Low-orbit constellation beam resource allocation method based on adaptive evolutionary algorithm

Abstract

The invention discloses a low-orbit constellation beam resource allocation method based on an adaptive evolutionary algorithm, and the method comprises the implementation steps: obtaining a user demand prediction result through a trained user demand prediction neural network, pre-judging user demand fluctuation in advance, and improving the real-time performance and accuracy of resource allocation. On the basis of a preference list, a Gal-Shapley algorithm is utilized to match satellite beams and users in a beam overlapping area, an objective function and a resource allocation constraint condition are defined according to a satellite communication system model, a self-adaptive evolutionary algorithm is proposed to jointly allocate power and bandwidth, and beam resource allocation optimization is achieved. According to the method, the capacity provided by the satellite is matched with the capacity required by the user, so that the user service coverage rate is improved, and meanwhile, the resource utilization rate of the low-orbit satellite is improved.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and further relates to a low-orbit constellation beam resource allocation method based on an adaptive evolutionary algorithm in the field of satellite communication technology. The present invention can be used to improve the service coverage of giant low-orbit constellations facing heterogeneous user needs. According to the user's traffic demand, the bandwidth and power resources of the constellation beam are dynamically managed based on an optimization algorithm, and the purpose of improving the service coverage is achieved by matching the provided traffic with the traffic required by the user. Background Art

[0002] With the growing demand for global communication services, building a giant satellite constellation network has become a key development trend to improve system performance and meet user needs. With the advantage of its low-Earth orbit, low-orbit satellites can provide high-bandwidth and low-latency services to global users, which makes them show great application potential in many fields such as the Internet of Things, intelligent transportation, telemedicine, and emergency communications. However, due to the characteristics of large angular velocity and high-speed movement of low-orbit satellites, users in the coverage area are constantly switched over time, and the distribution of user terminals in the satellite coverage area is usually uneven. These uneven distributions lead to significant differences in user needs for different beams, especially during peak hours, when some areas may face excessive resource use, while other areas cannot fully utilize available resources. This unbalanced user distribution and dynamic traffic demand make satellite communications difficult to provide services. At the same time, due to the increase in the number of satellites and users, it brings about problems of high computational complexity and low resource utilization. Therefore, how to design an effective beam resource management strategy to match the provided traffic with the traffic required by users to achieve improved service coverage while ensuring on-board resource utilization has become one of the important issues that need to be urgently solved in current satellite communication systems.

[0003] Xi'an Institute of Space Radio Technology disclosed a multi-beam control method for a low-orbit mobile communication satellite constellation in its patent application "A multi-beam control method for a low-orbit mobile communication satellite constellation" (application date: June 18, 2021, application number: CN 202110678636.1, authorization announcement number: CN113612512 B). Based on the characteristics of the low-orbit satellite constellation, this method proposes a dynamic multi-beam control method that can adjust the beam direction, power allocation and spectrum resources of each satellite according to dynamic factors such as satellite position and communication period. During the operation of the satellite constellation, the communication frequency of each satellite is allocated and managed to avoid co-frequency interference in the overlapping coverage area of ​​the satellite group on the ground at any time, control the beam coverage status of each satellite, reduce the mutual interference of signals in the satellite coverage area, and analyze the C / I of the satellite coverage area at the same time. By adopting a new constellation beam management and control method, the problem of mutual interference of signals in the satellite coverage area can be reduced while ensuring seamless coverage of the ground surface. However, this method still has two shortcomings. First, the proposed method does not take into account the real-world scenario of uneven distribution of ground users. In practical applications, there are often large differences in the distribution of ground users, such as the gap between cities and remote areas. Second, the traffic demand of users in different regions also varies. Beam adjustment based solely on factors such as the satellite's orbital position and communication period may not achieve optimal resource allocation.

[0004] Chongqing University of Posts and Telecommunications disclosed a method for allocating resources in a multi-beam satellite communication system based on a multi-agent A3C algorithm in its patent application "Multi-beam satellite communication system resource allocation method based on a multi-agent A3C algorithm" (application date: August 30, 2023, application number: 202311105149.1, application publication number: CN116896407 A). The specific steps of the method are: the first step is to establish a low-orbit satellite communication system model and a ground service model; the second step is to model the beam association variable scheme and the power allocation variable scheme, and determine the satellite load balancing scheme through satellite channel modeling, and model the system utility function and constraints; the third step is to model the system state, action and reward, construct and train the multi-agent A3C network, and determine the system beam hopping scheme and resource allocation strategy based on the multi-agent A3C algorithm. The disadvantage of this method is that the objective function in the proposed method only considers the user's satisfaction, which may lead to insufficient allocation of satellite resources, resulting in low utilization of satellite resources. In addition, the algorithm training process requires a large amount of historical data and computing resources. The resource allocation strategy training process may be affected by factors such as the dynamic changes of the satellite coverage area over time and the convergence delay, resulting in a more complicated training process and a slower convergence speed, which affects the real-time and accuracy of the resource allocation strategy. Summary of the invention

[0005] The purpose of the present invention is to propose a low-orbit constellation beam resource allocation method based on an adaptive evolutionary algorithm in view of the deficiencies in the above-mentioned prior art, so as to solve the problems of low user service coverage and satellite resource utilization due to the uneven distribution of ground users and differentiated user needs, as well as the lack of real-time and accuracy of beam resource allocation due to dynamic changes of satellite systems and algorithm convergence delays.

[0006] The technical idea of ​​the present invention to achieve the purpose is that the present invention establishes a satellite communication system model based on the basic parameters of the low-orbit constellation communication link. For the problem of insufficient real-time and accuracy of beam resource allocation caused by dynamic changes and convergence delays of the satellite system, the user demand prediction neural network is trained using a training set constructed by historical user terminal positions and user service demand information to obtain user demand prediction results, and user demand fluctuations are predicted in advance to improve the real-time and accuracy of resource allocation. For the problem of low user service coverage and satellite resource utilization, the Gale-Shapley algorithm is used in the beam domain to match the overlapping area satellite and the user terminal based on the preference list to obtain a matching list. In the user domain, based on the user demand prediction results and the matching list, combined with the defined system model objective function and resource allocation constraints, the adaptive evolutionary algorithm is used to jointly allocate power and bandwidth to determine the optimal beam resource allocation strategy. It ensures that the satellite communication capacity provided by the satellite beam matches the user demand, improves the user's service coverage and minimizes the waste of satellite resources.

[0007] To achieve the above object, the specific implementation steps of the present invention include the following:

[0008] Step 1: Establish a satellite communication system model based on the orbit information of the low-orbit constellation of the beam resources to be allocated and the basic parameters of the communication link;

[0009] Step 2: Based on historical user terminal locations and user service demand information, a training set of sample data is constructed, and a user demand prediction neural network trained with the training set is used to obtain a user demand prediction result;

[0010] Step 3, obtaining the user terminal location information in the overlapping area of ​​beams between different satellites, using the Gale-Shapley algorithm to match the satellites in the overlapping area with the user terminals, and generating a matching list;

[0011] Step 4, defining the objective function and resource allocation constraints according to the satellite communication system model;

[0012] Step 5: Based on the user demand prediction results and the matching list, the power and bandwidth are jointly allocated through an adaptive evolutionary algorithm to determine the optimal beam resource allocation strategy.

[0013] Furthermore, the satellite communication system model specifically includes:

[0014] The first step is to establish the gain of each channel: h b,u (t) = a b,c (t)·G b,u (θ t )·G b,u (θ r ), where h b,u (t) represents the channel gain of the bth beam reaching the uth user at the tth time slot, b=1,2,...,B, B represents the total number of beams of a single satellite, u=1,2,...,U, U represents the total number of user terminals, t=1,2,...,T, T represents the total number of time slots of length 50ms into which the total time of the low-orbit constellation system to be allocated is equally divided, and the ground area served by the satellite system in each time slot remains unchanged, a b,c (t) represents the free space loss from the bth beam to the cth user at the tth time slot, G b,u (.), G b,u (.) represent the transmission gain and receiving antenna gain between the bth beam and the uth user, θ t and θ r Respectively represent the angles of the receiving end and the transmitting end;

[0015] The second step is to establish the signal-to-interference-noise ratio from each beam to each user: Among them, SINR b,u represents the signal-to-interference-noise ratio from the bth beam to the uth user, i b,u represents the channel allocation coefficient from the bth beam to the uth user, p b,u represents the transmission power of the bth beam allocated to the uth user, p j,u represents the transmission power of the jth beam allocated to the uth user, N0 represents the noise power spectrum density, B total represents the total bandwidth of each satellite, h j,u represents the channel gain of the jth beam reaching the uth user at the tth time slot;

[0016] The third step is to establish the satellite communication capacity provided by each satellite to each user: in, represents the communication capacity provided by the bth beam of the nth satellite to the nth user, B n,b,u represents the bandwidth allocated to the u-th user by the b-th beam of the n-th satellite, n=1,2,...,N, and N represents the total number of satellites in the LEO constellation.

[0017] Furthermore, the training set of the sample data includes the location information of all user terminals in the low-orbit constellation system where resources are to be allocated and the label data of user service requirements; the user terminal location information is a three-dimensional data matrix, the rows of the user terminal location information matrix represent the number of time slots before the user demand time slot to be predicted, the columns represent the number of each user terminal, and the depth is the three-dimensional position coordinates of the user terminal in the Cartesian coordinate system in each time slot; the label data of the user service requirements is a two-dimensional matrix, the rows of the matrix represent the number of user demand time slots to be predicted in the low-orbit constellation system where resources are currently to be allocated, the columns represent the number of each user terminal, and each element value in the matrix represents the user service requirement.

[0018] Furthermore, the structure of the user demand prediction neural network is composed of a first fully connected layer, a hidden layer, and a second fully connected layer connected in series in sequence; wherein, according to the size of the user terminal location information, the input feature number of the first fully connected layer is set to 3×V, and the value of V is equal to the total number of user terminals U; the number of input channels is equal to the number of time slots of the user demand to be predicted; the number of output neurons is set to 9×V; the hidden layer is composed of three first hidden layers, a second hidden layer, and a third hidden layer with the same structure connected in series in sequence; each hidden layer is composed of a first fully connected layer, a first activation function layer, a second fully connected layer, and a second activation function layer connected in series in sequence; the output neurons of the first and second fully connected layers of the first hidden layer are connected in series in sequence. The number of neurons in the first and second fully connected layers of the second hidden layer is set to 8×V and 7×V respectively; the number of output neurons in the first and second fully connected layers of the second hidden layer is set to 6×V and 5×V respectively; the number of output neurons in the third hidden layer is set to 4×V, and the number of output neurons in the second fully connected layer is set to 3×V; the number of input channels of the first fully connected layer and the second fully connected layer of the first to third hidden layers is set to 1, the first activation function layer of the first to third hidden layers is implemented by the Tanh activation function, and the second activation function is implemented by the sigmoid activation function; according to the output feature tensor of the third hidden layer, the number of input features of the second fully connected layer is set to 3×V, the number of input channels is 1, and the number of output neurons is set to V.

[0019] Furthermore, the training of the user demand prediction neural network refers to inputting the training set into the user demand prediction neural network, using the gradient descent method to iteratively update the network parameters until the network loss function converges, thereby obtaining a trained user demand prediction neural network.

[0020] The network loss function is: Among them, L(.) represents the loss function, θ represents the network optimization parameter of the user demand prediction neural network in the current iteration, M represents the number of users to be allocated resources in the current iteration, Indicates the user demand predicted by the user demand prediction neural network in the current iteration, Representation and Corresponding to actual user needs.

[0021] Furthermore, the steps of matching satellites and user terminals in overlapping areas of different satellite beams using the Gale-Shapley algorithm are as follows:

[0022] The first step is to establish a user terminal preference list and a satellite preference list of a beam overlap area respectively; the user terminal preference list includes a matching satellite number covering each user within each beam overlap area; the satellite preference list includes terminal numbers of all users covered by each satellite within each beam overlap area; each user terminal in the satellite preference list is arranged in ascending order according to the number, and the smallest satellite number is stored in the user terminal preference list;

[0023] The second step is to calculate the signal strength of each satellite to each user in each beam overlap area: Among them, ML q,n,u represents the signal strength of the nth satellite to the uth user in the qth beam overlap area, d n,u Indicates the distance from the nth satellite to the uth user in the beam overlap area;

[0024] The third step is to compare the nth user terminal with all satellites covering the uth user terminal in the order of arrangement in each beam overlap area. u,i satellites and the signal strength of the n'th satellite to the uth user stored in WL(u). If Then use the nth u,i The number of satellites n u,i Replace the satellite number stored in the preference list WL(u) as n', delete the u-th user terminal number in the preference list WS(n) of the n-th satellite, and add the deleted u-th user terminal number to the n-th satellite preference list WS(n). u,i If the preference list of satellites is Then no operation is performed, and the satellite channel gain for the u-th user terminal finally stored in WL(u) is added to the matching list;

[0025] In the fourth step, each user terminal in each overlapping area is traversed and the third step is performed on each user terminal to obtain a matching list for each overlapping area composed of a two-dimensional matrix. The rows of the matching list represent the satellite numbers of each in the beam overlapping area, and the columns represent the numbers of each user terminal in the beam overlapping area. Each element value in the matrix represents the matching channel gain between the satellite and the user terminal.

[0026] Furthermore, the objective function and resource allocation constraints are defined according to the satellite communication system model:

[0027] The objective function is: Among them, USC represents the maximum unmet service coverage minimization function of user demand in the low-orbit constellation beam, that is, to minimize and The gap, represents the satellite communication capacity provided by the bth beam in the nth satellite to the uth user, represents the communication capacity required by the u-th user;

[0028] The resource allocation constraints include: channel allocation status constraint, bandwidth constraint, angle constraint from user to beam center, maximum allocation power constraint, and beam service user number constraint; the channel allocation status constraint is that the channel allocation status of each beam for each user can only be allocated or not allocated; the bandwidth constraint is that the total transmission bandwidth allocated to each beam for users within its coverage area does not exceed the maximum transmission bandwidth of the beam; the angle constraint from user to beam center is that the angle from each beam to the beam center for each user within its coverage area is less than 3dB of the beam; the maximum allocation power constraint is that the total transmission power allocated to all users within the coverage area of ​​each beam does not exceed the maximum available transmission power of the beam; the beam service user number constraint is that the number of users served by each beam does not exceed the maximum number of users allowed to be served by the beam.

[0029] Furthermore, the steps of jointly allocating power and bandwidth by using an adaptive evolutionary algorithm are as follows:

[0030] The first step is to randomly generate a set of resource pre-allocation strategy matrices R = {r1, r2, …, r Z} as the parent, where Z represents the total number of randomly generated resource pre-allocation strategy matrices;

[0031] The second step is to establish the fitness function of the genetic algorithm: f(r z )=-USC(r z ), where f(r z ) represents the fitness value of the z-th resource pre-allocation strategy matrix predicted by the trained neural network, represents the z-th resource pre-allocation strategy matrix, and denote the z1th power allocation matrix and the z2th bandwidth allocation matrix, respectively. The power allocation matrix and the bandwidth allocation matrix respectively contain the power and bandwidth allocated by each satellite beam to each user within its coverage area in the zth resource pre-allocation strategy. The superscript T denotes a transposition operation.

[0032] The third step is to establish the crossover probability function and mutation probability function as follows:

[0033]

[0034] Among them, P c 、p m They represent the crossover probability function and the mutation probability function respectively. represent the maximum and minimum values ​​of the crossover probability, respectively. They represent the maximum and minimum values ​​of the mutation probability, exp represents the exponential operation with the natural constant e as the base, c0 represents the linear parameter, and its value is 9.903438, f max Indicates the maximum fitness value in the current parent generation. represents the larger fitness value of the two resource allocation strategy vectors involved in the crossover, f m represents the fitness value of the current resource allocation strategy vector, f avg represents the average fitness value of all resource allocation strategy vectors of the parent generation, f z Represents the fitness value of the current z-th resource allocation strategy vector;

[0035] The fourth step is to select the parent generation based on the fitness value using the selection probability to obtain the selected parent generation, calculate the crossover probability using the crossover probability function, perform a pairwise crossover on the resource allocation strategy vectors of the selected parent generation to obtain the crossover parent generation, and then use the mutation probability function to calculate the mutation probability, perform a mutation operation on the resource allocation strategy vector of the crossover parent generation, and finally generate a new generation of resource pre-allocation strategy matrix R' as the new parent generation, and calculate the fitness of each resource allocation strategy vector in R' for the new parent generation based on the user demand predicted by the user demand prediction neural network;

[0036] In the fifth step, the fourth step is repeated until the fitness function of the adaptive evolutionary algorithm converges and the optimal resource allocation strategy is obtained.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] First, the present invention uses a training set constructed by using historical user terminal positions and user service demand information to train a user demand prediction neural network to obtain user demand prediction results and timely predict changes in user demand. In the beam domain, the Gale-Shapley algorithm is used to match satellites and user terminals in the overlapping area based on a preference list to obtain a matching list. In the user domain, based on the user demand prediction results and the matching list, combined with the defined objective function and resource allocation constraints, the power and bandwidth are jointly allocated through an adaptive evolutionary algorithm to determine the optimal beam resource allocation strategy. The proposed objective function comprehensively considers the satellite communication capacity provided by the satellite beam and the matching degree of user demand and the utilization rate of satellite resources, and overcomes the problems of low user service coverage, low satellite resource utilization, and insufficient real-time and accuracy of beam resource allocation caused by differentiated distribution of user demand, dynamic changes of satellite systems, and algorithm convergence delay in the prior art. While ensuring the user's service coverage and resource utilization, the real-time and accuracy of allocation are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of a satellite communication system according to an embodiment of the present invention;

[0040] Figure 2 It is a flow chart of the beam resource allocation method based on the adaptive evolutionary algorithm of the present invention;

[0041] Figure 3 It is a simulation result diagram of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings.

[0043] Reference Figure 1 , the satellite communication system of the present invention is further described.

[0044] The low-orbit satellite communication system used in this example includes 720 low-orbit multi-beam satellites with an orbital altitude of 1200km. The satellites use the Ka band (17.7~21.2GHz). Each satellite's onboard multi-beam transmitter transmits a total of 16 beams. The total system bandwidth is 40MHz, and the satellite's maximum transmitting antenna gain is 58.5dBi.

[0045] Reference Figure 2 , the specific steps of the beam resource allocation method of the present invention are further described.

[0046] Step 1: Establish a satellite communication system model based on the orbit information of the low-orbit constellation of the beam resources to be allocated and the basic parameters of the communication link.

[0047] In an embodiment of the present invention, a satellite communication scenario is built using Python according to the parameters of the low-orbit constellation, and after the beam is generated, the orbit information and basic parameters of the communication link of the low-orbit constellation are obtained.

[0048] The satellite communication system model establishment process is as follows:

[0049] The first step is to establish the gain of each channel as:

[0050] h b,u (t) = a b,c (t)·G b,u (θ t )·G b,u (θ r )

[0051] Among them, h b,u (t) represents the channel gain of the bth beam reaching the uth user in the tth time slot, b = 1, 2, ..., 16, u = 1, 2, ..., U, t = 1, 2, ..., T, T represents the total time of the low-orbit constellation system to be allocated resources divided into the total number of time slots of length 50 ms, and the ground area served by the satellite system in each time slot remains unchanged, a b,c (t) represents the free space loss from the bth beam to the cth user at the tth time slot, G b,u (.), G b,u (.) represent the transmission gain and receiving antenna gain between the bth beam and the uth user, θ t and θ r Represent the angles of the receiving end and the transmitting end respectively.

[0052] The second step is to establish the signal-to-interference-noise ratio from each beam to each user:

[0053]

[0054] Among them, SINR b,u represents the signal-to-interference-noise ratio from the bth beam to the uth user, i b,u represents the channel allocation coefficient from the bth beam to the uth user, p b,u represents the transmission power of the bth beam allocated to the uth user, p j,u represents the transmission power of the jth beam allocated to the uth user, N0 represents the noise power spectrum density, B total represents the total bandwidth of each satellite, h j,u It represents the channel gain of the j-th beam reaching the u-th user in the t-th time slot.

[0055] The third step is to establish the satellite communication capacity provided by each satellite to each user:

[0056]

[0057] in, represents the communication capacity provided by the bth beam of the nth satellite to the nth user, B n,b,u represents the bandwidth allocated to the u-th user by the b-th beam of the n-th satellite, n=1,2,...,N, and N represents the total number of satellites in the LEO constellation.

[0058] Step 2: Based on the historical user terminal location and user business demand information, a training set of sample data is constructed, and the user demand prediction neural network trained with the training set is used to obtain the user demand prediction result.

[0059] The steps of predicting user demand by the user demand prediction neural network are as follows:

[0060] Step 1. Obtain user service demands within the beam coverage range according to the satellite communication system scenario. The user service demands of an embodiment of the present invention are the product of the traffic concurrency rate of the current user area and the population density in the user area. The traffic concurrency rate refers to the 2021 traffic concurrency rate dataset, and the population density mainly refers to the grid population (GPW) v4 dataset. Through gridding, each 1°×1° area is regarded as a cell, and the population data is regarded as the population density of the cell.

[0061] Step 2: Perform data processing on the user terminal locations and user service demand information in the five time slots before the current time slot to obtain a training set.

[0062] In the third step, the training set is input into the user demand prediction neural network, and the network parameters are iteratively updated using the gradient descent method until the network loss function converges to obtain a trained user demand prediction neural network, and the user demand prediction neural network is used to obtain the predicted user demand.

[0063] Step 3: Obtain the user terminal location information in the overlapping area of ​​beams between different satellites, use the Gale-Shapley algorithm to match the satellites in the overlapping area with the user terminals, and generate a matching list.

[0064] The steps of matching the overlapping area satellites with the user terminals based on the Gale-Shapley algorithm are as follows:

[0065] Step 1: Establish a user terminal preference list and a satellite preference list in the beam overlap area respectively.

[0066] Step 2: Calculate the signal strength of each satellite to each user within each beam overlap area.

[0067] Step 3: In each beam overlap area, compare the signal strength of each satellite to the user in turn according to the satellite arrangement order to complete the matching list of users and satellites.

[0068] Step 4: Define the objective function and resource allocation constraints based on the satellite communication system model.

[0069] The objective function in the embodiment of the present invention is:

[0070]

[0071] Among them, USC represents the maximum unmet service coverage minimization function of user demand in the low-orbit constellation beam, that is, to minimize and The gap, represents the satellite communication capacity provided by the bth beam in the nth satellite to the uth user, represents the communication capacity required by the u-th user.

[0072] In the embodiment of the present invention, resource allocation constraints include: channel allocation state constraints, bandwidth constraints, angle constraints from users to beam center, maximum allocation power constraints, and beam service user number constraints.

[0073] The channel allocation state constraint means that the channel allocation state of each beam to each user can only be allocated or not allocated.

[0074] The bandwidth constraint is that the total transmission bandwidth allocated to users within the coverage area of ​​each beam does not exceed the maximum transmission bandwidth of the beam.

[0075] The angle constraint from the user to the beam center is that the angle from each user within the coverage area of ​​each beam to the beam center is less than 3 dB of the beam.

[0076] The maximum allocated power constraint is that the total transmission power allocated in each beam to all users within the coverage area of ​​the beam does not exceed the maximum available transmission power of the beam.

[0077] The constraint on the number of users served by the beam is that the number of users served by each beam does not exceed the maximum number of users allowed to be served by the beam.

[0078] Step 5: Based on the user demand prediction results and the matching list, the power and bandwidth are jointly allocated through an adaptive evolutionary algorithm to determine the optimal beam resource allocation strategy.

[0079] The steps of jointly allocating power and bandwidth by the adaptive evolutionary algorithm in the embodiment of the present invention are as follows:

[0080] Step 1: Based on the resource allocation constraints, a set of resource pre-allocation strategy matrices R = {r1, r2, ..., r Z} as the parent, where Z represents the total number of randomly generated resource pre-allocation strategy matrices.

[0081] Step 2: Establish the fitness function of the adaptive evolutionary algorithm:

[0082] f(r z )=-USC(r z )

[0083] Among them, f(r z ) represents the fitness value of the z-th resource pre-allocation strategy matrix predicted by the trained neural network, represents the z-th resource pre-allocation strategy matrix, and They represent the z1th power allocation matrix and the z2th bandwidth allocation matrix respectively. The power allocation matrix and the bandwidth allocation matrix respectively contain the power and bandwidth allocated by each satellite beam to each user within its coverage area in the zth resource pre-allocation strategy. The superscript T represents the transposition operation.

[0084] Step 3: Establish the crossover probability function and mutation probability function as follows:

[0085]

[0086] Among them, P c 、p m They represent the crossover probability function and the mutation probability function respectively. represent the maximum and minimum values ​​of the crossover probability, respectively. They represent the maximum and minimum values ​​of the mutation probability, exp represents the exponential operation with the natural constant e as the base, c0 represents the linear parameter, and its value is 9.903438, f max Indicates the maximum fitness value in the current parent generation. represents the larger fitness value of the two resource allocation strategy vectors involved in the crossover, f m represents the fitness value of the current resource allocation strategy vector, f avg represents the average fitness value of all resource allocation strategy vectors of the parent generation, f z Represents the fitness value of the current z-th resource allocation strategy vector.

[0087] Step 4: Use the selection probability to select the parent based on the fitness value to obtain the selected parent, use the crossover probability function to calculate the crossover probability, perform pairwise crossover on the resource allocation strategy vectors of the selected parent to obtain the crossover parent, then use the mutation probability function to calculate the mutation probability, perform mutation operation on the resource allocation strategy vector of the crossover parent, and finally generate a new generation of resource pre-allocation strategy matrix R' as the new parent, and calculate the fitness of each resource allocation strategy vector in R' for the new parent based on the user demand predicted by the user demand prediction neural network.

[0088] Step 6: Repeat step 4 of this step until the fitness function of the adaptive evolutionary algorithm converges and the optimal resource allocation strategy is obtained.

[0089] The effects of the present invention are further described below in conjunction with the simulation experiments of the present invention.

[0090] 1. Simulation experiment conditions:

[0091] The software platforms for the simulation experiment of the present invention are: Windows 11 operating system and Matlab2018b.

[0092] In the simulation experiment scenario of the present invention, the low-orbit constellation is the OneWeb constellation, the orbital altitude is 1200km, the number of beams per satellite is 16, and the user demand traffic mainly refers to the Grid Population (GPW) v4 dataset and the 2021 business concurrency rate dataset.

[0093] 2. Analysis of simulation content and results.

[0094] The simulation experiment of the present invention optimizes and simulates three different algorithms respectively, and obtains the user satisfaction standard deviation and beam resource utilization rate of the three algorithms under different total business demand conditions. Then, the software Matlab2018b is used to draw a line chart of the user satisfaction standard deviation of the three beam resource allocation algorithms under different total business demand conditions, as shown in the figure. Figure 3 (a) shows the beam resource utilization histogram of the three beam resource allocation algorithms under different total service demands. Figure 3 (b) as shown.

[0095] The three beam resource allocation algorithms are:

[0096] The first one is a low-orbit constellation beam resource allocation algorithm based on random allocation. The user demand prediction result is obtained by training the user demand prediction neural network proposed in step 2 of the embodiment of the present invention. The satellites and user terminals in the overlapping area of ​​different satellite beams are matched using the Gale-Shapley algorithm proposed in step 3. The beam resources are randomly allocated to each user terminal using a random function.

[0097] The second is a low-orbit constellation beam resource allocation algorithm based on a genetic algorithm. The user demand prediction neural network proposed in step 2 of the embodiment of the present invention is used to obtain the user demand prediction result. The Gale-Shapley algorithm proposed in step 3 is used to match the satellites and user terminals in the overlapping area of ​​different satellite beams. The natural selection process is simulated based on a genetic algorithm to randomly generate multiple beam resource allocation strategies as a beam resource allocation strategy set, and the fitness of each strategy is evaluated. The strategy with higher fitness is selected for crossover and mutation to generate a new generation of strategy sets. Repeat the evaluation and selection until the convergence condition is reached to determine the optimal beam resource allocation strategy.

[0098] The third type is a low-orbit constellation beam resource allocation algorithm based on an adaptive evolutionary algorithm proposed in the present invention. The user demand prediction neural network training proposed in step 2 of the embodiment of the present invention is used to obtain the user demand prediction result. The Gale-Shapley algorithm proposed in step 3 is used to match the satellites and user terminals in the overlapping area of ​​different satellite beams. Then, the adaptive evolutionary algorithm proposed in step 5 of the present invention is used to jointly allocate power and bandwidth to determine the optimal beam resource allocation strategy.

[0099] Combine the following Figure 3 The simulation effect of the present invention is further described.

[0100] Figure 3 The horizontal axis in (a) represents the total traffic demand of users in Gbps, and the vertical axis represents the standard deviation of user satisfaction. Figure 3 The horizontal axis in (b) represents the total user traffic demand in Gbps, and the vertical axis represents the satellite resource utilization rate, wherein the light gray rectangular box corresponds to the histogram of the LEO constellation beam resource allocation algorithm based on random allocation, the dark gray rectangular box corresponds to the histogram of the LEO constellation beam resource allocation algorithm based on the genetic algorithm, and the black rectangular box corresponds to the histogram of the LEO constellation beam resource allocation algorithm based on the adaptive evolutionary algorithm proposed in the present invention.

[0101] Depend on Figure 3It can be seen that when the total user traffic demand is lower than 800Gbps, because the resources are relatively abundant, the three algorithms are basically the same, and the standard deviation of user satisfaction of different algorithms is not much different overall. However, when the total user traffic demand is higher than the value of 800Gbps, since the satellite beam system resources are no longer abundant, the difference in user demand becomes larger, and the user service coverage of the low-orbit constellation beam resource allocation algorithm based on the adaptive evolutionary algorithm is significantly higher than the other two, and the resource utilization rate is also better. When the total user traffic demand reaches 1200Gbps, the method of the present invention increases the resource utilization rate to 88.78%. When the total user traffic demand is higher than 1320Gbps, the total user traffic demand begins to be greater than the total resources of the satellite beam system, and the effect of improving resource utilization begins to weaken. Figure 3 The displayed results indicate that the method proposed by the present invention improves the user service coverage as much as possible while also improving the satellite resource utilization.

[0102] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0103] The contents not described in detail in this specification belong to the common knowledge of those skilled in the art.

Claims

1. A low-orbit constellation beam resource allocation method based on an adaptive evolutionary algorithm, characterized in that: The objective function model of low-orbit satellite beam resource allocation is constructed. Through the adaptive adjustment algorithm, user needs and channel conditions are predicted to allocate beam bandwidth and power. The steps of the allocation method are as follows: Step 1: Establish a satellite communication system model based on the orbit information of the low-orbit constellation of the beam resources to be allocated and the basic parameters of the communication link; Step 2: Based on historical user terminal locations and user service demand information, a training set of sample data is constructed, and a user demand prediction neural network trained with the training set is used to obtain a user demand prediction result; Step 3, obtaining the user terminal location information in the overlapping area of ​​beams between different satellites, using the Gale-Shapley algorithm to match the satellites in the overlapping area with the user terminals, and generating a matching list; Step 4, defining the objective function and resource allocation constraints according to the satellite communication system model; Step 5: Based on the user demand prediction results and the matching list, the power and bandwidth are jointly allocated through an adaptive evolutionary algorithm to determine the optimal beam resource allocation strategy.

2. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 1, characterized in that: The satellite communication system model described in step 1 specifically includes: The first step is to establish the gain of each channel as: h b,u (t)=a b,c (t)·G b,u (θt)·G b,u (θ r ) Among them, h b,u (t) represents the channel gain of the bth beam reaching the uth user at the tth time slot, b=1,2,...,B, B represents the total number of beams of a single satellite, u=1,2,...,U, U represents the total number of user terminals, t=1,2,...,T, T represents the total number of time slots of length 50ms into which the total time of the low-orbit constellation system to be allocated is equally divided, and the ground area served by the satellite system in each time slot remains unchanged, a b,c (t) represents the free space loss from the bth beam to the cth user at the tth time slot, G b,u (.), G b,u (.) represent the transmission gain and receiving antenna gain between the bth beam and the uth user, θ t and θ r Respectively represent the angles of the receiving end and the transmitting end; The second step is to establish the signal-to-interference-noise ratio from each beam to each user: Among them, SINR b,u represents the signal-to-interference-noise ratio from the bth beam to the uth user, i b,u represents the channel allocation coefficient from the bth beam to the uth user, p b,u represents the transmission power of the bth beam allocated to the uth user, p j,u represents the transmission power of the jth beam allocated to the uth user, N0 represents the noise power spectral density, B total represents the total bandwidth of each satellite, h j,u represents the channel gain of the jth beam reaching the uth user at the tth time slot; The third step is to establish the satellite communication capacity provided by each satellite to each user: in, represents the communication capacity provided by the bth beam of the nth satellite to the nth user, B n,b,u represents the bandwidth allocated to the u-th user by the b-th beam of the n-th satellite, n=1,2,...,N, and N represents the total number of satellites in the LEO constellation.

3. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 1, characterized in that: The training set of the sample data in step 2 includes the historical terminal positions, user service demand information, and label data of user service demands of all users in the low-orbit constellation system where resources are to be allocated; the user terminal position information is a three-dimensional data matrix, the rows of which represent the number of time slots before the time slot of the user demand to be predicted, the columns represent the number of each user terminal, and the depth is the three-dimensional position coordinates of the user terminal in the Cartesian coordinate system in each time slot; the label data of the user service demand is a two-dimensional matrix, the rows of which represent the number of time slots of the user demand to be predicted in the low-orbit constellation system where resources are to be allocated, the columns represent the number of each user terminal, and each element value in the matrix represents the user service demand.

4. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 3, characterized in that: The structure of the user demand prediction neural network in step 3 is composed of a first fully connected layer, a hidden layer, and a second fully connected layer connected in series in sequence; wherein, the input feature number of the first fully connected layer is set to 3×V according to the size of the user terminal location information, and the value of V is equal to the total number of user terminals U; the number of input channels is equal to the number of time slots of the user demand to be predicted; the number of output neurons is set to 9×V; the hidden layer is composed of three first hidden layers, a second hidden layer, and a third hidden layer with the same structure connected in series in sequence; each hidden layer is composed of a first fully connected layer, a first activation function layer, a second fully connected layer, and a second activation function layer connected in series in sequence; the output neurons of the first and second fully connected layers of the first hidden layer are connected in series. The numbers of output neurons of the first and second fully connected layers of the second hidden layer are set to 8×V and 7×V respectively; the number of output neurons of the first and second fully connected layers of the second hidden layer are set to 6×V and 5×V respectively; the number of output neurons of the third hidden layer is set to 4×V, and the number of output neurons of the second fully connected layer is set to 3×V; the number of input channels of the first fully connected layer and the second fully connected layer of the first to third hidden layers are both set to 1, the first activation function layer of the first to third hidden layers are all implemented using the Tanh activation function, and the second activation function is all implemented using the Sigmoid activation function; according to the output feature tensor of the third hidden layer, the number of input features of the second fully connected layer is set to 3×V, the number of input channels is 1, and the number of output neurons is set to V.

5. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 4, characterized in that: The training of the user demand prediction neural network in step 3 refers to inputting the training set into the user demand prediction neural network, using the gradient descent method to iteratively update the network parameters until the network loss function converges, and obtaining a trained user demand prediction neural network.

6. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 5, characterized in that: The network loss function is as follows: Among them, L(.) represents the loss function, θ represents the network optimization parameter of the user demand prediction neural network in the current iteration, M represents the number of users to be allocated resources in the current iteration, Indicates the user demand predicted by the user demand prediction neural network in the current iteration, Representation and Corresponding to actual user needs.

7. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 2, characterized in that: The steps of using the Gale-Shapley algorithm in step 3 to match satellites and user terminals in overlapping areas of different satellite beams are as follows: The first step is to establish a user terminal preference list and a satellite preference list of the beam overlap area respectively; the user terminal preference list includes the matching satellite number covering each user within each beam overlap area; The satellite preference list includes terminal numbers of all users covered by each satellite within each beam overlap area; Arrange each user terminal in the satellite preference list in ascending order according to the number, and store the smallest satellite number in the user terminal preference list; The second step is to calculate the signal strength of each satellite to each user in each beam overlap area: Among them, ML q,n,u represents the signal strength of the nth satellite to the uth user in the qth beam overlap area, d n,u Indicates the distance from the nth satellite to the uth user in the beam overlap area; The third step is to compare the nth user terminal with all satellites covering the uth user terminal in the order of arrangement in each beam overlap area. u,i satellites and the signal strength of the n'th satellite to the uth user stored in WL(u). If Then use the nth u,i The number of satellites n u,i Replace the satellite number stored in the preference list WL(u) as n', delete the u-th user terminal number in the preference list WS(n) of the n-th satellite, and add the deleted u-th user terminal number to the n-th satellite preference list WS(n). u,i If the preference list of satellites is Then no operation is performed, and the satellite channel gain for the u-th user terminal finally stored in WL(u) is added to the matching list; In the fourth step, each user terminal in each overlapping area is traversed and the third step is performed on each user terminal to obtain a matching list for each overlapping area composed of a two-dimensional matrix. The rows of the matching list represent the satellite numbers of each in the beam overlapping area, and the columns represent the numbers of each user terminal in the beam overlapping area. Each element value in the matrix represents the matching channel gain between the satellite and the user terminal.

8. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 1, characterized in that: The objective function and resource allocation constraints defined in step 4 according to the satellite communication system model are as follows: The objective function is: Among them, USC represents the maximum unmet service coverage minimization function of user demand in the low-orbit constellation beam, that is, to minimize and The gap, represents the satellite communication capacity provided by the bth beam in the nth satellite to the uth user, represents the communication capacity required by the u-th user; The resource allocation constraints include: channel allocation state constraints, bandwidth constraints, angle constraints from user to beam center, maximum allocation power constraints, and beam service user number constraints; The channel allocation state constraint is that the channel allocation state of each beam to each user can only be allocated or not allocated; The bandwidth constraint is that the total transmission bandwidth allocated to users within the coverage area of ​​each beam does not exceed the maximum transmission bandwidth of the beam; The angle constraint from the user to the beam center is that the angle from each user within the coverage area of ​​each beam to the beam center is less than 3 dB of the beam; The maximum allocated power constraint is that the total transmission power allocated in each beam to all users within the coverage area of ​​the beam does not exceed the maximum available transmission power of the beam; The constraint on the number of users served by the beam is that the number of users served by each beam does not exceed the maximum number of users allowed to be served by the beam.

9. The method for allocating low-orbit constellation beam resources based on an adaptive evolutionary algorithm according to claim 8, characterized in that: The steps of jointly allocating power and bandwidth by adaptive evolutionary algorithm described in step 5 are as follows: The first step is to randomly generate a set of resource pre-allocation strategy matrices R = {r1, r2, ..., r Z } as the parent, where Z represents the total number of randomly generated resource pre-allocation strategy matrices; The second step is to establish the fitness function of the adaptive evolutionary algorithm: f(r z )=-USC(r z ) Among them, f(r z ) represents the fitness value of the z-th resource pre-allocation strategy matrix predicted by the trained neural network, represents the z-th resource pre-allocation strategy matrix, and denote the z1th power allocation matrix and the z2th bandwidth allocation matrix, respectively. The power allocation matrix and the bandwidth allocation matrix respectively contain the power and bandwidth allocated by each satellite beam to each user within its coverage area in the zth resource pre-allocation strategy. The superscript T denotes a transposition operation. The third step is to establish the crossover probability function and mutation probability function as follows: Among them, P c 、p m They represent the crossover probability function and the mutation probability function respectively. represent the maximum and minimum values ​​of the crossover probability, respectively. They represent the maximum and minimum values ​​of the mutation probability, exp represents the exponential operation with the natural constant e as the base, c0 represents the linear parameter, and its value is 9.903438, f max Indicates the maximum fitness value in the current parent generation. represents the larger fitness value of the two resource allocation strategy vectors involved in the crossover, f m represents the fitness value of the current resource allocation strategy vector, f avg represents the average fitness value of all resource allocation strategy vectors of the parent generation, f z Represents the fitness value of the current z-th resource allocation strategy vector; The fourth step is to select the parent generation based on the fitness value using the selection probability to obtain the selected parent generation, calculate the crossover probability using the crossover probability function, perform a pairwise crossover on the resource allocation strategy vectors of the selected parent generation to obtain the crossover parent generation, and then use the mutation probability function to calculate the mutation probability, perform a mutation operation on the resource allocation strategy vector of the crossover parent generation, and obtain the final new generation resource pre-allocation strategy matrix R' as the new parent generation. For the new parent generation, based on the user demand predicted by the user demand prediction neural network, calculate the fitness of each resource allocation strategy vector in R'; In the fifth step, the fourth step is repeated until the fitness function of the adaptive evolutionary algorithm converges and the optimal resource allocation strategy is obtained.

Images