile agent-based applications. In 25

Computer Modeling and Simulation (EMS), 2010 Fourth UKSim European 490

Symposium on, pages 460–465. IEEE, 2010. [16] J. Shao. Study on wsn topology control based on location[d]. 2010. [17] T. Wehs, M. Janssen, C. Koch, and G. Von Cölln. System architecture for data communication and localization under harsh environmental conditions in maritime automation. In Industrial Informatics (INDIN), 2012 10th

495

IEEE International Conference on, pages 1252–1257. IEEE, 2012. [18] H. Wu, Q. Zhang, S. Nie, W. Sun, and X. Guan. An energy distribution and optimization algorithm in wireless sensor networks for maritime search and rescue. International Journal of Distributed Sensor Networks, 2013, 2013.

500

[19] J. Zhang, Y. Xiang, X. Liu, W. Wang, and Y. Tang. An energy-efficient distributed algorithm for virtual backbone construction with cellular structure in wsn. International Journal of Distributed Sensor Networks, 2012, 2012. [20] Radicchi, Filippo and Castellano, Claudio and Cecconi, Federico and

505

Loreto, Vittorio and Parisi, Domenico. Defining and identifying communities in networks. In Proceedings of the National Academy of Sciences of the United States of America, volume 101. number 9. pages 2650–2663. National Acad Sciences 2004. [21] N. Aschenbruck, R. Ernst, E. Gerhards-Padilla, M. Schwamborn BonnMo-

510

tion - a Mobility Scenario Generation and Analysis Tool. In 3rd International ICST Conference on Simulation Tools and Techniques (SIMUTools ’10), Torremolinos, Malaga, Spain, 2010. [22] María Gabriela Calle Torres Energy Consumption in Wireless Sensor Networks using GSP.

515

[23] Aaron Carroll, Gernot Heiser An Analysis of Power Consumption in a Smartphone. 26

[24] J. C. Reyes-Guerrero, Gordan Sisul and Luis A. Mariscal Measuring and Estimating the Propagation Path Loss and Shadowing Effects for Marine Wireless Sensor Networks at 5.8 GHZ In 20th Telecommunications forum 520

TELFOR, Serbia, Belgrade, November 20-22, 2012 [25] http://sarl.ece.gatech.edu/wp-content/uploads/sites/488/2016/ 09/20-Log-normalShadowing.pdf [26] Marco Zuniga and Bhaskar Krishnamachari Analyzing the Transitional Region in Low Power Wireless Links In IEEE, 2004

525

[27] Abtin Keshavarzian, Elif Uysal-Biyikoglu Falk Herrrnann, Arati Manjeshwar Energy-efficient Link Assessment in Wireless Sensor Networks In IEEE, 2004

27

1

Jun Wang received his Ph.D. in Computer Science and Engineering from the University of Cincinnati in 2002. He received the B.Eng. degree in Computer Engineering from Wuhan Technical University of Surveying and Mapping (now Wuhan University) and M.Eng. degree in Computer Engineering from Huazhong University of Science and Technology. He is a Full Professor of Computer Science and Engineering, and Director of the Computer Architecture and Storage Systems (CASS) Laboratory at the University of Central Florida, Orlando, FL, USA. He has won Deans Research Professor Award in 2013, and was named Charles N. Millican Faculty Fellow in EECS during 20102012. He is the recipient of National Science Foundation Early Career Award 2009 and Department of Energy Early Career Principal Investigator Award 2005. He has authored over 60 publications in premier journals such as IEEE Transactions on Computers, IEEE Transactions on Parallel and Distributed Systems, and leading HPC and systems conferences such as HPDC, EuroSys, ICS, Middleware, FAST, IPDPS. He has graduated 9 Ph.D. students who upon their graduation were employed by major US IT corporations (e.g., Google, Microsoft, etc.). He has served as numerous US NSF grant panelists and US DOE grant panelists and TPC members for many premier conferences such as IPDPS, ICPP, HiPC. He currently serves in the editorial board for the IEEE Transactions on Cloud Computing since 2016, IEEE Transactions on Parallel and Distributed Systems in 2012-2014, and was associate editor for International Journal of Parallel, Emergent and Distributes Systems (IJPEDS) during 20102012. He serves in the steering committee for the Second International Workshop on Energy Consumption and Reliability of Storage Systems (ERSS 2012), California, USA. He is an IEEE ScalaCom2012 program committee vice chair. He serves as a program co-chair (storage track) for the 7th IEEE conference on Network, Architecture and Storage (NAS), June 2012. He has co-chaired the 1st International Workshop on Storage and I/O Virtualization, Performance, Energy, Evaluation and Dependability (SPEED 2008) held together with HPCA. He is a Senior Member of IEEE.

Huafeng Wu received Huafeng Huafeng the Wu undergraduate is a profes- degreePh.D fromSupervisor, Navigation Executive Technology of Jimei sor, Member of University (Junior preparatory classes), in 1997, the Shanghai Institute of Electronics and Shangmasters degree in trafficDr. information engineering hai Shuguang Scholar. Wu received his undergraduate degree fromMaritime Navigation Tech- in and control from Dalian University, nology of Jimei (Junior prepara- ap2004, and the University PhD degree in computer tory classes) in 1997;from and Fudan in 2004,he got in plication technology University, master degree of TrafficPhD Information Engi2008. He is a professor, supervisor, execuneering and Control in Dalian Maritime Unitive member of Shanghai Institute of Electronics versity; He graduated fromScholar. Fudan University and Shanghai Shuguang In 2008-2009 as Ph.D. of Computer Applicationresearch Technol- with he awas engaged in postdoctoral ogy in 2008; In 2008 -2009he was postdoctoral research Jiao Carnegie Mellon University; thenengaged he wasinstudy with Shanghai at Carnegie Mellon University; Then he study in and Shanghai Jiaohe is Tong University as a visiting scholar in was 2012-2013; currently Tongdeputy University as aofvisiting in 2012-2013; And the director Humanscholar Resource Division and a currently full professor he the is the Deputy Director HumanShanghai ResourceMaritime Division and asso- He in Merchant Marine of College University. ciateenthusiasm professor offorMerchant in Shanghai Maritime and has the basicMarine theoryCollege of computer, communication University. Research AreasDr. Wufield hasofenthusiasm for the network and its application in the transportation and basic maritime, theory of wireless computer,sensor communication and network and its application including network (WSN), radio frequency identification in the field of transportation and maritime, including wireless sensor systems (RFID), cloud computing and Internet-of-Things, and so on. network (WSN), radio frequency systems (RFID), cloudsuch He has published more than 70identification papers in authoritative journals computing and Internet-of-Things, and so on. He has published as the International Journal of Distributed Sensor Networks, Journal more than 70 papers in top authoritative journals such assuch International of Communication, and International conference as the IEEE Journal of Distributed Sensor Networks, Journal of communication, ICDCS and IEEE PerCom, most of them are indexed by SCI, EI. He and top International conference such as the IEEE ICDCS and IEEE presides over the national Natural Science Fund of China (NSFC) PerCom, most of them are indexed by SCI, EI. He Presides over the projects, key project of natural science fund of Shanghai, Shanghai national Natural Science Fund of China (NSFC) projects, key project ShuGuang plan talent project, applied basic research project of ministry of natural science fund of Shanghai, Shanghai ShuGuang plan of transport, Shanghai Education Research and Innovation Key Project, talent project, applied basic research project of ministry of transport, selection and training of outstanding young teachers in universities in Shanghai Education Research and Innovation Key Project, selection Shanghai project, and he young also participates the United and training of outstanding teachers in in universities in States Shang- NSF project, national project, in ministry of science andproject, technology hai project, and he863 alsoplan participates the United States NSF special national natural scienceand fundtechnology project, key project of nationalprogram, 863 plan the project, ministry of science special Shanghai science and natural technology commission and PuJiang talent program, the national science fund project, key project of plan and so on;science and he and has technology been authorized 3 national patents, 6 Shanghai commission and invention PuJiang talent utility model plan and so patents. on; And he has been authorized 3 national invention patents, 6 utility model patents.

Raghavendra Rao Ananta received the BTech degree in Electronics and Communication Engineering from India, in 2012. He is pursuing the Master degree in Computer Engineering from the Electrical Engineering and Computer Science Department, University of Central Florida, since 2016. His research focuses on energyefficiency computing and file/storage systems, wireless sensor network, mobile computing, Internet of Things.

Vamsee Reddy Kommareddy received the BE degree in Electronics and Communication Engineering from India in 2013. He is pursuing the Master degree in Computer Engineering from the Electrical Engineering and Computer Science Department, University of Central Florida, since 2016. His research focuses on file/storage systems, wireless sensor network, mobile computing, Internet of Things and energy-efficiency 14 computing.

Rui Wang received the BS degree from Xi’an Jiaotong University City College, in 2011. She is pursuing the Master degree in computer engineering from the Electrical Engineering and Computer Science Department, University of Central Florida, since 2016. Her research interests include machine learning and approximate computing.

Dr. Prasant Mohapatra is a Professor in the Department of Computer Science, serves as the Dean of Graduate Studies, and Vice-Provost of Graduate Education of the University of California, Davis. He is the former Tim Bucher Family Endowed Chairman of the department. In the past, he has held Visiting Professor positions at AT&T, Intel Corporation, Panasonic Technologies, Institute of Infocomm Research (I2R), Singapore, and National ICT Australia (NICTA), University of Padova, Italy, Korea Advanced Institute of Science and Technology (KAIST), and Yonsei University, South Korea. He is the Editor-in-Chief of the IEEE Transactions on Mobile Computing, and has served on the editorial boards of the IEEE Transactions on Computers, IEEE Transactions on Mobile Computing, IEEE Transaction on Parallel and Distributed Systems, ACM WINET, and Ad Hoc Networks. He has been on the program/organizational committees of several international conferences. Dr. Mohapatra is the recipient of an Outstanding Engineering Alumni Award from Penn State University, an Outstanding Research Faculty Award from the College of Engineering at the University of California, and the HP Labs Innovation Research Award winner for three years. He is a Fellow of the IEEE and AAAS. Dr. Mohapatra’s research interests are in the areas of wireless networks, mobile communications, sensor networks, Internet protocols, and QoS. Dr. Mohapatra’s research has been funded through grants from the National Science Foundation, Department of Defense, Intel Corporation, Siemens, Panasonic Technologies, Hewlett Packard, Raytheon, Huawei Technologies, and EMC Corporation.

Highlights of “Prediction based Opportunistic Routing for Maritime Search and Rescue Wireless Sensor Network” •

For highly dynamic marine search and rescue wireless sensor network, we propose a new routing algorithm which makes some amendments to the existing opportunistic routing algorithm and implements a light weight time series based prediction method to select the node to forward the information.

•

We implemented our proposed routing protocols on NS2 simulator and conducted extensive experiments for evaluation, including Overall packets delivery rate and Average energy consumption per delivered packet.

•

NS2 based simulation Results: 30% more Packet Delivery Ratio is achieved compared to AODV protocol. Using AODV protocol 60% efficiency was achieved. Our approach gathered 90% efficiency. But an additional 2% more energy is consumed by the nodes. We feel additional 2% energy consumption to improve delivery greatly by 30% is a good tradeoff.