Professional Documents
Culture Documents
. The RN data
transmission rate is donated by
i
which can be defined by
following equation:
=
=
n
j
j i
0
(1)
And d
j
is the agency nodes data transmission rate. In the
following figure 1 shows that there are four paths to form a 2-
connected fault tolerant network. In this case, when the AN0
data is transmitted by the path1, the RN 4 transfers data from
the Agency Node 0 to Base-station, if the data transmission
rate of Agency Node 0 is
0
. According to the equation (1),
1
=
4
=
0
. When the AN0 data and AN1 data is transmitted
via the RN3, and the
3
=
0
+
1
. And then
( ) 5 1 i i
can be
obtained, if the
( )
5 1 i i
, and the RN transmission
capacity can satisfy this constrained condition [20].
Fig. 1. Communication Capacity demo
B. The Edge between Two Adjacent Nodes
Normally, in wireless communication, two nodes may not
be feasible which is based on two reasons. Firstly, the
communication distance between two nodes is too great.
Secondly, the line-of-sight (l-o-s) adjacency is required
between so many wireless links [20]. Therefore, a distance
matrix is defined in here, D=(d
ij
: i, j V) where d
ij
is the
communication distance between node
i
and node
j
. The
calculation of distance obeys the Euclidean distance formula.
And then this distance matrix can be regarded as the cost
matrix, C. On the other hand, the local factors are considered in
non-Euclidean distances situations. And the d
max
denotes the
maximum communication distance. The Boolean line-of-sight
(l-o-s) matrix is defined by [=(ij:i,jV),when the line-of-sight
is between nodes i and j,
ij
=1and
ij
=0 otherwise. The edge
viability matrix Z=(Z
ij
: i, jV) is defined by the equation (2)
[20]:
=
=
otherwise
d d
z
ij ij
ij
: 0
, 1 & : 1
max
(2)
C. The Path between A Given SN and Base-station Node
In some researches, the shortest communication path is an
evaluation criterion for relay placement [21]. However, the
shortest communication path criterion doesnt consider the
linear distance of SNs and the Basestation Node. Because
when the linear distance of SNs and the Basestation Node is
too great, the placement of the RNs will become not
reasonable. Therefore, this paper introduces the distance factor
of SN and Basestation Node as the one constraint, when this
distance factor is the smaller the better performance for RNs
placement [21]. Due to there are two independent paths
between SN and Basestation Node, and the network distance
factor can be expressed by equation (3):
=
+
=
n
i i
path i path i
BN SN d
L L
factor Dis
1
2
2
) 2 , (
2
) 1 , (
) , ( 2
(3)
In the equation (3),
2
) 2 , (
2
) 1 , ( path i path i
L L +
represents the
distance squared value of two independent paths between SN
and Basestation Node, and ) , (
2
BN SN d
i
represents the linear
distance squared value of SN
i
and BN.
V. THE PROPOSED APPROACH
The proposed approach uses simple greedy heuristics
algorithm to search the MST which are for separated sub-
networks connection with the minimum number of RNs. This
approach is implemented in a iterating computer program
consists two main tasks: 1) Firstly, the RNs should be
identified. The sub-networks can be considered as vertices of a
graph and the border sub-networks is identified by the convex
polytope computation. The proposed approach in order to find
RNs to connect each sub-network. These links connect the sub-
networks and adjacent candidate relay nodes, which are based
on the highest node degree among the candidate relay nodes of
subnetworks. And then each selected node can be connected to
each sub-network. In all following rounds, the identified RNs
and sub-networks which have not been selected in previous
rounds become the vertices. And the convex polytope is again
calculated to determine the respective border sub-networks and
new RNs are obtained in a new round. This step is circulated
until the considered vertices can connect to the Base-station. 2)
Secondly, the selected RNs are for sub-networks and Base-
station connection and the selected candidate nodes are for
inter-sub-network 2-connected topology with the optimal total
number of the RNs placement. And the main processes are
described in the next subsection.
A. The main algorithm process
In the RN placement process includes the following steps:
Step 1: Input the SNs position information, and select an
Agency Node for each sub-network;
Step 2: Input the Basestation node position information;
Step 3: Based on the Agency Nodes and Basestation Node
position information for convex polytope computation, and
then obtain the border of convex polytope;
Step 4: Based on the nodes of border to find the adjacent
RNs, until the RNs can connect BN;
Step 5: Algorithm complete.
The details if each step are described in the following
subsections.
B. Agency Nodes Selection
In first step, the Agency Nodes selection is concerned.
Every Sensor Node is considered as vertices in a graph and
each sub-network border is computed by the convex polytope.
After the sub-network border is identified and some
predetermined convergence time, the vertices of convex
polytope would be selected as an agency node which lost
connection to the most number of neighbors, as mentioned
above section 3.
C. Identify the Boundary Nodes and Relay Nodes
In this step, the proposed approach operates in the first
round (r=0), a set of sub-networks in a wind farm, which
buildups the smallest convex hull that includes other sub-
networks. Assuming the sub-networks is regarded as terminals,
the convex hull of all sub-networks and Base-station Node is
utilized to determine boundary sub-networks. Additionally, at
least three non-collinear sub-networks are assumed, and then
the convex hull CH0 can be confirmed and formed a closed
polygon by the Graham algorithm [22].
This paper refers to paper [6] case which is re-link seven
segments problem. As shown in Figure 2, the small stars
represents the AN and the big star represents the Base-station,
the del represents the selected RN and the red squares represent
the candidate RNs. A candidate node is selected initially as RN,
which is the node of highest degree among the adjacent
candidate RN nodes of agency node. And then a connection is
established between RN and AN. From the selected RNs, some
new candidate nodes are selected which are adjacent to the
maximum degree of un-spanned candidate nodes. This process
is repeated until the every selected RN can connect to the Base-
station, as illustrated in figure 2. This can be achieved by
Prims algorithm, the effectiveness has been proved by Vic
Grout [20]. The figure 2 shows the inter-subnetwork topology
which is a minimum spanning tree of six ANs (i=05), one
Base-station and seventeen RNs.
Fig. 2. The convex hull CH0
In figure 2, the improved ADD Algorithm constructs a core
network, N, from an empty connection set. And the edge
matrix can be defined in here, where G= (V, E), the link matrix
is ) , : ( V j i
N
ij
N
= , and the
N j i
N j i
N
ij
) , ( : 0
) , ( : 1
; the
edge matrix is ) , : ( V j i e E
ij
= and
=
E j i
E j i
R
or R
e
ij
) , ( : 0
) , ( :
2
,
; the network degree
vector is
) : ( V i
N
i
N
=
and
=
n
i
N
ij
N
i
.The
temporary spanning vector
) : ( V i s
N
i
N
s
=
is utilized, if
0 =
N
i
s initialization is made for all V i ; if 1 =
N
i
s , i is
included. This improved ADD algorithm has some advantages,
such as offer more path selection, relieve potential traffic
bottleneck, and offer more candidate positions for backup
nodes or supplementary nodes. The pseudo code of the ADD
Algorithm is shown as:
ADD
{Initialization}
For all iV do
0 =
N
i
s
For all i, j V do
0 =
N
ij
1 =
N
j
s }
Find i such that
) ( max
N
j j j
N
i i
v v = where 1 =
N
j
s }
D. MST identification
In this step, the MST identification is processed in multiple
rounds. In the first round, each Agency Node presents each
sub-network and considered as vertices in a graph, and then
inter-subnetwork border is computed by the convex hull.
Subsequently, the MST is calculated in adjacent candidate
nodes based on the highest node degree, ANs and Base-station
Node connected by the relay nodes. In the following rounds,
the identified MST and Agency Nodes which have not been
calculated in prior rounds become the vertices of convex hull.
And then the convex hull is computed again for respective
border vertices, the new MST is then found. This procedure
will be circulated until vertices can communicate to each other
in a round.
Finally, the RNs placement for subnetwork connection and
MST are used to form a connected inter-subnetwork. To satisfy
the network fault-tolerance requirements, each sensor has two
independent paths to link two BSNs or RNs, and then the RNs
form a 2-connected network structure with the BSNs.
VI. PERFORMANCE ANALYSIS
The effectiveness validation of the proposed algorithm will
be made in this part. The simulation environment is based on C
language, a variable number of partitioned sub-networks are
considered in a WSN. And some published algorithms are
compared with the proposed algorithm. Some parameters are
considered in network for variable characteristics, for instance,
the communication radius of relays which affects the
performance of algorithm and the number of the subnetworks
which affects the connectivity requirement [6]. Four metrics
are used in this paper, the number of RNs, the MST tree length,
the number of the searched candidate nodes, and the load of the
key RNs.
A. Simulation Results
In the simulations, the ADD algorithm is used for finding
MST in each round from candidate nodes in the convex hull.
For candidate MST distribution is modeled as a grid
distribution in a limited area (2000m*2000m). And then a
comparison of performance is made for the MST, ORC and the
proposed algorithm. The following figures show network
topologies which are formed by these algorithms.
Fig. 3. Connected Network Topology by proposed Algorithm (RNs=17)
Fig. 4. The ORC algorithm (RNs=16)
Fig. 5. The Prims algorithm (RNs=18)
B. Analysis
a) The number of the RNs: The figure 3 shows topology
of the proposed algorithm, which is obtained through the
minimum spanning tree based on the improved ADD
algorithm. The total number of RNs is 17. The figure 4 and the
figure 5 show the topologies of ORC algorithm and Prim
algorithm, and the total number of the RNs is 16 and 18
respectively. The proposed algorithm takes one more RN than
ORC algorithm, and one less RN than Prims.
b) The path length of the Tree: In this case, the path
length of the tree is about 4240m which based on proposed
algorithm. And the ORC algorithm and the Prim are about
4280m and 4340m respectively. Apparently, the proposed
algorithm reduces the path length for the minimum spanning
tree. Therefore, the entire network power consumption will be
saved for the communication.
c) The traffic load in the key RNs: There are three RNs
connected with the Basestation in the proposed algorithm and
there are two RNs in the ORC algorithm. In proposed
algorithm, these three RNs undertake the one-sixth, one-sixth
and four-sixth of the network traffic load respectively. In
figure 4, one RN undertakes two-sixth traffic load and another
one RN undertakes four-sixth traffic load in ORC algorithm.
Therefore, the RN will run out of the energy quickly, which
undertakes four-sixth traffic load in the network. These issues
will lead to unstable network. Apparently, the proposed
algorithm improves the traffic load balance and the reliability
of network.
d) The time complexity of Algorithm: The time
complexity of ORC is
))) ) / ( 2 ( ) (log( (
2
R LP Size N N N
R seg Seg Rnd
+
.
Rnd
N
represents the number of rounds;
Seg
N represents the
number of segments; R represents the communication radius
of RN; LP
R
represents the smallest rectangle which also
includes the largest triangle. The time complexity of Prims
algorithm is ) log ( n n and the ADD algorithm has the
similar complexity which is confirmed in paper [20].
However, in a 1000 nodes network, the time complexity of
ADD algorithm is only one-sixth of ) log ( n n . Additionally,
due to the improved ADD algorithm, the number of boundary
nodes which is confirmed in any convex hull and the number
of selected candidate nodes in each round have to be less or
equal to the number of subnetworks, N
sub
. In every round, the
number of the searched candidate nodes for finding MST is
less or equal to eight. Therefore, the time complexity of the
improved ADD algorithm is )) (log( (
Sub Sub Rnd
N N N ,
which is determined by the number of subnetworks. N
Rnd
is
the number of rounds, and the N
Sub
is the number of
subnetworks. Obviously, the proposed algorithm reduces the
time complexity for the relay node placement algorithm.
VII. CONCLUSION
In practical application, due to the obstacle or the great
distance between two nodes of Wireless Sensor Network,
therefore, the Relay Nodes have to be placed into the network.
In this paper, the Relay Nodes Placement problem has been
researched for disjoint subnetworks for wind farm application.
The improved ADD algorithm is applied for the Relay Nodes
placement. The final simulation results show that the
effectiveness of the improve ADD algorithm. The improved
ADD algorithm can save the energy for the communication
and improve the traffic load balance, the reliability of network,
and reduce the time complexity.
REFERENCES
[1] Jennifer Yick, Biswanath Mukherjee, and Dipak Ghosal, Wireless
sensor network survey, Computer Networks, 52, pp.2292-2330, April
2008
[2] Jillian Cannons, Laurence B. Milstein, and Kenneth Zeger, An
algorithm for wireless relay placement, IEEE Transactions on Wireless
Communication, Vol. 8, NO. 11, November 2009.
[3] J. Tang, B. Hao, A. Sen, Relay node placement in large scale wireless
sensor networks, Computer Communication 29, pp.490501, 2006
(special issue on wireless sensor networks).
[4] E.L. Lloyd, G. Xue, Relay node placement in wireless sensor
networks, IEEE Transactions on Computers, 56 (1),pp.134138, 2007
[5] X. Cheng, D.-z. Du, L.Wang, B. Xu, Relay sensor placement in
wireless sensor networks, Wireless Networks, 14 (3),pp.347355, 2008
[6] S. Lee, M. Younis, Optimized relay node placement for connecting
disjoint wireless sensor networks, Computer Networks, 56, pp.2788-
2804, 2012
[7] A. Efrat Sa ndor, P. Fekete, P. Gaddehosur, J. Mitchell, V. Polishchuk,
and J. Suomela, Improved Approximation Algorithms for Relay
Placement, Proc. 16th Ann. European Symp. Algorithms (ESA 08),
pp. 356-367, 2008
[8] B. Hao, J. Tang, and G. Xue, Fault-Tolerant Relay Node Placement in
Wireless Sensor Networks: Formulation and Approximation, Proc.
Workshop High Performance Switching and Routing (HPSR), pp. 246-
250, 2004
[9] J. Pan, Y.T. Hou, L. Cai, Y. Shi, and S.X. Shen, Topology Control for
Wireless Sensor Networks, Proc. ACM MobiCom, pp. 286-299, 2003
[10] Quanhong Wang, Kenan Xu, G. Takahara, and H. Hassanein, Device
placement for heterogeneous wireless sensor networks: minimum cost
with lifetime constraints, IEEE Transactions on Wireless
Communications. Vol. 6, NO.7, July 2007.
[11] Dejun Yang, S. Misra, Xi Fang, Guoliang Xue, and Junshan Zhang,
Two-tiered constrained relay node placement in wireless sensor
networks: computational complexity and efficient approximations,
IEEE Transactions on Mobile Computing, Vol. 11, NO.8, August 2012.
[12] J. Bredin, E. Demaine, M. Hajiaghayi, and D. Rus, Deploying Sensor
Networks with Guaranteed Fault Tolerance, IEEE/ACM Trans.
Networking, vol. 18, no. 1, pp. 216-228, Feb. 2010.
[13] W. Zhang, G. Xue, and S. Misra, Fault-tolerant relay node placement in
wireless sensor networks: Problems and algorithms, in Proc. IEEE
INFOCOM, pp. 16491657,2007
[14] J. Bredin, E. Demaine, M. Hajiaghayi and D. Rus; Deploying sensor
networks with guaranteed capacity and fault tolerance, Mobihoc05;
pp. 309-319, 2005
[15] Xiaofeng Han, Xiang Cao, E. L. Lloyd, and Chien-Chung Shen, Fault-
tolerant relay node placement in heterogeneous wireless sensor
networks, IEEE Transactions on Mobile Computing, Vol. 9, NO.5 May
2010.
[16] G. Lin, G. Xue, "Steiner tree problem with minimum number of Steiner
points and bounded edge-length," Information Processing Letters, 69,
pp.5357, 1999
[17] H. Liu, P.J. Wan, and X.H. Jia, Fault-Tolerant Relay Node Placement
in Wireless Sensor Networks, Proc. Intl Computing and Combinatorics
Conf. (COCOON), pp. 230-239, 2005.
[18] S. Misra, S. D. Hong, G. Xue and Jian Tang, Constrained relay node
placement in wireless sensor networks: formulation and approximation,
IEEE/ACM Transactions on Networking, Vol. 18, NO.2, April 2010.
[19] Mark de Berg, O. Cheong, M. Kreveld and M. Overmars,
Computational geometry: Algorithms and Applications, Third Edition,
Springer-Verlag Berlin Heidelberg, 2008, pp.243-254
[20] Vic Grout, Principles of Cost Minimisation in Wireless Networks,
Journal of Heuristics, Vol. 11, No.2, pp.115-133, March 2005
[21] Wang Zhu, Wang Qi, Wei Debao, Wang Ling, Relay node placement
and addition algorithms in wireless sensor netorks, Acta Phys Sin,
Chinese Physical Society, Vol. 61, No. 12, 120505, pp.1-10, 2012 (In
Chinese)
[22] R.L. Graham, An efcient algorithm for determining the convex hull of
a nite planar set, Information Processing Letters, 1, pp.132133, 1972