978-1-4577-0351-5/11/$26.00 c 2011 IEEE

Recycled ID Assignment for Relocation of Hopping Sensors

Moonseong Kim

Information and Communications Examination Bureau

Korean Intellectual Property Office

Daejeon, 302-701, Korea

Email: moonseong@kipo.go.kr

Matt W. Mutka

Department of Computer Science and Engineering

Michigan State University

East Lansing, MI 48824, USA

Email: mutka@cse.msu.edu

Abstract—Redundant mobile sensors might be moved inorder to cover sensing holes or replace power-exhausted sen-sors. Within rugged terrains, the use of hopping sensors maybe more suitable than wheeled mobile sensors. Since WSNcommunication is data-centric, globally unique ID allocationthat is used for MANETs is usually not applicable for WSNs. Arecent study classifies the locally unique ID allocation schemefor WSNs into two representative types: a proactive and areactive scheme. In the reactive scheme, energy preservationis improved because ID conflict resolution is delayed untildata communication is needed. Although a typical reactive IDassignment scheme is used, local uniqueness cannot be guar-anteed if hopping sensors are relocated. In order to overcomethe weakness, we propose the recycled ID assignment schemefor relocation of hopping sensors. Simulation results show thatthe proposed recycled ID assignment scheme outperforms thetypical reactive scheme for relocation of hopping sensors.

Keywords-wireless sensor netowrks; hopping sensors; reloca-tion of sensors; sensing holes; ID assignment;

I. INTRODUCTION AND PREVIOUS WORK

Redundant sensors are scattered initially over the entire

field using a well planned deployment; thus, if a sensing

hole is detected, a relocation of mobile sensors is needed

to recover the sensing hole. That is, mobile sensor nodes

could move to a specific emergent area, or replace energy-

exhausted sensor nodes. Recently, the discussion of reloca-

tion of hopping sensors has been further accelerated by [1]-

[3]. However, there is no research to support ID assignment

for the efficient communication after some hopping sensors

move to the desired areas. As a matter of course, the research

for supporting hopping sensor mobility could consider a

Mobile Ad-hoc NETwork (MANET). However, although a

WSN is similar to a MANET, because both are multi-hop

wireless communication, they are different in terms of data

communication architectures. Generally, a MANET is an

IP-based network; thus, every node has to need a globally

unique ID such as an IP address, and any node can try to

connect with any other node for initiating communication.

Unlike a MANET that is address-centric, the communi-

cation for a WSN is usually data-centric. Actually, a com-

munication of WSN is interested in the location and the re-

quested/collected data, but does not care about the address of

a sensor node. In addition, if all data is encapsulated within

a TCP/UDP/IP packet, it wastes energy and bandwidth,

since the payload length in the data packet is usually small.

Thus, customized network protocols may be adopted to save

the communication overhead and energy consumption. For

example, Directed Diffusion (DD [4]) might be considered

a representative reactive dissemination protocol for WSNs

instead of IP-based data communications. In order to collect

the desired data, a sink node first broadcasts an Interest

message that could be flooded to the entire network. Every

relay node records the upstream node as the next hop that

is back toward the sink. After the desired data is found out

at a source node and some paths are built, a reply message

is sent back from the source to the sink along the reverse

paths. Finally, the sink node can select a suitable path using

Reinforce and can unicast. One of the assumptions in DD is

that every node has a locally unique ID instead of a globally

unique ID.

Although there are many auto-configuration algorithms

proposed for ID assignment in MANETs, the goal of an

assignment is globally unique; hence, it is not appropriate for

ID assignment for WSNs. The authors of [5] propose the use

of Huffman coded address, called proactive ID assignment,

for locally unique ID assignment for WSNs. The size of

the proposed ID is not fixed and is usually less than that

of a fixed address format. Even if the use of a Huffman

coded ID decreases the average address field length, the

authors admit that there is no benefit in a practical dense

WSN. Moreover, a priori Huffman code table may not be

optimal, the Huffman code table must be stored in the

memory of the sensor node during all its lifetime, and it

would be much easier for designers to allocate the fixed-

length. Hence, the authors of [6] propose the use of fixed-

length address assignment, called reactive ID assignment, for

locally unique ID assignment for WSNs. They prove that the

proposed reactive ID assignment outperforms the proactive

one. However, since there is frequent movement of sensors

in hopping sensor networks, the conflict probability may be

high when the locally unique ID assignment is considered. In

this paper, therefore, we address the ID assignment problem

for hopping sensor networks.

The remainder of this paper is organized as follows.

Section II presents details of the proposed scheme. Section

III evaluates our proposal by employing a simulation; and

finally, Section IV concludes this paper.

II. ID ASSIGNMENT FOR HOPPING SENSOR NETWORKS

In this section, we discuss the ID assignment policy to

resolve the ID conflicts occurred when the hopping sensors

are relocated. In general, there are three types of hopping

strategies, , direct, relayed, and cascaded hopping [2].

Cluster heads are assumed to be capable of jumping high

and have redundant energy. The cluster heads cooperate

with each other in order to cover and maintain the entire

WSN. After a hopping sensor is appointed as a cluster

head, the cluster head can broadcast to its neighbor hopping

sensors included to the cluster. If it does not receive an Ack

message from its neighbor, then the cluster may be a sensing

hole; otherwise, the neighbor hopping sensors can send the

Ack message to the cluster head using jumping as high as

possible.

Figure 1. An example of relocation of hopping sensors between aconsumer and supplier

Here, we assume that the cluster head detects the sensing

hole through self-verification, and then the cluster head of

the sensing hole is called the consumer. The consumer

broadcasts the Interest message, including the number of

required sensors, to its neighbor cluster heads like the DD

paradigm. In order to communicate and avoid ID conflict

with each cluster head, the mentioned reactive ID assign-

ment [6] could be applied. After the candidate paths are

constructed between the consumer and some suppliers, the

consumer could choose one or more suitable suppliers,

as shown in Fig. 1. After the requested hopping sensors

are completely moved to the neighbor cluster, each moved

hopping sensor must first try to communicate with the

current cluster head. They are also able to communicate with

each other hopping sensor with/without jumping, , multi-

hop communication. Here, first and most importantly, the

resolution to avoid ID conflict is needed; thus, the reactive

ID assignment is also used.

The proposed address format is as follows briefly:

Random number for communication between clusters

Random number for intra communication . Let be

the address format. For instance, if the hopping sensors are

in the same cluster, then all of the sensors’ addresses are

same. To satisfy 1-hop uniqueness as address uniqueness

among direct neighbor clusters, the range of should be

greater than 3 bits ( 6=2.58), as shown in Fig. 1, if

the cluster model is considered as a hexagon. Moreover, to

satisfy 2-hop uniqueness, the range of should be greater

than 5 bits ( 18=4.16). Finally, the range of should

be greater than , where is the average number of

hopping sensors per each cluster initially deployed, , the

density of hopping sensor nodes per a cluster.

(a) The typical reactive ID assignment

(b) The proposed recycled ID assignment

Figure 2. Scenarios for description of the recycled scheme reactivescheme under relayed movement

Fig. 2 shows a simple scenario to compare with the previ-

ous reactive scheme for relocation of hopping sensors under

relayed movement. The cluster head is currently recording

the addresses of inner hopping sensors. As illustrated in

Fig. 2 (a), there is a duplicated ID after a hopping sensor

is moved. According to the typical reactive ID assignment

scheme, a resolve process may be started. On the other

hand, we generally use relayed or cascaded movement for

relocation of hopping sensors. This means that the number

of incoming hopping sensors is likely more than or equal to

the number of outgoing hopping sensors. Therefore, if the

cluster head can keep the addresses of outgoing hopping

sensors and reallocate the ones to the incoming hopping

sensors, then the overall overhead and the spent time could

be reduced, as described in Fig. 2 (b). However, there is

possibility that a sensor cannot move out. Since the proposed

recycled ID allocation is applied, there must be an ID

conflict. In order to resolve the problem, the failure sensor

should regenerate its address properly.

III. PERFORMANCE EVALUATION

We analyze some numerical results that can be used for

comparing the performance of the proposed recycled ID

assignment and the typical reactive ID assignment schemes.

We simulate them in perspective of the impact of ID conflict

after the requested hopping sensors are relocated. Hence, for

simplicity, we do not take into account the current cluster

address of . We assume that the size of is set

to 6 bits, =64; therefore, the number of hopping sensors

initially deployed per each cluster could be assumed to 30

and 40. Each ID is randomly and distinctly chosen for each

hopping sensor initially.

0 10 20 30 40 50

0

10

20

30

40

50

60

Number of incoming hopping sensors (%)

Reactive ID assignmentRecycled ID assignment

Figure 3. Communication overhead in terms of the number of incomingsensors

0 10 20 30 40 50

0

20

40

60

80

100

120

140

Number of incoming hopping sensors (%)

Reactive ID assignment w. pf=0.0

Reactive ID assignment w. pf=0.2

Reactive ID assignment w. pf=0.8

Recycled ID assignment w. pf=0.0

Recycled ID assignment w. pf=0.2

Recycled ID assignment w. pf=0.8

Figure 4. Communication overhead for the case of relocation failure

Fig. 3 shows the total number of packets sent, as the

requested hopping sensors migrate to the new cluster. 30

hopping sensors initially deployed per each cluster. The x-

axis represents the number of incoming hopping sensors. For

instance, 10% means that 3 hopping sensors (30 10%=3)

are requested from the direct neighbor cluster. As might be

expected, the advantage of using the proposed recycled ID

assignment versus the typical reactive ID assignment can be

clearly seen in these graphs.

In Fig. 4, we take into account that means the probabil-

ity of failure when the hopping sensor moves out under the

relayed hopping environment. Here, the number of hopping

sensors initially deployed per a cluster is 40. We easily note

that the recycled ID assignment scheme outperforms the

typical reactive scheme as the failure probability increases.

As a result, in rugged terrains, the hopping sensor is suitable;

and the proposed recycled ID assignment scheme is also

appropriate for communications.

IV. CONCLUSION

Mobile sensors are able to move to a specific emer-

gent area, or replace power-exhausted sensors. The use

of hopping sensors is more adaptable to the rugged area

than wheeled mobile sensors. Since the communication for

WSNs is data-centric, it is sufficient to assign locally unique

IDs for WSNs. In this paper, we have studied about the

ID assignment problem for relocation of hopping sensors.

Actually, under the relayed or cascaded movement for the

relocation, the number of incoming hopping sensors may

be more than or equal to the number of outgoing hopping

sensors. Therefore, if the addresses of outgoing hopping

sensors are recycled for the incoming hopping sensors, then

the overall overhead can be reduced. We have proved that the

performance of the proposed recycled ID assignment scheme

is superior to the typical reactive scheme for relocation of

hopping sensors.

ACKNOWLEDGMENT

This research was supported in part by NSF (USA) Grants

No. OCI-0753362 and CNS-0721441.

REFERENCES

[1] Z. Cen and M. W. Mutka, “Relocation of Hopping Sensors,”IEEE ICRA 08, pp. 569-574, May 2008.

[2] M. Kim and M. W. Mutka, “Multipath-based RelocationSchemes Considering Balanced Assignment for Hopping Sen-sors,” IEEE/RSJ IROS 09, pp. 5095-5100, October 2009.

[3] M. Kim, M. W. Mutka, and H. Choo, “On Relocation ofHopping Sensors for Rugged Terrains,” IEEE ICCSA 10, pp.203-210, March 2010.

[4] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, andF. Silva, “Directed Diffusion for Wireless Sensor Networking,”IEEE/ACM Transactions on Networking, vol. 11, no. 1, pp. 2-16, February 2003.

[5] C. Schurgers, G. Kulkarni, and M. B. Srivastava, “DistributedOn-demand Address Assignment in Wireless Sensor Net-works,” IEEE Transactions on Parallel and Distributed Sys-tems, vol. 13, no. 10, pp. 1056-1065, October 2002.

[6] H. Zhou, M. W. Mutka, and L. M. Ni, “Reactive ID Assign-ment for Wireless Sensor Networks,” International Journal ofWireless Information Networks, vol. 13, no. 4, pp. 317-328,October 2006.