You are on page 1of 2

2016

Fault-tolerant Connectivity Restoration of a Partitioned Wireless Sensor Network


Sookyoung Lee, Meejeong Lee*
Ewha Womans University
sookyoung,lmj@ewha.ac.kr
Abstract
Wireless sensor networks (WSNs) deployed in harsh environments are susceptible to simultaneous failure of multiple
collocated nodes which partitions the network into multiple disjoint segments. Connectivity restoration among the
segments is essential in order to make the WSN operate. In addition, provisioning spare connections would be
desirable to proactively avoid potential segmentation caused by a single node failure in the repaired topology. In this
paper, we present an optimized fault-tolerant connectivity restoration algorithm that establishes two-vertex disjoint
paths between every pair of segments by exploiting relay nodes while minimizing the number of the required relays.
The proposed algorithm opts to identify a simple cycle in which every segment is visited exactly once. Then the
least count of relays are populated with respect to its communication range along the steinerized edges of the
resulting bi-connected topology. The performance of the proposed algorithm is validated through simulation
experiments.
Keywords-Wireless sensor networks, Partitioning, Fault-tolerance, Connectivity Restoration, Relay Node Placement

load balancing
connectivity.

. Introduction
In recent years, growing interest has been drawn to
WSNs due to their numerous applications, especially
those serving in hostile environments such as searchand-rescue and battlefield reconnaissance [1]. In such a
harsh setup, a WSN is exposed to hazardous conditions
such as explosives, avalanches, or landslides and
susceptible to simultaneous failure of multiple collocated
nodes which causes the network to split into disjoint
segments.
The focus of this paper is on restoring connectivity in
the partitioned network and additionally tolerating future
single node failures in the recovered topology by
exploiting relay nodes. The rationale for such an
objective is that shortly after the recovery a node may
be damaged due to aftermath or unexploded bombs, etc.
In addition, exhaustion of the onboard energy supply
and electronic breakdown increase the probability of a
single node failure. Therefore, forming a bi-connected
inter-segment topology will help in reducing the
probability of network re-partitioning after restoration.
Assume the network nodes are stationary, the
topology cannot be autonomously reconfigured and relay
nodes need to be deployed to achieve our goal. In order
to reduce the restoration overhead, the relay count opts
to be minimized. Therefore, the problem tackled in the

paper is fundamentally how to establish an intersegment topology in which there are two-vertex disjoint
paths between every pair of segments by populating the
least relay count. In order to meet the bi-connectivity
goal with the least cost, we strive to form a single
cycle of all segments which is a Hamiltonian cycle
where no cut-vertex exists. Finding a Hamiltonian cycle
is a known NP-Hard problem [2]. In order to address
such complexity, we pursue a heuristic approach that
forms a 2-connected network topology of distinct
segments which provides route alternatives and enables

as

well

as

restores

inter-segment

. Fault-tolerant Connectivity Restoration Algorithm


The proposed algorithm consists of three phases.
(1) We first group the segments in rounds by
calculating nested convex hulls based on which the
segments are divided in terms of their relative proximity
to the periphery of an area of interest. Then, the
segments lying on each convex hull are categorized in a
group denoted as OSi. Then, an acyclic digraph G=(V,E)
is formed by finding directed inter-segment edges
between adjacent groups OSi and OSi+1. Figure 1-(a)
shows the example which contains four directed graph
G for a total of 27 segments, each of which is rooted at
segments 1, 2, 3 and 4 respectively. Thus, segments are
divided into four groups, OSi, i = 1,,4.
(2) Since the convex hull does not identify well
boundaries, if the number of the involved points is large,
in the second phase, a concave hull of segments, Cave
is calculated using the directed edges inwards toward a
center. Then a Hamiltonian cycle of segments that is
represented as a bi-connected undirected graph, G2c, is
formed based on Cave by visiting exactly once each of
the internal segments, in short iNodes that reside in
Cave. Figure 1-(b) shows the identified Cave and the
formed bi-connected topology G2c is seen in Figure 1(c).
(3) In the 3rd phase, we opt to disentangle the
periphery of the polygon formed by G2c, by examining
every tuple of three subsequent nodes in V2c. Then
Steiner points (SPs) are computed on each edge e in the
updated E2c where relays will be placed. The edge
steinerization is performed with respect to a radio range,
R of a relay and the number of SPs equals 2+
((length(e)-2R)/R-1). Finally we form a two-vertex
disjoint network topology by populating the least count
of relays at the identified SPs. Figure 1-(d) shows a
final topology after raveling edges seen in Figure 1-(c),

1018

2016

Figure 1: The proposed algorithm categorizes segments (a),


computes a concave hull, Cave, (b) based on which a biconnected graph G2c is formed (c), and then ravels some
tangled edges in G2c and places the least count of relays (d).

where every pair of segments has two disjoint routing


paths.
. Performance Evaluation
In order to evaluate the performance of the proposed
approach,
we
have
implemented
a
simulation
environment in C. The environment articulates the effect
of damage on a single WSN that originally covers an
area. Each segment in the partitioned network is
represented as a gateway node. In simulation, varying
numbers of segments are evenly distributed in the
quadrants of 1200m1200m. The network characteristics
are varied by R and the number of segments (Nseg) and
the performance metric is the number of relays (NRN).
The effectiveness of the proposed algorithm is
compared to three competing approaches such as the
best known solution for providing 2-vertex connectivity
referred to as 2C-MSS [3], CRAFT which bi-connects
segments based on an inner ring at the center of an
area and bi-connects segments to the ring using SPs
[4]. The third one is 2C-SpiderWeb [5], which forms a
topology that resembles a spider web with segments
located at its perimeter and adds extra edges to make it
2-vertex connected. These baseline approaches address
the same problem tackled in the paper using different
solution strategies.
In Figure 2-(a) all approaches populate more relays
to bi-connect segments as R decreases since the internode communication range primarily affects the number
of nodes to form a connected topology regardless of the
level of connectivity. As R varies between 50m and
120m, our algorithm populates fewer relays than
baselines including 2C-MSS that is the best known
heuristic in the literature for minimizing the relay count
for bi-connectivity. Such a distinct performance is

because our approach utilizes the closest inter-segment


edge to the center of the area in determining the
established links in the bi-connected topology. Moreover,
we consider the proximity among all segments during
the formation of bi-connectivity while 2C-MSS does not
consider the location of multiple segments in a collective
manner.
In Figure 2-(b) which shows the performance of our
algorithm under varying Nseg, all approaches deploy more
relays when the segment count grows. This is very
much expected since the inter-segment connectivity
requirement becomes stronger as Nseg increases.
Nonetheless, the proposed algorithm is more scalable
than the baselines because it considers the intersegment proximity inward towards the center of the
area while CRAFT and 2C-SpiderWeb take into account
inter-node distance among the partial set of the
segments and 2C-MSS does not factor in the intersegment distance in forming the bi-connected topology.
IV. Conclusion
In this paper, we propose the connectivity recovery
algorithm with the least number of relays, which
provides two-vertex disjoint inter-segment paths and
thus proactively prevents re-partitioning after restoration.
The proposed approach strives to bi-connect segments
by forming a Hamiltonian cycle where every node is
visited only once. Simulation results show that our
algorithm outperforms competing approaches in the
literature, by populating fewer relays. The best
performance of the algorithm is achieved for networks
with large number of segments when the communication
range of a relay is small.
ACKNOWLEDGMENT
This research was supported by Basic Science Research Program
through the National Research Foundation of Korea(NRF) funded by
the Ministry of Education (2015R1D1A1A01057095). This work was
also supported by the ICT R&D program of MSIP/IITP. [B0126-151051, A Study on Hyper Connected Self-Organizing Network
Infrastructure Technologies for IoT Service].

Reference
[1] F. Akyildiz, W. Su , Y. Sankarasubramaniam, and E.
Cayirci, Wireless sensor networks: a survey , Computer
Networks, Vol. 38, pp. 393-422, 2002
[2] M. R. Garey, D. S. Johnson, and R. E. Tarjan, The
Planar Hamiltonian Circuit Problem is NP-Complete,
SIAM Journal of Computer, Vol. 5, No. 4, pp. 704-714,
1976
[3]A. Kashyap, S. Khuller and M. Shayman, Relay placement
for higher order connectivity in wireless sensor networks ,
Proc. of IEEE INFOCOM, Barcelona Spain, April 2006.
[4] S. Lee, M. Younis, and M. Lee, Connectivity Restoration
in a Partitioned Wireless Sensor Network with Assured
Fault Tolerance, Journal of Ad Hoc Networks, vol. 24, pp.
1-19, January 2015
[5] F. Senel, M. Younis, and K. Akkaya, Bio-inspired Relay
Node Placement Heuristics for Repairing Damaged Wireless
Sensor Networks, IEEE Trans. on Vehicular Technology,
Vol. 60, No 4, pp. 1835 1848, April 2011

Figure 2: Performance comparison of the proposed algorithm to


baselines in terms of the number of the required relays to bi-1019
connect segments. (a) Nseg = 24 and (b) R = 50.

You might also like