Network deployment method for heterogeneous data collection of UAV clusters based on multi-objective optimization algorithm

Abstract

The present invention belongs to the field of multi-objective optimization and relates to a method for deploying heterogeneous data acquisition networks for drone clusters based on a multi-objective optimization algorithm. This method uses spatial geometry knowledge to calculate a realistic model of the impact of obstacles on wireless sensor signal propagation, and combines this with real-world environments to optimize coverage and minimize energy consumption. The NDDEA algorithm is improved, and the improved multi-objective optimization algorithm is used to simulate the deployment of heterogeneous data acquisition networks for drone clusters. The present invention can effectively improve the coverage of heterogeneous data acquisition networks deployed for drone clusters, reduce the energy consumption of heterogeneous data acquisition networks for drone clusters, and better align with general reality. Furthermore, compared with other multi-objective evolutionary algorithms, the set of solutions optimized by this method exhibits better distribution and convergence.

Description

Technical Field

[0001] The present invention relates to the field of multi-objective optimization, and in particular to a method for deploying heterogeneous data acquisition networks of unmanned aerial vehicle clusters based on a multi-objective optimization algorithm. Background Art

[0002] The Chinese market is one of the largest IoT markets in the world, boasting a vast user base and significant growth potential. From smart homes to smart cities, from the industrial internet to intelligent transportation, there is vast market demand across all sectors. my country's IoT sector is booming, with IoT technology widely applied across various industries, including smart homes, smart cities, industrial automation, intelligent transportation, and healthcare. These applications not only improve production efficiency and quality of life, but also provide new impetus for industrial upgrading and transformation.

[0003] The Internet of Things (IoT) requires massive amounts of data, and drones are increasingly being incorporated into the IoT as data collection nodes. Drones offer numerous advantages in this area. First, they possess significant flexibility and maneuverability, enabling them to fly freely in diverse environments and easily reach inaccessible or dangerous areas, such as high altitudes, mountainous areas, and waterways, thus achieving comprehensive data collection coverage. Drones are also cost-effective, often offering lower costs for data collection than traditional human or ground-based equipment. They can cover larger areas and complete tasks in a shorter timeframe, improving efficiency and reducing costs. Furthermore, drones can transmit collected data in real time, enabling users to make informed decisions based on real-time insights. This real-time monitoring and feedback capability is crucial for applications requiring rapid response, such as disaster monitoring and rescue. Drones have a wide range of applications in IoT data collection. In agriculture, drones can be equipped with various sensors and cameras to capture and monitor farmland. By collecting information such as soil moisture, vegetation indices, and meteorological data, they help farmers implement precision agriculture management and improve crop production efficiency and quality. In environmental monitoring, drones can monitor environmental parameters such as air, water quality, and soil. They can hover or cruise in the air, equipped with various sensors to collect environmental data and help monitor conditions such as air pollution, water quality changes, and soil erosion. In the field of disaster monitoring and rescue, drones can quickly reach the disaster area after a natural disaster and conduct aerial reconnaissance and monitoring. They can be equipped with infrared cameras, smoke sensors, and other equipment to help rescue workers understand the disaster situation, search for survivors, and provide real-time data support for rescue operations. In summary, the application of drones in IoT data collection has broad application prospects, providing efficient and accurate data support for various industries and promoting the development of intelligent and information-based systems.

[0004] When performing data collection tasks, drone swarms face a variety of environments. For example, when natural disasters occur and walls or building materials collapse, the transmission characteristics of the communication signals of drone data collection nodes will change due to the influence of geological disasters, and the signal transmission distance will be significantly shortened. In order to maintain good communication and detection during data collection tasks, it is necessary to consider improving the method of deploying drone swarm data collection networks. To this end, this paper analyzes the impact of obstacles in the environment on drone swarm data collection networks, proposes performance evaluation indicators for drone swarm data collection network deployment problems, and establishes a mathematical model for drone swarm data collection network deployment optimization problems.

[0005] At present, drone data collection networks can be applied in multiple fields, such as environmental monitoring, smart agriculture, disaster relief, etc. Due to different application requirements in different fields, the capabilities of sensor nodes carried by drones vary, and heterogeneous networks are receiving increasing attention. At the same time, research on the deployment of drone data collection networks is still mainly at the academic level and has not yet been applied on a large scale in actual scenarios. To this end, the present invention proposes a heterogeneous network deployment model for drone data collection, which adopts different coverage according to different areas to be collected, with maximizing coverage and minimizing energy consumption as optimization goals, and proposes an improved multi-objective optimization algorithm. The method of selecting individuals for sub-critical layers and critical layers is adopted, and the vector angle and minimum neighborhood distance between the selected individuals and the associated vector are used as fitness, so as to increase the uniformity and diversity of the population solution distribution and obtain multiple different network deployment schemes. Summary of the Invention

[0006] The purpose of the present invention is mainly to address the current situation in which the research on the deployment of drone cluster data acquisition networks in existing work is still mainly at the academic level and has not yet been applied on a large scale in actual scenarios, and the shortcomings of the deployment of drone data acquisition networks in ideal environments. By combining a multi-objective optimization algorithm, the deployment of heterogeneous networks for drone data acquisition is realized in simulated real-world scenarios. The core idea of ​​the present invention: for the study of heterogeneous network deployment for drone data acquisition under obstacle interference, different perception radii are set according to the different execution functions of drone data acquisition, with maximizing coverage and minimizing energy consumption as optimization goals. In order to improve the diversity of population solutions, the present invention proposes an improved multi-objective optimization algorithm, which adopts a method for selecting individuals in sub-critical layers and critical layers, and uses the vector angle and minimum neighborhood distance between the individual to be selected and the associated vector as fitness, so as to increase the uniformity and diversity of the distribution of population solutions.

[0007] The technical solution of the present invention:

[0008] A method for deploying a heterogeneous data collection network for a drone cluster based on a multi-objective optimization algorithm includes the following steps:

[0009] Step 1: Collect 3D data of real scenes and build simulation scenes; randomly generate a set of UAV data collection network deployment points S, S = {s1, s2, ..., s N}; monitoring area T is the set of target points to be covered, T={t1,t2,....,t M The obstacle area R is the area that blocks the propagation of electromagnetic waves. The obstacle is a rectangular obstacle.

[0010] Step 2: Perform boundary detection on the drone data collection location in the simulation scene;

[0011] Step 3: Based on the characteristics of different types of drone data collection nodes, spatial geometry is used to calculate the extent to which obstacles hinder electromagnetic wave signal propagation at these nodes. Existing drone data collection network deployment methods mostly assume the absence of obstacles. However, in real-world network deployment scenarios, the impact of obstacles on drone data collection networks must be considered. Obstacles significantly reduce the transmission distance of communication signals. Using spatial geometry, the length l of the signal transmitted by the drone sensor through the obstacle is calculated. The greater the length, the greater the signal attenuation.

[0012] Step 4: Calculate the coverage and energy consumption of the drone data collection network based on the impact of obstacles on electromagnetic wave signal propagation and the actual drone data collection behavior;

[0013] Step 5: Convert the deployment problem of the UAV cluster heterogeneous data collection network into a multi-objective optimization problem, with maximizing coverage and minimizing energy consumption as the optimization goals;

[0014] Step 6: Using an improved multi-objective optimization algorithm, a method for selecting individuals in the sub-key layer and key layer is adopted. The vector angle and minimum neighborhood distance between the selected individual and the associated vector are used as fitness. This increases the uniformity and diversity of the population solution distribution, further optimizing the network coverage and energy consumption in the UAV cluster heterogeneous data collection network deployment model.

[0015] Step 7: Set the simulation scenario, UAV data collection nodes, and parameters of the improved multi-objective optimization algorithm, and conduct a simulation experiment on the UAV data collection network deployment.

[0016] The deployment height of the drone data acquisition node is not a fixed height, and different heights are selected depending on whether it is deployed on an obstacle; when the drone data acquisition network deployment point in step 2 falls outside the monitoring area, the position of the drone is corrected so that the drone is located at a data acquisition network deployment point that falls within the monitoring area.

[0017] Step 3 includes the following steps:

[0018] Find the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary, and then find the length l that the electromagnetic wave signal emitted by the drone data acquisition penetrates the obstacle. Since the obstacle is a rectangular parallelepiped, and the plane where the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary is parallel to plane yoz, in the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary, x is the horizontal coordinate of the known obstacle boundary. The next step is to find the y and z coordinates based on x.

[0019] 301) When calculating the y coordinate, the electromagnetic wave signal transmitted from the drone data collection node to the monitoring point is projected onto the xoy coordinate plane;

[0020] Among them, the coordinates of the drone data collection node are A (x1, y1, z1), the coordinates of the monitoring point are B (x2, y2, z2), connect AB, and intersect with the obstacle boundary at point D (x, y, z), draw a straight line through point A parallel to the x-axis, and draw a straight line through point B parallel to the y-axis, the two straight lines intersect at point C, connect AC, and intersect with the obstacle boundary at point E, construct similar triangles ΔAED ~ ΔABC, and we get The solution is:

[0021] 302) When calculating the z coordinate, the electromagnetic wave signal transmitted from the drone data collection node to the monitoring point is projected onto the xoz coordinate plane;

[0022] Same as 301), construct similar triangles ΔAED~ΔABC, and we get Right now The solution is:

[0023] 303) Through steps 301) and 302), the boundary point D (x, y, z) at which the electromagnetic wave signal of the drone data acquisition node penetrates the obstacle is calculated, and then the distance between point A and point D is calculated, that is, the length l of the signal emitted by the drone data acquisition node penetrating the obstacle. This length l affects the attenuation degree of the signal. The longer the length, the greater the attenuation degree of the signal.

[0024] Step 4 includes the following steps:

[0025] 401) UAV data collection node electromagnetic wave signal loss

[0026] The electromagnetic wave signal of the UAV data collection node will produce a certain loss during the propagation process. This loss consists of two parts: one is path loss, and the other is signal attenuation caused by obstacles.

[0027] The path loss P of the electromagnetic wave signal of the UAV data collection node PL1 The formula is as follows:

[0028]

[0029] Signal attenuation caused by obstacles P PL2 The formula is as follows:

[0030]

[0031] Therefore, the propagation loss P of the electromagnetic wave signal of the UAV data collection node is PL The formula is as follows:

[0032]

[0033] Where d is the Euclidean distance between the UAV data collection node and the monitoring point; d0 is the reference distance; n is the path loss exponent, which represents the growth rate of path loss with distance; n1 is the path loss exponent in free space; and n2 is the path loss exponent in other environments.

[0034] 402) Design of coverage indicators

[0035] The grid model method is used to calculate the coverage area of ​​the UAV data collection network by calculating the intersection points of the grid covered by the UAV data collection nodes;

[0036] Set a UAV data collection node s i The perception radius is R cov , the coordinates of the drone data collection node are (x s ,y s ,z s ), in three-dimensional space, for any point t in the monitoring area j , point t j The coordinates of (x t ,y t ,z t ), p(s i ,t j ) represents the drone data collection node s i Detect the target point t j The probability of UAV data collection node model is:

[0037]

[0038] Among them, d(s i ,t j ) is the UAV data collection node s i With point t j The Euclidean distance between them; α is the signal attenuation factor; PPL is the propagation loss of electromagnetic wave signals at the UAV data collection node, R i Represents the signal coverage radius of different communication nodes.

[0039] The coverage rate refers to the proportion of drone data collection node network coverage points to the grid points in the entire monitoring area, reflecting the coverage range of the drone data collection node network. The number of points covered by the drone data collection node network is expressed as:

[0040]

[0041] Among them, w1 is the weight of the UAV data collection node deployed in the area without obstacle coverage, and w2 is the weight of the UAV data collection node deployed in the area with obstacle coverage. Both are set hyperparameters. i ,t j ) refers to the drone data collection node s i Detect the target point t j probability.

[0042] In summary, coverage is defined as follows:

[0043]

[0044] Among them, Num is the number of points covered by the UAV data collection node network, and M is the number of all monitoring points in the target area.

[0045] 403) Design of energy consumption indicators

[0046] The energy consumed by a UAV data collection node consists of two parts: one part is used for monitoring information and data processing within the cluster, which is represented by E col The other part is used to transmit data information between sensor nodes, represented by E tra express.

[0047] UAV data collection nodes i , monitor the energy E consumed by n packets moni (s i ,t j ,n), expressed as follows:

[0048] E moni (s i ,t j ,n)=k l nd 2 (s i ,t j ) (7)

[0049] Among them, k lThe coefficient of energy attenuation of the monitoring data signal of the UAV data collection node, d(s i ,t j ) represents the drone data collection node s i With the target monitoring point t j The Euclidean distance between the UAV data collection node and the target monitoring point is proportional to the energy consumption. The longer the Euclidean distance between the UAV data collection node and the target monitoring point, the greater the energy consumed by the monitoring data.

[0050] Generally speaking, the higher the frequency of the electromagnetic wave of the drone data collection node, the shorter the wavelength, the shorter the propagation radius, and the greater the energy consumption of data processing. The energy consumption of data processing is inversely proportional to the propagation radius. Therefore, the drone data collection node s i , the energy consumed by processing n packets E data (s i ,n), expressed as:

[0051]

[0052] Among them, θ represents the energy consumption coefficient of the sensor processing data in the UAV data collection node, R i Represents the propagation radius of different types of sensors in the UAV data collection node.

[0053] The total energy consumption of monitoring information and data processing in the cluster E col (s i ,t j ,n) is:

[0054] E col (s i ,t j ,n)=E moni (s i ,t j ,n)+E data (s i ,n) (9)

[0055] UAV data collection nodes i Transmit n data packets to the drone data collection node s p Energy consumed E tra (s i ,s p ,n), expressed as follows:

[0056] E tra (s i ,s p ,n)=k2nd 2 (s i ,s p ) (10)

[0057] Among them, k2 represents the energy attenuation coefficient when communicating between UAV data collection nodes, d(s i ,s p ) represents the communication distance between drone data collection nodes. Similarly, the energy consumption is proportional to this distance.

[0058] In summary, the energy consumption E in the heterogeneous network of drone data collection nodes is sum , expressed as:

[0059]

[0060] The multi-objective optimization problem in step 5 has the optimization objectives of maximizing coverage and minimizing energy consumption. Therefore, the problem of deploying heterogeneous networks of drone data collection nodes is transformed into a multi-objective optimization problem, with maximizing coverage and minimizing energy consumption as the optimization objectives. In multi-objective optimization problems, the goal is usually to minimize the value of the objective function. Therefore, the coverage and energy consumption of the heterogeneous network of drone data collection nodes are transformed as follows:

[0061]

[0062] The algorithm in step 6) adopts a method of selecting individuals in the sub-key layer and the key layer, and uses the vector angle and minimum neighborhood distance between the selected individual and the associated vector as fitness, thereby increasing the uniformity and diversity of the distribution of the population solution, and optimizing the two objectives of coverage and energy consumption in step 5); each UAV data acquisition node network deployment plan is an individual, and multiple individuals are a population; the values ​​of the two optimization objectives are obtained under each deployment plan as the individual solution, thereby calculating the vector angle and population density; this algorithm obtains the coverage, energy consumption and convergence and distribution of the UAV data acquisition node network by inputting the number of population individuals and the number of iterations, thereby solving the problem of heterogeneous data acquisition network deployment of UAV clusters and obtaining a deployment plan; it includes the following steps:

[0063] The basic framework of the algorithm is as follows: First, the population P is initialized, and genetic mutation crossover is performed on N individuals to generate offspring R. Then, the parents and offspring are used as the new population S, and environmental selection is performed on them to generate a new population P until the iteration ends;

[0064] 602) Environment selection process: Based on the proposed algorithm, the parent population P and the offspring R are input, and their union is the population S. Through the environment selection operation, N individuals are selected from the dual population S as the new population. First, the individuals in the population S are normalized, and then the individuals in the population S are divided into different layers through the non-dominated sorting operation, and the key layer F is determined. lIn order to increase the diversity of the population, the individuals before the sub-critical layer are added to the new population, and the individuals in the sub-critical layer and the critical layer are crossover, mutation, and selection. According to the niche method, K (K = N-|P|) individuals are selected to join the new population P.

[0065] 603) Correlation vector: Calculate F l The solutions in are associated with the vector angles of the solutions in population P, and then the two solutions with the smallest vector angles are associated;

[0066] 604) Neighborhood Distance: The neighborhood of a candidate is represented by the neighborhood of the candidate's association vector. The neighborhood distance is the minimum distance between the candidate and any individual in the neighborhood. Neighborhood distance in a population refers to the distance between each individual and its nearest neighbor. It is an important indicator for measuring the similarity and density between individuals in a population.

[0067] When calculating the neighborhood distance, the following formula can be used:

[0068] D i =min{dist(x i ,x j )|j=1,2,...K} (14)

[0069] Among them, D i represents individual x i The distance within the neighborhood, K represents the number of individuals in the neighborhood.

[0070] 605) Fitness function: F l The vector angle and neighborhood distance of the solution in the layer and the solution in the associated P are used as the fitness of the solution. The solution with the largest fitness enters the new population P first. The fitness formula is as follows:

[0071] F(j)=A(j)-D(j) (15)

[0072] Among them, A(j) represents the solution f j The vector angle with the associated vector in the target population P, θ(f j ) is the solution f j The vector with the smallest vector angle with the vectors in population P is the associated vector, and the corresponding expression is as follows:

[0073] A(j)=min{θ(f j )|f j ∈F∧(flag(f j )==false)} (16)D(j) represents the solution f j The minimum distance to individuals in the neighborhood of population P is as follows:

[0074] D(j)=min{d ij |i=μ-1,μ+1} (17)

[0075] Where μ is the solution f j The subscript of the associated vector.

[0076] Through the analysis of the fitness function, we can see that the larger the value of A(j), the better the solution f j The greater the dissimilarity with the associated vector, the greater the diversity of the population, and the smaller the value of D(j), indicating that the solution f j The greater the uniformity of the distribution in the neighborhood area.

[0077] The step 7) comprises the following specific steps:

[0078] 701) Setting the length, width, and height of the simulation scene and the position and height of obstacles according to the real environment;

[0079] 702) Setting the perception radius, communication radius and number of sensors of the drone data collection node signal;

[0080] 703) The improved multi-objective optimization algorithm is set with different numbers of individuals and iterations to conduct multiple simulation experiments.

[0081] Compared with the prior art, the present invention has the following beneficial effects:

[0082] 1. This invention considers the deployment of a heterogeneous data acquisition network for drone clusters in the presence of obstacles. In most data acquisition environments, walls or terrain may obstruct electromagnetic wave propagation, affecting drone communication signals and significantly reducing signal transmission distance. This invention considers the impact of obstacles on drone sensor signals in a heterogeneous drone data acquisition network in three-dimensional space, which is more realistic.

[0083] 2. This invention designs coverage indicators based on real-world scenarios. Different interest levels are set for different areas, and higher coverage is set for areas with high interest levels, which better meets actual data collection requirements.

[0084] 3. This invention designs energy consumption indicators based on actual scenarios. Considering that the UAV has a limited flight time, it is necessary to reduce energy consumption as much as possible to improve the UAV's flight time, which is more in line with reality.

[0085] 4. This paper, based on a multi-objective optimization algorithm (NDDEA), selects individuals from sub-critical and critical layers, using the vector angle and minimum neighborhood distance between the selected individuals and the associated vector as fitness metrics. This increases the uniformity and diversity of the population solution distribution. This improved algorithm is used to perform multi-objective optimization on heterogeneous UAV data collection network deployments. Experimental results show that compared with other multi-objective optimization algorithms, the proposed method achieves a more uniform distribution of optimized solutions and better convergence. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 : This is an example diagram of the y coordinate of the signal passing through an obstacle in the present invention; (a) is a coordinate plane mapping diagram; (b) is a similar triangle diagram.

[0087] Figure 2 : This is an example diagram of the z coordinate of the signal passing through an obstacle in the present invention; (a) is a coordinate plane mapping diagram; (b) is a similar triangle diagram.

[0088] Figure 3 This is a framework diagram of the adaptive strategy based on vector angle and population density in the present invention.

[0089] Figure 4 The improved NDDEA algorithm of the present invention is iterated 200 times for the network deployment problem of heterogeneous data collection of drone clusters, and 200 experiments are run independently. The distribution diagram of the non-dominated solutions of the population obtained at the 1st, 20th, 50th, 100th, 150th, and 200th iterations.

[0090] Figure 5 The average HV value obtained by independently running the improved NDDEA algorithm of the present invention 20 times for the UAV cluster heterogeneous data acquisition network deployment problem.

[0091] Figure 6 It is the PD value of the population solution obtained by the NDDEA algorithm for the network deployment problem of heterogeneous data collection of drone clusters. DETAILED DESCRIPTION

[0092] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0093] This embodiment, based on computer simulation, designs performance metrics for coverage and energy consumption of drone data collection network deployments in obstructed environments. It also uses a modified NDDEA algorithm to perform multi-objective optimization of heterogeneous drone data collection network deployments. By mathematically modeling the data collection environment with obstacles, a multi-objective optimization algorithm based on vector angles and minimum neighborhood distance as fitness metrics is proposed. This multi-objective optimization algorithm is then used to optimize heterogeneous drone data collection network deployments in three-dimensional environments. This invention provides a new solution for drone data collection network deployment.

[0094] The present invention provides a method for deploying a drone data collection network based on a multi-objective optimization algorithm, which includes the following steps:

[0095] Step 1: Build a simulation environment and determine the monitoring area and obstacle area.

[0096] The present invention takes civil air defense facilities as the research object. The civil air defense facilities are 186 meters long, 21 meters wide and 3.5 meters high. The monitoring area is gridded to simulate the existence of three obstacles in the civil air defense facilities. The obstacles are regarded as rectangular obstacles, with obstacles at 30-33 meters in length, 66-69 meters and 170-173 meters in length respectively. The obstacles are 3 meters high.

[0097] Step 2: Boundary detection. If the drone's deployment position exceeds the deployment area, the drone's position is corrected. For example, if the drone's x-coordinate is less than 0, the x-coordinate is set to 0. If the x-coordinate is greater than 186, the x-coordinate is set to 186. Similarly, the drone's y-coordinate and z-coordinate are also corrected. At the same time, if the drone is deployed on an obstacle, the drone's z-coordinate is set to the height of the obstacle. Examples of the y-coordinate and z-coordinate of the signal passing through the obstacle are as follows: Figure 1 and Figure 2 shown.

[0098] Step 3: Consider the impact of obstacles on drone signals in the real environment. If there is an obstacle between the signal sent by the drone data collection node and the monitoring point, use spatial geometry to calculate the length l that the drone data collection node sends the signal to penetrate the obstacle. The length l affects the signal propagation distance.

[0099] Step 4: Calculate the coverage and energy consumption indicators of the drone data collection network deployment in the deployment scenario.

[0100] 4.1) Drone Data Collection Signal Loss. Wireless signals experience a certain amount of loss during propagation. This loss is primarily composed of two components: path loss and signal attenuation caused by obstacles. The loss of drone data collection signals is calculated using the wireless signal propagation loss formula and substituted into the coverage and energy consumption indicator formulas.

[0101] 4.1) Calculate the coverage in the UAV data collection network.

[0102] The coverage rate refers to the percentage of drone data collection network coverage points in the entire monitoring area grid points, reflecting the coverage range of the drone data collection network. The number of points covered by the drone data collection network can be expressed as:

[0103]

[0104] Among them, a covered point can be expressed as:

[0105]

[0106] Assuming that M represents the total number of all grid points in the monitoring area, the coverage CovDegree can be expressed as:

[0107]

[0108] 4.2) Calculate the energy consumption in the UAV data collection network.

[0109] The energy consumed by a UAV data collection node consists of two parts: the energy used for monitoring information and data processing within the cluster, represented by E col The other part is used to transmit data information between UAV data collection nodes, and is represented by E tra express.

[0110] UAV data collection nodes i , monitoring the energy consumed by n packets, expressed as follows:

[0111] E moni (s i ,t j ,n)=k l nd 2 (s i ,t j ) (twenty one)

[0112] Among them, k l The coefficient of energy attenuation of the monitoring data signal of the UAV data collection node, d(s i ,t j ) represents the drone data collection node s i With the target monitoring point t j The energy consumption is proportional to the distance between the drone data collection node and the target monitoring point. The farther the distance between the drone data collection node and the target monitoring point, the greater the energy consumed by the monitoring data.

[0113] Generally speaking, the higher the frequency of the electromagnetic wave of the drone data collection node, the shorter the wavelength, the shorter the propagation radius, and the greater the energy consumption of data processing. The energy consumption of data processing is inversely proportional to the propagation radius. Therefore, the drone data collection node s i The energy consumed in processing n packets is expressed as:

[0114]

[0115] The energy consumption of monitoring information and data processing within the cluster is:

[0116] Ecol (s i ,t j ,n)=E moni (s i ,t j ,n)+E data (s i ,n) (23)

[0117] Among them, θ represents the energy consumption coefficient of the sensor processing data in the UAV data collection node, R i Represents the propagation radius of different types of sensors in the UAV data collection node.

[0118] UAV data collection nodes i Transmit n data packets to the drone data collection node s p The energy consumed is expressed as follows:

[0119] E tra (s i ,s p ,n)=k2nd 2 (s i ,s p ) (twenty four)

[0120] Among them, k2 represents the energy attenuation coefficient when communicating between UAV data collection nodes, d(s i ,t p ) represents the communication distance between drone data collection nodes. Similarly, the energy consumption is proportional to this distance.

[0121] In summary, the energy consumption in the heterogeneous network of drone data collection nodes is expressed as:

[0122]

[0123] Step 5: Problem Formulation. Convert the UAV data collection network deployment problem into a multi-objective optimization problem:

[0124]

[0125] Step 6: Using the algorithm proposed by the present invention, i.e., using a method for selecting individuals in the sub-key layer and the key layer, as Figure 3 As shown, the vector angle between the individual to be selected and the associated vector and the minimum neighborhood distance are used as fitness to increase the uniformity and diversity of the population solution distribution and optimize the two objectives of coverage and energy consumption in step 5).

[0126] Step 7: Simulation experiment on the deployment of drone data collection network for civil air defense facilities.

[0127] 7.1) The simulation scene was designed to be 186 meters long, 21 meters wide, and 3.5 meters high, based on the actual environment. Three obstacles were set up: rectangular obstacles at 30-33 meters, 66-69 meters, and 170-173 meters, with each obstacle standing 3 meters high.

[0128] 7.2) Set the drone signal perception radius to 10 meters, the communication radius to 25 meters, and the number of drones to 13;

[0129] 7.3) Perform multiple simulation experiments with different numbers of individuals and iterations, and select the experimental data with the best optimization effect;

[0130] 7.4) Conduct comparative experiments on the improved algorithm with various multi-objective evolutionary algorithms to verify the effectiveness of the improved algorithm.

[0131] The comparison algorithms are NSGA-II algorithm, CCMO algorithm, and HypE algorithm.

[0132] This study uses NDDEA, NSGA-II, CCMO, and HypE algorithms to compare the results of the evolution process of the BT1-BT3, ZDT1, and ZDT2 series problems. The BT series function is a standard test function in multi-objective optimization problems. The specific form of the BT series function is as follows:

[0133] (1) BT1:

[0134]

[0135] where x∈[0,1] 30 , I k ={j|mod(j,2)=k-1,j=2,...,n}, k=1,2,y j =x j -sin(jπ / (2n)),j=2,...,n.

[0136] (2) BT2:

[0137]

[0138] Where x∈[0,1] 30 ,I1,I2,y j ,j=2,...,n。

[0139] (3) BT3:

[0140]

[0141] Where x∈[0,1] 30 ,I1,I2,y j ,j=2,...,n。

[0142] The ZDT problem is a set of commonly used multi-objective optimization test problems, which are often used to test the performance of multi-objective optimization algorithms.

[0143] The ZDT1 problem is a binary decision variable problem, and its objective function is as follows:

[0144]

[0145] Among them, x i ∈[0,1], n is the number of decision variables.

[0146] The objective function of the ZDT2 problem is as follows:

[0147]

[0148] Among them, x i ∈[0,1], n is the number of decision variables.

[0149] Table 1 shows the median and interquartile range (IQR, marked in brackets) of the IGD, HV, and Δ metric values ​​throughout the algorithm’s evolution for the test problem, where the best performance is highlighted in bold black.

[0150] Table 1

[0151]

[0152]

[0153] Table 1 shows that, according to the Δ metric, the proposed NDDEA algorithm performs best for the BT1 and BT2 test problems. According to the IGD metric, the proposed NDDEA algorithm performs best for the BT1, BT2, and BT3 test problems. Similarly, according to the HV metric, the NDDEA algorithm performs best for the BT1, BT2, BT3, and ZDT1 test problems. Therefore, a comparison of these three metrics shows that the proposed NDDEA algorithm is able to stably optimize multiple test problems throughout the evolutionary process.

[0154] Table 2 shows the median and interquartile range (IQR, marked in brackets) of the Δ, IGD, and HV index values ​​obtained from the last 500 experiments of the iterative process of the NDDEA, NSGA-II, CCMO, and HypE algorithms for the BT1-BT3, ZDT1, and ZDT2 series of problems, where the best performance is shown in bold black.

[0155] Table 2

[0156]

[0157]

[0158] Table 2 shows that, according to the Δ metric, the proposed NDDEA algorithm performs well for the BT2 test problem. According to the IGD metric, the NDDEA algorithm performs best for the BT1, BT2, and BT3 test problems. Similarly, according to the HV metric, the NDDEA algorithm also performs best for the BT1, BT2, and BT3 test problems. Therefore, in the final evolutionary process, the proposed NDDEA algorithm performs best for optimizing multiple test problems, effectively improving the convergence, uniformity, and diversity of the population solution.

[0159] Step 8: Perform a visual analysis of the experimental data and display the optimization effect diagrams of the improved algorithm and the comparative algorithm for the two objectives of UAV data collection network deployment, as well as the convergence and distribution index diagrams of the solutions.

[0160] Figure 4 The improved algorithm of the present invention was iterated 200 times for the deployment problem of heterogeneous data collection networks for drone clusters, and 200 independent experiments were run. The distribution diagram of the non-dominated population solutions obtained at the 1st, 20th, 50th, 100th, 150th, and 200th iterations is shown. As can be seen from the figure, as the number of iterations increases, the improved algorithm of the present invention can effectively optimize both objectives, gradually increasing coverage and reducing energy consumption. Furthermore, the figure also shows that the distribution of population solutions becomes increasingly uniform. This shows that the improved algorithm of the present invention can optimize the deployment problem of heterogeneous data collection networks for drone clusters. Figure 5 This is the average HV value obtained by independently running the improved algorithm of the present invention 20 times for the UAV cluster heterogeneous data acquisition network deployment problem. As can be seen in the figure, as the population continues to iterate, the HV value of the population solution obtained by the algorithm gradually increases and finally stabilizes at 0.801. The larger the HV value, the better the convergence and distribution of the population solution obtained in the experiment. Figure 6 This is the PD value of the population solution obtained by the improved algorithm of the present invention for the network deployment problem of heterogeneous data acquisition of drone clusters. It can be seen from the figure that with the increase of the number of iterations, the PD value of the population solution gradually increases, indicating that the diversity of the population solution gradually increases. Therefore, the improved algorithm of the present invention can well optimize the network deployment problem of heterogeneous data acquisition of drone clusters and improve the diversity of the population solution.

Claims

1. A method for deploying heterogeneous data collection networks for drone clusters based on a multi-objective optimization algorithm, characterized in that: The following steps are involved: Step 1: Build a simulation scene based on the 3D data collected from the real scene; randomly generate a set of different types of drone deployment points S, S = {s1, s2, ..., s N }; monitoring area T is the set of target points to be covered, T={t1,t2,....,t M The obstacle area R is the area that blocks the propagation of electromagnetic waves. The obstacle is a rectangular obstacle. Step 2: Perform boundary detection on the drone position in the simulation scene; Step 3: Based on the characteristics of different types of UAV data collection nodes, spatial geometry is used to calculate the degree of obstruction of obstacles on the propagation of electromagnetic wave signals of UAV data collection nodes; Step 4: Calculate the coverage rate and energy consumption index of the drone data collection nodes based on the impact of obstacles on electromagnetic wave signal propagation and the actual drone data collection behavior; Step 5: Convert the deployment problem of the UAV cluster heterogeneous data collection network into a multi-objective optimization problem, with maximizing coverage and minimizing energy consumption as the optimization goals; Step 6: Using an improved multi-objective optimization algorithm, a method for selecting individuals in the sub-key layer and key layer is adopted. The vector angle and minimum neighborhood distance between the selected individual and the associated vector are used as fitness. This increases the uniformity and diversity of the population solution distribution, further optimizing the network coverage and energy consumption in the UAV cluster heterogeneous data collection network deployment model. Step 7: Set the simulation scenario, UAV data collection nodes, and parameters of the improved multi-objective optimization algorithm, and conduct a simulation deployment experiment on the UAV data collection network; The step 4 includes the following steps: 401) UAV data collection node electromagnetic wave signal loss The electromagnetic wave signal of the UAV data collection node will produce loss during the propagation process. This loss consists of two parts: one is path loss, and the other is signal attenuation caused by obstacles. The path loss P of the electromagnetic wave signal of the UAV data collection node PL1 The formula is as follows: Signal attenuation caused by obstacles P PL2 The formula is as follows: Therefore, the propagation loss P of the electromagnetic wave signal of the UAV data collection node is PL The formula is as follows: Where d is the Euclidean distance between the UAV data collection node and the monitoring point; d0 is the reference distance; n is the path loss exponent, which represents the growth rate of path loss with distance; n1 is the path loss exponent in free space; and n2 is the path loss exponent in other environments. 402) Design of coverage indicators The grid model method is used to calculate the coverage area of ​​the UAV data collection node network by calculating the intersection points of the grid covered by the UAV data collection nodes; Set a UAV data collection node s i The perception radius is R cov , the coordinates of the drone data collection node are (x s ,y s ,z s ), in three-dimensional space, for any point t in the monitoring area j , point t j The coordinates of (x t ,y t ,z t ), p(s i ,t j ) represents the drone data collection node s i Detect the target point t j The probability of UAV data collection node model is: Among them, d(s i ,t j ) is the UAV data collection node s i With point t j The Euclidean distance between them; α is the signal attenuation factor; P PL is the propagation loss of electromagnetic wave signals at the UAV data collection node, R i Represents the signal coverage radius of different communication nodes; The coverage rate refers to the proportion of drone data collection node network coverage points to the grid points in the entire monitoring area, reflecting the coverage range of the drone data collection node network. The number of points covered by the drone data collection node network is expressed as: Among them, w1 is the weight of the UAV data collection node deployed in the area without obstacle coverage, and w2 is the weight of the UAV data collection node deployed in the area with obstacle coverage. Both are set hyperparameters; p(s i ,t j ) refers to the drone data collection node s i Detect the target point t j probability; In summary, coverage is defined as follows: Among them, Num is the number of points covered by the UAV data collection node network, and M is the number of all monitoring points in the target area; 403) Design of energy consumption indicators The energy consumed by a UAV data collection node consists of two parts: one part is used for monitoring information and data processing within the cluster, which is represented by E col The other part is used to transmit data information between UAV data collection nodes, and is represented by E tra express; UAV data collection nodes i , monitor the energy E consumed by n packets moni (s i ,t j ,n), expressed as follows: E moni (s i ,t j ,n)=k l nd 2 (s i ,t j ) (7) Among them, k l The coefficient of energy attenuation of the monitoring data signal of the UAV data collection node, d(s i ,t j ) represents the drone data collection node s i With the target monitoring point t j The Euclidean distance between the data collection node and the target monitoring point is proportional to the energy consumption. The longer the Euclidean distance between the data collection node and the target monitoring point, the greater the energy consumed by the monitoring data. The higher the frequency of the electromagnetic wave of the drone data collection node, the shorter the wavelength and the shorter the propagation radius, and the greater the energy consumption of data processing. The energy consumption of data processing is inversely proportional to the propagation radius. Therefore, the drone data collection node s i The energy consumed by processing n packets is E data (s i ,n), expressed as: Among them, θ represents the energy consumption coefficient of the sensor processing data in the UAV data collection node, R i represents the propagation radius of different types of sensors in the UAV data collection node; The total energy consumption of monitoring information and data processing in the cluster E col (s i ,t j ,n) is: E col (s i ,t j ,n)=E moni (s i ,t j ,n)+E data (s i ,n) (9) UAV data collection nodes i Transmit n data packets to the drone data collection node s p Energy consumed E tra (s i ,s p ,n), expressed as follows: E tra (s i ,s p ,n)=k2nd 2 (s i ,s p ) (10) Among them, k2 represents the energy attenuation coefficient when communicating between UAV data collection nodes, d(s i ,s p ) represents the distance between the UAV data collection nodes for communication. Similarly, the energy consumption is proportional to this distance; In summary, the energy consumption E in the heterogeneous network of drone data collection nodes is sum , expressed as: The multi-objective optimization problem in step 5 has the optimization objectives of maximizing coverage and minimizing energy consumption. Therefore, the problem of deploying heterogeneous networks of drone data collection nodes is transformed into a multi-objective optimization problem, with maximizing coverage and minimizing energy consumption as the optimization objectives. In multi-objective optimization problems, the goal is to minimize the value of the objective function, so the coverage and energy consumption of the heterogeneous networks of drone data collection nodes are transformed as follows:

2. The method for deploying heterogeneous data acquisition networks of drone clusters based on a multi-objective optimization algorithm according to claim 1 is characterized in that: The deployment height of the drone data acquisition node is not a fixed height, and different heights are selected depending on whether it is deployed on an obstacle; when the deployment point of the drone data acquisition node in step 2 falls outside the monitoring area, the position of the drone data acquisition node is corrected so that the drone is located at a data acquisition network deployment point that falls within the monitoring area.

3. The method for deploying heterogeneous data acquisition networks of drone clusters based on a multi-objective optimization algorithm according to claim 1 is characterized in that: Step 3 includes the following specific steps: Find the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary, and then find the length l that the electromagnetic wave signal emitted by the drone data acquisition node penetrates the obstacle. Since the obstacle is a rectangular parallelepiped, and the plane where the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary is parallel to plane yoz, in the coordinates (x, y, z) of the point where the electromagnetic wave signal from the drone data acquisition node passes through the obstacle boundary, x is the horizontal coordinate of the known obstacle boundary. The next step is to find the y and z coordinates based on x. 301) When calculating the y coordinate, the electromagnetic wave signal transmitted from the drone data collection node to the monitoring point is projected onto the xoy coordinate plane; Among them, the coordinates of the drone data collection node are A (x1, y1, z1), the coordinates of the monitoring point are B (x2, y2, z2), connect AB, and intersect with the obstacle boundary at point D (x, y, z), draw a straight line through point A parallel to the x-axis, and draw a straight line through point B parallel to the y-axis, the two straight lines intersect at point C, connect AC, and intersect with the obstacle boundary at point E, construct similar triangles △AED ~ △ABC, and we get The solution is: 302) When calculating the z coordinate, the electromagnetic wave signal transmitted from the drone data collection node to the monitoring point is projected onto the xoz coordinate plane; Same as 301), construct similar triangles △AED~△ABC, and we get Right now The solution is: 303) Through steps 301) and 302), the boundary point D (x, y, z) at which the electromagnetic wave signal of the drone data acquisition node penetrates the obstacle is calculated, and then the distance between point A and point D is calculated, that is, the length l of the signal emitted by the drone data acquisition node penetrating the obstacle. The length l affects the degree of signal attenuation. The longer the length, the greater the degree of signal attenuation.

4. The method for deploying heterogeneous data acquisition networks of drone clusters based on a multi-objective optimization algorithm according to claim 1 is characterized in that: The improved multi-objective optimization algorithm in step 6 adopts a method for selecting individuals in the sub-key layer and the key layer, and uses the vector angle and the minimum neighborhood distance between the selected individual and the associated vector as fitness, thereby increasing the uniformity and diversity of the distribution of the population solution, and optimizing the two objectives of coverage and energy consumption in step 5; each drone data acquisition node network deployment plan is an individual, and multiple individuals are a population; the values ​​of the two optimization objectives are obtained under each deployment plan as the individual solution, thereby calculating the vector angle and population density; the algorithm obtains the coverage, energy consumption and convergence and distribution of the drone data acquisition node network by inputting the number of population individuals and the number of iterations, thereby solving the problem of heterogeneous data acquisition network deployment of drone clusters and obtaining a deployment plan; the algorithm includes the following steps: The basic framework of the algorithm is as follows: First, the population P is initialized, and genetic mutation crossover is performed on N individuals to generate offspring R. Then, the parents and offspring are used as the new population S, and environmental selection is performed on them to generate a new population P until the iteration ends; 602) Environment selection process: Based on the proposed algorithm, the parent population P and the offspring R are input, and their union is the population S. Through the environment selection operation, N individuals are selected from the dual population S as the new population; first, the individuals in the population S are normalized, and then the individuals in the population S are divided into different layers through the non-dominated sorting operation, and the key layer F is determined. l In order to increase the diversity of the population, the individuals before the sub-critical layer are added to the new population, and the individuals in the sub-critical layer and the critical layer are crossover, mutation, and selection. According to the niche method, K (K = N-|P|) individuals are selected to join the new population P; 603) Correlation vector: Calculate F l The solutions in are associated with the vector angles of the solutions in population P, and then the two solutions with the smallest vector angles are associated; 604) Neighborhood distance: The neighborhood range of an individual to be selected is expressed as the neighborhood of the association vector of the individual to be selected; the neighborhood distance is the minimum distance between the individual to be selected and the individuals in the neighborhood range; the neighborhood distance in a population refers to the distance between each individual and its nearest neighbor in the population; When calculating the neighborhood distance, the following formula is used: D i =min{dist(x i ,x j )∣j=1,2,…K} (14) Among them, D i represents individual x i In the distance within the neighborhood, K represents the number of individuals in the neighborhood; 605) Fitness function: F l The vector angle and neighborhood distance of the solution in the layer and the solution in the associated P are used as the fitness of the solution. The solution with the largest fitness enters the new population P first. The fitness formula is as follows: F(j)=A(j)-D(j) (15) Among them, A(j) represents the solution f j The vector angle with the associated vector in the target population P, θ(f j ) is the solution f j The vector with the smallest vector angle with the vectors in population P is the associated vector, and the corresponding expression is as follows: A(j)=min{θ(f j )|f j ∈F∧(flag(f j )==false)} (16) D(j) represents the solution f j The minimum distance to individuals in the neighborhood of population P is as follows: D(j)=min{d ij |i=μ-1,μ+1} (17) Where μ is the solution f j The subscript of the associated vector.

5. The method for deploying heterogeneous data acquisition networks of drone clusters based on a multi-objective optimization algorithm according to claim 1 is characterized in that: The step 7 comprises the following steps: 701) Setting the length, width, and height of the simulation scene and the position and height of obstacles according to the real environment; 702) Setting the perception radius, communication radius and number of sensors of the drone data collection node signal; 703) The improved multi-objective optimization algorithm is set with different numbers of individuals and iterations to conduct multiple simulation experiments.

Images