Frequency spectrum access method between high and low orbit satellites based on frequency spectrum prediction

Abstract

The invention discloses a high-orbit and low-orbit inter-satellite spectrum access method based on spectrum prediction, and belongs to the field of inter-satellite spectrum access. In order to solve the technical problems that an existing access method cannot adjust access conditions in real time to carry out beam switching, so that a channel which is being used by a high-orbit satellite is accessed, a GSO satellite is interfered, the minimum switching frequency is not considered, and the stability of continuous communication cannot be ensured. The method is technically characterized in that firstly, the occupation condition of a channel is predicted to serve as data input for access, and meanwhile, an available channel for LEO constellation access to a GSO satellite beam is dynamically adjusted through a deep reinforcement learning method. In order to solve the problem of complete data transmission of single satellite data in a scene, an optimization target is set to minimize satellite switching times, and a DQN algorithm is utilized to realize an overall long-term optimal access effect.

Description

Technical Field

[0001] The present invention belongs to the field of inter-satellite spectrum access, and particularly relates to a method for high-low orbit inter-satellite spectrum access based on spectrum prediction in this field. Background Art

[0002] The connection time between a high-speed moving low-earth orbit (LEO) satellite and ground devices such as ground terminals and gateway stations is short, and a single satellite cannot effectively transmit complete continuous data. On the contrary, since a geostationary earth orbit (GSO) satellite is stationary relative to the ground, it can be used as a relay satellite for LEO satellites for data forwarding. LEO satellites can access the idle frequency bands of GSO satellites to achieve efficient data transmission.

[0003] Currently, the research on high-low orbit satellite spectrum access methods is relatively scarce, and most methods are static, and the beam switching strategy must be planned in advance. For example, the existing patent document with the document number CN117856876B discloses a high-low orbit inter-satellite distributed cooperative communication system and method. The system includes: a GSO satellite; a pair of low-orbit communication units installed on the GSO satellite for sending synchronization signals, cooperative relay timing tables, and phase synchronization completion identification information to a pair of high-orbit communication units, and receiving data transmission requests, test signals, and data to be transmitted from the pair of high-orbit communication units; a cooperative relay calculation unit installed on the GSO satellite for calculating the cooperative relay timing table; multiple LEO satellites; a pair of high-orbit communication units installed on the LEO satellites for sending data transmission requests, performing signal synchronization tests, sending test signals, and sending data to be transmitted; an inter-satellite interconnection unit installed on the LEO satellites for data transmission between LEO satellites. This existing technology can solve the problems of weak transmission capacity of a single LEO satellite and discontinuous transmission caused by the relative motion between high-low orbit satellites, and improve the capacity and efficiency of satellite communication.

[0004] However, with the dynamic change of communication requirements, it is necessary to adjust the access conditions in real time for beam switching. The existing methods mainly have the following problems:

[0005] (1) The existing access methods usually regard the channel utilization situation of satellite beams as known data, and cannot cope with the frequency band interference or emergencies that may be encountered by LEO satellites during actual access. The future channel frequency usage situation is likely to change, resulting in accessing the channels being used by GSO satellites and causing interference to GSO satellites;

[0006] (2) The current main optimization goal of access is the overall longest communication time, without considering the minimum number of handovers, so the stability of continuous communication cannot be guaranteed. Summary of the Invention

[0007] The technical problem to be solved by the present invention is:

[0008] The object of the present invention is to provide a high-low orbit inter-satellite spectrum access method based on spectrum prediction, so as to solve the technical problems existing in the existing access methods, such as the inability to adjust the access conditions in real time for beam switching, resulting in interference to the GSO satellite by accessing the channels currently used by the high-orbit satellite, and the failure to consider the minimum number of handovers, thus being unable to ensure the stability of continuous communication.

[0009] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0010] A high-low orbit inter-satellite spectrum access method based on spectrum prediction, the improvement lies in that it includes the following steps:

[0011] Step 1, perform normalization processing on the historical spectrum data of the high-orbit GSO satellite, and divide the spectrum data into a training set and a test set:

[0012] During the historical time of t0 ∈ [1, T0], for the historical spectrum data of the GSO satellite in the past L channels perform normalization processing, and the data range is scaled to the interval [0, 1]. The normalization formula is as follows:

[0013]

[0014] In the above formula, x l ′ represents the normalized spectrum result of the l-th channel, l ∈ [1, L], max(x l ) and min(x l ) represent the maximum and minimum values of the l-th channel;

[0015] Divide the normalized spectrum data into a training set and a test set according to the ratio of 8:2;

[0016] Step 2: Construct a multi-channel spectrum prediction model based on CNN-LSTM, set the initial parameter values of the model, use the training set to optimize the hyperparameters of the model to obtain the optimal model; apply the test set to the trained optimal model, perform anti-normalization processing on the predicted values to make them in the same order of magnitude as the true values, and output the prediction results, prediction error curves, and fitting curves of the true values and predicted values to evaluate the prediction quality of the optimal model;

[0017] Step 3: Obtain other factor variables during the access process:

[0018] Including the remaining visible time and coverage of the low-orbit LEO satellite in the GSO satellite beam;

[0019] Step 4: Establish an optimization model:

[0020] Define the total number of LEO satellites as \(K = \{1, 2, \ldots, K\}\), and the set of available GSO satellite beams in the system as \(N=\{1, 2, \ldots, N\}\). Denote the coverage of GSO satellite beam \(n\) for LEO satellite \(k\) at time \(t\) as:

[0021]

[0022] The set of GSO satellite beams covered within the time interval \([t\) u , \(t\) u+1 ) is:

[0023]

[0024] The service situation of LEO satellite \(k\) in beam \(n\):

[0025]

[0026] Divide the total bandwidth of each beam into \(L\) equal-bandwidth channels. The channel usage situation is obtained from step 2. During the transmission of each LEO satellite, each user can only use one channel, and it is required that the idle channel situation of the entire beam does not exceed the number of channels \(L\). The overall channel budget constraint is:

[0027]

[0028] Regard the LEO satellite handover problem as an optimization problem of selecting a GSO satellite beam to serve LEO satellite \(k\) from in each coverage area. The average number of LEO satellite handovers is as follows:

[0029]

[0030] HO k Denote the number of satellite beam handovers of LEO satellite \(k\). If the service relationship between GSO satellite beam \(n\) and LEO satellite \(k\) changes as the coverage area changes from \([t\) u , \(t\) u+1 ) to \([t\) u+1 , \(t\) u+2 ), then HO k increases by 1;

[0031] Establish the following optimization model:

[0032]

[0033] In the above formula, denotes two states of user association between LEO satellite \(k\) and GSO satellite \(n\);

[0034] Step 5: Design an optimized access algorithm. Use the remaining visible time of LEO obtained in Step 1 and the spectrum prediction results in Step 3 as the input data of the DQN model, perform empirical learning, and optimize and update the network parameters.

[0035] Step 6: Use the predicted future spectrum usage results in Step 2 and other factor variables in Step 3 as the input of the DQN model, and the DQN model outputs the satellite frequency band to be accessed.

[0036] Furthermore, Step 2 specifically includes the following steps:

[0037] Step 21: Initialize the CNN-LSTM multi-channel spectrum prediction model:

[0038] First, organize the historical spectrum data into a form that can be utilized by the model and import it into the model.

[0039] Secondly, the CNN in the CNN-LSTM model analyzes the correlation features between the utilization of frequency bands and other frequency bands. Subsequently, the LSTM analyzes the temporal features of the spectrum. Finally, the FC layer maps the high-dimensional feature vectors output by the LSTM layer to a lower-dimensional representation space, retains the effective information, and a Dropout layer is added after the CNN network layer.

[0040] Finally, output the prediction results of the model.

[0041] Step 22: Use the training set to train the model:

[0042] Input the training set in Step 1 into the CNN-LSTM network to predict the spectrum utilization of this beam in the future time period, and continuously update the hyperparameters of the network through backpropagation.

[0043] Step 23: Use the test set to verify the model and perform data denormalization for verification and comparison:

[0044] Use the test set in Step 1 as the input of the model, and the output results of the model are as follows:

[0045]

[0046] T1 represents the predicted future total time, and X 1 represents the historical spectrum data of L channels in the future T1 time slots, represents the spectrum prediction data of the l-th channel of the GSO beam channel received by the spectrum sensing satellite at the future t1 moment.

[0047] Furthermore, Step 5 specifically includes:

[0048] Convert the optimized access problem into a multi-agent reinforcement learning optimization problem based on stochastic games. The key definitions of multi-agent reinforcement learning are as follows:

[0049] Agent: The LEO satellite \(k\in K\) takes actions at each step, causing a transition in the coverage state;

[0050] State: Denotes the state of the \(k\)-th LEO satellite agent at time \(t\), consisting of the covering satellite Satellite available channels and the remaining visible time of the satellite ;

[0051] Action: Denotes the action of the \(k\)-th LEO satellite agent, indicating whether LEO satellite \(k\) is served by GSO satellite beam \(n\) at time \(t\);

[0052] Reward: Denotes the reward of the \(k\)-th LEO satellite agent, which is used to describe the immediate reward after executing the action in the state. Assuming that the agent does not know the reward functions of other agents, but they can obtain the actions of other agents, the reward function is defined as follows: When the LEO satellite agent \(k\) selects the GSO satellite beam of the covering satellite, but the satellite beam does not serve the LEO satellite, an immediate handover occurs.

[0053]

[0054] When the LEO satellite agent \(k\) selects the GSO satellite beam of the covering satellite, but the beam channel of the GSO satellite is overloaded.

[0055] When the LEO satellite agent \(k\) selects the GSO satellite beam that covers the LEO beam and serves the LEO satellite, and the satellite channel is sufficient for the LEO satellite accessing it.

[0056]

[0057] When the LEO satellite agent \(k\) selects a GSO satellite beam that covers the LEO beam and serves the LEO satellite, and the satellite channel is sufficient for the LEO satellite accessing it.

[0058] Furthermore, a deep neural network is constructed using convolutional layers and fully connected layers. The deep neural network consists of six 2D convolutional layers, a flattening layer, and a fully connected layer. The six convolutional layers are used to extract features from the state information. The size of the convolutional kernel for each convolutional layer is set to 3x3, and the activation function is the LeakyReLU function. The output quantities of the filters in the six convolutional layers are sequentially set to 8, 16, 32, 64, 16, and 8. The flattening layer flattens the feature matrix learned by the convolutional layers into a vector and inputs it into the fully connected layer to map out the Q-values of each behavior. The number of neurons in the fully connected layer is the same as the number of behaviors in the behavior space.

[0059] The beneficial effects of the present invention are as follows:

[0060] Through a method for high and low orbit inter-satellite spectrum access based on spectrum prediction, the present invention first uses the predicted occupancy of the channel as data input for access, and at the same time uses the deep reinforcement learning method to dynamically adjust the available channels for LEO constellations to access GSO satellite beams. For the problem of complete transmission of data for a single satellite in the scenario, the optimization objective is set to minimize the number of satellite handovers, and the DQN algorithm is used to achieve the overall long-term optimal access effect. The present invention is suitable for the dynamic changes in communication requirements, can adjust the access conditions in real time for beam switching, cope with the frequency band interference or emergencies that may be encountered during the actual access of LEO satellites, and can avoid interfering with GSO satellites by accessing the channels currently used by high orbit satellites when the frequency usage of the channel changes. The present invention can also ensure the stability of continuous communication.

[0061] The method disclosed by the present invention can obtain the abnormal frequency usage results of the spectrum by predicting the future time of the known spectrum and comparing it with the known results. The spectrum prediction used can predict the future channel utilization of the satellite beam to cope with the frequency band interference or emergencies that may be encountered during the access of LEO satellites, avoid accessing the channels currently used by high orbit satellites, and prevent interference with the communication quality.

[0062] The method of the present invention uses the abnormal frequency usage results of the spectrum and the remaining visible time as constraint conditions. For the problem of complete transmission of data for a single satellite in the scenario, the optimization objective is set to minimize the number of satellite handovers to ensure the stability of the continuous communication quality and achieve the overall long-term optimal dynamic access effect. Description of the Drawings

[0063] Figure 1 is the flowchart of the method of the present invention;

[0064] Figure 2 is the multi-channel spectrum prediction model diagram based on CNN-LSTM;

[0065] Figure 3It is a schematic diagram of optimizing access by deep reinforcement learning;

[0066] Figure 4 It is a DQN network diagram. Specific implementation manners

[0067] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the following further details the present invention in conjunction with the appended Figures 1-4 drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0068] Embodiment 1. This embodiment discloses a method for high and low orbit inter-satellite spectrum access based on spectrum prediction. First, the occupancy situation of the predicted channel is used as data input, and at the same time, the deep reinforcement learning method is used to dynamically adjust the available channels for the LEO constellation to access the GSO satellite beam. Aiming at the problem of complete transmission data of single satellite data in the scenario, the optimization objective is set to minimize the number of satellite handovers, and the DQN algorithm is used to achieve the overall long-term optimal access effect.

[0069] This method proposes to use the CNN-LSTM network to perform anomaly prediction on the planned spectrum to prevent spectrum anomaly access problems in future accesses, improve the anti-interference ability, and the designed spectrum prediction network CNN-LSTM model can achieve high-accuracy prediction of spectrum data and detect spectrum occupancy anomalies. The spectrum prediction result is used as the condition for deep reinforcement learning access, and the optimization objective of deep reinforcement learning is set to minimize the number of satellite handovers. The DQN algorithm is used for deep reinforcement learning to ensure the continuity of communication services.

[0070] This method adopts the deep learning method. In the scenario where the low-orbit satellite constellation and the high-orbit satellite coexist, the low-orbit satellite uses the high-orbit satellite as a relay to perform dynamic communication with the ground gateway station. The LEO satellite calculates the channel occupancy situation of the future GSO satellite beam through spectrum prediction, and obtains its remaining visible time under the GSO beam by calculating its own ephemeris. The channel occupancy situation and the remaining visible time are used as constraint conditions, and the deep reinforcement learning method is used to iteratively optimize repeatedly to give the optimal access strategy for the LEO satellite to access the GSO satellite beam, so as to achieve the optimization objective of minimizing the number of handovers between satellites during the entire access process and ensure the continuity of communication services.

[0071] As Figure 1 shown, it specifically includes the following steps:

[0072] Step 1, normalize the historical spectrum data of the high-orbit GSO satellite, and divide the spectrum data into a training set and a test set:

[0073] The normalized data can accelerate the model convergence speed and reduce the training time. Specifically, within the historical time of \(t_0\in[1,T_0]\), the historical spectrum data of GSO satellites for the past \(L\) channels is normalized, and the data range is scaled to the interval \([0,1]\). The normalization formula is as follows:

[0074]

[0075] In the above formula, \(x\) l ' represents the spectrum result after normalization for the \(l\)-th channel, \(l\in[1,L]\), \(\max(x\) l ) and \(\min(x\) l ) represent the maximum and minimum values of the \(l\)-th channel;

[0076] The normalized spectrum data is divided into a training set and a test set in a ratio of 8:2;

[0077] Step 2: Construct a multi-channel spectrum prediction model based on CNN-LSTM, set the initial parameter values of the model, use the training set to optimize the hyperparameters of the model to obtain the optimal model; apply the test set to the trained optimal model, perform inverse normalization on the predicted values to make them in the same order of magnitude as the true values, and output the prediction results, prediction error curves, and fitting curves of the true values and predicted values to evaluate the prediction quality of the optimal model;

[0078] Step 21, initialize the CNN-LSTM multi-channel spectrum prediction model:

[0079] As Figure 2 shown, the basic structure of the CNN-LSTM multi-channel spectrum prediction model includes an input, a CNN-LSTM model, and an output;

[0080] First, organize the historical spectrum data into a form that can be utilized by the model and import it into the model;

[0081] Secondly, the CNN in the CNN-LSTM model analyzes the correlation features of the frequency band utilization situation with other frequency bands, and then the LSTM analyzes the temporal features of the spectrum. The final FC layer maps the high-dimensional feature vectors output by the LSTM layer to a lower-dimensional representation space, retains the effective information, and at the same time adds a Dropout layer after the CNN network layer to enhance the robustness and prevent the model from overfitting;

[0082] Finally, output the prediction results of the model;

[0083] Step 22, use the training set to train the model:

[0084] Input the training set of step 1 into the CNN-LSTM network to predict the spectrum utilization of the beam in the future time period, and continuously update the hyperparameters of the network through backpropagation;

[0085] Step 23, use the test set for model verification and perform data denormalization for verification and comparison:

[0086] Use the test set in step 1 as the input of the model, and the output results of the model are as follows:

[0087]

[0088] T1 represents the total predicted future time, and X 1 represents the historical spectrum data of L channels in T1 time slots in the future, represents the spectrum prediction data of the l-th channel of the GSO beam channel received by the spectrum sensing satellite at time t1 in the future.

[0089] Step 3: Obtain other factor variables during the access process:

[0090] including the remaining visible time and coverage of LEO satellites in the GSO satellite beam;

[0091] Use the LEO on-board computing module to calculate the remaining visible time of LEO for the GSO satellite beam according to its own TLE data and the GSO satellite beam position. Since the GSO satellite is relatively stationary with respect to the earth and its beam position is fixed, the visible time range and coverage of LEO in each GSO beam can be calculated using the position relationship between LEO and GSO.

[0092] Step 4: Establish an optimization model to achieve the optimal result of minimizing the long-term interruption times, aiming to minimize the average handover times while improving the channel utilization efficiency of the LEO satellite system:

[0093] Define the total number of LEO satellites as K = {1, 2,..., K}, and the set of available GSO satellite beams in the system as N = {1, 2,..., N}, represents the coverage of GSO satellite beam n for LEO satellite k at time t:

[0094]

[0095] The set of GSO satellite beams covered within the time [t u , t u+1 ) is:

[0096]

[0097] The service situation of LEO satellite k in beam n:

[0098]

[0099] Divide the total bandwidth of each beam into L channels with equal bandwidth. The channel frequency usage situation is obtained from step 2. And during the transmission process of each LEO satellite, each user can only use one channel, and it is required that the number of idle channels used in the whole beam cannot be greater than the number of channels L. The overall channel budget constraint is as follows:

[0100]

[0101] Regard the LEO satellite handover problem as an optimization problem of selecting a GSO satellite beam to serve LEO satellite k in each coverage area. The average number of LEO satellite handovers is as follows: from

[0102]

[0103] HO k represents the number of satellite beam handovers of LEO satellite k. If the service relationship between GSO satellite beam n and LEO satellite k changes as the coverage area changes from [t u , t u+1 ) to [t u+1 , t u+2 ), then HO k increases by 1;

[0104] Establish the following optimization model to achieve the optimal result of the long-term interruption times, aiming to optimize the service association index Minimize the average number of handovers within a time period T, and at the same time improve the channel utilization efficiency of the LEO satellite system. The whole optimization problem is described as follows:

[0105]

[0106] In the above formula, represents two states of user association between LEO satellite k and GSO satellite n; L is the constraint of the total channels of the satellite beam.

[0107] Step 5: Design an optimized access algorithm. Use the remaining visible time of the LEO satellite obtained in step 1 and the spectrum prediction result in step 3 as the input data of the DQN model, conduct empirical learning, and optimize and update the network parameters;

[0108] Transform the optimized access problem into an optimization problem of multi-agent reinforcement learning based on stochastic game. The LEO satellite handover optimization problem is essentially a generalized sum K-agent game problem because the agents in the system have both cooperative and competitive relationships. As Figure 3As shown below, the key definitions of Multi-Agent Reinforcement Learning (MARL) are as follows:

[0109] Agent: The LEO satellite \(k\in K\) takes actions at each step and causes a transition in the coverage state;

[0110] State: Denotes the state of the \(k\)-th LEO satellite agent at time \(t\), which consists of the covering satellites Available satellite channels And the remaining visible time of the satellite Composed of;

[0111] Action: Denotes the action of the \(k\)-th LEO satellite agent, Indicates whether LEO satellite \(k\) is served by the GSO satellite beam \(n\) at time \(t\);

[0112] Reward: Denotes the reward of the \(k\)-th LEO satellite agent, which is used to describe the instantaneous reward after taking the action in the state Execute Action, Is the joint action of all agents at time \(t\). Assume that an agent does not know the reward functions of other agents, but they can obtain the actions of other agents. The reward function is defined as follows:

[0113]

[0114] When the LEO satellite agent \(k\) selects the GSO satellite beam of the covering satellite, but the satellite beam does not serve the LEO satellite, an immediate handover occurs,

[0115] When the LEO satellite agent \(k\) selects the GSO satellite beam of the covering satellite, but the beam channel of the GSO satellite is overloaded,

[0116] When the LEO satellite agent \(k\) selects a GSO satellite beam that covers the LEO beam and serves the LEO satellite, and the satellite channel is sufficient for the LEO satellite accessing it,

[0117] Represents the visible time. Only when there is no immediate handover and the satellite's channel budget is sufficient, the reward value is a positive integer. Otherwise, it is a negative integer, and the positive integer of the reward is equal to the remaining visible time because the longer the remaining visible time, the smaller the possibility of future handover.

[0118] To better learn the potential relationship function between state information and each behavior and reduce the computational complexity, a deep neural network is constructed using convolutional layers and fully connected layers. As Figure 4 shown, the deep neural network consists of six 2D convolutional layers, a flattening layer, and a fully connected layer. The six convolutional layers are used to extract features from the state information. The convolutional kernel size of each convolutional layer is set to 3x3, and the activation function is the LeakyReLU function. The output quantities of the filters in the six convolutional layers are sequentially set to 8, 16, 32, 64, 16, and 8. The flattening layer flattens the feature matrix learned by the convolutional layers into a vector and inputs it into the fully connected layer to map the Q-values of each behavior. The number of neurons in the fully connected layer is the same as the number of behaviors in the behavior space.

[0119] Step 6: Use the predicted future spectrum usage result in Step 2 and other factor variables in Step 3 as the input to the DQN model, and the DQN model outputs the satellite frequency band to be accessed.

[0120] It has been verified that the method proposed in the present invention solves the technical problems proposed in the present invention. The method described in the present invention has been practically applied to verify the claimed technical effects and practicality of the present invention.

[0121] The method described in the present invention has been verified by simulation experiments and practical applications, and the claimed technical effects of the present invention have been verified.

[0122] The algorithm (method) proposed in the present invention is the underlying technical core of the present invention. Based on the algorithm (method) proposed in the present invention, a high-low orbit inter-satellite spectrum access system based on spectrum prediction is developed using a programming language. The system has program modules corresponding to the steps of the above technical solution and executes the steps in the above high-low orbit inter-satellite spectrum access method based on spectrum prediction when running.

[0123] The computer program of the developed system (software) is stored on a computer-readable storage medium. The computer program is configured to implement the steps of the above high-low orbit inter-satellite spectrum access method when called by a processor. That is, the present invention is materialized on a carrier to become a computer program product.

[0124] A high-low orbit inter-satellite spectrum access device, the device includes at least one processor, and a memory communicatively connected to the at least one processor. Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above high-low orbit inter-satellite spectrum access method.

[0125] The various embodiments of the systems and techniques described herein can be implemented in digital electronic circuitry, integrated circuit systems, dedicated ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general purpose programmable processor that receives data and instructions from a storage system, at least one input device, and at least one output device, and transmits the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0126] The computational programs (also referred to as programs, software, software applications, or code) in the present invention include machine instructions for a programmable processor and can implement these computational programs using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used in the present invention, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., magnetic disks, optical disks, memories, programmable logic device PLD) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0127] It should be understood that various forms of the flow shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in this application can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this application can be achieved, all within the scope of protection of the present invention.

Claims

1. A spectrum access method for high-orbit and low-orbit satellites based on spectrum prediction, characterized in that: The method comprises the following steps: Step 1: Normalize the historical spectrum data of high-orbit GSO satellites and divide the spectrum data into training set and test set: In the historical time t0∈[1,T0], the historical spectrum data of GSO satellites for the past L channels After normalization, the data range is scaled to the interval [0, 1]. The normalization formula is as follows: In the above formula, x l ′ represents the normalized spectrum result of the lth channel, l∈[1,L], max(x l ) and min(x l ) represents the maximum and minimum values ​​of the lth channel; The normalized spectrum data is divided into training set and test set in a ratio of 8:2; Step 2: Build a multi-channel spectrum prediction model based on CNN-LSTM, set the initial parameter values ​​of the model, use the training set to optimize the model's hyperparameters, and obtain the optimal model; apply the test set to the trained optimal model, denormalize the predicted value to make it at the same level as the true value, and output the prediction result, prediction error curve, and fitting curve of the true value and the predicted value to evaluate the prediction quality of the optimal model; Step 3: Obtain other factor variables during the access process: Including the remaining visibility time and coverage of low-orbit LEO satellites in the GSO satellite beam; Step 4: Build the optimization model: The total number of LEO satellites is defined as K = {1, 2, ..., K}, and the set of GSO satellite beams available in the system is N = {1, 2, ..., N}. Indicates the coverage of GSO satellite beam n for LEO satellite k at time t: At time [t u ,t u+1 The set of GSO satellite beams covered in ) is: LEO satellite k is served in beam n: The total bandwidth of each beam is divided into L channels of equal bandwidth. The channel usage is obtained from step 2. During the transmission process of each LEO satellite, each user can only use one channel, and the idle channel situation of the entire beam cannot be greater than the number of channels L. The overall channel budget constraint is: The LEO satellite switching problem is regarded as switching from one satellite to another within each coverage area. In the optimization problem of selecting a GSO satellite beam to serve LEO satellite k, the average number of LEO satellite switching times is as follows: HO k represents the number of satellite beam switching times of LEO satellite k. If the service relationship between GSO satellite beam n and LEO satellite k increases as the coverage area changes from [t u ,t u+1 ) changes to [t u+1 ,t u+2 ),So HO k Increase by 1; The following optimization model is established: In the above formula, Indicates the two states of user association between LEO satellite k and GSO satellite n; Step 5: Design an optimized access algorithm, use the LEO remaining visible time obtained in step 1 and the spectrum prediction result in step 3 as the input data of the DQN model, conduct experience learning, and optimize and update network parameters; In step 6, the future spectrum usage results predicted in step 2 and other factor variables in step 3 are used as inputs to the DQN model, and the DQN model outputs the satellite frequency band to be accessed.

2. The high-orbit and low-orbit inter-satellite spectrum access method based on spectrum prediction according to claim 1, characterized in that: Step 2 specifically includes the following steps: Step 21, initialize the CNN-LSTM multi-channel spectrum prediction model: First, organize the historical spectrum data into a form that can be used by the model and import it into the model; Secondly, the CNN in the CNN-LSTM model analyzes the frequency band utilization and the associated features of other frequency bands, and then the LSTM analyzes the temporal features of the spectrum. The final FC layer maps the high-dimensional feature vector output by the LSTM layer to a lower-dimensional representation space to retain valid information. At the same time, a Dropout layer is added after the CNN network layer. Finally, output the model’s prediction results; Step 22: Use the training set to train the model: Input the training set in step 1 into the CNN-LSTM network to predict the spectrum utilization of the beam in the future time period, and continuously update the network's hyperparameters through back propagation; Step 23: Use the test set to verify the model and perform data denormalization for verification and comparison: Taking the test set in step 1 as the input of the model, the output of the model is as follows: T1 represents the predicted total future time, X1 represents the historical spectrum data of L channels in the future T1 time slots, It represents the spectrum prediction data of the lth channel of the GSO beam channel received by the spectrum sensing satellite at the future time t1.

3. The high-orbit and low-orbit inter-satellite spectrum access method based on spectrum prediction according to claim 1, characterized in that: Step 5 specifically includes: The optimization access problem is transformed into a multi-agent reinforcement learning optimization problem based on random games. The key definition of multi-agent reinforcement learning is as follows: Agent: LEO satellite k∈K, takes actions at each step and causes the transition of coverage state; State: represents the state of the kth LEO satellite agent at time t, which is represented by the coverage satellite Satellite available channels and the remaining visible time of the satellite composition; Action: represents the action of the kth LEO satellite agent, Indicates whether LEO satellite k is served by GSO satellite beam n at time t; Reward: represents the reward of the k-th LEO satellite agent, which is used to describe the state implement The instantaneous reward after the action, assuming that the agent does not know the reward function of other agents, but they can obtain the actions of other agents, the reward function is defined as follows: When the LEO satellite agent k selects the GSO satellite beam that covers the satellite, but the satellite beam does not serve the LEO satellite, an instant switch occurs. When the LEO satellite agent k selects the GSO satellite beam that covers the satellite, but the beam channel of the GSO satellite is overloaded, When the LEO satellite agent k selects a GSO satellite beam that covers the LEO beam and serves the LEO satellite, and the satellite channel is sufficient for the LEO satellite it accesses, 4. The high-orbit and low-orbit inter-satellite spectrum access method based on spectrum prediction according to claim 3, characterized in that: A deep neural network is constructed using convolutional layers and fully connected layers. The deep neural network consists of 6 2D convolutional layers, an expansion layer and a fully connected layer. The 6 convolutional layers are used to extract features from state information. The convolution kernel size of each convolutional layer is set to 3x3, and the activation function is the LeakyReLU function. The filter outputs in the 6 convolutional layers are set to 8, 16, 32, 64, 16 and 8 respectively. The expansion layer expands the feature matrix learned by the convolutional layer into a vector, which is input into the fully connected layer to map the value Q value of each behavior. The number of neurons in the fully connected layer is consistent with the number of behaviors in the behavior space.

5. A high-orbit and low-orbit inter-satellite spectrum access system based on spectrum prediction, characterized in that: The system has a program module corresponding to the steps of any one of claims 1 to 4 above, and executes the steps in the high- and low-orbit inter-satellite spectrum access method based on spectrum prediction when running.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to implement the steps of a high-orbit and low-orbit inter-satellite spectrum access method based on spectrum prediction according to any one of claims 1 to 4 when called by a processor.

Images