Hopping Enhanced Sensors for Efficient SensorNetwork Connectivity and Coverage

Fernando J. CintronComputer Science and Engineering

Michigan State University3115 Engineering Building

East Lansing, MI 48824-1226 USAEmail: cintrong@cse.msu.edu

Matt W. MutkaComputer Science and Engineering

Michigan State University3115 Engineering Building

East Lansing, MI 48824-1226 USAEmail: mutka@cse.msu.edu

Abstract—This work presents the organization and operationof a hopping sensor wireless network. It defines two decentralizedalgorithms for the discovery of isolated nodes, aggressive andsmart discovery, to propagate base station connectivity. Bothalgorithms rely on the usage of hopping sensors on the boundaryof a cluster, without prior topology information. We studiedthe efficiency of the algorithms in terms of network energyconsumption and connectivity propagation. A cluster to cluster(C-to-C) packet forwarding scheme, which relies upon boundaryhopping sensor gateway nodes, is defined and simulated, andproves to have remarkable low energy consumption. In-fieldmeasurements were performed to obtain real data to serve asinput parameters for simulations. Simulation results indicatesignificant savings in network wide energy consumption (57%for aggressive discovery and 79% for smart discovery).

I. INTRODUCTION

Network connectivity in a wireless sensor network (WSN)is crucial for network operation. When wireless sensor nodesare randomly deployed in a field, there may be areas that lacksensor nodes. These areas tend to create communication holes,and in the worst case, break the topology into disconnectedclusters of nodes, i.e., disconnected groups of interconnectednodes.

The problem of communication holes in WSN can beaddressed by different approaches. An intuitive approach isto increase the node density, i.e., amount of nodes per unitarea. This approach provides a feasible solution for noninvasive tasks, which can be performed with relatively cheaptechnology. However, leading edge technology demandingapplications are expensive, given that the price per node tendsto be higher. On the other hand, node relocation provides theopportunity to use those nodes that are redundant in someareas to cover communication holes. However, relocation is notpossible in difficult to navigate terrains such as river stream,lakes, or harmful areas. As an alternative to reach fartherwireless nodes, signal transmission power can be amplified.Nevertheless, many topology control protocols for WSN usemaximum transmission power of the nodes at their initialphases, and still experience connectivity problems, due to thepresence of communication holes.

This work is supported in part by NSF Grants No. OCI-0753362, CNS-0721441 and CNS-0551464.

Complete network connectivity within a random wirelesssensor network topology is difficult to guarantee. For thescope of this work, network connectivity is defined asthe percentage of nodes able to reach a base station (BS),through a direct or indirect path. Figure 1 depicts a WSNtopology of 200𝑚 𝑥 200𝑚 area with a single BS located atthe middle of the bottom edge; nodes with connectivity aremarked with diamond shapes, and nodes without connectivityare marked with asterisks. All the nodes are transmitting atmaximum power, however due to communication holes, only31% network connectivity is achieved.

0 20 40 60 80 100 120 140 160 180 2000

20

40

60

80

100

120

140

160

180

200

Fig. 1. WSN Topology at ground level (31% connectivity)

Hopping sensors, while providing relocation capabilities,have an increase in communication range when airborne. Atypical sensor node (IRIS Crossbow) was measured to commu-nicate up to 10𝑚 at surface level when maximum transmissionpower was used, whereas it reaches 40𝑚 when elevated toa height of 1𝑚, without having packet losses [1]. As aresult, hopping sensors can be used to increase connectivityin wireless sensor networks. However, a topology of hoppingsensor nodes configured to enhance network connectivity bymeans of jumping can significantly reduce the network lifespan, if not well managed. This work presents an approach

978-1-4244-7489-9/10/$26.00 ©2010 IEEE 119

to deal with communication holes, to enhance the networkconnectivity using selected hopping sensors nodes located atthe boundaries of clusters of nodes. With the help of selectedhopping sensors on disconnected clusters, 100% of networkconnectivity is achieved in the same topology (Figure 1), andis depicted in Figure 2. In the figure, wireless links establishedby hopping sensors are depicted as dotted lines.

0 20 40 60 80 100 120 140 160 180 2000

20

40

60

80

100

120

140

160

180

200

Fig. 2. WSN Topology with gateway hopping sensors (100% connectivity)

The main contributions of this paper include two decen-tralized algorithms that use the hopping capabilities of sensornodes for the discovery of neighbor clusters, Boundary NodeAggressive Neighbor Discovery (aggressive) and BoundaryNode Smart Discovery (smart). Both algorithms rely on the us-age of boundary hopping sensor nodes, without prior topologyinformation. An energy balanced packet forwarding schemewith the use of hopping sensors gateways is presented forcluster-to-cluster (C-to-C) communication.

Hopping sensors related work and existing techniques forboundary nodes identification are reviewed in Section II.Topology organization is described in Section III. The dis-covery of neighbor node’s clusters algorithm is presented inSection IV. Section V enhances the discovery algorithm tobe energy and communication range area aware. Support datafrom hopping sensor experiments and simulation results arepresented on Section VI. Section VII concludes this work.

II. RELATED WORK

Hopping sensors nodes provide a unique opportunity tocover areas where wheeled sensors cannot perform. Researchinvolving hopping sensors has been focused on their re-location, energy impact, and more recently on their com-munication advantages over other sensors. Cen, et al. [2]discussed hopping sensor relocation, considering the impactof air disturbance and movement accuracy. Similarly, Pei, etal. [3] provide a relocation algorithm for hopping sensors inan obstacle abundant terrain. Cintron, et al. [1] introduced an

energy balancing network model to maximize coverage withthe use of hopping sensors.

The results presented by Cintron et al. [1] demonstrate anenhance in wireless communication range by relying on thechanges in elevation while a hopping sensor jumps. Utilizingthis we can have a selected set of sensors to discover neigh-boring cluster of nodes. Cluster’s boundary nodes providethe highest successful discovery probability because of theirposition relative to others. Therefore, selecting only boundarynodes to perform discovery, increases the probability of dis-covery, and reduces the network energy cost. The identificationof boundary nodes has been addressed by Sahoo et al. [4], andFayed and Mouftah [5].

Two kinds of boundary detection algorithms were intro-duced by Sahoo et al. [4]: sequential (SBNS) and distributed(DBNS). Furthermore, using these boundary nodes, a targetdetection protocol is used to know the entry and the exit ofsingle target through the monitoring region. In selecting theboundary nodes, the two algorithms have similar performance.The main differences between them is the lower communica-tion cost of DBNS compared to SBNS due to the less amountof information exchange and the faster selection procedure inDBNS regardless the increase of network size.

Fayed and Mouftah [5] proposed a local convex view(lcv) as a means to identify nodes close to the boundary.It is a localized algorithm and can establish neighborhoodcoordinates if location information is not prior available. Tocheck if a node is on the boundary, a node is verified tolie on the convex hull built with its one-hop neighborhood.The most important result is that lcv seems unaffected byposition estimation error because of its geometric properties.However, it has some difficulties identifying boundary nodesunder special scenarios where there are areas without sensors(communication holes) within the topology.

Boundary nodes have shown to be beneficial in wirelessnetwork topologies. They have been used to help other nodesin the localization process on wireless sensor networks [6].Another approach presented by Aissani et al. [7] employsboundary nodes to avoid routing packets to dead-end routes.Han and Poor [8] introduced a method based on coalitiongames, in where boundary nodes are used to cooperate withnodes in the middle of the network to enhance networkperformance and connectivity.

Research to define protocols and automated behavior onhopping sensors is still an open field of study. This workintroduces a novel approach to overcome network connectivityproblems due to communication holes in a hopping sensornetwork. First, the wireless sensor network topology organi-zation is established by identifying node’s clusters. Second, thediscovery of disconnected neighbor node’s cluster is performedwith boundary nodes, to propagate connectivity. Finally, C-to-C packet forwarding is established and distributed amongboundary gateway hopping sensors.

120

III. HOPPING SENSOR NETWORK

This section introduces the hopping sensor network topol-ogy. It also presents a distributed wireless sensor networkorganization to establish a functional network, without initialtopology information.

A. Topology

The deployment phase and characteristics of the geographicarea impacts a wireless sensor network topology greatly. Acommon method of sensor deployment is to locate sensorsprecisely in the field to guarantee sensor connectivity with abase station, while maximizing the covered area. This methodbecomes infeasible when the network scales to cover difficult-to-reach or hazardous areas, e.g., volcano monitoring [9]. Asan alternative, sensor motes are deployed from the air; atechnique sometimes referred as air-dropped-sensors, wheresensors are carried and dropped by a helicopter, or an airplaneto the target area. This technique gives the opportunity to reachthose difficult-to-reach or hazardous areas. However, there ischance for sensors to land into segregated groups of sensorsthat may not have a connectivity path to the base station.

We define a cluster of sensor nodes as a group of in-terconnected nodes relying on their wireless communicationrange at ground surface level, and packet routing capabilities.A cluster can have direct connectivity, i.e., one or morenodes from the cluster can reach directly the base station,and provide connectivity to others within the cluster. Similarly,the connectivity propagation concept applies between clusters.A cluster with direct connectivity will provide connectivityto its neighbor clusters, and those will continue in turn toform a cluster-to-cluster (C-to-C) path to base station. C-to-C connectivity propagation will be discussed further inSection IV.

With the existence of clusters in the topology, we canlabel those sensor nodes that lay on the edge (frontier) ofa cluster as boundary nodes. These nodes will help in thediscovery of neighbor clusters and form bridge nodes, i.e.,those boundary nodes that can successfully communicate withnodes from other clusters, by employing “hopping sensorairborne communication” [1]. There may be more than onebridge node that shares the same intended cluster receiver.Moreover, a cluster may be able to reach base station troughdifferent clusters. In order to maximize network functionallifetime, a cluster can distribute the packet forwarding taskamong bridge nodes, by selecting a subset of bridge nodes asgateways to perform C-to-C packet forwarding.

B. Distributed Organization

Once sensor nodes are deployed, a topology organizationhas to be built. This process is organized in the phasesdescribed below.

1. Node Discovery: Every node will discover and exchangeneighborhood information, such as, node ID (NID), connectiv-ity with base station, cluster ID (CID) belonging to (initiallyits own NID), and obtain or compute additional informationas needed, e.g., distance with respect to one-hop neighbors.

2. Cluster Boundary Nodes: Every cluster has a set ofboundary nodes. Therefore, the identification of boundarynodes is performed locally on each cluster. The discussion ofthe algorithm to discover boundary nodes is out of the scopeof this work, for insight details please refer to the related workSection II.

3. Isolated Cluster Discovery: The discovery of isolatedclusters, i.e., those clusters without direct connectivity, isperformed by the boundary nodes on each cluster. A discoveryalgorithm for this procedure is discussed in Section IV, andenhanced in Section V.

As sensor nodes deplete their energy, the remaining sensornodes in the network iterate through these phases to adapt totopology changes.

IV. BOUNDARY NODE AGGRESSIVE NEIGHBOR

DISCOVERY

This section presents a decentralized algorithm that utilizesall the hopping sensor boundary nodes (BNs) from clusters,to jump and discover surrounding neighbor clusters.

In order to propagate the connectivity on the wirelessnetwork, this process begins on the cluster with direct con-nectivity, and continues on other clusters as they becomediscovered, in a sequential manner.

Every BN within a cluster will perform the cluster dis-covery, which consists of hopping (jumping), transmitting adiscovery message and listening for any reply while airborne.In order for a BN to perform cluster discovery, it has to followthe following phases: Pre-hop, Hopping, Post-hopping.

Pre-hop: Before a BN hops, it has to coordinate with itsone-hop BN neighbors, to avoid these BNs to hop at the sametime and interfere with each other.

Every BN waits for a random amount of time before tryingto hop and discover. Once this time is expired, the BNsends a Hop Advertise packet to its one-hop BN neighbors,including its NID. Those BNs receiving this packet, reply witha Hop Ack packet, including their NID and the NID of theBN initiating the discovery. These BNs constrain themselvesto perform hopping discovery while the current discovery istaking place. Upon receive of these Hop Ack packets, the BNdiscovery initiator verifies if all its BN neighbors replied andproceeds to hop.

One or more BN neighbors may fail to generate theHop Ack packet, due to a packet loss or due to competitionfor hopping at the same time. To address the case of missingacknowledgement due to packet loss, the BN initiator waits arandom amount of time and resends a Hop Advertise packetto those missed BN neighbors, who in turn will reply with aHop Ack packet.

A similar procedure applies if the BNs were competing fordiscovery. Each competing BN is set to wait before attemptingto re-advertise, to provide a chance for the other competitor togenerate the Hop Advertise packet first. Upon receipt of theadvertisement packet, any competitor yields and replies with aHop Ack packet. The winning BN proceeds to hop, and thoseBN neighbors that exclusively received the advertisement

121

packet from BN competitors and not from the winning BN,will reset their wait times to start competing among them.

Hopping: While the BN is airborne, it transmits a Discov-ery Msg packet, including its CID and connectivity informa-tion, i.e., direct connectivity, or the number of intermediaryclusters to reach the base station. Any BN belonging to anothercluster that is able to receive this packet will attempt toreply with a Discovery Reply packet, including its CID andconnectivity information.

Post-hopping: BNs on each cluster with discovery infor-mation become bridge nodes, and advertise it to the BNsof the cluster with a Bridge Advertise packet, including itsNID, node remaining energy, neighbor CID, and connectivityinformation of the neighbor cluster. In addition, clusters thatare discovered will start to perform discovery, to propagateconnectivity.

After the current discovery process finishes on a BN, the BNneighbors resume their wait timers for discovery. Those BNsthat performed the discovery will not compete for discoveryuntil a new discovery process is triggered, i.e., in the eventa cluster gets divided due to communication holes resultingfrom missing nodes. It is worth noting that after a hoppingnode jumps, it may be possible (depending on the geographicarea) to land in a position within the communication range ofthe two or more clusters at surface level; in such case, clustersare merged into one.

V. BOUNDARY NODE SMART NEIGHBOR DISCOVERY

This section enhances the discovery algorithm to findsurrounding neighbor clusters with a subset of BNs. Thesubset selection of BNs is intended to balance the energyconsumption in order to prolong network life.

The approach relies on the fact that once a BN is selectedto perform discovery, its one-hop BN neighbors will not con-tribute much to expand the discovered area. Figure 3 depictsthe gain in discovered area by only selecting BN that are notone-hop neighbors of the BNs already selected to performjump discovery. The communication range of BN hoppingsensors at ground level is represented by a solid line circle,while the communication range when hopping (jumping) isrepresented with a dotted line circle.1 From Figure 3 it is alsodepicted the resulting missed area by not selecting the one-hopBN between the selected BNs. However, the contribution tothe communication area of the missed area, in most of thecases, is minimal.

In order to provide a BN subset selection, the process beginsfrom the cluster having direct connectivity, where all BNscompete to perform discovery, similarly to the AggressiveNeighbor Discovery, discussed in Section IV. The main dif-ferences are on the BN behavior during the Pre-hopping andPost-hopping phases.

In the Pre-hopping phase, once a BN sends aHop Advertise packet, all its one-hop BNs will not compete

1Figure 3 is meant for visualization purpose only and is not drawn to scale.Hopping communication range has been measured to increase by a factor of6 when hopping, this is discussed further in Section VI-A.

Fig. 3. Discovered Area Gain of BNs Subset Selection for Discovery

for further discovery until another discovery event is required.BNs that did not receive an advertisement packet will continuecompeting for discovery by repeating the process. In thismanner, a subset of BNs is selected to perform the discovery.

During the Post-hopping phase, if the discoverer BN doesnot reach any other cluster, it sends a Discovery Null packet toits one-hop BN neighbors, and stops competing to do furtherdiscoveries until a new discovery event is triggered. Uponreceipt of this packet, one-hop BN neighbors will reset theirwait timer to compete for discovery. On the other hand, ifa cluster is discovered, the BN advertises the discovery toall the BNs with the Bridge Advertise packet. Those one-hopBN neighbors of the discoverer BN will stop competing fordiscovery. For the discovered cluster, the set of BNs able tosuccessfully receive the Discovery Msg packet become bridgenodes. These nodes advertise their new status to all the BNswith the Bridge Advertise packet. BNs that are a one-hopneighbors of BNs that became bridge nodes will refrain fromdiscovery until a new discovery process is triggered.

VI. EXPERIMENTAL DATA AND SIMULATION

In order to simulate and test the discovery algorithmsand the C-to-C packet forwarding performance, experimentswith IRIS Crossbow sensors were performed to obtain realcommunication range values on an outdoor field.

A. Experimental Data

IRIS Crossbow sensors were used in conjunction with ahopping emulator mechanism, able to produce a vertical jumpof 1 meter height, with an accuracy error of ±0.05𝑚. TheIRIS sensor was programmed to transmit packets of 36 bytes at20𝑚𝑠𝑒𝑐 rate intervals. Maximum transmission power (3𝑑𝐵𝑚)was used. An IRIS sensor attached to a base station boardinterface, connected to a laptop computer, was programmedas a sink to collect the packets.

When the hopping sensor is at ground level, it was foundto have an average wireless communication range of 10

122

meters. Therefore, we study the transmission of packets froma hopping sensor (while hopping to a height of 1 meter) toa sink from 25 meters horizontal distance onward, to avoidpossible reception at ground level.

Fig. 4. Airborne Transmitted Packets over Distance at 1 Meter Hop

Figure 4 shows the amount of packets received from thehopping sensor while airborne for each horizontal distance.Since the hopping height is constant, as the sensor is fartheraway from the intended receiver there is less available “time offlight” for the hopping sensor to communicate [1]. Hence, wecan observe a decrement in the amount of packets successfullytransmitted as the distance between sensors increases.

Experiment Analysis: IRIS sensors use 2.4𝐺𝐻𝑧 IEEE802.15.4 standard, and have a 250kbps data rate radio. Onepacket of 36 bytes would take 1.2𝑚𝑠𝑒𝑐 to be transmittedplus an additional 20𝑚𝑠𝑒𝑐 wait time before transmitting thenext packet, as we configured. With the hopping emulationmechanism, the time to perform 1 meter hop (measured fromgoing up, to touching ground once) is approximately 1 second,with some additional time due to the bouncing effect beforelanding, and hopping mechanism inaccuracy.

Let us consider 1 second transmission at groundlevel, for each packet taking 21.2𝑚𝑠𝑒𝑐; then, 47 packets(1𝑠𝑒𝑐/21.2𝑚𝑠𝑒𝑐) can be transmitted. The improvement incommunication range on ground level for IRIS sensors can bequantified by comparing the obtained hopping and transmittingexperiment results. Results show a 200% improvement on thecommunication range, i.e., a communication range increasefrom 10 to 30 meters, considering the packet throughput(greater or equal to 47 packets on average), with one meterhop height.

B. Simulation Setup and Results

Simulations were performed to test the discovery algorithmsin terms of energy consumption and connectivity propagationfor a hopping sensor network. The life span of the networkwas tested with different configurations of C-to-C packetforwarding. Experimental data results were used as input forthe communication ranges, 10 meters ground level commu-nication, and 30 meters for airborne communication. Theairborne communication range limit of 30 meters was selectedas a conservative measure to account other factors, such asterrain and weather conditions, that can affect communication.

Topology: Twenty (20) sensor network topologies weregenerated representing a geographic area of 200𝑚 𝑥 200𝑚.Nodes were positioned following a random uniform distribu-tion. Each node was simulated to provide sensing coverage toa circular area with a 7 meters radius. The number of nodesin the network was selected to provide a sensing coverage of3 nodes per unit area (1 square meter), resulting in 781 nodes(all hopping sensors). A base station node was located in themiddle of one of the edge of the square topology acting as asink.

Energy: Each node was simulated to have the same amountof energy as the hopping sensor prototype built by Zhao etal. [10]. It uses a 3.7𝑉 , 100𝑚𝐴ℎ LiPo battery, for a totalstored energy of 1332𝐽 . Their prototype can make jumps upto a height of 0.353 meters, consuming 882.6𝑚𝐽 of energy foreach jump. A similar design of a hopping sensor able to jump1 meter height, is estimated to consume 1.775𝐽 of energy. Forsimulation purposes, we considered an energy consumption of2𝐽 for the jump operation. Other node’s energy consumptionoperations were determined using IRIS data sheet [11] asfollow: energy to transmit a packet of 36 bytes at maximumtransmission power (3𝑑𝐵𝑚) is 0.09𝑚𝐽 , and the energy toreceive a packet is rounded to be the same as to transmit,since the different is relative small.

Discovery Algorithm Performance: To study the discoveryalgorithms performance, we analyzed energy consumptionand connectivity propagation on the 20 network topologiesafter initial discovery. The algorithms were compared withall hopping sensors randomly employing individual discovery(referred as ALL). Table I presents the mean connectivityof network after the discovery phase, and the mean networkenergy cost in Joules for each algorithm. It can be appreciatedthat as the discovery algorithms become smarter with the se-lection of nodes to perform discovery, the energy consumptionreduces without compromising connectivity.

TABLE IDISCOVERY ALGORITHM PERFORMANCE

Algorithm Connectivity % Energy Consumption (J)ALL 100 1562.0AGGRESSIVE 100 670.5SMART 100 330.6

Network Performance: The network performance wasanalyzed over 5000 time period events for each of the networktopologies. For every time period, a percentage of the nodeswith connectivity was randomly selected to generate packets.Simulations were run separately for five different percentagesof nodes generating packets (communication loads): 20%,10%, 5%, 2%, and 1%.

Boundary C-to-C packet forwarding was employed withhopping sensor acting as gateway nodes to reach the basestation. A delay transmission protocol was employed forhopping sensors airborne transmission. Packets arriving atgateways (nodes that need to jump and transmit) were buffereduntil transmission buffer became full, but not for more than 3time periods after arrival to the node, then transmitted airborne.

123

(a) Remaining Network Energy for Different Communication Loads (b) Remaining Network Connectivity for Different Communication Loads

(c) Mean Packet Delay Time for Different Communication Loads

Fig. 5. Network Performance Evaluation Metrics After 5000 Time Periods Simulation

Two configurations were simulated to evaluate the networkperformance employing C-to-C packet forwarding: single gate-way per cluster, and multiple gateways per clusters. Theywere compared with packet forwarding without gateways (no-cluster organization). Single gateway per clusters is an energypreserving approach, in where every discovered cluster elects asingle gateway node to forward packets to neighbor clusters,to reach the base station. Shortest path and remaining nodeenergy were used as the gateway selection criteria, smallestNID was used to break any tie. Multiple gateways per clusteris an extension of the single gateway approach, but less energyconservative. It provides alternative routes to nodes in a clusterto obtain shorter paths length to reach base station troughdifferent gateways. With packet forwarding without gateways,every hopping sensor node takes advantage of reaching thefarthest nodes with airborne communication to attempt shortestpaths routes to base station. It is worth noting that if ahopping sensor is not able to shorten the path with airbornecommunication, then it employs ground level communication.

Each C-to-C packet forwarding configuration was evaluatedwith three metrics:

1) Remaining Network Energy, mean of the remainingenergies in all the nodes, at the end of simulation, with

respect to the total initial network energy.2) Remaining Network Connectivity, percentage of nodes

left with connectivity to BS.3) Mean Packet Delay Time, mean of the time periods that

each packet required to arrive at BS.

Figures 5(a) and 5(b) present the remaining network energyand connectivity results, respectively, for the different com-munication loads after 5000 time periods. Single gateway (S.GTW) C-to-C packet forwarding shows to have a remarkablelower energy consumption, due to the fact that for any timeperiod the number of sensors jumping to forward packets inthe network is bound by the number of clusters without directconnectivity, thus, consuming less energy. Multiple gateways(M. GTWs) C-to-C packet forwarding consumed more energy,but proved to be more efficient than packet forwarding withoutgateways (No GTWs). Connectivity results show to be highlycorrelated with network energy for a hopping sensor topology;as more nodes deplete their energy, other dependent nodes areleft without connectivity.

A trade-off in energy efficiency and packet delay time isevident in Figure 5(c). Packet delay time for Single GTWshowed to be constant and slightly higher than the others,for communication loads 1%, 2%, and 5%, whereas rapidly

124

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000

10

20

30

40

50

60

70

80

90

100

Periods

Net

wor

k E

nerg

y %

S. GTWM. GTWsNo GTWs

(a) 20% Nodes Generating Packets

0 500 1000 1500 2000 2500 3000 3500 4000 4500 500010

20

30

40

50

60

70

80

90

100

Periods

Net

wor

k E

nerg

y %

S. GTWM. GTWsNo GTWs

(b) 10% Nodes Generating Packets

0 500 1000 1500 2000 2500 3000 3500 4000 4500 500030

40

50

60

70

80

90

100

Periods

Net

wor

k E

nerg

y %

S. GTWM. GTWsNo GTWs

(c) 5% Nodes Generating Packets

0 500 1000 1500 2000 2500 3000 3500 4000 4500 500060

65

70

75

80

85

90

95

100

PeriodsN

etw

ork

Ene

rgy

%

S. GTWM. GTWsNo GTWs

(d) 2% Nodes Generating Packets

0 500 1000 1500 2000 2500 3000 3500 4000 4500 500075

80

85

90

95

100

Periods

Net

wor

k E

nerg

y %

S. GTWM. GTWsNo GTWs

(e) 1% Nodes Generating Packets

Fig. 6. Network Energy Consumption During Simulation Time Periods for Different Packet Loads

increasing for higher communication loads, due to networksaturation and bottleneck effect created from gateways. Mul-tiple GTWs is more resilient to delay incurred with increasedcommunication load, since it provides alternative routes tonodes within each cluster.

The energy consumption rates of the network for the dif-ferent communication loads are presented in Figure 6. SingleGTW packet forwarding shows the smallest energy consump-tion ratio, followed by Multiple GTWs packet forwarding.For 20% communication loads (Figure 6(a)) is noted thatafter 3000 time periods there a is an increase in the energyconsumption rate for Multiple GTWs. This change in rateis due to more gateways being used concurrently due topacket traffic on the network. The energy consumption rateis reduced later (around 3500 time period), which is a resultof the successful reconfiguration of the network to allocate

new gateways. With packet forwarding without gateways (NoGTWs), most of the network nodes have depleted their energyby this time. However, the network energy is not shown toreach zero since for simulation purposes, a node was considerdead after reaching the threshold energy value of 5𝐽 . A similarpattern is seen for 10% communication loads (Figure 6(b))with slightly lower energy consumption rate. The same in-crease in energy consumption rate is seen again for MultipleGTWs and No GTWs, but occurring later in time (around4250 time period) due to the less packet traffic generation.The energy consumption rates for lower communication loads(Figures 6(c), 6(d), and 6(e)) were comparatively lower to theothers, as expected.

125

VII. CONCLUSION

Hopping sensor network topologies provide a viable solu-tion to overcome communication holes. Both of the discoveryalgorithms defined, aggressive and smart, showed to increasethe total network connectivity to 100% for each topologyscenario. Simulation results show a great reduction in theenergy consumption for the boundary node neighbor discoveryalgorithms, a 79% network energy consumption savings forsmart discovery, whereas a 57% energy savings for aggressivediscovery. Results from C-to-C packet forwarding protocolemploying hopping sensors as gateways show remarkable lowenergy consumption, which prolongs network useful life.

REFERENCES

[1] F. J. Cintron, K. Pongaliur, M. W. Mutka, and L. Xiao, “Energybalancing hopping sensor network model to maximize coverage,” in TheIEEE 18th International Conference on Computer Communications andNetworks (ICCCN 2009), August 2009, pp. 1–6.

[2] Z. Cen and M. Mutka, “Relocation of hopping sensors,” in IEEEInternational Conference on Robotics and Automation, May 2008, pp.569–574.

[3] Y. Pei, F. J. Cintron, M. W. Mutka, J. Zhao, and N. Xi, “Hopping sensorrelocation in rugged terrain,” in The IEEE/RSJ International Conferenceon Intelligent RObots and Systems (IROS 2009), October 2009.

[4] P. Sahoo, K.-Y. Hsieh, and J.-P. Sheu, “Boundary node selection andtarget detection in wireless sensor network,” in Wireless and OpticalCommunications Networks. WOCN ’07. IFIP International Conferenceon, July 2007, pp. 1–5.

[5] M. Fayed and H. T. Mouftah, “Localised convex hulls to identifyboundary nodes in sensor networks,” Int. J. Sen. Netw., vol. 5, no. 2,pp. 112–125, 2009.

[6] X. Du, D. Mandala, W. Zhang, C. You, and Y. Xiao, “A boundary-nodebased localization scheme for heterogeneous wireless sensor networks,”in Military Communications Conference. MILCOM 2007. IEEE, Oct.2007, pp. 1–7.

[7] M. Aissani, A. Mellouk, N. Badache, and M. Djebbar, “A new approachof announcement and avoiding routing voids in wireless sensor net-works,” in Global Telecommunications Conference. GLOBECOM 2008.IEEE, Dec. 2008, pp. 1–5.

[8] Z. Han and H. Poor, “Coalition games with cooperative transmission:A cure for the curse of boundary nodes in selfish packet-forwardingwireless networks,” Communications, IEEE Transactions on, vol. 57,no. 1, pp. 203–213, January 2009.

[9] W.-Z. Song, R. Huang, M. Xu, A. Ma, B. Shirazi, and R. LaHusen, “Air-dropped sensor network for real-time high-fidelity volcano monitoring,”in MobiSys ’09: Proceedings of the 7th international conference onMobile systems, applications, and services. New York, NY, USA: ACM,2009, pp. 305–318.

[10] J. Zhao, R. Yang, N. Xi, B. Gao, X. Fan, M. W. Mutka, and L. Xiao,“Development of a miniature self-stabilization jumping robot,” in TheIEEE/RSJ International Conference on Intelligent RObots and Systems(IROS 2009), October 2009.

[11] “Iris wireless measurment system datasheet,” Crossbow. [Online].Available: http://www.cse.msu.edu/rgroups/elans/docs/iris-datasheet.pdf

126