High-energy-efficiency data collection method for space-air-ground integrated large-scale internet of things

Abstract

The invention discloses a high-energy-efficiency data collection method for a space-air-ground integrated large-scale internet of things, and the method comprises the steps: firstly providing an energy consumption minimization problem model, considering the CS data reconstruction precision requirement, the flight energy consumption of an unmanned plane, and the energy consumption balance distribution of the internet of things, and decoupling an energy consumption minimization problem into two sub-problems; for the node selection sub-problem, network energy consumption balance and unmanned aerial vehicle flight distance minimization targets under CS reconstruction precision constraint are realized through a CSNS algorithm; for the energy consumption iterative optimization sub-problem, a non-convex problem is converted into a convex problem by adopting continuous convex approximation, variable relaxation, variable substitution and other technologies through a two-stage joint optimization algorithm, and the goal of energy consumption minimization of the space-air-ground integrated large-scale Internet of Things network system is achieved by joint optimization of the flight path, the flight time and node scheduling of the unmanned aerial vehicle. The effectiveness and superiority of the method are proved through the embodiment.

Description

Technical Field

[0001] The present invention relates to the field of air-to-ground communication technology, and in particular to a high-energy-efficiency data collection method for a large-scale, air-ground-integrated Internet of Things network. Background Art

[0002] In recent years, with the gradual maturity of 6G network technology, the application demand for the Internet of Everything and seamless network coverage has become increasingly strong. It is estimated that by 2030, the number of sensor nodes that need to access the network per square kilometer worldwide will reach tens of millions, which poses a great challenge to the access capacity of the current ground mobile communication network. In addition, for remote areas such as wilderness, deserts, forests, oceans and military battlefields, the high construction and maintenance costs make it difficult for traditional ground mobile networks to cover. Therefore, it has become a significant problem to achieve large-scale node data backhaul in remote areas. Some researchers use satellites to backhaul data from ground sensor nodes (SNs), but the high-density access of nodes and the high energy consumption caused by long-distance communication are not suitable for most IoT application scenarios. To solve the above problems, the space-ground integrated IoT (SAG-IoT) network architecture, which uses unmanned aerial vehicles (UAVs) as air-based platforms to assist in ground IoT network data collection and forwards the collected data to remote servers via low-orbit satellites, has attracted the research interest of researchers. Due to its high mobility and flexible deployment capabilities, UAVs not only increase the coverage of the network and improve the capacity of the IoT network, but also can build efficient LoS links with nodes to improve the link quality and communication rate of wireless communications. At the same time, since there is no need to generate and maintain network routes between nodes, the hardware cost and energy consumption of the nodes can be further reduced. Therefore, the addition of UAVs has greatly expanded the application scope of IoT networks and provided new solutions for efficient data collection of large-scale nodes and coverage monitoring in remote areas.

[0003] In the data collection research of SAG-IoT network, researchers proposed different solutions for different application scenarios and optimization goals. For IoT networks with large-scale deployment and long-term monitoring tasks, since both device nodes and UAVs are energy-constrained devices, energy constraints pose a severe challenge to network system performance. Therefore, research on reducing system energy consumption has become a hot topic of widespread concern in academia and industry.

[0004] In recent years, a large number of researchers have conducted research on the energy consumption problem of UAV-assisted data collection in SAG-IoT networks. In response to the energy consumption optimization problem of UAV, Zeng et al. studied the energy consumption minimization problem of rotor UAV in the process of node data collection, and used the "fly-hover-communication" scheme and the "fly-while-communication" scheme for data collection respectively. The original problem was transformed into a convex optimization problem through the successive convex approximation (SCA) technology to solve the minimum energy consumption of UAV. In response to the energy consumption optimization problem of IoT nodes, Zhan et al. studied the node energy consumption problem under the time-varying channel fading model, and used the joint optimization of node sleep scheduling and UAV trajectory to minimize the maximum node energy consumption under the constraint of reliable data transmission. Baek et al. studied the node energy consumption problem under the constraints of data collection and UAV flight distance, and found a set of UAV hovering points and flight trajectories according to the node energy state to maximize the minimum residual energy of the nodes in the entire network. In the scenario of UAV-assisted base stations providing mobile data collection services, Mondal et al. studied the problem of minimizing the energy consumption of network nodes under sparse coverage of ground base stations, and proposed a reinforcement learning algorithm that jointly optimizes UAV trajectories, node access strategies, and transmission power. Yang et al. used Lyapunov optimization theory to transform the random optimization problem into a single-time-slot online optimization problem. On the premise of ensuring the stability of the data queue and the UAV single-time-slot energy consumption queue, they jointly optimized the network resource allocation strategy and UAV trajectory to minimize the weighted total energy consumption of all network nodes.

[0005] The above studies only reduce network energy consumption from a single perspective, and do not consider the overall energy consumption optimization problem of UAV and nodes. Sun et al. minimized node communication energy consumption and UAV energy consumption by jointly optimizing the three-dimensional mobile trajectory of fixed-wing UAV and the node access matching strategy, but only considered small-scale network scenarios. In large-scale IoT long-term monitoring network applications, the high network life and long-distance UAV flight requirements make the system energy consumption optimization problem more complicated. Zhu et al. proposed a deep reinforcement learning scheme of sequential model strategy based on clustered ground network topology, which minimized the overall energy consumption of the system by jointly optimizing the UAV access order and cluster head selection, but the establishment and maintenance of the ground network cluster topology brought additional energy consumption to the nodes.

[0006] It should be noted that in large-scale, long-term IoT monitoring networks, the perception data of each node in the network has significant correlation characteristics in the spatiotemporal dimension. It is of significant advantage to use the correlation characteristics between data to reduce the energy consumption of network data collection. Compressed sensing (CS) technology can simultaneously perform data compression and sampling processes, and achieve efficient compression of correlation data at a sampling rate lower than the Nyquist rate, which greatly reduces the data collection energy consumption of the network. It has been widely used in data collection in IoT networks. D. Ebrahimi et al. used CS technology to compress the data of nodes in a network cluster to reduce the energy consumption of data transmission within the cluster. UAVs visited the cluster head nodes in turn to collect data. This study only reduced the network energy consumption within the cluster, and the energy consumption of network topology maintenance still existed. In addition, energy-constrained UAVs do not have the ability to traverse and collect data from large-scale network nodes. To solve this problem, by introducing sparse CS technology, it is possible to recover the perception data of the entire network from an incomplete perception data set, thereby realizing data collection of the entire network. However, the perception data reconstructed by sparse CS technology has a certain degree of error compared with the original data, and this error is seriously dependent on the design of the observation matrix.

[0007] Since the system energy consumption is closely related to UAV flight control and node scheduling strategy, and the existing related research has not fully utilized the flexible flight advantages of UAV to study the joint optimization problem of UAV flight time, UAV flight trajectory and node scheduling, it is of great significance to carry out corresponding research on minimizing the energy consumption of SAG-IoT network data collection in large-scale IoT node deployment scenarios. Summary of the invention

[0008] The purpose of the present invention is to provide a high-energy-efficiency data collection method for a large-scale integrated air-ground-space Internet of Things network, which jointly optimizes UAV flight trajectory, UAV flight time and node scheduling, and can effectively reduce the energy consumption of UAV-SNs system for large-scale IoT network data collection.

[0009] To achieve the above object, the technical solution adopted by the present invention is:

[0010] A high-energy-efficiency data collection method for a large-scale space-ground integrated Internet of Things network comprises the following steps: S1, build a CS data collection model and introduce a reconstruction error constraint C1 to limit the upper limit of the CS data reconstruction error; S2, construct UAV energy consumption model and node energy consumption model, decompose UAV flight trajectory by UAV trajectory discrete method, and introduce UAV flight trajectory constraint conditions; S3, construct the communication model of data transmission between UAV and nodes and introduce communication constraints; S4, construct an energy consumption minimization problem model, with UAV flight trajectory, UAV flight time, and node scheduling strategy as optimization variables; S5, node selection strategy: taking the energy consumption balance of network nodes and minimization of UAV flight distance as optimization goals, constructing a cost evaluation model for node selection, introducing the mean square error of the least squares estimate to evaluate the CS reconstruction accuracy, converting the constraint C1 into a threshold constraint on the mean square error of the least squares estimate, and then solving the optimization path matrix through the strategy optimization algorithm to select M nodes participating in data collection; S6, iterative optimization of energy consumption: Using the node selection strategy of S5, the energy consumption minimization problem of S4 is transformed into an energy consumption iterative optimization sub-problem, and an energy consumption iterative optimization model is constructed. A two-stage joint optimization algorithm is adopted. First, given the UAV flight trajectory, the optimal UAV flight time and node scheduling strategy are solved; secondly, given the optimized UAV flight time and node scheduling strategy, the optimal UAV flight trajectory is solved; finally, by alternately iterating the above process until the algorithm converges, the three optimization variable values ​​that minimize the energy consumption of UAV-SNs are obtained.

[0011] Further, in S2, the constraint conditions include constraint C2: limiting the flight trajectory of the UAV to start from the starting point and land at the end point, limiting the maximum flight distance of a single path segment of the UAV;

[0012] In S3, the constraints include constraint C3: limiting that at most only one node can perform data transmission with the UAV at a time; constraint C4: limiting the minimum data transmission task amount of each node selected to transmit data; constraint C5: limiting the communication distance between the UAV and the node to not exceed the effective communication range.

[0013] Furthermore, in S4, the energy consumption minimization problem model is: P: stC1,C2,C3,C4,C5 Among them, A represents the node scheduling strategy, Q represents the UAV flight trajectory, T represents the UAV flight time, λ is the energy balance factor, which is used to adjust the deviation caused by the UAV energy consumption and node energy consumption not being at the same level; ω (is the weight coefficient for measuring the importance of UAV energy consumption and node energy consumption, E s Represents the node energy consumption, E uav Represents UAV flight energy consumption.

[0014] Furthermore, in S5, a cost evaluation model for node selection is constructed: f(k)=ω0U r (k)+(1-ω0)D(k) Among them, 0≤ω0≤1 is the weight factor for measuring the node access rate and node distance cost; D(k) is the node SN k The distance cost objective function, U r (k) is the node SN k The access cost objective function.

[0015] Furthermore, in S5, the least squares estimate of the reconstructed signal is: The mean square error of the least squares estimate of the reconstructed signal is: Among them, Φ is the CS observation matrix, Ψ∈ is the sparse basis matrix, is the path matrix, S is the sparse signal, σ 2 is the variance, Tr(·) represents the trace of the matrix; The constraint threshold of the mean square error of the least squares estimate is: in, The path matrix corresponding to the Gaussian random node selection strategy.

[0016] Furthermore, in S5, the node selection subproblem S1P is transformed into an optimization problem S1P1 of selecting M nodes from K nodes and minimizing the sum of the costs of the M nodes under the CS reconstruction accuracy constraint: The reconstruction error is limited to not exceed the constraint threshold η.

[0017] Furthermore, in S5, the CSNS strategy optimization algorithm is used to select M nodes participating in data collection, and the specific steps are as follows: S51, the binary variable Relaxed to a continuous variable between 0 and 1, the non-convex problem S1P1 is transformed into a convex problem S1P2, S52, using convex optimization algorithm, solve S1P2 and obtain the optimal path matrix S53, for the set Sort the values ​​​​in descending order to obtain the sequence S54, according to the sequence The index of the first M values ​​of the node set is constructed

[0018] Furthermore, in S6, a continuous convex approximation algorithm is used to transform the non-convex problem into a series of convex sub-problems through variable relaxation and substitution, and the optimal solution of the energy consumption minimization problem is approximated iteratively until the algorithm converges.

[0019] Furthermore, in S5, solving the optimization path matrix is ​​completed in advance by the remote server at the UAV take-off and landing point before each data collection; in S6, the two-stage joint optimization algorithm is executed in advance on the remote server.

[0020] Furthermore, the IoT network includes a ground network responsible for sensing and monitoring regional data, an unmanned aerial platform responsible for data collection, and a satellite network responsible for transmitting data back to a remote data processing server. The unmanned aerial platform acts as a mobile aerial base station and periodically takes off from a starting point to perform flight missions with a specific flight trajectory, flight time, and node scheduling strategy, and uses CS sparse sampling to collect data from IoT nodes across the entire network.

[0021] By adopting the above technical solution, the present invention can achieve the following beneficial effects:

[0022] 1. The present invention establishes a UAV-SNs energy consumption minimization problem model based on sparse CS data collection technology, proposes an energy consumption optimization mechanism for UAV-assisted data collection, solves the difficulties faced by non-convexity and lack of mathematical closed-form expression, and decouples the difficult-to-solve energy consumption minimization problem into a CS node selection strategy optimization subproblem and an energy consumption iterative optimization subproblem, which not only meets the CS data reconstruction accuracy requirements, but also considers the balanced distribution of UAV flight energy consumption and IoT network energy consumption, and realizes low-energy UAV-assisted data collection.

[0023] 2. The present invention aims to balance IoT network energy consumption and minimize UAV flight distance, establishes a multi-objective optimization problem model based on CS reconstruction accuracy constraints, uses the CSNS algorithm to solve the multi-objective optimization problem, obtains the CS observation matrix, and optimizes the path matrix. The edge server at the UAV take-off and landing point or the remote high-performance server can complete it in advance before each data collection, without consuming additional energy of the UAV or network nodes. The present invention adds a threshold constraint on the reconstruction error, the reconstruction error is small, and the reconstruction accuracy is improved.

[0024] 3. The energy consumption optimization mechanism proposed in the present invention will adaptively adjust the flight speed of the UAV according to the communication task requirements to minimize energy consumption. This has obvious advantages over the strategy of using a fixed flight time in most current UAV trajectory optimization processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of data collection for the integrated air-space-ground Internet of Things network;

[0026] Figure 2 is the Pareto frontier under different M values ​​in the embodiment;

[0027] Figure 3 Comparison of CS reconstruction accuracy of different algorithms in the embodiments;

[0028] Figure 4 Comparison of UAV flight distances of different algorithms in the embodiments;

[0029] Figure 5 The convergence of the TSJO algorithm under different weight coefficients in the embodiment;

[0030] Figure 6 The energy consumption comparison of UAV and SNs under different weight coefficients in the embodiment;

[0031] Figure 7 Comparison of energy consumption of UAV-SNs system under different mechanisms in the embodiments;

[0032] Figure 8 The flight trajectory and flight time of the UAV under different communication task loads in the embodiment;

[0033] Fig. 9 Schematic diagram of the communication scheduling strategy for M nodes under four communication task loads in the embodiment. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] In order to reduce the energy consumption of the UAV-SNs system for data collection in a large-scale SAG-IoT network, this embodiment provides a high-efficiency data collection method for a large-scale integrated air-ground-space IoT network, wherein the SAG-IoT network includes a ground network responsible for sensing and monitoring regional data, a UAV mobile aerial platform responsible for data collection, and a satellite network responsible for transmitting data back to a remote data processing server. The UAV has the location information of all ground node deployments, the UAV onboard energy can support a round of data collection tasks, and the UAV can be provided with energy replenishment at the UAV take-off and landing points.

[0036] like Figure 1 As shown in the figure, a rotor UAV acts as a mobile aerial base station and periodically takes off from a starting point to perform flight missions with a specific UAV flight trajectory, UAV flight time, and node scheduling strategy. It uses CS sparse sampling to collect data from IoT nodes in the entire network, and transmits the collected data to a remote server via a low-orbit satellite. The server uses the CS reconstruction algorithm to reconstruct the data of nodes in the entire network.

[0037] The high energy efficiency data collection method described in this embodiment specifically includes the following steps: S1, build a CS data collection model and introduce a reconstruction error constraint C1 to limit the upper limit of the CS data reconstruction error; S2, construct UAV energy consumption model and node energy consumption model, decompose UAV flight trajectory by UAV trajectory discrete method, and introduce UAV flight trajectory constraint conditions; S3, construct the communication model of data transmission between UAV and nodes and introduce communication constraints; S4, construct an energy consumption minimization problem model, with UAV flight trajectory, UAV flight time, and node scheduling strategy as optimization variables; S5, node selection strategy: taking the energy consumption balance of network nodes and minimization of UAV flight distance as optimization goals, constructing a cost evaluation model for node selection, introducing the mean square error of the least squares estimate to evaluate the CS reconstruction accuracy, converting the constraint C1 into a threshold constraint on the mean square error of the least squares estimate, and then solving the optimization path matrix through the strategy optimization algorithm to select M nodes participating in data collection; S6, iterative optimization of energy consumption: Using the node selection strategy of S5, the energy consumption minimization problem of S4 is transformed into an energy consumption iterative optimization sub-problem, and an energy consumption iterative optimization model is constructed. A two-stage joint optimization algorithm is adopted. First, given the UAV flight trajectory, the optimal UAV flight time and node scheduling strategy are solved; secondly, given the optimized UAV flight time and node scheduling strategy, the optimal UAV flight trajectory is solved; finally, by alternately iterating the above process until the algorithm converges, the three optimization variable values ​​that minimize the energy consumption of UAV-SNs are obtained.

[0038] In step S1, the specific process of constructing the CS data collection model is as follows:

[0039] Assume that K sensing nodes are deployed in the monitoring area of ​​L×L, and the horizontal position of node SNk is represented by w k =(x k ,y k ), the perception data vector of the whole network perception node is Z = (z1, z2, ..., z K ) T Due to the spatial correlation between the sensing data in the monitoring area, the data vector Z is The following is expressed as: Z=ΨS,||S||0≤s (1) Where, .S. is the sparse signal after the data vector Z is sparsely transformed, with only s (s<<K) non-zero elements. Then the data vector collected by the UAV is It can be expressed as Y=ΦZ=ΦΨS=ΘS (2) in is the observation matrix, is the perception matrix. According to CS theory, if the perception matrix Θ satisfies the isometry constraint, the sparse signal S can be successfully reconstructed by solving the convex optimization problem in (3), where represents the l0 norm.

[0040] Since different CS sparse sampling node selections will cause different reconstruction errors, in order to prevent the CS reconstruction error during data collection from exceeding the maximum threshold ε allowed by the system, a reconstruction error constraint C1 is introduced, where G(·) represents the reconstruction function of the sparse signal S.

[0041] In step S2, the specific process of constructing the UAV energy consumption model and the node energy consumption model is as follows:

[0042] UAV energy consumption includes UAV flight energy consumption and UAV communication energy consumption. Compared with UAV flight energy consumption, UAV communication energy consumption can usually be ignored. UAV flight power can be expressed as a function of UAV flight speed v(t), where t is a continuous time variable. Therefore, UAV flight energy consumption is expressed as: Where T is the total flight time of the UAV, P uav (v(t)) represents the flight power of the UAV.

[0043] Assuming that the UAV flies at a fixed altitude H during the entire data collection process, the flight trajectory of the UAV is described as a time-varying two-dimensional horizontal coordinate q(t) = (x u (t),y u (t)).

[0044] This application uses the UAV trajectory discrete method to decompose the UAV flight trajectory. Assuming that the UAV flight trajectory is decomposed into N discrete line segments, the corresponding N+1 track points are represented as where q(0) = q I ,q(N)=q F Respectively represent the starting point and end point of the UAV flight. For the nth discrete line segment, the UAV flight time is represented as T(n), and the total flight time of the UAV is The UAV flight trajectory constraint C2 is introduced, which means that the UAV flight trajectory starts from the starting point and lands at the end point, and at the same time limits the maximum flight distance of a single path segment of the UAV. Constraint C2 is: Among them, Δ max and V maxare the maximum flight distance and maximum flight speed allowed by the UAV in each line segment. Assuming that the length of each discrete line segment is small enough, the flight speed v(n) of the UAV in a single path segment can be regarded as a fixed value, which is expressed as v(n) = ||q n+1 -q n || / T(n)=Δ(n) / T(n). After the UAV flight trajectory is discretized, the energy consumption formula of the UAV is expressed as: Among them, UAV flight power P uav The mathematical model of (v(n)) is: Among them, P0 and P1 are the blade profile power and induced power of the UAV in the hovering state, U tip is the tip angular velocity of the blade, v0 represents the average induced velocity of the rotor in hover, d0 and u are the fuselage drag ratio and the ratio of the rotor area to the rotor area, ρ and A are the air density and the blade rotor area, respectively.

[0045] For node energy consumption, this embodiment only considers its communication energy consumption. The node adopts the "sleep-wake" scheduling mechanism to participate in the data collection process, and defines the binary access scheduling matrix A = (a k (n)) K×N , a k (n) = 1 means that node k establishes a communication connection with the UAV in the nth path segment, otherwise, a k (n) = 0. Assume that the node communicates with a fixed power P k For data transmission, the node energy consumption is expressed as:

[0046] In step S3, a communication model for data transmission between the UAV and the node is constructed, and the specific process is as follows:

[0047] In order to analyze the air-ground link during data collection, a probabilistic LoS channel model is used to describe the large-scale fading between air-ground links. The probability of a LoS link between node k and the UAV is: Among them, a, b are parameters related to the environment, θ k (n) represents the elevation angle between the UAV and node k, which is defined as:

[0048] Then the probability of the existence of an NLoS link is Due to the randomness of the probability of LoS links and NLoS links, the channel gain of the air-to-ground link is a random variable. Therefore, by averaging the two random variables, the expected channel gain is obtained as: Where β<1 is the attenuation factor of the NLoS channel, represents the regularized LoS probability taking into account the NLoS channel attenuation effect, h0 represents the reference channel gain when the distance is 1m, and d k (n) is the distance between UAV and the kth node, and ||·|| represents the Euclidean norm.

[0049] In this embodiment, the mean approximation method is used to approximate the regularized LoS probability Right now in represents the regularized LoS probability obtained by using the average value of the communication angle. Furthermore, the communication rate between the UAV and the kth node can be expressed as: Where B is the channel bandwidth, N0 is the noise power.

[0050] In the process of UAV-assisted data collection, in order to avoid signal interference, it is assumed that the communication between nodes and UAVs uses time division multiple access technology for data transmission, and in any route segment, the UAV communicates with at most one node, and the constraint C3 is constructed:

[0051] Define the node set participating in sparse sampling as To ensure that the data of each node can be fully transmitted, the communication task of any node in the constraint set κ is not less than Q K , that is, to limit the minimum data transmission task amount of each selected data transmission node, and construct constraint C4:

[0052] In addition, since the performance of the compressed sensing data collection algorithm is extremely sensitive to link packet loss, in order to ensure the reliability of data transmission in the air-to-ground link, the maximum communication coverage angle of the UAV is considered to be θ, and the communication distance between the UAV and the node is set not to exceed the effective communication range, and constraint C5 is constructed:

[0053] In step S4, the energy consumption minimization problem model is constructed as follows:

[0054] With the help of the efficient reconstruction capability of CS technology, this application applies sparse CS technology to the UAV data collection process. UAV reconstructs the node data of the entire network by collecting only part of the node data, reducing the UAV flight distance, thereby greatly reducing the energy consumption of UAV-SNs. The UAV-SNs energy consumption minimization problem model for each round of data collection in the SAG-IoT network can be expressed as:

[0055] Among them, the optimization variable of problem P is the UAV flight trajectory UAV flight time and node scheduling strategy The parameter λ is the energy consumption balance factor, which is used to adjust the deviation caused by the UAV energy consumption and node energy consumption not being at the same level; the parameter ω (0≤ω≤1) is the weight coefficient for measuring the importance of UAV energy consumption and node energy consumption.

[0056] In problem P, constraint C1 limits the upper limit of CS data reconstruction error; constraint C2 indicates that the UAV flight trajectory starts from the starting point and lands at the end point, and at the same time limits the maximum flight distance of a single path segment of the UAV; constraint C3 ensures that at most only one node transmits data to the UAV at a time; constraint C4 limits the minimum data transmission task amount of each node selected to transmit data; constraint C5 ensures that the communication distance between the UAV and the node does not exceed the effective communication range.

[0057] For the energy minimization problem P, since the CS reconstruction accuracy and optimization variables in constraint C1 cannot be mathematically expressed in closed form, and the objective function and constraint C4 are non-convex, solving problem P is extremely challenging. After an in-depth analysis of problem P, it was found that the reason for the lack of mathematical closed-form expression between CS reconstruction accuracy and optimization strategy is that there are great difficulties in integrating CS technology with UAV-assisted data collection technology. In order to solve the above-mentioned technical integration difficulties, this application designs a reasonable CS sparse sampling node selection strategy, which not only meets the CS data reconstruction accuracy requirements, but also takes into account the balanced distribution of UAV flight energy consumption and IoT network energy consumption.

[0058] This application adopts the UAV-assisted data collection energy consumption optimization mechanism based on sparse CS technology, and decouples the original problem P into two sub-problems. First, in order to solve the problem of missing mathematical closed-form expression between CS reconstruction accuracy and optimization strategy, a node selection sub-problem S1P is defined. By optimizing the node selection strategy, the network node energy consumption balance and UAV flight distance minimization goals under the CS reconstruction accuracy constraint can be achieved; secondly, using the node selection strategy obtained by sub-problem S1P, the original problem P is converted into an energy consumption iterative optimization sub-problem S2P, and the approximate solution to the UAV-SNs energy consumption minimization problem P is achieved by solving problem S2P. The two sub-problems and their corresponding solution processes are described in detail below.

[0059] In step S5, the node selection strategy is as follows:

[0060] Since the selection of CS nodes not only directly affects the CS reconstruction accuracy, but is also related to the UAV flight energy consumption and the energy consumption balance of network nodes, the CS node selection strategy is modeled as a multi-objective optimization problem. The energy consumption balance of network nodes and the minimization of UAV flight distance are taken as optimization objectives, and a cost evaluation model for node selection is constructed.

[0061] For the network node energy consumption balance target, define a single node SN k The access cost objective function is: Among them, J r (k) represents the number of times node k is selected until the rth round of data collection, Indicates that J in round r r (k) is the maximum value. Choose U r The smaller the (k) value of the node, the more balanced the network energy consumption.

[0062] For the UAV flight distance target, define node SN k To the starting point q I With the end point q F The distance between the lines is: Where q=(q F -q I ) / ||q F -q I ||, then node SN k The distance cost objective function can be defined as: D(k) = d(k) / d max , where d max is the maximum value of d(k). The smaller the D(k) value is, the shorter the UAV flight distance is. To describe the selection cost of each node with respect to the above two optimization objectives, the classic linear weighted method is used to define the node SN k The cost function is: f(k)=ω0U r (k)+(1-ω0)D(k) (19) Among them, 0≤ω0≤1 is the weight factor for measuring the node access rate and node distance cost.

[0063] The above process gives a cost evaluation model for node selection. Next, we will model and analyze the CS reconstruction accuracy constraint. Assume that the set of nodes selected to participate in data collection is represented as where k i represents the ID number of the node, then the CS observation matrix Φ can be designed as: Among them, each row in Φ has only one non-zero element. Since the node selection strategy is closely coupled with the UAV flight trajectory, for the convenience of design, the path matrix is ​​defined It is a diagonal matrix with the following diagonal elements:

[0064] According to matrix transformation theory, the path matrix The relationship with the observation matrix Φ can be expressed as

[0065] In order to realize the constraint on CS reconstruction accuracy in constraint C1, the mean square error (MSE) of the least squares estimate is introduced to evaluate the CS reconstruction accuracy. The least squares estimate of the reconstructed signal is: Then the mean square error of the least squares estimate of the reconstructed signal is: Among them, σ 2 is the variance, and Tr(·) represents the trace of the matrix.

[0066] Therefore, the constraint C1 can be transformed into a threshold constraint on MSE. Given that the data reconstruction performance of Gaussian random measurement matrices has been verified many times, the constraint threshold of MSE can be set to in, is the path matrix corresponding to the Gaussian random node selection strategy, and constraint C1 is replaced by formula (24).

[0067] Through the above modeling, the node selection subproblem S1P can be transformed into an optimization problem S1P1, which is to select M nodes from K nodes and minimize the sum of the costs of the M nodes under the condition of satisfying the CS reconstruction accuracy constraint, as shown in formula (25). Among them, constraints (25a) and (25b) limit the number of nodes involved in data collection to M, and constraint (25c) limits the reconstruction error to no more than the maximum allowable threshold η. For constraint (25c), due to is a diagonal matrix and If non-negative, is a positive definite symmetric matrix. Since when X is a positive definite symmetric matrix, Tr(X -1 ) is a convex function with respect to X, so constraint (25c) is a convex constraint. However, constraint (25a) is a binary variable, so problem S1P1 is a non-convex optimization problem.

[0068] In order to make the above non-convex problem easier to solve, this application proposes a CSNS (CS-based Node Selection) strategy optimization algorithm, which converts binary variables Relaxation is a continuous variable between 0 and 1, then the problem S1P1 is transformed into the following optimization problem:

[0069] Through the above relaxation process, problem S1P1 is transformed into a convex problem S1P2. Using the classic convex optimization algorithm, the problem can be effectively solved and the optimal path matrix is ​​obtained.

[0070] It should be noted that the optimal path matrix obtained by solving problem S1P2 is The value is not necessarily a binary integer. According to the heuristic principle, this application selects the largest first M values ​​and uses the corresponding index numbers as the selected M nodes. The specific process is implemented according to the following algorithm: enter: ω0,Ψ, Output: node set κ 1: for all k∈{1,2,…,K}do 2: Calculate the cost function f(k) according to (19); 3: end for 4: Calculate the reconstruction error constraint threshold 5: Solve the convex problem (S1P2) and express the optimal solution as 6: Pair collection Sort the values ​​​​in descending order to obtain the sequence 7: Build a node set based on the index of the first M values ​​of the sequence

[0071] Since the above optimization problem can be solved by CVX toolbox, according to the analysis, the computational complexity of problem S1P2 is O(K 3.5 ). In addition, the solution of problem S1P2 can be completed in advance by the edge server at the UAV take-off and landing point or the remote high-performance server before each data collection, without additional energy consumption of the UAV or network nodes.

[0072] Then proceed to step S6, iterative optimization of energy consumption: transform the energy consumption minimization problem of S4 into an iterative optimization sub-problem of energy consumption, construct an iterative optimization model of energy consumption, and adopt a two-stage joint optimization algorithm. First, given the UAV flight trajectory, solve the optimal UAV flight time and node scheduling strategy; second, given the optimized UAV flight time and node scheduling strategy, solve the optimal UAV flight trajectory; finally, by alternately iterating the above process until the algorithm converges, obtain the three optimization variable values ​​that minimize the energy consumption of UAV-SNs. The specific process is as follows:

[0073] By solving the node selection subproblem S1P, the constraint C1 in the original problem P is satisfied. Using the obtained node selection optimization strategy, the original problem P is transformed into the energy consumption iterative optimization subproblem S2P, and its mathematical model is expressed as (27):

[0074] Since the optimization objective function in the S2P1 problem is a non-convex function with respect to the optimization variables {A, Q, T}, in order to transform it into an easy-to-handle convex function, a slack variable z(n) is introduced, and its mathematical expression is:

[0075] Then problem S2P1 is transformed into problem S2P2 shown in (29), where a new constraint (29b) is added to ensure that problem S2P2 is equivalent to problem S2P1. stC2-C5(29a)

[0076] However, constraints C4 and (29b) are non-convex constraints on the optimization variables, and constraint C3 has binary integer variables. Problem S2P2 is a mixed integer non-convex optimization problem, which cannot be solved by traditional optimization methods. In addition, since there are three optimization variables {A, Q, T}, it is difficult to find the global optimal solution. Therefore, a two-stage joint optimization (TSJO) algorithm is proposed, which uses block coordinate descent technology to deal with the S2P2 problem and reduce the complexity of solving the problem.

[0077] 1) Optimize node scheduling and flight time

[0078] Given the node scheduling strategy A and the UAV flight time T optimization problem under the UAV flight trajectory Q, the optimization problem S2P2 is reconstructed as: S2P3: stC3-C5

[0079] In order to make the above optimization problem into a convex optimization problem that is easy to handle, the relaxation binary variable a k (n) is a continuous variable At the same time, the introduction of auxiliary variables Then constraint C3 is rewritten as:

[0080] The constraint C5 is changed to τ k (n)||w k -q(n)||≤Htanθ·T(n), If the slack variable The optimal solution of is not a binary variable, using Alternative As the final binary value, represents the integer closest to x. Due to the limitation of the optimization objective, the UAV can only maintain the longest communication time with the node closest to it to reduce the energy consumption of UAV-SNs. and The difference is almost negligible.

[0081] In addition, to handle the non-convexity of constraint C4, the slack variable H shown in (32) is introduced k (n) Replace the non-convex part of the left side of constraint C4.

[0082] So constraint C4 is rewritten as Similarly, to ensure that the relaxed problem is equivalent to the original problem, (32) needs to be added as a new constraint to the optimization problem, and the problem S2P3 is transformed into:

[0083] Since constraints (33f) and (33g) are non-convex constraints on the optimization variables, in order to further transform them into convex constraints, this application uses the SCA algorithm to iteratively approximate the optimal solution of the original problem until the algorithm converges. Specifically, for constraint (33f), given the local point z at the lth iteration (l) (n), the global lower bound on the right side is obtained through the first-order Taylor expansion: Then constraint (33f) can be rewritten as:

[0084] Similarly, for constraint (33g), given the local point of the lth iteration is The global lower bound on the left side is obtained through the first-order Taylor expansion: Then constraint (33g) can be rewritten as: By dealing with the non-convex constraints (33f) and (33g) above, problem S2P4 is transformed into:

[0085] The above process transforms the non-convex problem S2P4 into a series of solvable convex optimization subproblems S2P5, by continuously solving the convex subproblems and iteratively updating the local point z according to the obtained solutions. (l) (n) and Until the difference between the optimization target values ​​of two adjacent iterations is less than or equal to the threshold δ1, the algorithm converges and the final optimized node scheduling strategy A and UAV flight time T are obtained.

[0086] 2) Optimize UAV flight trajectory

[0087] This part studies the optimization problem of UAV flight trajectory Q under given node scheduling strategy A and UAV flight time T. The optimization problem S2P2 is reconstructed as: stq(0) = q I ,q(N)=q F (39a)

[0088] Since constraints (39d) and (39e) are non-convex constraints, the above problem is a non-convex optimization problem. To deal with non-convex constraints, the slack variable I in (40) is introduced. k (n) Then constraint (39d) is rewritten as: Therefore, problem S2P6 is transformed into:

[0089] In the above problem, constraints (39e) and (40) are still non-convex constraints. Similarly, this application adopts the SCA method to transform the non-convex problem into a series of convex sub-problems for iterative solution. For constraint (39e), given the local point z at the lth iteration (l) (n) and q (l) (n), the global lower bound on the right side is obtained through the first-order Taylor expansion: in, Then constraint (39e) is rewritten as: Similarly, for constraint (40), after its right side is processed by first-order Taylor series expansion, the constraint is rewritten as: Through the above processing transformation, problem S2P7 is transformed into:

[0090] Problem S2P8 is a convex subproblem. By continuously updating the local point z (l) (n) and q (l) (n), and iteratively solve the convex subproblem until the difference between the target values ​​of two consecutive iterations is less than or equal to the threshold δ2, then the algorithm converges and obtains the optimized UAV flight trajectory Q.

[0091] 3) Two-stage joint optimization

[0092] Based on the problem processing and solution in the above two stages, this section will study the joint optimization method of node scheduling strategy, UAV flight time and UAV flight trajectory, and give the specific implementation process of the two-stage joint optimization (TSJO) algorithm. Specifically, in the i-th iteration, firstly, the UAV flight trajectory Q is given (i-1) , by solving a series of convex subproblems S2P5, we get the optimized node scheduling strategy and UAV flight time Secondly, based on the optimization results obtained above and Solve a series of convex subproblems S2P8 to obtain the optimized UAV flight trajectory Then take the optimized UAV flight trajectory As the initial trajectory of a new round of iteration, repeat the above process, and so on, until the difference between the optimization objective function values ​​of two adjacent iterations is less than or equal to the given threshold δ, the algorithm cycle ends, and the final optimization variable {A opt ,T opt ,Q opt}, the TSJO algorithm is as follows: 1: Initialization: Get the initial feasible A (0) ,T (0) ,Q (0) and the stop threshold δ. 2: Iteration indicator function i = 0. 3:Repeat 4: Given Q (i) Solve the problem in (38), the optimal solution is expressed as A (i+1) ,T (i+1) . 5: Given A (i+1) ,T (i+1) Solve the problem in (46), the optimal solution is expressed as Q (i+1) . 6: Calculate the value of the objective function in (27) based on the optimization variables 7: Update i = i + 1. 8: Until the decrease in the objective function value in (27) is lower than the given threshold δ.

[0093] Since both steps 4 and 5 in Algorithm 2 use SCA technology to approximate the non-convex problem by solving a series of convex subproblems, the computational complexity of steps 4 and 5 is O(L1(KN) 3.5 ) and O(L2N 3.5 ), where L1 and L2 are the number of times the convex subproblem is solved in steps 4 and 5 respectively. Assuming that the number of loops of the TSJO algorithm is L0, the total computational complexity of Algorithm 2 can be expressed as O(L0(L1(KN) 3.5 +L2N 3.5 )). Similarly, Algorithm 2 can be executed in advance on the remote server, and does not require the UAV to perform real-time calculations, thus avoiding additional energy consumption of the system.

[0094] To further illustrate the effectiveness and superiority of the present invention, the following comparative analysis is performed.

[0095] In the simulation experiment, it is assumed that 50 nodes are randomly and evenly deployed in the two-dimensional monitoring area of ​​400×400m 2 Within the range, a rotor UAV is used to collect data from nodes in the network deployment area. The locations of the UAV take-off and landing points are respectively located at q I =[0,200] and q F = [400, 200], the flight altitude is fixed at H = 100m, and the maximum speed of the UAV is V max =30m / s, the maximum communication coverage angle of UAV is θ=π / 3, and the flight energy consumption parameter settings of UAV are shown in Table 1. In addition, the path of UAV is discretized into N=500 path segments, and the communication power of the node is set to P k = 0.5W, the communication bandwidth between UAV and node B = 1MHz, and the channel noise power variance is set to σ 2 =-110dBm, the reference power at 1m is h0=-50dB, the attenuation factor of the NLoS channel is β=0.2, and the energy balance parameter λ=2000. Table 1 UAV flight energy consumption parameters parameter value parameter value <![CDATA[Blade profile power P0]]> 79.8W <![CDATA[UAV fuselage drag ratio d0]]> 0.6 <![CDATA[Induced power of blade P1]]> 88.6W Ratio of rotor area to rotor area 0.05 <![CDATA[Tip angular velocity of blade U tip > 120m / s Blade rotor area A <![CDATA[0.503m 2 ]]> <![CDATA[Average induced velocity v0 of the rotor]]> 4.03m / s Air density ρ <![CDATA[1.225kg / m 3 ]]>

[0096] A. Performance of the CSNS Algorithm

[0097] According to the objective function of the CSNS algorithm, in order to minimize the total flight distance of the UAV, the UAV needs to frequently visit nodes near the line connecting the starting point and the end point of the UAV. This method will turn the area near the line into a hotspot for access, thereby reducing the balance of network energy consumption. Therefore, the UAV flight distance and energy consumption balance objectives are essentially in a competitive relationship. In order to quantitatively describe the competitive relationship between the two, the network energy consumption balance objective is defined as where k i Indicates the ID number of the node selected to participate in data collection. When the node selection strategy makes the objective function U smaller, the network energy consumption tends to be more balanced. For the total flight distance of the UAV, when the nodes participating in data collection are determined, the total flight distance target D of the UAV can be calculated by the nearest neighbor algorithm in the Traveling Salesman Problem (TSP).

[0098] Under different weights ω0∈[0,1], by executing the CSNS algorithm to solve problem S1P2, a series of corresponding sparse CS node selection optimization strategies can be obtained. Based on the above optimization strategy set, the network energy consumption balance target U and the UAV total flight distance target D can be further calculated. Figure 2 The Pareto frontier between the energy consumption balance target U and the UAV total flight distance target D is given, where the sampling interval of the weight ω0 is set to 0.1, with a total of 11 sampling points. There is some overlap in the data of the sampling points in the experimental results. Figure 2 It can be seen that under different M values, as ω0 increases, the UAV flight distance D gradually increases, while the energy consumption balance index U gradually decreases, verifying the competitive relationship between D and U. Therefore, the weight ω0 needs to be appropriately reduced under the application requirements that are sensitive to the UAV flight energy consumption, and when there is a large deviation in the energy consumption distribution in the network, the weight ω0 needs to be increased in time to prevent the energy void problem caused by the exhaustion of energy of some nodes. In the subsequent simulation, without loss of generality, ω0 = 0.5 is selected.

[0099] In order to further compare and analyze the CS reconstruction accuracy and UAV flight distance of the CSNS algorithm, this application sets the random algorithm and the greedy algorithm as two comparative benchmark algorithms, where the random algorithm uses a random node selection method from the network, and the greedy algorithm uses a node selection method with the shortest UAV flight distance as the optimization goal. The real data collected from the GreenOrbs system sensor node is used as the simulation data for the comparative experiment, the DFT basis is selected as the sparse basis of the data, the OMP algorithm is the CS reconstruction algorithm, and the relative error of the reconstructed data is defined as in For the CS reconstruction data of K nodes, all comparative experimental data are the average of 1000 independent repeated experiments.

[0100] Figure 3 The CS reconstruction accuracy performance of different node selection strategies under different M values. As can be seen from the figure, with the increase of the number of participating nodes (M value), the reconstruction error corresponding to each scheme gradually decreases. When M exceeds 6, the reconstruction errors of the three algorithms tend to be stable. Under different M values, the CSNS algorithm has the smallest reconstruction error, while the reconstruction error under the greedy algorithm is the largest. This is because the threshold constraint on the reconstruction error is added to the CSNS algorithm, and the greedy algorithm uses the shortest UAV flight distance greedy strategy for non-uniform CS data sampling, so its corresponding reconstruction error is the largest. Figure 4 The total flight distance of UAV under different algorithms is given. As the value of M increases, the UAV flight distance increases to varying degrees. When M=6, the CSNS algorithm is about 45% lower than the random algorithm in UAV flight distance. Since the reconstruction error of the greedy algorithm does not meet the application requirements, its corresponding UAV flight distance is only used as a theoretical benchmark lower limit.

[0101] B. Performance of the TSJO algorithm

[0102] This section analyzes the performance of the TSJO algorithm for solving the energy consumption iterative optimization subproblem S2P. Set the number of node selections M = 6, and the transmission task volume of the node is Q K =50Mbits. In order to evaluate the convergence of the alternating iterative optimization process in the TSJO algorithm, the first-order difference value E of the energy consumption of the UAV-SNs system is defined as D (i)=E(i)-E(i+1), first-order difference value E D The smaller (i) is, the smaller the change in UAV-SNs energy consumption after the TSJO algorithm is optimized in the i-th iteration, and the more the algorithm converges. Figure 5 The convergence of the TSJO algorithm under different weight coefficients ω is shown. According to the experimental results, the TSJO algorithm can converge quickly. When the number of iterations is greater than 8, the first-order difference value tends to zero, and the energy consumption of UAV-SNs remains almost unchanged. When the number of iterations is less than 3, the larger the energy consumption weight coefficient ω, the greater the change in the first-order difference value. Since the larger the weight coefficient ω, the greater the proportion of UAV energy consumption in the energy consumption of the UAV-SNs system, the result reflects that the alternating iteration process of the TSJO algorithm has a greater impact on the energy consumption of the UAV, and a relatively small impact on the energy consumption of the SNs.

[0103] To further evaluate the impact of the weight coefficient ω on energy consumption performance, Figure 6The comparison of UAV energy consumption and node energy consumption under different weight coefficients ω is shown, where the left ordinate is node energy consumption and the right ordinate is UAV energy consumption. According to the experimental results, it can be found that with the increase of weight ω, UAV energy consumption gradually decreases and node energy consumption gradually increases. This is because the increase of weight ω makes the optimization goal of system energy consumption more focused on reducing UAV energy consumption, so the system scheduling strategy will give priority to shortening the UAV flight distance. However, this strategy will make the UAV fly away from the node, thereby increasing the communication distance and reducing the communication rate, so the node energy consumption increases accordingly. Therefore, in practical applications, for application scenarios with large coverage, since UAV needs to fly long distances, the system application is more sensitive to UAV energy consumption, and the weight ω can be appropriately increased during application; in application scenarios with low node energy reserves, the system application is more sensitive to node energy consumption. In order to extend the network life, a smaller ω value needs to be selected.

[0104] C. Performance of energy consumption optimization methods

[0105] First, the energy consumption performance of the energy consumption optimization method described in this application is evaluated, and two benchmark mechanisms TSP-Random and TSP-CSNS are used for comparison. Both benchmark mechanisms use the commonly used TSP method to perform data collection. The UAV starts from the starting point and flies to the nodes in turn to hover over the nodes to perform data collection until it reaches the end point. The difference between the two lies in the different node selection algorithms used. The TSP-Random mechanism uses a node random selection algorithm, while the TSP-CSNS mechanism uses the CSNS algorithm designed in this application. The simulation experiment sets the data collection volume of the node to Q K =50Mbits, and the energy consumption weight coefficient is set to ω=0.5. Figure 7 The UAV-SNs energy consumption of the three mechanisms under different M values ​​is given. The results show that under the same M value, the UAV-SNs energy consumption of the data collection mechanism proposed in this application is significantly better than the other two benchmark mechanisms. For the TSP-CSNS mechanism, since the CSNS algorithm is used to optimize node selection, it has lower UAV-SNs energy consumption than the TSP-Random mechanism. This result also further verifies the effectiveness of the CSNS algorithm.

[0106] Figure 8 and Fig. 9 The proposed energy consumption optimization mechanism is given respectively for UAV flight trajectory, UAV flight time and node scheduling strategy under different transmission task loads. Figure 8 The experimental results show that when Q K =10Mbits, the UAV flight trajectory is far from the node location, and this flight strategy reduces the total flight distance of the UAV. kAs the communication task Q increases, the UAV flight trajectory gradually approaches the node, and the UAV flight distance also gradually increases. K = 200Mbits, the UAV needs to fly almost directly above the node to collect data in order to increase the communication rate and ensure the effective execution of the data collection task. Figure 8 The total flight time T of the UAV under four communication transmission requirements is also given, which are 38.9s, 41.3s, 50.7s and 99.2s respectively. Although intuitively, with the increase of communication tasks, the total flight distance of the UAV does not increase much, but the flight time increases significantly. This reflects that the energy consumption optimization mechanism proposed in this application will adaptively adjust the flight speed of the UAV according to the communication task requirements to minimize energy consumption. This has obvious advantages compared to the fixed flight time strategy used in most current UAV trajectory optimization processes.

[0107] Fig. 9 The communication scheduling strategies of M nodes under four communication task loads are shown. It should be noted that since the total flight time of UAVs under different communication task loads is different, the time ranges of the horizontal axes in different subgraphs are not consistent. Figure 8 From the UAV flight trajectory in the problem, it can be found that when the communication task volume is determined, the closer the UAV flight trajectory is to the node, the shorter the node data transmission time is. This is because the shortening of the communication distance increases the communication rate, thus shortening the communication time. In addition, due to the C4 constraint in problem S2P1, as Q K As the node's data transmission time increases, the node's data transmission time also increases.

[0108] In summary, this application establishes a UAV-SNs energy consumption minimization problem model based on sparse CS data collection technology, and proposes an effective UAV-assisted data collection energy consumption optimization mechanism. Through analysis, the difficult-to-solve energy consumption minimization problem is decoupled into a CS node selection strategy optimization subproblem and an energy consumption iteration optimization subproblem, and the CSNS algorithm and TSJO algorithm are designed to solve the above two subproblems, thereby obtaining an approximate solution to the UAV-SNs energy consumption minimization problem. The comparative experimental results verify the effectiveness and superiority of the energy consumption optimization method proposed in this application.

[0109] Finally, it should be noted that the parts of the present invention that are not described in detail are all prior art. Those of ordinary skill in the art can understand that the above are only preferred examples of the invention and are not intended to limit the invention. Although the invention is described in detail with reference to the aforementioned examples, those of ordinary skill in the art can still modify the technical solutions recorded in the aforementioned examples, or replace some of the technical features therein with equivalents. Any modifications, equivalent replacements, etc. made within the spirit and principles of the invention should be included in the scope of protection of the invention.

Claims

1. A high-efficiency data collection method for a large-scale space-ground integrated Internet of Things network, characterized in that: The following steps are involved: S1, build a CS data collection model and introduce a reconstruction error constraint C1 to limit the upper limit of the CS data reconstruction error; S2, construct UAV energy consumption model and node energy consumption model, decompose UAV flight trajectory by UAV trajectory discrete method, and introduce UAV flight trajectory constraint conditions; S3, construct the communication model of data transmission between UAV and nodes and introduce communication constraints; S4, construct an energy consumption minimization problem model, with UAV flight trajectory, UAV flight time, and node scheduling strategy as optimization variables; S5, node selection strategy: taking the energy consumption balance of network nodes and minimization of UAV flight distance as optimization goals, constructing a cost evaluation model for node selection, introducing the mean square error of the least squares estimate to evaluate the CS reconstruction accuracy, converting the constraint C1 into a threshold constraint on the mean square error of the least squares estimate, and then solving the optimization path matrix through the strategy optimization algorithm to select M nodes participating in data collection; S6, iterative optimization of energy consumption: Using the node selection strategy of S5, the energy consumption minimization problem of S4 is transformed into an energy consumption iterative optimization sub-problem, and an energy consumption iterative optimization model is constructed. A two-stage joint optimization algorithm is adopted. First, given the UAV flight trajectory, the optimal UAV flight time and node scheduling strategy are solved; secondly, given the optimized UAV flight time and node scheduling strategy, the optimal UAV flight trajectory is solved; finally, by alternately iterating the above process until the algorithm converges, the three optimization variable values ​​that minimize the energy consumption of UAV-SNs are obtained.

2. The high-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 1 is characterized in that: In S2, the constraint conditions include constraint C2: limiting the flight trajectory of the UAV to start from the starting point and land at the end point, limiting the maximum flight distance of a single path segment of the UAV; In S3, the constraints include constraint C3: limiting that at most only one node can perform data transmission with the UAV at a time; constraint C4: limiting the minimum data transmission task amount of each node selected to transmit data; constraint C5: limiting the communication distance between the UAV and the node to not exceed the effective communication range.

3. The high-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 2 is characterized in that: In S4, the energy consumption minimization problem model is: Among them, A represents the node scheduling strategy, Q represents the UAV flight trajectory, T represents the UAV flight time, λ is the energy balance factor, which is used to adjust the deviation caused by the UAV energy consumption and node energy consumption not being at the same level; ω is the weight coefficient for measuring the importance of UAV energy consumption and node energy consumption, E s Represents the node energy consumption, E uav Represents UAV flight energy consumption.

4. The high energy efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 1 or 3, characterized in that: In S5, a cost evaluation model for node selection is constructed: f(k)=ω0U r (k)+(1-ω0)D(k) Among them, 0≤ω0≤1 is the weight factor for measuring the node access rate and node distance cost; D(k) is the node SN k The distance cost objective function, U r (k) is the node SN k The access cost objective function.

5. The high-energy-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 4 is characterized in that: In S5, the least squares estimate of the reconstructed signal is: The mean square error of the least squares estimate of the reconstructed signal is: Among them, Φ is the CS observation matrix, Ψ∈ is the sparse basis matrix, is the path matrix, S is the sparse signal, σ 2 is the variance, Tr(·) represents the trace of the matrix; The constraint threshold of the mean square error of the least squares estimate is: in, The path matrix corresponding to the Gaussian random node selection strategy.

6. The high-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 5 is characterized in that: In S5, the node selection subproblem S1P is transformed into an optimization problem of selecting M nodes from K nodes and minimizing the sum of the costs of the M nodes under the CS reconstruction accuracy constraint. The reconstruction error is limited to not exceed the constraint threshold η.

7. The high-energy-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 6 is characterized in that: In S5, the CSNS strategy optimization algorithm is used to select M nodes participating in data collection. The specific steps are as follows: S51, the binary variable Relaxed to a continuous variable between 0 and 1, the non-convex problem S1P1 is transformed into a convex problem S1P2, S52, using convex optimization algorithm, solve S1P2 and obtain the optimal path matrix S53, for the set Sort the values ​​​​in descending order to obtain the sequence S54, according to the sequence The index of the first M values ​​of the node set is constructed 8. The high-energy-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 7 is characterized in that: Therefore, in S6, a continuous convex approximation algorithm is used to transform the non-convex problem into a series of convex sub-problems through variable relaxation and substitution, and the optimal solution to the energy consumption minimization problem is approached iteratively until the algorithm converges.

9. The high-energy-efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 8, characterized in that: In S5, solving the optimization path matrix is ​​completed in advance by the remote server at the UAV take-off and landing point before each data collection; in S6, the two-stage joint optimization algorithm is executed in advance on the remote server.

10. The high energy efficiency data collection method for a large-scale space-ground integrated Internet of Things network according to claim 1, characterized in that: The IoT network includes a ground network responsible for sensing and monitoring regional data, a UAV mobile aerial platform responsible for data collection, and a satellite network responsible for transmitting data back to a remote data processing server. The UAV acts as a mobile aerial base station and periodically takes off from a starting point to perform flight missions with a specific flight trajectory, flight time, and node scheduling strategy. The CS sparse sampling method is used to collect data from IoT nodes across the entire network.

Images