Disclosure of Invention
In order to solve the technical problem, the invention provides a node dormancy scheduling method based on data similarity in a wireless sensor network. The method utilizes a similarity function to obtain the similarity of the perception data of all the active nodes in the cluster, and carries out cluster analysis on neighbor nodes with high similarity to obtain redundant nodes. And then, screening out the dormant nodes under the condition of avoiding the perception blind area. Meanwhile, a node dormancy scheduling method is provided, so that the collection of redundant data is greatly reduced, and the network energy consumption is balanced.
In order to achieve the purpose, the invention adopts the technical method that: a node dormancy scheduling method based on data similarity in a wireless sensor network is characterized in that: the method comprises the following steps:
step 1: after network clustering is completed, sensor nodes periodically acquire sensing data; one cycle of collection is a cycle, and each cycle is divided into two task stages: a collection phase and a scheduling phase;
step 2: a collection stage: the active sensor nodes in the cluster periodically collect sensing data, after each period is finished, the nodes transmit sensing data information and a neighbor table to a cluster head together, and the cluster head transmits the data to the BS after aggregating the data;
and step 3: scheduling stageSection (2): the BS constructs a fuzzy matrix according to the perception data in different sub-periods, then carries out cluster analysis by utilizing a similarity threshold and a node neighbor table to obtain a redundant node set of each sub-period and records the redundant node set as each sub-period
;
And 4, step 4: obtaining a final redundant node set according to the redundant node set combination set obtained in the step 3 and recording the final redundant node set as a final redundant node set
;
Taking the neighbor condition of the node as a merging standard during merging;
and 5: to is directed at
Screening out the dormant nodes by adopting a corresponding dormant node selection method under the clustering condition;
step 6: and after the sleeping node is selected, the BS broadcasts the relevant information to carry out sleeping scheduling.
The specific method of the collection stage in the step 2 is as follows:
a collection stage: dividing each period into 3 sub-periods, wherein each sub-period comprises data collection and data transmission, and the stage is carried out in clusters and among clusters; after the active sensor nodes in each sub-period cluster acquire sensing data, a data matrix is formed, and the sensing data and the node neighbor table are sent to the cluster head together; and after all the sub-periods are finished, the cluster head aggregates the information and sends the information to the BS in a single-hop or multi-hop mode.
The specific process of the scheduling stage and the cluster analysis in the step 3 is as follows:
3.1) scheduling phase: the method comprises the steps of node clustering analysis, dormant node selection and dormant node scheduling, wherein the step is carried out at a BS; the BS constructs a corresponding fuzzy equivalent matrix according to the perception matrix of each sub-period to perform cluster analysis, and each sub-period obtains a group of redundant node sets; then, determining a final redundant node through the 3 groups of redundant nodes, and finally screening out dormant nodes according to a dormant node selection method to perform dormant scheduling;
3.2) sensor node
iIn the first place
pThe perception data matrix formed after the end of the sub-period is recorded as
Then it is first
pThe matrix formed by the perception data collected by the active nodes in the cluster in the sub-period is recorded as
Expressed as follows:
wherein:mfor the number of active sensors in a cluster,krepresenting a sub-periodkThe number of sensory data collected by the sensors within a time slot;
3.3) according to the fuzzy clustering analysis theory, the sensor node can be knowniAnd sensor nodejThe similarity between the perception data is calculated by formula (1):
wherein:
x ih representing sensor nodes
iIn the first place
hThe data collected for each of the time slots is,
representing nodes
iMean of data of
;
Fuzzy similarity matrix
Expressed as follows:
3.4) further extending the fuzzy similar matrix into a fuzzy equivalent matrix: from fuzzy similarity matrices
RStarting, sequentially squaring:
when equation (2) occurs for the first time:
thenR k The solved fuzzy equivalent matrix; fuzzy similarity matrixSCalculating to obtain fuzzy equivalent matrix according to the formulaE(S) Expressed as follows:
clustering principle: given a similarity threshold by a particular network environment
If the similarity between the sensing data of any two adjacent nodes is greater than
Then a set of redundant nodes is obtained
And the redundant nodes are grouped into one type;
according to the clustering principle, the node clustering can be divided into two types: a base class and an extension class; the extension class may extend to different representations having the same clustering characteristics.
3.4) the basic classes and diffusion classes are specifically:
3.4.1) base classes: redundant node
The redundant nodes are grouped into one type, namely all the redundant nodes are neighbors;
3.4.2) extension class:
3.4.2.1) order
AAll the nodes in the network are adjacent to each other,
Ball nodes in the network are adjacent nodes;
Anode in and
Bthe node in (1) is not a neighbor node but has a common node; clustering the nodes respectively to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: the common redundant nodes are contained by different sets;
3.4.2.2) order
Respectively clustering the nodes to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: all nodes in the common class are respectively contained by different sets;
3.4.2.3) order
Respectively clustering the nodes to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: some of the nodes in the common class are each contained by a different set.
The method for selecting the dormant node in the step 5 specifically comprises the following steps:
5.1) base class 3.4.1) sleep node selection method
Suppose that each sub-period sensor node sends to the cluster head in each roundlBit data, known from an energy decay model, of the sensor nodeiNumber of wheels capable of workingR i Comprises the following steps:
wherein:
is the initial energy of the node and is,
in order to obtain the radio frequency energy consumption coefficient,
is the energy consumption coefficient of the amplifier under the free space model,
the distance from the node to the cluster head; then it is first
rTime of the turn, node
iIn the first place
pRemaining energy required to maintain an active state for a sub-period
Comprises the following steps:
wherein:
ris the current number of working rounds of the active node,
the current number of working rounds of the dormant node,
pis the first in the wheel
pA sub-period; first, the
pAverage residual energy of sub-periodic redundant nodes
Comprises the following steps:
wherein:kthe number of redundant nodes in the current cluster is obtained; when the current residual energy of the redundant node is lower than the average residual energy, namely:
selecting the redundant node as a dormant node; wherein:
residual energy for redundant nodes;
5.2) selection method of dormant node in extension type 3.4.2.1)
Separately computing collections
A,
BAverage residual energy of medium redundant node
(ii) a For redundant node
iAnd if and only if
Time, node
iAs a sleeping node, in addition to the above, the node
iNot as a sleeping node; respectively take
The corresponding redundant node in the node is taken as a dormant node;
representation collection
AOther nodes than the common node; when more than one public redundant node is needed, judging whether each node is taken as a dormant node or not;
5.3) selection method of dormant node in extension type 3.4.2.2)
Respectively calculate
(ii) a Respectively take
The corresponding redundant node in the node is taken as a dormant node;
5.4) selection method of dormant node in extension type 3.4.2.3)
Respectively calculate
(ii) a For redundant node
iAnd if and only if
Time, node
iMust be taken as a sleeping node, except for the above cases
Not as a sleeping node; judging redundant nodes by adopting the method
jWhether to act as a sleeping node; respectively take
The corresponding redundant node in (b) acts as a sleeping node.
The sleep scheduling described in step 6 is specifically implemented as follows:
6.1): when the network is initialized, each sensor node maintains a round counter Count, and the initial value is 0; adding 1 to the Count value every time data collection is carried out;
6.2): the sensor node collects sensing data through the collection method in the step 2, and finally sends the sensing information and neighbor table information to the cluster head together;
6.3): the cluster head selects a dormant node through the clustering analysis and the dormant node screening method provided in the step 3-5, and finally, the BS broadcasts the ID number of the dormant node;
6.4): after receiving the message, the node performs the following two checks:
6.4.1) judging whether the node is a dormant node, if so, entering a dormant state. The node in the dormancy is awakened after the preset dormancy time and enters the active state again;
6.4.2) judging whether the dormant node is a neighbor node thereof, if so, modifying the node Mode value in the neighbor table to S;
6.5) after one round, the cluster head checks whether the current residual energy is lower than a threshold value, if so, the clustering algorithm is executed again; otherwise, continuing the next round of data collection. Repeat execution 6.2) until the network lifecycle is over.
The beneficial effects created by the invention are as follows: aiming at the problems of high redundancy of sensing data and unbalanced network energy in a wireless sensor network, the invention provides a node dormancy scheduling method based on data similarity. The method utilizes a similarity function to obtain the similarity of the perception data of all the active nodes in the cluster, and carries out cluster analysis on neighbor nodes with high similarity to obtain redundant nodes. And then, screening out the dormant nodes under the condition of avoiding the perception blind area. By scheduling the sleeping nodes, the collection of redundant data is greatly reduced. Simulation results show that the method can effectively balance network energy consumption and greatly prolong the life cycle of the network.
Detailed Description
Step 1: after network clustering is completed, sensor nodes periodically acquire sensing data; one cycle of collection is a cycle, and each cycle is divided into two task stages: a collection phase and a scheduling phase.
Step 2: a collection stage: the active sensor nodes in the cluster periodically collect sensing data, after each period is finished, the nodes send sensing data information and the neighbor table to the cluster head together, and the cluster head aggregates the data and sends the aggregated data to the BS.
The specific method comprises the following steps: dividing each period into 3 sub-periods, wherein each sub-period comprises data collection and data transmission, and the stage is carried out in clusters and among clusters; after the active sensor nodes in each sub-period cluster acquire sensing data, a data matrix is formed, and the sensing data and the node neighbor table are sent to the cluster head together; and after all the sub-periods are finished, the cluster head aggregates the information and sends the information to the BS in a single-hop or multi-hop mode.
And step 3: and a scheduling stage: the BS constructs a fuzzy matrix according to the perception data in different sub-periods, then carries out cluster analysis by utilizing a similarity threshold and a node neighbor table to obtain a redundant node set of each sub-period and records the redundant node set as each sub-period
。
The scheduling stage and the cluster analysis specifically comprise the following processes:
3.1) scheduling phase: the method comprises the steps of node clustering analysis, dormant node selection and dormant node scheduling, wherein the step is carried out at a BS; the BS constructs a corresponding fuzzy equivalent matrix according to the perception matrix of each sub-period to perform cluster analysis, and each sub-period obtains a group of redundant node sets; and then determining a final redundant node through the 3 groups of redundant nodes, and finally screening out the dormant nodes according to a dormant node selection method to perform dormant scheduling.
3.2) sensor node
iIn the first place
pThe perception data matrix formed after the end of the sub-period is recorded as
Then it is first
pThe matrix formed by the perception data collected by the active nodes in the cluster in the sub-period is recorded as
Expressed as follows:
wherein:mfor the number of active sensors in a cluster,krepresenting a sub-periodkThe number of sensory data collected by the sensors within a time slot;
3.3) according to the fuzzy clustering analysis theory, the sensor node can be knowniAnd sensor nodejThe similarity between the perception data is calculated by formula (1):
wherein:
x ih representing sensor nodes
iIn the first place
hThe data collected for each of the time slots is,
representing nodes
iMean of data of
;
Fuzzy similarity matrix
Expressed as follows:
3.4) further extending the fuzzy similar matrix into a fuzzy equivalent matrix: from fuzzy similarity matrices
RStarting, sequentially squaring:
when equation (2) occurs for the first time:
thenR k The solved fuzzy equivalent matrix; fuzzy similarity matrixSCalculating to obtain fuzzy equivalent matrix according to the formulaE(S) Expressed as follows:
clustering principle: given a similarity threshold by a particular network environment
If the similarity between the sensing data of any two adjacent nodes is greater than
Then a set of redundant nodes is obtained
And the redundant nodes are grouped into one type;
according to the clustering principle, the node clustering can be divided into two types: a base class and an extension class; the extension class may extend to different representations having the same clustering characteristics.
The basic class and the diffusion class are specifically as follows:
3.4.1) base classes: redundant node
The redundant nodes are grouped into one type, namely all the redundant nodes are neighbors;
3.4.2) extension class:
3.4.2.1) order
AAll the nodes in the network are adjacent to each other,
Ball nodes in the network are adjacent nodes;
Anode in and
Bthe node in (B) is not a neighbor node butThere is a common node; clustering the nodes respectively to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: the common redundant nodes are contained by different sets;
3.4.2.2) order
Respectively clustering the nodes to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: all nodes in the common class are respectively contained by different sets;
3.4.2.3) order
Respectively clustering the nodes to obtain a redundant node set
;
Such clustering can be extended to a variety of situations, and the common features of such clustering are: some of the nodes in the common class are each contained by a different set.
And 4, step 4: obtaining a final redundant node set according to the redundant node set combination set obtained in the step 3 and recording the final redundant node set as a final redundant node set
;
And taking the neighbor condition of the node as a merging standard during merging.
And 5: to is directed at
And (4) screening out the dormant nodes by adopting a corresponding dormant node selection method under the clustering condition.
The method for selecting the dormant node specifically comprises the following processes:
5.1) base class 3.4.1) sleep node selection method
Suppose that each sub-period sensor node sends to the cluster head in each roundlBit data, known from an energy decay model, of the sensor nodeiNumber of wheels capable of workingR i Comprises the following steps:
wherein:
is the initial energy of the node and is,
in order to obtain the radio frequency energy consumption coefficient,
is the energy consumption coefficient of the amplifier under the free space model,
the distance from the node to the cluster head; then it is first
rTime of the turn, node
iIn the first place
pRemaining energy required to maintain an active state for a sub-period
Comprises the following steps:
wherein:
ris the current number of working rounds of the active node,
the current number of working rounds of the dormant node,
pis the first in the wheel
pA sub-period; first, the
pAverage residual energy of sub-periodic redundant nodes
Comprises the following steps:
wherein:kthe number of redundant nodes in the current cluster is obtained; when the current residual energy of the redundant node is lower than the average residual energy, namely:
selecting the redundant node as a dormant node; wherein:
residual energy for redundant nodes;
5.2) selection method of dormant node in extension type 3.4.2.1)
Separately computing collections
A,
BAverage residual energy of medium redundant node
(ii) a For redundant node
iAnd if and only if
Time, node
iAs a sleeping node, in addition to the above, the node
iNot as a sleeping node; respectively take
The corresponding redundant node in the node is taken as a dormant node;
representation collection
AMiddle-removing public sectionOther nodes outside the point; when more than one public redundant node is needed, judging whether each node is taken as a dormant node or not;
5.3) selection method of dormant node in extension type 3.4.2.2)
Respectively calculate
(ii) a Respectively take
The corresponding redundant node in the node is taken as a dormant node;
5.4) selection method of dormant node in extension type 3.4.2.3)
Respectively calculate
(ii) a For redundant node
iAnd if and only if
Time, node
iMust be taken as a sleeping node, except for the above cases
Not as a sleeping node; judging redundant nodes by adopting the method
jWhether to act as a sleeping node; respectively take
The corresponding redundant node in (b) acts as a sleeping node.
Step 6: and after the sleeping node is selected, the BS broadcasts the relevant information to carry out sleeping scheduling.
The sleep scheduling is specifically implemented as follows:
6.1): when the network is initialized, each sensor node maintains a round counter Count, and the initial value is 0; adding 1 to the Count value every time data collection is carried out;
6.2): the sensor node collects sensing data through the collection method in the step 2, and finally sends the sensing information and neighbor table information to the cluster head together;
6.3): and (5) selecting the dormant nodes by the cluster heads through the cluster analysis and the dormant node screening method provided in the step 3-5. Finally, the BS broadcasts the ID number of the sleeping node;
6.4): after receiving the message, the node performs the following two checks:
6.4.1) judging whether the node is a dormant node, if so, entering a dormant state. The node in the dormancy is awakened after the preset dormancy time and enters the active state again;
6.4.2) judging whether the dormant node is a neighbor node thereof, if so, modifying the node Mode value in the neighbor table to S;
6.5) after one round, the cluster head checks whether the current residual energy is lower than a threshold value, if so, the clustering algorithm is executed again; otherwise, continuing the next round of data collection; repeat execution 6.2) until the network lifecycle is over.
Example 1:
the detailed description of the system model, the clustering protocol, the dormancy scheduling, the specific application process and the verification process based on the protocol is as follows:
network model
The method is suitable for wireless sensor network application of periodically collecting data. The sensor nodes in the network model are randomly deployed in a monitoring area, and the density of the network nodes is large enough to cover the whole area. The sensor node properties are as follows:
1) the sensor nodes are static nodes, and the positions of the sensor nodes cannot be changed once the sensor nodes are deployed. But its topology may change due to node energy exhaustion or node failure, among other reasons.
2) All sensor nodes are of the same type, namely, the sensor nodes have the same initial energy and communication radiusRAnd radius of perceptionr s And has certain computing power. Each node has a unique ID.
3) A node may perceive all nodes within its communication radius that are in an active state.
4) The communication power of the node is adjustable, and the transmitting power can be adjusted according to the distance.
5) The sensor node calculates the distance from the sensor node to the adjacent node by an RSSI ranging method.
6) The collision problem is not considered in the data transmission process among the nodes.
Energy consumption model
The same wireless communication energy consumption model as the LEACH protocol is employed herein. Sensor node transmissionlThe distance between two nodes is d, and the transmission energy consumption formula is as follows:
wherein
E elec In order to obtain the radio frequency energy consumption coefficient,
and
the amplifier energy consumption coefficients under a free space model and a multipath fading model respectively,
depending on the particular network environment and the amplifier power consumption coefficient,
. Receiving
lThe energy consumption formula of the bit data is as follows:
in addition to transmitting and receiving data, the cluster head also aggregates data from active sensor nodes in the cluster. The formula for calculating the energy consumption is as follows:
EDA is a data fusion coefficient, and n is the number of active nodes in the cluster.
Clustering protocol
The network is clustered using Distributed heterogeneous Clustering-routing protocol (DEBUC). DEBUC adopts a cluster head competition algorithm based on time, and the broadcasting time depends on the residual energy of a candidate cluster head and the residual energy of a neighbor node. Meanwhile, the cluster size close to the base station is smaller by controlling the competition range of the candidate cluster heads at different positions. In this way, the energy consumption of intra-cluster and inter-cluster communication between nodes at different positions in the network can be mutually compensated.
Cluster analysis and redundant node selection
Sensor node
iIn the first place
pThe perception data matrix formed after the end of the sub-period is recorded as
Then it is first
pThe matrix formed by the perception data collected by the active nodes in the cluster in the sub-period is recorded as
Expressed as follows:
whereinmFor the number of active sensors in a cluster,krepresenting a sub-periodkThe amount of sensory data collected by the sensor during a time slot.
Definition 1 similarity: and measuring the similarity degree between the sensor nodes sensing data. The values are between [0,1], 0 represents that the perception data are completely different, and 1 represents that the perception data are completely the same. The closer the value is to 1, the higher the degree of similarity and vice versa.
According to the fuzzy clustering analysis theory, the sensor nodeiAnd sensor nodejThe similarity between the perception data is calculated by formula (1):
wherein
x ih Representing sensor nodes
iIn the first place
hThe data collected for each of the time slots is,
representing nodes
iThe mean value of the data of (a),
;
fuzzy similarity matrix
Expressed as follows:
to facilitate cluster analysis, the fuzzy similarity matrix is further extended to a fuzzy equivalence matrix. From fuzzy similarity matrices
RStarting, sequentially squaring:
when equation (2) occurs for the first time:
thenR k Is the fuzzy equivalence matrix that is being sought. Calculating the fuzzy equivalent matrix by the formula (2) and the formula (3)E(S) Expressed as follows:
define 2 neighbor table: storingA data structure in a sensor node, the structure of which is shown in table 1. Wherein each row records information of one neighbor node of the node. The ID in each row represents the unique identification of the neighbor node;
representing the remaining energy of the node; distance represents the Distance of a node to its neighbor; mode represents the node state: w denotes an active state and S denotes a sleep state. The node neighbor table is shown in table 1.
Table 1 node neighbor table
Defining 3 redundant nodes: and the data similarity is larger than a given similarity threshold and the nodes are adjacent to each other.
Clustering principle: given a similarity threshold by a particular network environment
If the similarity between the sensing data of any two adjacent nodes is greater than
Then a set of redundant nodes is obtained
And the redundant nodes are grouped into one type.
According to the clustering principle, the node clustering can be divided into two types: a base class and an extension class. The extension class may extend to different representations having the same clustering characteristics. Node clustering is shown in FIGS. 2a-2i
(1) Basic classes
Redundant node
The clustering is a kind, i.e. all redundant nodes are neighbors of each other, as shown in fig. 2 a.
(2) Extension class
1) Order to
AAll the nodes in the network are adjacent to each other,
Ball nodes in the network are adjacent nodes.
ANode in and
Bthe node in (1) is not a neighbor node but there is a common node. Clustering the nodes respectively to obtain a redundant node set
As shown in fig. 2 b.
Such clustering can be extended to a variety of cases, as shown in fig. 2c and 2 d. Common features of such clustering are: the common redundant nodes are contained by different sets.
2) Order to
Respectively clustering the nodes to obtain a redundant node set
As shown in fig. 2 e.
Such clustering can be extended to a variety of cases, as shown in fig. 2f and 2 g. Common features of such clustering are: all nodes in a common class are each contained by a different set.
3) Order to
Respectively clustering the nodes to obtain a redundant node set
As shown in fig. 2 h.
Such clustering can be extended to a variety of cases, as shown in FIG. 2 i. Common features of such clustering are: some of the nodes in the common class are each contained by a different set.
Dormant node selection
(1) Method for selecting dormant nodes in basic class
Assuming that each sub-period sensor section of each roundPoint to cluster head transmissionlBit data, known from an energy decay model, of the sensor nodeiNumber of wheels capable of workingR i Comprises the following steps:
wherein:
is the initial energy of the node and is,
in order to obtain the radio frequency energy consumption coefficient,
is the energy consumption coefficient of the amplifier under the free space model,
is the distance of the node to the cluster head. Then it is first
rTime of the turn, node
iIn the first place
pRemaining energy required to maintain an active state for a sub-period
Comprises the following steps:
wherein
rIs the current number of working rounds of the active node,
the current number of working rounds of the dormant nodes.
pIs the first in the wheel
pAnd (4) a sub-period. First, the
pAverage residual energy of sub-periodic redundant nodes
Comprises the following steps:
whereinkIs the number of redundant nodes in the current cluster. When the current residual energy of the redundant node is lower than the average residual energy, namely:
the redundant node is selected as the sleeping node. Wherein
The remaining energy of the redundant node.
(2) Selection method of dormant node in extension class 1)
In FIG. 2b, the sets are computed separately
A,
BAverage residual energy of medium redundant node
. For redundant node
iAnd if and only if
Time, node
iMust be taken as a sleeping node, except for the above cases
iNot necessarily as a sleeping node. Respectively take
The corresponding redundant node in (2) is taken as a dormant node.
Representation collection
AOther than the common node. When there is more than one common redundant node, each node needs to be judged whether to be used as a dormant node. It is noted that the independent consideration of whether the common node is a sleeping node is to prevent the occurrence of a blind sensing zone.
(3) Selection method of dormant node in extension class 2)
In FIG. 2e, calculate separately
. Respectively take
The corresponding redundant node in (2) is taken as a dormant node.
(4) Selection method of dormant node in extension class 3)
In FIG. 2h, calculate separately
. For redundant node
iAnd if and only if
Time, node
iMust be taken as a sleeping node, except for the above cases
The node does not need to be used as a dormant node, and redundant nodes can be judged in the same way
jWhether to act as a sleeping node. Respectively take
The corresponding redundant node in (b) acts as a sleeping node.
Dormancy scheduling method
In the method, node scheduling is performed according to a round, and only node scheduling in one round is described here. It is assumed that all nodes are active when the network is initially running. The specific process is as follows:
step 1: when the network is initialized, each sensor node maintains a round counter Count, and the initial value is 0. Each time data collection is performed, the Count value is increased by 1.
Step 2: after the member nodes in each sub-period cluster of each round acquire the sensing data, the sensing information and the neighbor table information are sent to the cluster head together.
And step 3: and after receiving the data of all the sub-periods, the cluster head aggregates the data and sends the data to the BS. And the BS constructs a fuzzy matrix according to the sensing data of each sub-period, and selects the sleep nodes according to the clustering analysis and sleep node screening methods provided in sections 4 and 5. Finally, the BS broadcasts the sleeping node ID number.
And 4, step 4: after receiving the message, the node performs the following two checks:
1) and judging whether the node is a dormant node or not, and if so, entering a dormant state. And awakening the node in the sleep state after the preset sleep time, and re-entering the active state.
2) And judging whether the dormant node is a neighbor node of the dormant node, and if so, modifying the Mode value of the node in the neighbor table to S.
And 5: after one round, the cluster head checks whether the current residual energy is lower than a threshold value, and if so, the clustering algorithm is executed again. Otherwise, continuing the next round of data collection. Repeat execution 6.2) until the network lifecycle is over.
Simulation experiment
To evaluate the performance of the dormancy method proposed herein, simulation experiments were performed in an OMNET environment, comparing the N3SDS method with SSTBC, ESSM. The parameter settings of the simulation environment are shown in table 2. Comparisons are made primarily from three aspects: 1) The survival rate of the nodes; 2) Network energy consumption; 3) The accuracy of the data.
Table 2 parameter settings for simulation environment
Parameter name
|
Parameter value
|
Parameter name
|
Parameter value
|
Simulation area
|
200m×200m
|
|
0.013pJ/bit/m4 |
Number of nodes
|
400 are
|
E
DA
|
5nJ/bit
|
E
init
|
0.5J
|
Period of time
|
10 s
|
E
elec
|
50nJ/bit
|
Data packet
|
4000 bits
|
|
100pJ/bit/m2 |
Radius of communicationR |
50m |
(1) Node survival rate
The survival rate of the nodes in the network reflects the consumption condition of the energy of the nodes in the network, and is one of the standards for evaluating the energy efficiency of the network.
Fig. 3 is a graph showing the comparison of the survival rates of the nodes in the three dormancy methods, wherein the number of failed nodes in the network is gradually increased and the number of available nodes is gradually decreased as the number of running rounds is increased. As can be seen from fig. 3, the network lifetime of N3SDS is extended by 28.2% and 72.4% compared to ESSM and SSTBC, respectively. At round 450, the first node begins to fail. This is because the N3SDS performs fuzzy clustering analysis on the sensing data of each sub-period, respectively, thereby selecting three sets of redundant node sets. When the final redundant node is determined, a merging standard is adopted, more redundant nodes can be selected as dormant nodes, the candidate number of the dormant nodes is increased, the network energy can be balanced more effectively, and the survival time of the corresponding nodes is longer.
Network energy consumption
(2) Network energy consumption
Network energy consumption refers to the energy consumed by all nodes in the network in each round.
Fig. 4 shows a comparison of network power consumption for three sleep methods. As can be seen from fig. 4, compared with ESSM and SSTBC, the energy consumption of the N3SDS network changes more slowly, and the total energy consumption of nodes is lower under the same round number. This is because the N3SDS calculates theoretical residual energy values for an active node and a dormant node among the redundant nodes, respectively, when selecting the dormant node, and then compares with the redundant node by using their average residual energy as a threshold, and all nodes below the threshold are dormant. Therefore, the selection of the dormant nodes is more reasonable, and more nodes enter the dormant state under the condition of ensuring the data accuracy. Therefore, the N3SDS can balance the network energy more effectively and reduce the network energy consumption.
(3) Data accuracy
The data accuracy refers to the proximity between the fused data received by the BS and the data actually acquired by the sensor node.
FIG. 5 is a graph showing a comparison of data accuracy for three sleep methods. It can be seen from fig. 5 that as the number of sensors increases, the data accuracy continues to improve. And exhibit a tendency to fluctuate as the data accuracy rises to a certain value. Data accuracy of N3SDS was improved by 5.4% and 18.1% compared to ESSM and SSTBC, respectively. And the total variation trend of the data accuracy of the N3SDS is more stable when the number of the nodes is between 200 and 350. This is because as the number of sensor nodes increases, the monitoring area is fully covered by the nodes. The N3SDS starts from data, selects redundant nodes based on the similarity of the data, and screens dormant nodes under the condition of avoiding a perception blind area, so that the residual active nodes can sufficiently realize the data collection of the same level, the reliability of the data is ensured, and the corresponding data is higher in accuracy and more stable.
Aiming at the problems of high redundancy of sensing data and unbalanced network energy in a wireless sensor network, a node dormancy scheduling method (N3SDS) based on data similarity is provided. The method utilizes a similarity function to obtain the similarity of the perception data of all the active nodes in the cluster, and carries out cluster analysis on neighbor nodes with high similarity to obtain redundant nodes. And then, screening out the dormant nodes under the condition of avoiding the perception blind area. By scheduling the sleeping nodes, the collection of redundant data is greatly reduced. Simulation results show that the N3SDS can effectively balance network energy consumption and greatly prolong the life cycle of the network.