Research on energy saving algorithm of field animal monitoring based on cluster sensor network 1

Abstract

Monitoring the diversity of wild animals is a core part of the research and protection of wild animals. Due to the harsh outdoor environment, researchers cannot squat in the deep forest for a long time. Therefore, designing a sensor network system for wildlife monitoring is of great value to wildlife research, protection, and management. When deploying a wildlife monitoring network in the wild environment, it is necessary to solve the problem of the effective use of energy. To this end, this paper proposes an energy-saving optimization method for node scheduling and a wake-up scheme based on a cultural genetic algorithm. This method achieves the purpose of energy saving by making redundant nodes fall asleep and waking up sleep nodes to repair the coverage blind area caused by dead nodes. Simulation results show that this method can activate fewer sensor nodes to monitor the required sensing area, and its performance is better than other known solutions.

Keywords

Cultural genetic algorithm wild animal monitoring wireless sensor network coverage control energy saving

1 Introduction

In a natural evolution, wild animals have formed a close symbiotic relationship similar to spider webs through the chain relationship of the food chain with the surrounding ecological environment. The extinction of a species will have a chain reaction, and the consequences may be the disappearance and collapse of an ecological system. Therefore, human beings need to rely on nature. We protect wild animals, eliminate poaching [1], protect biodiversity, and protect the ecological environment in which they live. We are protecting ourselves [2]. How to effectively detect the ecological environment of wild animals, obtain relevant activity trajectory data of wild animals, and then realize the scientific research, protection, and management of wild animals, has always been a research hotspot.

There are two traditional solutions for wildlife monitoring: follow the shots on-site or arrange many cameras. However, these two methods have certain drawbacks. Because the living environment of wild animals is a complex and harsh primitive ecosystem, there are specific difficulties and dangers for the staff to follow the shooting on the spot, and it is often the case that the wild animals cannot be effectively tracked, resulting in the omission of the collected image data. And the detection time for wild animals lasts from a few months to as long as several years. The task of shooting on-site is a massive challenge to the staff’s body and mind. However, arranging many cameras in the environment where wild animals live will increase the cost. Because the trajectory of wild animals is uncertain and the range of activities is extensive, cameras in specific locations may not collect any image data about wild animals at all. The camera utilization rate is low, and the image data, camera, and battery working conditions cannot be obtained in time. So, the information transmission is lagging, and the management and maintenance workload is enormous. There are many shortcomings in the traditional wildlife monitoring schemes of following scene shooting or arranging multiple cameras, making the research work challenging to carry out smoothly.

Therefore, based on the above analysis, how to adopt new solutions to effectively monitor the living environment of wild animals, collect relevant image data and activity trajectory data, and then conduct research and investigation on the diversity of wild animals to realize the protection of wild animals, is one of the urgent problems that need to be solved at present.

The wireless sensor network has many characteristics such as large scale, integration, low cost, and self-organizing network. It is very advantageous to realize the monitoring of wild animals [3, 4]. Therefore, designing an intelligent wildlife monitoring system based on wireless sensor networks for wildlife protection provides a modern means for the research, protection, and management of wildlife, while also saving human resources and reducing work complexity[5, 6]. In the actual field environment, the sensor nodes realize the monitoring of wild animals through cooperation to obtain the image data of the relevant wild animals. However, this still faces two problems: How to repair the coverage loopholes caused by dead nodes and maintain high coverage of the wildlife monitoring area, how to use the wireless sensor node scheduling algorithm to realize the effective use of the energy of the sensor node, and then extend the lifetime of the entire monitoring network.

In response to the above problems, this article focuses on the intelligent wildlife monitoring system’s target coverage monitoring and energy-saving algorithms based on wireless sensor networks.

At present, the research on energy consumption and optimal coverage of wireless sensor networks has made certain achievements. Roselin proposed an efficient connection coverage algorithm [7], aiming to find the maximum number of intersection coverage sets while ensuring coverage and connectivity while minimizing sensor redundancy around the target. Compared with disjoint covering sets, intersecting covering sets can prolong the lifetime of the network.

Abdulhalim proposed a multi-layer genetic algorithm [8]. The goal of the algorithm is to find the maximum number of intersecting covering sets. While ensuring complete coverage, it also ensures that the minimum number of sensors is allocated to each intersection covering set. The algorithm uses a post-heuristic operation, in which unassigned sensors can be used to enhance the coverage of each group of coverage sensors or simplify coverage settings. Senouci [9] proposed a sensor node deployment algorithm PSODA based on particle swarm optimization to solve the deterministic deployment of node coverage in wireless sensor networks. Yeasmin [10] researched the problem of k-degree coverage and proposed the k-coverage verification protocol and the k- coverage sleep scheduling protocol. Wen Tao [11] and Liu Caixia [12] proposed a redundant node sleep scheduling algorithm, the former adopts different scheduling strategies for boundary nodes and internal nodes, and the latter divides the functions of intermediate nodes, which makes the redundant nodes go to sleep directly after deployment, and effectively extends the network lifetime. Danratchadakorn et al. [13] proposed a distributed and localized sleep scheduling protocol based on the coverage maximization and sleep scheduling (CMSS) protocol. The protocol assumes that sensor nodes are randomly deployed in grid cells, and the primary purpose is to satisfy all the target points while reducing the number of active nodes as much as possible. Torkestani [14] proposed a heuristic algorithm based on an automatic learning machine to find the approximate optimal solution of the connected dominating set problem extended by the minimum weight constrained by the agent equivalence. Lin [15] proposed two advanced deployment algorithms named EDA-1 and EDA-2, which maximize the coverage of heterogeneous directional mobile networks. Xu Daoqiang [16] proposed the partial coverage algorithm of the probability model, using the probability model to screen the coverage nodes that meet the conditions. Xu Mengying [17] proposed an improved chaotic immune hybrid frog leaping algorithm, which immunized and mutated individuals with high adaptability and low adaptability, thereby improving the optimal local solution and the optimal global solution, which is effectively increased the number of targets successfully monitored. Kim et al. [18] proposed a sensor node activation method based on natural heuristic algorithm. Zameni [19] proposed a node deployment algorithm for target coverage in a wireless sensor network, using genetic algorithms to build a grid on a subset of locations, and then proposed integer linear programming to install sensors at each grid location, reducing the number of sensors. Xu Qinshuai [20] proposed a network coverage optimization method based on a hybrid strategy to improve the Antlion algorithm, which effectively improved the node coverage of wireless sensor networks. Yun Xu [21] proposed a new routing protocol — Improved Distributed Energy Effective Clustering (I-DEEC), which is based on the node scheduling of genetic algorithm and makes redundant nodes sleep through proper scheduling. When it is exhausted, it replaces the dead node, which effectively improves network’s life.

In terms of cost, Naureen [28] proposed a novel positioning and mobility modeling system to aggregate the collected data and upload it to the base station. Maroto [29] proposed to install some of the animals with a global spot system collar and some with a low-cost Bluetooth tag. A path planning method based on Markov decision [30, 31] is proposed to better distribute the sensors by obtaining the trajectory of the animal.

The above methods significantly improve the efficiency of finding the optimal solution, reduce the redundancy of data transmission when the sensor repeatedly covers the target, and improve the coverage rate of the monitoring target and the utilization rate of the sensor node. However, there are also problems of falling into local optimality, premature stagnation, and failure to consider network fault tolerance, resulting in multiple coverages or partial coverages of the target area. This paper proposes an energy-saving coverage control mechanism based on a cultural genetic algorithm, which converges faster than traditional EGDG, LEACH, and PEGASIS algorithms and solves the problem of node coverage control well. While realizing coverage control, redundant nodes can be quickly located and set to a dormant state to save network energy. At the same time, a wake-up scheme is proposed, which can quickly replace dead nodes and maintain coverage.

2 Optimization of cultural genetic algorithm based on energy-saving control

The cultural genetic algorithm is like the genetic algorithm. The difference is that the cultural genetic algorithm integrates a local search operator; its purpose is to use one or more heuristic searches to improve the crossover and mutation solutions of genetic algorithms for specific problems. Therefore, compared with a genetic algorithm, cultural genetic algorithm has faster convergence speed and better performance. This article optimizes based on cultural genetic algorithm and introduces a wake-up scheme to make node scheduling more efficient and reasonable.

2.1 Cultural genetic algorithm

The cultural genetic algorithm includes two parts: genetic algorithm and local search operator, as shown in Fig. 2-1. The initial population of the algorithm is usually generated randomly or in a certain way. After each iteration of the genetic algorithm, the cultural genetic algorithm will use the local search operator to find a better solution, and iteratively, until the algorithm meets certain conditions and terminates. It finds the minimum sensor nodes required to monitor the target point in the area so that the new population finally obtained is very close to the optimal global solution.

Fig. 2-1

Flow chart of cultural gene algorithm.

Compared with the problem of k-degree coverage in sensor networks (each target point in the sensing area is covered by at least k or more sensor nodes), this paper only considers 1-degree coverage. That means that at least one or more nodes cover each target point in the sensing area. Usually, the redundant nodes selected by the node scheduling algorithm must enter the sleep state to save node energy to achieve the maximum area coverage and maintain the minor energy loss. If a node runs out of energy and becomes a dead node, one or more sleep nodes near the dead node must be awakened immediately to replace the dead node and continue to monitor. This kind of node scheduling algorithm achieves 100 % coverage of the sensing area as much as possible.

2.1.1 Perceptual coverage model

It was given a two-dimensional plane target area R with an acreage of L_x × L_y, as shown in Fig. 2-2. There are N sensor nodes in area R, defined as {C = c₁, c₂, c₃, … , c_N }, where c_i = { x_i, y_i, r_s } i ∈ [1, N]. {x_i, y_i} is the position coordinates of the wireless sensor node. r_s is the sensing radius of the coverage area, and there are a total of M target points to be observed in area R, defined as P ={ p₁, p₂, p₃, … … , p_M },where the coordinates of node p_j are {x_j, y_j}. In addition, a binary variable $R_{i, j}$ is introduced to indicate whether the sensor node c_i covers the target point p_j [22], which $R_{i, j}$ is defined as follows:

Fig. 2-2

Observation area coverage model.

$R_{i, j} = {\begin{matrix} 1, if {(x_{i} - x_{j})}^{2} + {(y_{i} - y_{j})}^{2} < r_{s}^{2} \\ 0, otherwise \end{matrix}$ (1)

That is when $R_{i, j} = 1$ , the sensing coverage radius r_s of the sensor nodes is greater than the straight distance between c_i and p_j, and the sensor nodes c_i can successfully cover the target point p_j, then the animal activities can be monitored in the area. However, the target p_j may be covered by v sensor nodes at the same time. For example, for a specific target point p_j, if it is covered simultaneously by two or more sensor nodes, such as (c₁, c₂, c₃, …, c_v, v ⩽ N), the union of the binary variables $R_{1, j}, R_{2, j}, R_{3, j}, \dots, R_{v, j}$ of the target point p_j can be obtained by Boolean algebra:

$R_{1, j} \lor R_{2, j \dots} \lor R_{v, j} = 1 - \bar{R_{1, J}} \land \bar{R_{2, J}} \dots \land \bar{R_{v, J}}$ (2)

Where $\bar{R_{1, J}}$ means that $R_{1, j}$ is negate, ∨ means Boolean union, ∧ means Boolean intersect.

2.1.2 Energy loss model

Figure 2-3 shows the energy loss model in the wireless sensor network [23, 24]. Since the energy consumed by the wireless communication of sensor nodes accounts for a relatively large proportion of the energy utilization of the entire network, other energy consumption of the nodes may not be considered.

Fig. 2-3

Energy loss model.

Among them E_{T
_x} is the total energy consumption of sending H bit data packets, and E_{R
_x} is the total energy consumption of receiving H bit data packets. It is assumed that the initial energy of all sensor nodes is the same, and other power sources supply the sensor sink node, so the sink node is not considered in this energy model. Figure 2-3 E_elec represents the energy loss per bit of the transmitting or receiving circuit, and E_amp represents the energy loss per bit of the power amplifier circuit, and β represents the path loss index. Therefore, when the transmitting end sends an H bit packet to the receiving end, the total energy loss [25] is:

$E_{T_{x}} (d, H) = E_{elec} \times H + E_{amp} \times H \times d^{β}$ (3)

The energy loss of the receiving end to receive the H bit data packet is:

$E_{R_{x}} (d, H) = E_{elec} \times H$ (4)

2.1.3 Genetic operation

Genetic operations include selection, crossover, and mutation [26]. In the selection operator, this article adopts the tournament selection method. Each round randomly selects a certain number of chromosomes from the population and then selects the chromosomes suitable for crossover from the selected chromosomes. We know that the alleles on the chromosomes represent the on/off state of the node from above. The higher the fitness value of the chromosome, the more reasonable the on/off scheduling of the node. Each round randomly selects ten chromosomes, taking the highest and second fitness as parent chromosomes. Then the two parent chromosomes produce child chromosomes by crossing.

The crossover rate needs to be set [27] to maintain the diversity of the population. The crossover rate R_c indicates the possibility of two-parent chromosomes to produce child chromosomes through the crossover. The crossover rate is usually set to a relatively large value, and more child chromosomes can be generated through the crossover operation to update the gene pool. In this paper, a single-node crossover is used to exchange the right parts of the two-parent chromosomes at random positions to obtain two new chromosomes.

In genetic operations, mutation operators generally occur after crossover operations and generate new chromosomes by changing the value of one or more alleles on the chromosome, thereby effectively maintaining the diversity of the population and speeding up the convergence of the solution speed. The mutation rate needs to be set to a small value; otherwise, the entire genetic algorithm will degenerate into a random search. Since different crossover rates and mutation rates have different effects on the solution, here, to simplify the study, crossover rates and mutation rates are set to specific values.

2.1.4 Gene coding

Table 2-1 shows the gene coding of the energy-saving coverage maximization mechanism. The table Λ represents the total number of chromosomes. The length of each chromosome is N, and different population sizes have different effects on the optimal solution of the algorithm. Here we simplify the research and set the population size to a fixed constant based on experience. Moreover, allele ℓ_i,j represents the state of the sensor node c_j in the chromosome i; 1 represents the node is working, and 0 is in a sleep state. For example, Fig. 2-4 shows that chromosome k has ten sensors; the states are 1010110010, which c₂, c₄, c₇, c₈, c₁₀ are set to sleep state and c₁, c₃, c₅, c₆, c₉ are set to work usually.

Table 2-1
Gene coding of energy-saving coverage optimization mechanism

Chromosomes\Alleles ℓ_i,1 ℓ_i,2 ℓ_i,3 ... ℓ_i,j ... ℓ_i,N

Chromosome 1 1 0 1 ... 0 ... 0

Chromosome 2 1 1 1 ... 0 ... 1

: : : : : : : :

Chromosome Λ 1 0 0 ... 1 ... 1

Chromosomes\Alleles	ℓ_i,1	ℓ_i,2	ℓ_i,3	...	ℓ_i,j	...	ℓ_i,N
Chromosome 1	1	0	1	...	0	...	0
Chromosome 2	1	1	1	...	0	...	1
:	:	:	:	:	:	:	:
Chromosome Λ	1	0	0	...	1	...	1

Fig. 2-4

Chromosome k (10 sensor node of wireless coverage six target point), the solid line indicates the node is open, the dotted line indicates the node is closed, and the black dot indicates the target point.

2.1.5 Fitness function

In the genetic algorithm, the fitness function evaluates the alleles on each chromosome and calculates the fitness value of the chromosome. The genetic algorithm determines whether the chromosome is good enough according to the size of the fitness value so that it can stay in the original population to produce the next generation population. Therefore, the superiority of each sensor node scheduling takes the fitness function as an essential reference index.

According to the aforementioned perceptual coverage model, the coverage vector of a sensor node c_i is defined as: $π_{i} = {R_{i, 1}, R_{i, 2}, \dots, R_{i, M}}$ , c_i ∈ C, and for another sensor node c_j ∈ C, the coverage vector is also defined: $π_{j} = {R_{j, 1}, R_{j, 2}, \dots, R_{j, M}}$ . The following formula can calculate the composite cover vector of c_i and c_j:

$\begin{matrix} ϖ (i, j) = π_{i} \lor π_{j} = [R_{i, 1} \lor R_{j, 1}, \\ R_{i, 2} \lor R_{j, 2}, \dots, R_{i, M} \lor R_{j, M}] \end{matrix}$ (5)

ϖ (i, j) indicates whether sensor nodes c_i and c_j cover each target point. Therefore, the composite coverage vector of chromosome k can be defined as:

$ϖ (k) = ((ℓ_{k, 1} \cdot π_{1}) \lor (ℓ_{k, 2} \cdot π_{2}) \dots \lor (ℓ_{k, N} \cdot π_{N}))$ (6)

In addition, define the calculation expression of chromosome k coverage [21]:

$ɛ^{k} = \frac{∥ ϖ (k) ∥^{2}}{M}$ (7)

Among them, ∥ϖ (k) ∥ ² indicates the number of target points that have been covered. The calculation expression that defines the node utilization rate of the chromosome k is:

$ℵ_{t}^{k} = \frac{\sum_{g = 1}^{N} ℓ_{k, g}}{N}$ (8)

Among them, $\sum_{g = 1}^{N} ℓ_{k, g}$ indicates the number of selected activated nodes. Define the fitness calculation expression of chromosome k as: $f_{c}^{k} = α_{1} \cdot {(ɛ^{k})}^{λ_{1}} - α_{2} \cdot {(ℵ_{t}^{k})}^{λ_{2}}$ (9)

Substitute the expressions of ɛ^k and $ℵ_{t}^{k}$ into the above equation: $f_{c}^{k} = α_{1} \cdot {(\frac{∥ ϖ (k) ∥^{2}}{M})}^{λ_{1}} - α_{2} \cdot {(\frac{\sum_{g = 1}^{N} ℓ_{k, g}}{N})}^{λ_{2}}$ (10)

Assuming that the weighting coefficients α₁ = α₂ = 1 and the index λ₁ = 2, λ₂ = 0.5, the boundary conditions are: 0 ⩽ ɛ^k ⩽ 1, $0 ⩽ ℵ_{t}^{k} ⩽ 1$ , $- 1 ⩽ f_{c}^{k} ⩽ 1$ . Note that when ɛ^k increases or $ℵ_{t}^{k}$ decreases, $f_{c}^{k}$ will increases. That is to say, the higher the coverage rate, the less sensor nodes used, the better. In other words, the larger the value of $f_{c}^{k}$ indicates that the node on/off schedule in the chromosome k is reasonable.

2.1.6 Local search operator

A local search operator is used here to speed up the convergence of the optimal solution and improve the superiority of the population generated by the genetic algorithm. For each population generation, the operator successively changes each “1” allele to “0” on each chromosome. And compares the fitness values of the chromosomes before and after the change to determine whether the changed chromosome should be retained. Let Pop_k = { b₁, b₂, ⋯ , b_p } _k is the population k, which is a collection of a series of chromosomes. The pseudo-code is as follows:

Algorithm 1: Local search algorithm
Input: The population is Pop_k = { b₁, b₂, ⋯ , b_p } _k, Let b_i become the i-th chromosome in Pop_k, and N is the length of chromosome b_i. According to the definition of alleles ℓ_i,j, a particular string of all alleles ℓ_i,j is chromosomes b_i, where ∀i1 ⩽ j ⩽ N.
Output: Optimized population Pop_k
1: For i = 1 to p do
2: For j = 1 to N do
3: If ℓ _i,j = 1 then
4: Array S_i stores the value of ℓ_i,j // S_i is
used to store the “1” allele on each chromosome
5: End If
6: End For
7: End For
8: For i = 1 to p do
9: Let m bethelengthofthearray S_i
10: For k = 1 to m do
11: Fitness1 = fit (b_i); //fit () is the fitness function
12: For the k-th element in S_i, set the allele l_i,tmp in b_i be equal to 0
13: Fitness 2 = fit (b_i); // Recalculate the fitness function
14: If Fitness 1 > Fitness 2 then
15: Let l_i,tmp be equal to one
16: End If
17: End For
18: End For

If the best quality solution has not changed in each of the following populations, the algorithm will terminate, to get the best scheduling scheme of sensor nodes. Note that it is very vital to determine the appropriate chromosome value in each generation of the population. If many chromosomes are selected, the convergence speed of the algorithm will be reduced.

2.2 Wake-up scheme

A wake-up scheme is proposed to repair the coverage loopholes caused by the exhaustion of node energy. In the cluster sensor network, the sensor nodes randomly form clusters. In each cluster, only one sensor node can be used as the cluster head. The cluster sensor network uses time division multiple access to realize the communication between the sensor nodes and the cluster head node. In addition, the cluster head node can directly communicate with the sensor sink node and is responsible for forwarding the information received by the cluster head node to the sensor sink node. After the cluster head node is selected, time division multiple access scheduling will be set up. After the setup phase is over, the sensor nodes will be in the steady-state phase, and the collected data packets will be forwarded to the cluster head node, and the cluster head node will merge the received information and then forward it to the sensor sink node. In a clustered sensor network, each round can be re-clustered. In each round of elections, each cluster will select a new cluster head node, and each sensor node can directly communicate with the sensor sink node.

According to the sensor node scheduling scheme, some sensor nodes will be switched to a sleep state during initial deployment, and the remaining sensor nodes are responsible for sensing tasks. When a sensor node is about to run out of energy, the wake-up scheme will wake up some neighboring sensor nodes in the next round to repair the coverage hole caused by the death of the node. It is assumed that the sensor sink node has enough energy to implement the energy-saving coverage control mechanism and wake-up scheme. When the sensor sink node receives a signal from a node about to run out of energy, it will decide which sensor nodes to activate in the next round.

Based on the previously defined coverage vector, the sensor coverage can be evaluated using bit Boolean operations. If a sensor node n_i run out of energy, the sensor sink node will re-evaluate the coverage of the cluster sensor network through the following calculations and find the uncovered target points.

$π_{uncovered}^{n_{i}} = π_{n_{i}}^{▵} \oplus π_{{ns}_{i}}^{\nabla}$ (11)

Among them, ⊕ represents an exclusive OR Boolean operation. $π_{n_{i}}^{Δ}$ represents the initial composite coverage vector of all sensor nodes in the open state, including the sensor node n_i. $π_{n_{i}}^{\nabla}$ represents the composite coverage excluding the sensor node n_i. $π_{uncovered}^{n_{i}}$ is a vector that represents the target points which n_i is not covered.

Regarding the algorithm of the wake-up scheme, the following definitions are introduced:

Definition 1. C is a collection of all sensor nodes, and nodes n_i ∈ C, i ∈ [1, N].

Definition 2. Let u_i be the set of all adjacent nodes of the node n_i,i ∈ [1, N], and define u_i ={ n_j ∈ C|distance (n_i, n_j) < λ · r_s, i ≠ j, j ∈ [1, N] }, where λ is a specific value. If the straight-line distance between the two sensor nodes is not greater than λ · r_s, then define the two nodes as adjacent nodes.

Definition 3. Definition $ψ_{i} = {n_{*} \in u_{i} | \sum_{\forall h} π_{*} \cap π_{uncovered}^{n_{i}} ⩾ 1}$ , where ψ_i represents sensor nodes with overlapping target points between $π_{uncovered}^{n_{i}}$ and π_*. π_* represents the composite coverage vector of n_i’s adjacent nodes. h is the vector element in $π_{*} \cap π_{uncovered}^{n_{i}}$ . Let S_{ψ_i,k} be the k-th subset of ψ_i. χ means the number of ψ_i subsets.

The sensor sink node will search n_i’s adjacent nodes and wake them up to repair the coverage loopholes caused by dead nodes n_i through the following algorithm. The pseudo-code is as follows:

Algorithm 2: Wake-up scheme
Input: variants: bestcom, bestval = 0; map, res = [0, 0, ⋯ , 0] ,
length = M;
Output: Optimal scheduling scheme bestcom;
1: If node n_i runs out of energy then
2: For k = 1 to χdo
3: For each element d in S_{ψ_i,k} do // Node d in
the k-th subset
4: $map = π_{d} \cup π_{uncovered}^{n_{i}}$ // The set of target points
covered by the k-th subset
5: End For
6: $res = map \cap π_{uncovered}^{n_{i}}$ // The set of target points not
covered by n_i that can be covered by the k-th subset
7: $If \sum_{γ \in res} γ + \frac{1}{size (S_{ψ_{i}, k}) + 1} > bestval then$
8: $bestval = \sum_{φ \in res} φ + \frac{1}{size (S_{ψ_{i}, k}) + 1};$
9: bestcom = S_{ψ_i,k} // The optimal sensor node set
is obtained
10: End If
11: End For
12: Output:bestcom;
13: End If

As mentioned above, the sensor sink node will activate the nodes searched by the wake-up scheme at the beginning of the next round. If a sensor node in the network is about to run out of energy, the cluster-head node adjacent to this node will send relevant information to the sensor sink node. At this time, the wake-up plan will be activated, then the coverage area will be re-evaluated, and the other adjacent node near the dead node will be awakened to repair the coverage blind area. In this way, the coverage control of the target point can be effectively achieved.

3 Experiment and analysis

3.1 Simulation parameters

The mechanism is simulated on the MATLAB platform, aiming at the performance of the energy-saving coverage control mechanism proposed in the paper in extending the network lifetime and maintaining the coverage rate of the monitoring area. Considering that the photos taken by the monitoring camera in the range of 1–20 m are explicit and the imaging results are relatively satisfactory, the sensing radius of the simulation parameter sensor node can be made 16.5 m.

Table 3-1
Energy cost

Node Status Current Description

Stand-by 2.36mA

Transmitting data 37.4mA 5 ms transmission time

Dormant 0.245mA

Node Status	Current	Description
Stand-by	2.36mA
Transmitting data	37.4mA	5 ms transmission time
Dormant	0.245mA

Table 3-2

Simulation parameters

(1) The sensor node’s sensing radius r_s=16.5 m.

(2) λ is the judgment parameter of adjacent nodes (the distance between nodes is less than 16.5 m).

(3) The population number η=20, the crossover rate R_c=0.6, and the mutation rate R_m=0.07.

(4) E_elec=50nJ/bit, E_amp=100 pJ/bit/m², E_DA=5nJ/bit/report, β=2.

(5) The sensor sinks node coordinates (50,200), far away from the sensing area and sufficient energy.

(6) The size of each data packet is 2000bit, H = 2000.

(7) The initial energy of the node is 0.25 J.

The architecture of the wireless sensor node is shown in Fig. 3-1. It consists of a lithium battery, a power management circuit, three sets of pyro-infrared sensor modules, and a CC2530 controller that includes Zigbee transceiver functions.

Fig. 3-1

Architecture diagram of wireless sensor node.

Two 3000mAh rechargeable lithium batteries provide the power of the node in parallel. The battery capacity guarantees that the node can work continuously for more than 60 days under normal working conditions. The power management circuit realizes the optimal energy-saving state through the controller to achieve the energy scheduling of the entire node. The sensor detection part of the node is composed of three groups of pyroelectric sensors and the corresponding Fresnel lens. Since the selected Fresnel lens has a detection angle of 120°∼140°, the use of three sets of pyroelectric sensors can cover a complete 360° area.

Due to the difference in wildlife distribution and topographic characteristics in the three specific monitoring sample areas in the West Dongting Lake National Nature Reserve in Hunan, the corresponding wildlife sensors are arranged in the nature reserve’s three specific monitoring sample areas, as shown in Fig. 3-2. The sensor detection network in the intelligent wildlife monitoring system is a cluster sensor network in which all sensor nodes have the same initial energy and can communicate with the sensor sink node and other sensor nodes in the detection area. The sensor sink node uses this mechanism to control the on/off status of sensor nodes.

Fig. 3-2

Three specific monitoring sample areas.

(1) Monitoring sample area a: After viewing the wildlife diversity monitoring data and field trips in this nature reserve, it is found that the area is affluent in wildlife diversity, where birds account for the most significant proportion. Because bird animals have a long flying distance, extensive range of activities, and long-life trajectory, the method of evenly arranging sensor nodes in the area is adopted to record more wild animal image data in a more extensive monitoring range. Figure 3-3 shows the simulation results of monitoring sample node deployment in the area. Among them, the size of the simulation parameter monitoring area is a square area of 100m×100 m, and 64 target points and 200 sensor nodes are evenly distributed in the area.

Fig. 3-3

Wireless sensor network for monitoring sample area a.

Figure 3-3 shows that when the wildlife monitoring network is uniformly deployed, through this energy-saving coverage control mechanism, the sensor sink node can quickly locate the corresponding redundant sensor nodes and shut it down to save energy. Figure 3-3(left) shows the initial uniform deployment of sensor nodes and target points in the monitoring sample area. Figure 3-3(right) shows the situation where the target point is covered after applying the node scheduling based on the cultural gene algorithm. Among them, the red dot is the target point, and the blue circle is the monitoring area of the sensor node. It shows that only 5.5% of the sensor nodes activated successfully covered all the target points in the monitoring area, achieving full coverage.

(2) Monitoring sample area b: After research and investigation, it is found that there is a river flowing through the monitoring sample b area. Near the water source, there are many footprints of wild animals such as birds and mammals. From this, it can be judged that arranging sensor detection network nodes along the direction of the river can most effectively obtain many precious wild animal image data. Figure 3-4 shows the simulation results of the deployment of sensor nodes in the monitoring sample b area. The simulation parameters monitoring area is a square area of 100m×120 m, and 20 target points and 100 sensor nodes are deployed evenly on both sides of the river. Figure 3-4 (left) shows that there are many redundant nodes in the monitoring network. Figure 3-4 (right) shows the coverage of target points after applying the node scheduling based on the cultural gene algorithm. 96% of the sensor nodes have been switched to a sleep state, and the sensor sink node only needs to activate 4% of the sensor nodes to successfully achieve full coverage of all target points near the river.

Fig. 3-4

Wireless sensor network for monitoring sample area b.

(3) Monitoring sample area c: This area is located at the edge of the nature reserve. The purpose of deploying an intelligent wildlife sensor network in this area is to use sensor nodes to monitor human activities and to monitor whether there is poaching by ‘allies’ or criminals who enter the nature reserve from the outside. Because the trajectory of human activities is largely unknown, the method of randomly deploying monitoring nodes is adopted. The simulation parameter monitoring area is a square area of 160m×160 m, and 64 target points and 150 sensor nodes are randomly deployed in this region. Figure 3-5 shows the simulation results. After applying the sensor node scheduling based on the cultural genetic algorithm, many redundant nodes initially deployed randomly are shut down and set to sleep state. Only by activating 18% of the sensor nodes, a degree of coverage of all target points be achieved, which saves the energy of the monitoring system while maintaining a high coverage rate.

Fig. 3-5

Wireless sensor network monitoring sample area c.

It can be seen from the above results that the algorithm can effectively activate the sensor nodes according to the topographic features and achieve the effect of complete coverage of the area. By making redundant nodes sleep, unnecessary energy consumption is avoided, costs are saved, and the life cycle of the network is prolonged.

3.2 Analysis of results

The following simulation results verify the feasibility of deploying an intelligent cluster sensor detection network in nature reserves. If the sensor sink node has enough energy to set the best scheduling plan for each sensor node, when the sensor node’s energy is exhausted, the sensor sink node will wake up adjacent sensor nodes to repair the coverage blind zone. Therefore, the sensor sink node can simultaneously affect the number of activated sensor nodes and the coverage rate in the detection area. In the case of 400 sensor nodes and 64 target points evenly distributed in a 100m×100 m monitoring area, this paper compares the performance of the energy-saving coverage control mechanism and the three algorithms of EGDG, LEACH, and PEGASIS. Figure 3-6 shows the coverage rate of a specific monitoring area under a given number of rounds in the wildlife monitoring network. It indicates the energy-saving coverage control mechanism maintained 100% coverage of the monitoring area before the 1500 rounds. However, after 3,500 rounds, the superiority of the energy-saving coverage control mechanism is reflected, and the network life cycle is about 1,000 rounds longer than that of EGDG and other algorithms. The performance of the EGDG, LEACH, and PEGASIS algorithms in maintaining the perceived coverage rate is far lower than the energy-saving coverage control mechanism.

Fig. 3-6

Perceived coverage under specific number of turns.

Figure 3-7 represents a death sensor node percentage at a particular number of rounds of the wildlife monitoring network. The number of death sensor nodes of the EGDG, LEACH, and PEGASIS algorithms reached 50% in 1950, 1120, and 300 rounds, respectively. The PEGASIS and LEACH algorithm nodes died so quickly is because these two algorithms did not apply the sensor node scheduling scheme. The energy-saving coverage control mechanism and the EGDG algorithm can apply the node scheduling plan to shut down the corresponding redundant nodes and save energy to extend the network life cycle of the wild-only detection system. It indicates that the energy-saving coverage control mechanism has better performance than the other three algorithms in extending the sensor network’s life cycle.

Fig. 3-7

Percentage of dead nodes under specific number of turns.

The above simulation results show that when deploying an intelligent wildlife sensor network in the wild environment, the application of energy-saving coverage mechanism can extend the life cycle of the system network while still maintaining a high perception coverage of the monitoring area. Through continuous monitoring, many clear images of wild animals can be obtained, which significantly improves the system’s service quality.

4 Conclusion

Aiming to deploy front-end sensor nodes of the intelligent wildlife monitoring system, an energy-saving coverage control mechanism is proposed, including sensor node scheduling and a wake-up scheme based on a cultural gene algorithm. This mechanism can quickly locate redundant nodes in a specific monitoring area, set them to sleep state, and repair coverage loopholes caused by dead nodes, to maintain a high coverage rate of specific monitoring areas. Simulation results show that the performance of the proposed mechanism is better than the EGDG, LEACH, and PEGASIS algorithms in terms of coverage rate and a lifetime of the sensor network, which effectively improves the service quality of the intelligent wildlife monitoring system.

References

Rongchang

and Yufeng

, Reflection and Reconstruction of Wildlife Protection Law under the Background of COVID-19 Outbreak, Journal of Jining University 41(03) (2020), 50–56.

Awuken

, Protecting wild animals means protecting human beings, New Agriculture 2021(03), 61–63.

Shan’an

, Chunhe

, Yuan

, et al., Design of Wildlife Image Monitoring System Based on Wireless Sensor Network, Modern Manufacturing Technology and Equipment 2017(3), 64–66.

Zhishu

, Xinhai

, Xuezhi

, et al., Discussion on the monitoring standard of forest wildlife by infrared camera in China, Biodiversity 22(6) (2014), 704–711.

Yingzai

, Xinlin

, Mingtong

, et al., The application of infrared camera technology in wildlife monitoring on the western slope of the south section of Gaoligongshan National Nature Reserve, Anhui Agricultural Sciences 048(004) (2020), 108–111.

Riqi

, Application of infrared camera technology in wildlife monitoring research in China, South China Agriculture 15(02) (2021), 198–199.

Roselin

, Latha

and Benitta

, Maximizing the wireless sensor networks lifetime through energy efficient connected coverage[J], Ad Hoc Networks 62(Jul.) (2017), 1–10.

Abdulhalim

M.F.

and Attea

B.A.

, Multi-layer Genetic Algorithm for Maximum Disjoint Reliable Set Covers Problem in Wireless Sensor Networks[J], Wireless Personal Communications 80(1) (2015), 203–227.

Senouci

M.R.

, Daoud

, Farouk

, et al., Static wireless sensor networks deployment using an improved binary PSO[J], International Journal of Communication Systems 29(5) (2016), 1026–1041.

10.

Yeasmin

, k-Coverage Problems and Solutions in Wireless Sensor Networks: A Survey[J], International Journal of Computer Applications 100(17) (2014), 1–6.

11.

Tao

, Dongqing

, Quan

, et al., Sleep scheduling algorithm for redundant nodes in wireless sensor networks, Journal of Communication 35(10) (2014), 67–80.

12.

Caixia

and Shude

, An efficient node dormancy scheduling algorithm for prolonging WSN lifetime, Journal of Guilin Institute of Aerospace Technology 25(01) (2020), 21–24.

13.

Danratchadakorn

and Pornavalai

, Coverage maximization with sleep scheduling for wireless sensor networks[C]//, International Conference on Electrical Engineering/electronics. IEEE, 2015, 1–6.

14.

Torkestani

J.A.

, An Energy-Efficient Topology Control Mechanism for Wireless Sensor Networks Based on Transmit Power Adjustment[J], Wireless Personal Communications: An Internaional Journal 82(4) (2015), 2537–2556.

15.

Lin

T.Y.

, Santoso

, Wu

K.R.

, et al., Enhanced Deployment Algorithms for Heterogeneous Directional Mobile Sensors in a Bounded Monitoring Area[J], IEEE Transactions on Mobile Computing 16(3) (2017), 744–758.

16.

Daoqiang

, Bo

, He

, Zhaojian

and Zhiguo

, A coverage algorithm for optimizing wireless sensor networks[J], Journal of China Academy of Electronics 14(08) (2019), 786–790.

17.

Mengying

, Yi

and Jie

, A target coverage optimization algorithm in wireless sensor networks, Application of Electronic Technology 46(7) (2020), 94–98.

18.

Kim

and Yoo

, Sensor Node Activation Using Bat Algorithm for Connected Target Coverage in WSNs, Sensors 20(13) (2020), 3733.

19.

Zameni

, Rezaei

and Farzinvash

, Two-phase node deployment for target coverage in rechargeable WSNs using genetic algorithm and integer linear programming[J], The Journal of Supercomputing 2020(12), 1–29.

20.

Qinshuai

, Qing

and Kangyuan

Wei

, Wireless sensor network coverage optimization based on improved ant-lion algorithm, Journal of Sensing Technology 32(2) (2019), 266–275.

21.

, Jiao

and Tian

, Energy-Efficient Connected-Coverage Scheme in Wireless Sensor Networks[J], Sensors 20(21) (2020), 6127.

22.

Zhang

H.H.

and Hou

J.C.

, Maintaining Sensing Coverage and Connectivity in Large Sensor Networks[J], Ad Hoc & Sensor Wireless Networks 1(1–2) (2005), 89–124.

23.

Behera

T.M.

, Samal

U.C.

and Mohapatra

S.K.

, Energy-efficient modified LEACH protocol for IoT application[J], IET Wireless Sensor Systems 8(5) (2018), 223–228.

24.

Elbhiri

, Saadane

, Fkihi

S.E.

, et al., Developed Distributed Energy-Efficient Clustering (DDEEC) for heterogeneous wireless sensor networks, 2010 5th International Symposium On I/V Communications and Mobile Network, IEEE (2010), 1–4.

25.

Kang

, Sohn

and Lee

S.H.

, Enhanced Message-Passing Based LEACH Protocol for Wireless Sensor Networks, Sensors 19(1) (2019), 75.

26.

Deb

, Pratap

, Agarwal

, et al., A fast and elitist multiobjective genetic algorithm: NSGA-II[J], IEEE Transactions on Evolutionary Computation 6(2) (2002), 182–197.

27.

Mohammed

H.J.

, Abdulsalam

, Abdulla

A.S.

, et al., Evaluation of genetic algorithms, particle swarm optimisation, and firefly algorithms in antenna design[C]//, 2016 13th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design (SMACD). IEEE, 2016, 1–4.

28.

Naureen

, Zhang

and Furber

, et al., A GPS-less Localization and Mobility Modelling (LMM) System for Wildlife Tracking[J], IEEE Access PP(99) (2020), 1–1.

29.

Maroto-Molina

, Navarro-García

, Príncipe-Aguirre

, et al., A Low-Cost IoT-Based System to Monitor the Location of a Whole Herd[J], Sensors 19(10) (2019), 2298.

30.

Vera-Amaro

, Angeles

and Luviano-Juarez

, Design and Analysis of Wireless Sensor Networks for Animal Tracking in Large Monitoring Polar Regions Using Phase-Type Distributions and Single Sensor Model[J], IEEE Access 7 (2019), 45911–45929.

31.

, Solmaz

, Rahmatizadeh

, et al., Internet of Things Applications: Animal Monitoring with Unmanned Aerial Vehicle[J]. arXiv preprint arXiv:1610.05287, 2016.