A low earth orbit satellite inter-satellite handover method based on movement prediction

Abstract

The present application relates to the technical field of low-orbit satellite communication, and particularly relates to a low-orbit satellite inter-satellite switching method based on movement prediction, which comprises the following steps: a satellite base station collects position information, service demand and satellite ephemeris information of a ground terminal, and identifies the type of the ground terminal according to the service demand; the position information is input into a pre-trained GWO-GRU neural network model to generate a predicted position of the ground terminal after a preset time interval; a predicted coverage persistence score is obtained according to the predicted position; four indexes, including reference signal received power, satellite load and switching overhead, of each candidate satellite in a candidate satellite set are calculated by modeling; the normalized weight of each index is dynamically allocated according to the type of the ground terminal, and the comprehensive score of each candidate satellite in the candidate satellite set is calculated by weighting; a candidate satellite with the highest score and meeting the access constraint is selected as a switching target to perform inter-satellite switching; and the present application effectively improves the communication stability and overall service performance of the low-orbit satellite network.

Description

Technical Field

[0001] This invention relates to the field of low-Earth orbit satellite communication technology, and specifically to a low-Earth orbit satellite inter-satellite handover method based on motion prediction. Background Technology

[0002] With the development of communication technology, non-terrestrial networks (NTN) have become an important direction for overcoming the coverage limitations of terrestrial networks. Among them, low Earth orbit (LEO) satellites have become a core component of NTN due to their advantages such as low propagation delay, superior signal strength, and economical deployment costs. However, the high-speed and periodic motion characteristics of LEO satellites cause their ground coverage to change dynamically, requiring frequent switching of communication links between satellites and user equipment (UEs) to ensure continuity. At the same time, large-scale LEO satellite constellations lead to overlapping service areas, and multiple users tend to concentrate on accessing the satellite with the best single attribute, causing problems such as overload and access congestion.

[0003] Current inter-satellite handover strategies for low-Earth orbit (LEO) satellites are mainly divided into two categories: single-attribute and multi-attribute. Single-attribute handover algorithms, such as those based on Received Reference Signal Power (RSRP) and remaining satellite service time, focus only on a single indicator and do not comprehensively consider the overall network performance, which can easily lead to problems such as ping-pong handover and load imbalance. Multi-attribute handover algorithms, while integrating indicators such as transmission link quality and satellite load to improve system throughput, still have the following shortcomings: First, they do not fully consider the trajectory dynamics of high-speed mobile terminals, simplifying terminals into low-speed or fixed nodes, resulting in insufficient adaptability in heterogeneous terminal scenarios. Second, they lack the design to use motion prediction to obtain the future location of terminals to assist handover decisions, making it impossible to accurately predict the continuity of satellite coverage for terminals. Third, they do not adapt handover weights to the personalized service needs of different terminals, making it difficult to balance throughput and communication continuity.

[0004] In summary, existing inter-satellite handover methods for low-Earth orbit satellites cannot meet the communication needs of high-speed mobile terminals, and suffer from problems such as numerous handovers, load imbalance, and poor communication stability. There is an urgent need for a multi-attribute inter-satellite handover method that combines mobility prediction and adapts to the personalized needs of terminals. Summary of the Invention

[0005] This invention addresses the shortcomings of existing low-Earth orbit (LEO) satellite inter-satellite handover methods by providing a LEO satellite inter-satellite handover method based on mobility prediction, achieving at least one of the following objectives: (1) accurately predicting the future location of high-speed mobile terminals, anticipating the coverage continuity of candidate satellites, reducing unnecessary inter-satellite handovers, and lowering handover overhead; (2) making handover decisions by comprehensively considering multi-dimensional network attributes, avoiding load imbalance problems caused by single-attribute algorithms; (3) dynamically adjusting handover weights according to the personalized service needs of different terminals, taking into account both throughput and communication continuity; (4) optimizing the handover frame structure, reducing satellite-to-ground signaling round-trip time, and improving handover efficiency.

[0006] To achieve the objectives of this invention, the following technical solution is adopted:

[0007] A method for inter-satellite handover of low-Earth orbit satellites based on motion prediction includes the following steps:

[0008] The satellite base station completes system initialization, collects the location information, service requirements and satellite ephemeris information of the ground terminals, and identifies the type of ground terminal based on the service requirements.

[0009] The location information is input into a pre-trained GWO-GRU neural network model, which generates the predicted location of the ground terminal after a preset time interval based on the historical trajectory of the ground terminal.

[0010] Construct a candidate satellite set, and assign a score to the remaining service time that the candidate satellites can provide to the ground terminal based on the predicted location to obtain the predicted coverage persistence score;

[0011] The model calculates four indicators for each candidate satellite in the candidate satellite set: reference signal received power, satellite load, and handover overhead.

[0012] The four indicators are normalized, and the normalization weights of each indicator are dynamically allocated according to the ground terminal type. The comprehensive score of each candidate satellite in the candidate satellite set is calculated by weighting.

[0013] The candidate satellites in the candidate satellite set are sorted from highest to lowest according to their comprehensive scores, and the candidate satellite with the highest score that meets the access constraints is selected as the handover target for inter-satellite handover.

[0014] A further improvement lies in the fact that the pre-training process of the GWO-GRU neural network model specifically includes:

[0015] Historical trajectory data of high-speed mobile terminals are collected, and after missing value completion and outlier filtering, the data is converted to the ECEF coordinate system. The processed trajectory feature values ​​are then standardized using Z-Score.

[0016] Using the historical ECEF coordinates of the high-speed mobile terminal within 2 minutes as input and the ECEF coordinates of the high-speed mobile terminal after 1 minute and 2 minutes as output, a training sample set is constructed, and the training sample set is divided into a training set and a test set according to the proportion.

[0017] Initialize the GRU neural network and set the hyperparameter search range of the GRU neural network;

[0018] Using 3D position prediction error as the fitness function, the hyperparameters of the GRU neural network are iteratively optimized using the Grey Wolf Algorithm (GWO).

[0019] The optimal hyperparameters obtained through iterative optimization are substituted into the GRU neural network to complete the training. After the accuracy is verified by the test set, the trained GWO-GRU neural network model is loaded into the satellite base station.

[0020] A further improvement lies in the following: the specific method for constructing the candidate satellite set includes:

[0021] Calculate the communication angle of each satellite relative to the ground terminal, and combine it with a preset minimum service elevation angle threshold. The formula for determining the validity of a satellite is as follows:

[0022]

[0023] in, For ground terminal identification, For satellite identification, For the current moment, for Time Satellite Compared to ground terminals The corner of faith for Candidate satellites at any time The valid determination value, This indicates that the satellite is an invalid candidate satellite. This indicates that the satellite is a valid candidate satellite; all valid candidate satellites are integrated into a candidate satellite set.

[0024] A further improvement lies in the following: The specific method for assigning a score to the remaining service time that candidate satellites can provide for ground terminals based on the predicted location, to obtain the predicted coverage persistence score, includes:

[0025] The satellite coverage status of the ground terminal at the predicted location is determined by the coverage status function, and then a step-by-step scoring is performed based on the coverage status result to obtain the predicted coverage persistence score.

[0026] The coverage state function is:

[0027] ;

[0028] The formula for the predicted coverage persistence score is as follows:

[0029] ;

[0030] in For ground terminal identification; For satellite identification; The current moment; For the current moment The moment 1 minute later; For the current moment The time 2 minutes later; It is a binary function of 0 and 1, and is the terminal. exist Satellite of Time Coverage status: 1 indicates coverage, 0 indicates no coverage; for Time Terminal Corresponding candidate satellites The predicted coverage persistence score indicates that the longer the satellite coverage lasts for the terminal and the more stable the link is.

[0031] A further improvement is that the calculation of the reference signal received power is achieved by constructing a total path loss model for the satellite-to-ground link, specifically including:

[0032] First, calculate the total path loss of the satellite-to-ground link, using the following formula:

[0033]

[0034] in, for Time Terminal With satellite Total path loss between; for Time Terminal With satellite Free space path loss between for Time Terminal With satellite The straight-line distance in space, For carrier frequency; for Time Terminal With satellite Atmospheric loss in communication links, for Time Satellite Compared to ground terminals The corner of faith; For Rice fading loss;

[0035] Then, the received power of the reference signal is calculated based on the total path loss, using the following formula:

[0036]

[0037] in, for Time Terminal Received satellite The reference signal received power; This refers to the satellite's launch power. for Time Satellite The transmit antenna gain; for Time ground terminal The receiving antenna gain.

[0038] A further improvement lies in the following: the quantitative modeling formula for the switching overhead is:

[0039]

[0040] in, for Time Terminal Switch to satellite Total switching overhead, Normalized data loss during a single handover refers to the effective data loss caused by the terminal pausing service transmission during the handover preparation phase. Normalized energy loss per handover refers to the additional energy consumed by the terminal when performing the handover operation.

[0041] A further improvement is that the satellite load is used to characterize the channel resource utilization of candidate satellites. The number of idle channels is used as an evaluation index for satellite load. The larger the value, the more idle channels the satellite has, the lower the load, and the more terminals can access the satellite. The formula is:

[0042]

[0043] in, for Candidate satellites at any time The number of idle channels; For satellite Total number of channels; for Time Satellite The number of occupied channels is collected in real time by satellite base stations.

[0044] A further improvement lies in: the dynamic allocation of normalized weights for each indicator satisfies the weight sum constraint condition. ,and ;in The normalized weights are used as reference signal received power. To predict the normalized weights for coverage persistence scores, For the normalized weights of the satellite payload, Normalized weights for switching overhead;

[0045] The ground terminal types are divided into throughput-sensitive users and continuity-sensitive users. For throughput-sensitive users, the weighting priority is as follows: For users with high continuity sensitivity, the weighting priority is as follows: .

[0046] A further improvement is that the weighted calculation of the comprehensive score of each candidate satellite within the candidate satellite set is performed using the following formula:

[0047]

[0048] in, for Time Terminal Corresponding candidate satellites The overall score ranges from [0,1]. A higher score indicates that the satellite is more suitable as a handover target for the terminal. , , , These are the Min-Max normalized values ​​of the reference signal received power, predicted coverage persistence score, satellite load, and handover cost, respectively. The handover cost is a negative indicator and is inverted after Min-Max normalization.

[0049] Further improvements include: the access constraints include: a single terminal can only access one satellite in a single time slot and can only occupy one channel of that satellite; the total number of terminals accessing a single satellite in a single time slot does not exceed its number of idle channels; if all satellite channels are occupied, the terminal's access request to that satellite is determined to be a failure and waits for the next communication time slot to make a new handover decision.

[0050] The inter-satellite handover is performed according to the optimized handover frame structure, which is divided into a signaling preparation phase and a service transmission phase, and the following transmission rules are established: the handover frame is placed at the beginning of each communication time slot, and the communication link status between the terminal and the satellite remains unchanged within a single time slot; if the signaling transmission fails during the signaling preparation phase, retransmission is immediately initiated until the signaling transmission is successful or the signaling preparation phase duration is exhausted; after the signaling transmission during the signaling preparation phase is successful, the remaining duration of the handover frame is converted into the service transmission phase; if it is determined that no handover is needed, the entire duration of the handover frame is used as the service transmission phase.

[0051] The beneficial effects of this invention are as follows:

[0052] The GWO-GRU neural network model constructed in this invention can accurately predict the position of a high-speed moving bottom terminal after a preset time interval. It can predict the coverage continuity of the terminal by the candidate satellite in advance and avoid frequent invalid handovers caused by sudden changes in satellite coverage. Compared with the single attribute handover algorithm based on maximum RSRP, the average number of handovers of the terminal is reduced by 36.3% when the channel capacity difference is less than 4%.

[0053] This invention makes handover decisions by comprehensively considering multiple dimensions of indicators, including reference signal received power, coverage persistence, satellite load, and handover overhead, to avoid load overload and access congestion caused by multiple users accessing a single satellite. Compared with a single-attribute handover algorithm based on the maximum idle channel, the system throughput is improved by 37.5% while maintaining a blocking rate of only 2.41%.

[0054] This invention categorizes terminals into throughput-sensitive and continuity-sensitive types based on their service requirements. It dynamically adjusts the weights of various decision indicators to meet the different needs of the two types of terminals, thereby achieving personalized switching decisions and solving the problem of insufficient adaptability of traditional algorithms to different terminal requirements.

[0055] On the one hand, this invention optimizes the handover frame structure by separating the signaling preparation phase from the service transmission phase and formulating targeted signaling transmission rules to reduce satellite-to-ground signaling round-trip time. On the other hand, it establishes a quantitative handover overhead model, incorporating data loss and energy consumption into handover decision indicators to reduce the actual cost of handover from the decision-making level. Attached Figure Description

[0056] Figure 1 This is a flowchart of a low-Earth orbit satellite inter-satellite handover method based on motion prediction according to the present invention;

[0057] Figure 2 This is a model framework diagram of the GRU neural network described in this invention. Detailed Implementation

[0058] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0059] Please refer to the attached document. Figure 1 Appendix Figure 2 This invention proposes a low-Earth orbit (LEO) satellite inter-satellite handover method based on mobility prediction. This method is applicable to non-terrestrial communication networks composed of LEO satellite constellations, high-speed ground mobile terminals (such as high-speed trains, drones, and mobile vehicle-mounted terminals), and satellite base stations, allowing selection of a ground region and a specific time period. Inside, there is a collection of satellites. And there are user sets It is assumed that the connection status between the satellite and the terminal remains unchanged in any time slot, and the network topology does not change.

[0060] The satellite base station is equipped with capabilities for receiving ephemeris information, collecting terminal information, performing GWO-GRU neural network model calculations, modeling multi-dimensional indicators, making handover decisions, and controlling handover, serving as the core control unit for inter-satellite handover. The ground terminal is capable of GNSS positioning, location information reporting, service requirement feedback, satellite-to-ground signaling interaction, and service data transmission, and supports stable communication even at high speeds. The low-Earth orbit satellite constellation is a geostationary constellation. In this embodiment, the basic parameters are set as follows: satellite orbital altitude 550km, carrier frequency... MHz, total number of channels per satellite 100, minimum service elevation angle threshold The total duration of the switching frame is 10ms, and the time step for collecting the terminal's historical trajectory data is 5s (i.e., 24 time steps of location data are included within 2 minutes).

[0061] The method of the present invention will now be described in detail:

[0062] like Figure 1 As shown, the low-Earth orbit satellite inter-satellite handover method based on motion prediction proposed in this embodiment includes the following steps:

[0063] Step S1: The satellite base station completes system initialization, collects the location information, service requirements and satellite ephemeris information of the ground terminals, and identifies the type of ground terminal according to the service requirements.

[0064] Step S2: Input the location information into the pre-trained GWO-GRU neural network model to generate the predicted location of the ground terminal after a preset time interval based on the historical trajectory of the ground terminal.

[0065] Step S3: Construct a candidate satellite set, and assign a score to the remaining service time that the candidate satellites can provide to the ground terminal based on the predicted location to obtain the predicted coverage persistence score.

[0066] Step S4: Model and calculate the four indicators of the reference signal received power (RSRP), satellite load and handover overhead for each candidate satellite in the candidate satellite set.

[0067] Step S5: Normalize the four indicators, dynamically allocate the normalization weight of each indicator according to the ground terminal type, and calculate the comprehensive score of each candidate satellite in the candidate satellite set by weighting.

[0068] Step S6: Sort the candidate satellites in the candidate satellite set from high to low according to the comprehensive score, and select the candidate satellite with the highest score that meets the access constraints as the handover target for inter-satellite handover.

[0069] Specifically, the core of the GWO-GRU neural network model training process is to use the Grey Wolf Algorithm (GWO) to iteratively optimize the hyperparameters of the Gated Recurrent Unit (GRU) neural network. The hyperparameters include the learning rate, the number of hidden units, and the Dropout ratio. The structure of the GRU model is as follows: input layer, first GRU layer, Dropout layer, second GRU layer, ReLU activation layer, fully connected layer, and output layer, thereby constructing a high-precision terminal location prediction model. All operations are completed on the server side, and the trained model is fixed and loaded into the satellite base station.

[0070] Specifically, the pre-training process of the GWO-GRU neural network model includes:

[0071] Historical trajectory data from ground-based high-speed mobile terminals is collected, including raw information such as latitude, longitude, altitude, speed, and timestamps. This data is then processed using linear interpolation to fill in missing values ​​and the 3σ principle to filter out outliers before being converted to the ECEF coordinate system (Geocentric-Earth-Fixed Rectangular Coordinate System, with the Earth's center of mass as the origin, the X-axis pointing to the intersection of the Prime Meridian and the equator, the Y-axis pointing to the equator at 90° East longitude, and the Z-axis pointing to the Earth's North Pole). The processed trajectory feature values ​​are then standardized using Z-Score, and the formula for Z-Score standardization is as follows:

[0072]

[0073] in, These are the original trajectory feature values ​​(such as the X, Y, and Z values ​​of the ECEF coordinates). This is the mean of the feature across all training samples. Let the standard deviation of this feature be the total number of training samples. The normalized eigenvalues ​​follow a normal distribution with a mean of 0 and a variance of 1.

[0074] The model's input window size is set to 24 time steps. The historical ECEF coordinates of the high-speed mobile terminal within 2 minutes are used as input, and the ECEF coordinates of the high-speed mobile terminal after 1 minute and 2 minutes are used as output. The preprocessed time-series trajectory data is used to construct an "input-output" training sample set, and the training sample set is divided into a training set and a test set in an 8:2 ratio.

[0075] Initialize the GRU neural network and set the hyperparameter search range of the GRU neural network. Specifically, this step is as follows: Construct a GRU neural network as follows: Figure 2 The GRU neural network model shown has a sequential hierarchical structure, and the specific parameter settings for each layer are as follows:

[0076] Input layer: 72 neurons, corresponding to an input dimension of 3×24.

[0077] First layer GRU: The initial number of neurons in the hidden layer is 128, the activation function is tanh, and the returned sequence is True.

[0078] Dropout layer: The initial random deactivation ratio is 0.2, used to prevent the model from overfitting.

[0079] The second GRU layer has an initial number of 64 hidden neurons, an activation function of tanh, and returns a false sequence.

[0080] ReLU activation layer: The ReLU activation function is used to solve the gradient vanishing problem and improve the model's nonlinear fitting ability.

[0081] Fully connected layer: 6 neurons, corresponding to the model output dimension. Output layer: uses a linear activation function to output the ECEF-normalized coordinates at terminals t+1min and t+2min, which are then denormalized to obtain the actual coordinate values.

[0082] At the same time, set the hyperparameter search range for the GRU model: learning rate Number of hidden units Dropout ratio The specific values ​​of the hyperparameters are determined by iterative optimization using the GWO algorithm.

[0083] Using the 3D position prediction error as the fitness function, the hyperparameters of the GRU neural network are iteratively optimized using the Grey Wolf Algorithm (GWO). The basic parameters of the GWO algorithm are set as follows: wolf population size 15, maximum number of iterations 30, search dimension 3 (corresponding to 3 hyperparameters), initial value of convergence factor 2, and convergence factor decay coefficient 0.05.

[0084] The formula for the 3D position prediction error is:

[0085]

[0086] in, These are the three-dimensional coordinates of the terminal in the ECEF coordinate system predicted by the model. These are the true three-dimensional coordinates of the terminal in the ECEF coordinate system. The value represents the 3D position prediction error of the GWO-GRU neural network model, expressed in meters.

[0087] The specific process of GWO hyperparameter optimization is as follows: Initialize the wolf pack positions, with each wolf pack individual corresponding to a set of GRU hyperparameter combinations; calculate the fitness value (i.e., 3D position prediction error) of each wolf pack individual; sort the wolf pack individuals according to their fitness values ​​and divide them into three levels: α, β, and δ (the top 3 individuals with the best fitness values), and the rest are ω wolves; update the positions of α, β, and δ wolves, guide the ω wolves to move towards the optimal position, and at the same time update the convergence factor to narrow the search range; determine whether the maximum number of iterations has been reached. If not, return to continue iterating; if it has been reached, terminate the optimization and output the optimal hyperparameter combination.

[0088] The optimal hyperparameters obtained through iterative optimization are substituted into the GRU neural network to complete the training. After the accuracy is verified by the test set, the trained GWO-GRU neural network model is loaded into the satellite base station as a motion detection model.

[0089] Understandably, this invention utilizes the Grey Wolf algorithm to iteratively optimize the hyperparameters of the GRU neural network, solving the problem of blindness in manually setting GRU hyperparameters, improving the accuracy of terminal location prediction, providing reliable prior information for coverage persistence prediction, and ensuring the accuracy and foresight of handover decisions.

[0090] Specifically, in step S1, the satellite base station completes system initialization and collects the location information, service requirements, and satellite ephemeris information of the ground terminal. The specific process is as follows:

[0091] The satellite base station loads all satellite IDs and ephemeris information of the current low-Earth orbit satellite constellation, and simultaneously loads a pre-trained GWO-GRU neural network model, initializing the minimum service elevation angle threshold. Parameters such as frame switching duration.

[0092] The ground terminal obtains the current ECEF three-dimensional coordinates through its own GNSS module, and at the same time reports its own service requirements (such as high-definition video transmission and voice communication) to the satellite base station.

[0093] Based on the service requirements of the terminals, the satellite base station divides the ground terminals into two categories: throughput-sensitive users (TSU, whose core requirements are high throughput and high signal quality) and continuity-sensitive users (CSU, whose core requirements are low handover frequency, high communication continuity, and low latency), and completes the terminal type labeling to provide a basis for subsequent dynamic weight allocation.

[0094] Specifically, based on the collected information, the satellite base station completes coordinate system one: the satellite base station receives the ephemeris information of neighboring satellites in the same constellation, extracts the ECEF three-dimensional coordinates of each satellite, and uniformly converts the position information reported by the ground terminal and the position information of all satellites to the ECEF coordinate system, eliminating the impact of coordinate system differences on distance and elevation angle calculations.

[0095] Satellite base stations associate and store the ID, location information, and type tags of ground terminals to form a terminal information table, providing unified data support for subsequent handover decisions.

[0096] In this implementation, the specific method for constructing the candidate satellite set in step S3 includes:

[0097] The satellite base station calculates the communication angle of each satellite relative to the ground terminal based on the ECEF three-dimensional coordinates of the ground terminal and the satellite. The calculation of the faith angle uses spatial geometric formulas in the ECEF coordinate system to ensure calculation accuracy; combined with a preset minimum service elevation angle threshold. The formula for determining the validity of a satellite is as follows:

[0098]

[0099] in, For ground terminal identification, For satellite identification, For the current moment, for Time Satellite Compared to ground terminals The corner of faith for Candidate satellites at any time The valid determination value, This indicates that the satellite is an invalid candidate satellite. This indicates that the satellite is a valid candidate satellite; all valid candidate satellites are integrated into a candidate satellite set; terminal exist The candidate satellite set at time is represented as ,in In this embodiment, the number of valid candidate satellites is... The value range is 3-8, which meets the actual scenario of multi-satellite coverage.

[0100] Specifically, in step S2, the method of inputting location information into a pre-trained GWO-GRU neural network model and generating the predicted location of the ground terminal after a preset time interval based on the historical trajectory of the ground terminal includes: the satellite base station inputs the historical ECEF coordinate data of the ground terminal within 2 minutes into the GWO-GRU neural network model, and the GWO-GRU neural network model outputs the predicted ECEF coordinates of the ground terminal after 1 minute and 2 minutes, which are used to provide a basis for calculating the coverage persistence score of the candidate satellite.

[0101] Understandably, the predicted coverage persistence score is used to characterize the candidate satellite's coverage capability for ground terminals over a future period of time. A higher score indicates that the satellite will cover the ground terminals for a longer period of time and that the link stability is higher. It is one of the core indicators for handover decision-making.

[0102] Specifically, in step S3, the method for assigning a score to the remaining service time that the candidate satellites can provide based on the predicted location to obtain the predicted coverage persistence score includes:

[0103] The satellite coverage status of the ground terminal at the predicted location is determined by the coverage status function, and then a step-by-step scoring is performed based on the coverage status result to obtain the predicted coverage persistence score.

[0104] The coverage state function is:

[0105] ;

[0106] The formula for the predicted coverage persistence score is as follows:

[0107] ;

[0108] in For ground terminal identification; For satellite identification; The current moment; For the current moment 1 minute later The time (min); For the current moment The last 2 minutes ( The time (min); It is a binary function of 0 and 1, and is the terminal. exist Satellite of Time Coverage status: 1 indicates coverage, 0 indicates no coverage; for Time Terminal Corresponding candidate satellites The predicted coverage persistence score indicates that the longer the satellite coverage lasts for the terminal and the more stable the link is.

[0109] The values ​​are 0.1, 0.5, and 1.0, and the physical meaning of the assigned score is as follows:

[0110] Ground terminal exist Min before detaching from the satellite The coverage is poor, the link is about to be interrupted, and the coverage continuity is the worst, so it is assigned the lowest score of 0.1.

[0111] Ground terminal exist to min was satellite Coverage, but min to min will be out of coverage, posing a risk of link interruption. Coverage continuity is moderate, so it is assigned a score of 0.5.

[0112] Ground terminal exist to continuously being monitored by satellites within min It has the highest coverage, link stability, and coverage continuity, and is assigned the highest score of 1.0.

[0113] Understandably, the Reference Signal Received Power (RSRP) is used to characterize the signal quality between the ground terminal and the candidate satellite. The higher the value, the better the signal quality, and it is a basic indicator for handover decisions.

[0114] Specifically, in step S4, the calculation of the reference signal received power (RSRP) is achieved by constructing a total path loss model for the satellite-to-ground link, which specifically includes:

[0115] First, calculate the total path loss of the satellite-to-ground link. The signal loss of the satellite-to-ground link consists of three parts: free space path loss, atmospheric loss, and Ricean fading loss. The total path loss is the sum of these three, and the calculation formula is as follows:

[0116]

[0117] in, for Time Terminal With satellite The total path loss between them, in decibels; for Time Terminal With satellite The free space path loss between them is the energy attenuation of electromagnetic waves propagating in ideal space; for Time Terminal With satellite The straight-line distance in space, For carrier frequency; for Time Terminal With satellite Atmospheric loss in communication links, for Time Satellite Compared to ground terminals Angle of faith (unit: °); This is due to Rice decay loss.

[0118] Specifically, free space path loss The calculation formula is:

[0119]

[0120] in, for Time Terminal With satellite The spatial straight-line distance is calculated using spatial geometric formulas in the ECEF coordinate system.

[0121] Specifically, atmospheric loss electromagnetic waves refer to the losses caused by absorption / scattering of oxygen and water vapor when passing through the Earth's atmosphere. They are negatively correlated with the angle of incidence, and the calculation formula is as follows:

[0122]

[0123] in, Zenith decay, The satellite-to-terminal communication angle is limited. To avoid the denominator being 0.

[0124] Specifically, Rice fading loss is a small-scale fading loss caused by the superposition of signals from the direct path and the ground / building reflection path, as well as the multipath effect of the satellite-to-ground link. In this embodiment, a fixed value of 2dB is taken to adapt to the satellite-to-ground communication scenario of low-orbit satellites.

[0125] After obtaining the total path loss, the reference signal received power is calculated based on the total path loss, using the following formula:

[0126]

[0127] in, for Time Terminal Received satellite Reference signal received power (unit: dBm); This refers to the satellite's launch power. for Time Satellite The transmit antenna gain; for Time ground terminal The receiving antenna gain.

[0128] Specifically, in this embodiment, the satellite load is used to characterize the channel resource utilization of candidate satellites. The number of idle channels is used as the evaluation index of satellite load. The larger the value, the more idle channels the satellite has, the lower the load, and the more terminals can access it. The formula is:

[0129]

[0130] in, for Candidate satellites at any time The number of idle channels; For satellite The total number of channels (100 in this embodiment); for Time Satellite The number of occupied channels is collected in real time by satellite base stations.

[0131] It is understood that this invention achieves load balancing of the satellite network by introducing satellite load indicators and setting access constraints, avoiding situations where some satellites are overloaded and others are idle, and effectively improving the overall network resource utilization rate of the low-Earth orbit satellite constellation.

[0132] It is understood that handover overhead is the comprehensive cost incurred by the terminal performing inter-satellite handover operations. This invention quantifies handover overhead as the sum of normalized data loss and normalized energy loss. Specifically, in this embodiment, the quantitative modeling formula for handover overhead is:

[0133]

[0134] in, for Time Terminal Switch to satellite Total switching overhead, Normalized data loss during a single handover refers to the effective data loss caused by the terminal pausing service transmission during the handover preparation phase. Normalized energy loss per handover refers to the additional energy consumed by the terminal when performing handover operations (such as searching for a new beam / satellite, sending a random access preamble sequence, receiving configuration information, etc.).

[0135] The present invention models the additional energy consumption as follows: during the handover process, the terminal power is the maximum transmit power corresponding to its class (such as CLASS_3), plus the additional power density required by the 3GPP specification; while in the data transmission phase, the average transmit power of the terminal is used, and the additional energy loss is defined as the power difference between these two states.

[0136] In this embodiment, the normalized weights of each indicator are dynamically allocated to satisfy the weight summation constraint: ,and ;in The normalized weights are used as reference signal received power. To predict the normalized weights for coverage persistence scores, For the normalized weights of the satellite payload, Normalized weights for switching overhead.

[0137] The weighting principle is that the importance of the indicator is positively correlated with the weight value. For throughput-sensitive users, the weight priority is as follows: For example, setting , , , For users with high continuity sensitivity, the weighting priority is as follows: For example, setting: , , , .

[0138] It is understood that the weight allocation of this invention is dynamically adjustable, and the specific value of each weight can be adjusted according to the actual satellite network status and terminal service requirements, as long as the total constraint is met.

[0139] Furthermore, since the four core decision indicators have different dimensions and value ranges, they are first subjected to Min-Max normalization to map all indicators to the [0,1] interval, ensuring that the contribution of each indicator to the overall score is of the same order of magnitude. The normalization formula is as follows:

[0140]

[0141] in, The normalized value of the indicator. The actual value of the indicator. This is the minimum value of the indicator within the candidate satellite set. This represents the maximum value of the indicator within the candidate satellite set.

[0142] Special handling: Due to switching overhead Since it is a negative indicator, it is inverted after normalization (i.e.) This converts it into a positive indicator (the larger the value, the lower the switching overhead), ensuring that the evaluation direction of all decision indicators is consistent.

[0143] After this step is completed, four normalized decision metrics are obtained: Normalized Reference Signal Received Power Normalized prediction coverage persistence score Normalized satellite payload Normalized switching overhead .

[0144] Specifically, in this embodiment, the weighted calculation of the comprehensive score of each candidate satellite in the candidate satellite set is performed using the following formula:

[0145]

[0146] in, for Time Terminal Corresponding candidate satellites The overall score ranges from [0,1]. A higher score indicates that the satellite is more suitable as a handover target for the terminal. , , , These are the Min-Max normalized values ​​of the reference signal received power, predicted coverage persistence score, satellite load, and handover cost, respectively. The handover cost is a negative indicator and is inverted after Min-Max normalization.

[0147] To ensure load balancing in the low-Earth orbit satellite network and avoid overloading of any single satellite, this invention establishes three core access constraints. All candidate satellites must meet these constraints to be included in the selection range for handover targets. Specifically, the access constraints include: a single terminal can only access one satellite in a single time slot and occupies only one channel of that satellite; the total number of terminals accessing a single satellite in a single time slot does not exceed its number of idle channels, i.e. ,in For satellite The number of terminals waiting to access the satellite; if all satellite channels are occupied, the terminal's access request to the satellite is deemed a failure, and it waits for the next communication time slot to make a new handover decision.

[0148] If the satellite with the highest overall score does not meet the access constraints (e.g., the number of idle channels is 0), then the satellite with the second highest score will be selected in turn until a candidate satellite that meets the constraints is found.

[0149] Specifically, in order to address the problem of interleaved signaling and service data transmission in traditional satellite-to-ground communication handover frames, this invention optimizes the handover frame structure. The inter-satellite handover is performed according to an optimized handover frame structure, which is divided into a signaling preparation phase and a service transmission phase. The total duration of the handover frame is 10ms, with the signaling preparation phase lasting 3ms and the service transmission phase lasting 7ms. The following four transmission rules are established to reduce satellite-to-ground signaling round-trip time and improve handover efficiency: 1. The handover frame is placed at the beginning of each communication time slot, and the communication link status between the terminal and the satellite remains unchanged within a single time slot; 2. If signaling transmission fails during the signaling preparation phase (e.g., satellite-to-ground link interference, signaling loss), retransmission is immediately initiated, with a maximum of 3 retransmissions, until signaling transmission is successful or the signaling preparation phase duration is exhausted; 3. After successful signaling transmission during the signaling preparation phase, regardless of whether a handover is determined, the remaining duration of the handover frame is converted to the service transmission phase for transmitting the terminal's service data; 4. If the signaling preparation phase determines that no handover is needed, the entire duration of the handover frame is used as the service transmission phase, improving the efficiency of service data transmission.

[0150] Specifically, the satellite base station completes the satellite-to-ground signaling interaction with the terminal and the target satellite through the optimized handover frame signaling preparation phase. This includes: the satellite base station sending a handover command to the terminal, containing information such as the target satellite's ID, channel number, and handover time; the terminal sending an acknowledgment signaling to the satellite base station after receiving the handover command; the satellite base station sending a terminal access command to the target satellite, notifying the target satellite to allocate a communication channel for the terminal; and the target satellite sending a channel ready signaling to the satellite base station after completing channel allocation. If a transmission failure occurs during the signaling interaction, retransmission is initiated according to the handover frame transmission rules, with a maximum of three retransmissions.

[0151] After successful signaling interaction, the handover frame enters the service transmission phase. The terminal switches its service data from the original serving satellite's communication channel to the target satellite's allocated channel, completing the inter-satellite handover. After the handover is completed, the satellite base station monitors the communication link status between the terminal and the target satellite in real time. If the link quality meets the communication requirements, the inter-satellite handover process is complete; if the link quality does not meet the requirements, the handover decision-making process is restarted in the next communication time slot.

[0152] To verify the effectiveness of the method of the present invention, this embodiment uses SKT satellite constellation simulation software to build a low-orbit satellite communication network simulation platform. The simulation parameters are consistent with the basic parameters of this embodiment. The simulation area is the Earth's latitude [33-43]° and longitude [110-121]°. The simulation duration is 10 minutes, and the number of terminals (UE) is 200 (100 TSU terminals and 100 CSU terminals).

[0153] The low-Earth orbit satellite inter-satellite handover method based on motion prediction of the present invention is compared with three existing mainstream handover methods (single-attribute algorithm based on maximum RSRP, single-attribute algorithm based on maximum idle channel, and multi-attribute handover algorithm without motion prediction).

[0154] The experimental results are as follows:

[0155] The single-attribute algorithm based on maximum RSRP has an average of 7.5 handovers per user, a throughput of 1.503 Mbps, and a satellite network congestion rate of 8.25%, indicating significant ping-pong handover and load imbalance.

[0156] The single-attribute algorithm based on the maximum idle channel has an average of 8.8 handovers per user, a throughput of 1.085 Mbps, and a blocking rate of 0, but the throughput is relatively low.

[0157] The multi-attribute handover algorithm without mobility prediction averages 6 handovers per user, has a throughput of 1.318 Mbps, a satellite network congestion rate of 3.12%, cannot accurately predict coverage continuity, and has mediocre performance.

[0158] The low-Earth orbit satellite inter-satellite handover method based on motion prediction in this invention has an average of 4.7 handovers per user, a throughput of 2.040 Mbps, and a blocking rate of only 2.41%, which significantly improves system throughput while reducing the number of handovers.

[0159] Specifically, the single-attribute algorithm based on the maximum RSRP ensures that the UE can guarantee the transmission rate with a certain transmission quality according to the RSRP threshold. Therefore, it has the largest transmission rate in this comparison. However, this method will cause UEs to tend to access the satellite with the largest RSRP, which will increase the number of handovers and make the satellite prone to congestion or even interruption, resulting in unbalanced resource allocation.

[0160] The single-attribute algorithm based on the maximum free channel can ensure that the UE's blocking rate is 0, thus avoiding congestion for the user. However, due to the distance and RSRP of the accessed satellite, the transmission rate is not high, and the number of handovers is also relatively high.

[0161] While multi-attribute handover algorithms without motion prediction can ensure a good throughput without excessive handovers, the remaining satellite service time cannot be mathematically calculated due to the unknown trajectory of high-speed ground mobile terminals. As a result, the remaining service time is used as one of the decision attributes, ultimately leading to less than ideal overall performance.

[0162] The low-Earth orbit satellite inter-satellite handover method based on mobility prediction in this invention provides prior information for handover, fully considering the service time available by candidate satellites, and using this as one of the decision criteria. This algorithm closely approximates the optimal performance between the single-attribute algorithm based on maximum RSRP and the single-attribute algorithm based on maximum idle channel. While maintaining a congestion rate of 2.41%, it makes the UE's transmission power close to that of the maximum RSRP algorithm. Furthermore, because mobility prediction can predict the service duration of candidate satellites in advance, it is easier to select satellites with longer service times, thus reducing the number of handovers compared to the baseline algorithm.

[0163] In summary, the LEO satellite inter-satellite handover method based on motion prediction of this invention overcomes the shortcomings of traditional single-attribute handover algorithms that do not comprehensively consider the overall network communication performance, and solves the problem that conventional multi-attribute handover algorithms cannot utilize the future location of the terminal to assist in handover. This invention demonstrates significant advantages in reducing the number of handovers, increasing system throughput, reducing blocking rates, and adapting to the needs of heterogeneous terminals. By obtaining the future location of the terminal through motion prediction, it provides prior information for handover decisions. Combined with multi-attribute decision-making principles, it also adapts to the personalized service needs of terminals, fully tapping the system's performance potential and effectively improving the communication stability and overall service performance of LEO satellite networks.

[0164] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for inter-satellite handover of low-Earth orbit satellites based on motion prediction, characterized in that: Includes the following steps: The satellite base station completes system initialization, collects the location information, service requirements and satellite ephemeris information of the ground terminals, and identifies the type of ground terminal based on the service requirements. The location information is input into a pre-trained GWO-GRU neural network model, which generates the predicted location of the ground terminal after a preset time interval based on the historical trajectory of the ground terminal. Construct a candidate satellite set, and assign a score to the remaining service time that the candidate satellites can provide to the ground terminal based on the predicted location to obtain the predicted coverage persistence score; The model calculates four indicators for each candidate satellite in the candidate satellite set: reference signal received power, satellite load, and handover overhead. The four indicators are normalized, and the normalization weights of each indicator are dynamically allocated according to the ground terminal type. The comprehensive score of each candidate satellite in the candidate satellite set is calculated by weighting. The candidate satellites in the candidate satellite set are sorted from highest to lowest according to their comprehensive scores, and the candidate satellite with the highest score that meets the access constraints is selected as the handover target for inter-satellite handover.

2. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The pre-training process of the GWO-GRU neural network model specifically includes: Historical trajectory data of high-speed mobile terminals are collected, and after missing value completion and outlier filtering, the data is converted to the ECEF coordinate system. The processed trajectory feature values ​​are then standardized using Z-Score. Using the historical ECEF coordinates of the high-speed mobile terminal within 2 minutes as input and the ECEF coordinates of the high-speed mobile terminal after 1 minute and 2 minutes as output, a training sample set is constructed, and the training sample set is divided into a training set and a test set according to the proportion. Initialize the GRU neural network and set the hyperparameter search range of the GRU neural network; Using 3D position prediction error as the fitness function, the hyperparameters of the GRU neural network are iteratively optimized using the Grey Wolf Algorithm (GWO). The optimal hyperparameters obtained through iterative optimization are substituted into the GRU neural network to complete the training. After the accuracy is verified by the test set, the trained GWO-GRU neural network model is loaded into the satellite base station.

3. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The specific methods for constructing the candidate satellite set include: Calculate the communication angle of each satellite relative to the ground terminal, and combine it with a preset minimum service elevation angle threshold. The formula for determining the validity of a satellite is as follows: ； in, For ground terminal identification, For satellite identification, For the current moment, for Time Satellite Compared to ground terminals The corner of faith for Candidate satellites at any time Valid judgment value, This indicates that the satellite is an invalid candidate satellite. This indicates that the satellite is a valid candidate satellite; all valid candidate satellites are integrated into a candidate satellite set.

4. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The specific methods for assigning scores to the remaining service time that candidate satellites can provide based on the predicted location to obtain the predicted coverage persistence score include: The satellite coverage status of the ground terminal at the predicted location is determined by the coverage status function, and then a step-by-step scoring is performed based on the coverage status result to obtain the predicted coverage persistence score. The coverage state function is: ; The formula for the predicted coverage persistence score is as follows: ; in For ground terminal identification; For satellite identification; The current moment; For the current moment The moment 1 minute later; For the current moment The time 2 minutes later; It is a binary function of 0 and 1, and is the terminal. exist Satellite of Time Coverage status: 1 indicates coverage, 0 indicates no coverage; for Time Terminal Corresponding candidate satellites The predicted coverage persistence score indicates that the longer the satellite coverage lasts for the terminal and the more stable the link is.

5. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The calculation of the reference signal received power is achieved by constructing a total path loss model for the satellite-to-ground link, specifically including: First, calculate the total path loss of the satellite-to-ground link, using the following formula: ； in, for Time Terminal With satellite Total path loss between; for Time Terminal With satellite Free space path loss between for Time Terminal With satellite The straight-line distance in space (unit: meters). Carrier frequency (unit: MHz); for Time Terminal With satellite Atmospheric loss in communication links, for Time Satellite Compared to ground terminals The corner of faith; For Rice fading loss; Then, the received power of the reference signal is calculated based on the total path loss, using the following formula: ； in, for Time Terminal Received satellite The reference signal received power; This refers to the satellite's launch power. for Time Satellite The transmit antenna gain; for Real-time ground terminal The receiving antenna gain.

6. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The quantitative modeling formula for the switching overhead is as follows: ； in, for Time Terminal Switch to satellite Total switching overhead, Normalized data loss during a single handover refers to the effective data loss caused by the terminal pausing service transmission during the handover preparation phase. Normalized energy loss per handover refers to the additional energy consumed by the terminal when performing the handover operation.

7. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The satellite load is used to characterize the channel resource utilization of candidate satellites. The number of idle channels is used as the evaluation index of satellite load. The larger the value, the more idle channels the satellite has, the lower the load, and the more terminals can access the satellite. The formula is: ； in, for Candidate satellites at any time The number of idle channels; For satellite Total number of channels; for Time Satellite The number of occupied channels is collected in real time by satellite base stations.

8. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The normalized weights of each indicator are dynamically allocated to satisfy the weight summation constraint: ,and ;in The normalized weights are used as reference signal received power. To predict the normalized weights for coverage persistence scores, For the normalized weights of the satellite payload, Normalized weights for switching overhead; The ground terminal types are divided into throughput-sensitive users and continuity-sensitive users. For throughput-sensitive users, the weighting priority is as follows: ; For users who are sensitive to continuity, the weighting priority is as follows: .

9. A method for inter-satellite handover of low-Earth orbit satellites based on motion prediction according to claim 8, characterized in that: The weighted calculation of the overall score of each candidate satellite in the candidate satellite set is performed using the following formula: ； in, for Time Terminal Corresponding candidate satellites The overall score ranges from [0,1]. A higher score indicates that the satellite is more suitable as a handover target for the terminal. , , , These are the Min-Max normalized values ​​of the reference signal received power, predicted coverage persistence score, satellite load, and handover cost, respectively. The handover cost is a negative indicator and is inverted after Min-Max normalization.

10. The low-Earth orbit satellite inter-satellite handover method based on motion prediction according to claim 1, characterized in that: The access constraints include: a single terminal can only access one satellite in a single time slot and can only occupy one channel of that satellite; the total number of terminals accessing a single satellite in a single time slot does not exceed its number of idle channels; if all satellite channels are occupied, the terminal's access request to that satellite is determined to be failed, and it waits for the next communication time slot to make a new handover decision. The inter-satellite handover is performed according to the optimized handover frame structure, which is divided into a signaling preparation phase and a service transmission phase, and the following transmission rules are established: the handover frame is placed at the beginning of each communication time slot, and the communication link status between the terminal and the satellite remains unchanged within a single time slot; if the signaling transmission fails during the signaling preparation phase, retransmission is immediately initiated until the signaling transmission is successful or the signaling preparation phase duration is exhausted; after the signaling transmission during the signaling preparation phase is successful, the remaining duration of the handover frame is converted into the service transmission phase; if it is determined that no handover is needed, the entire duration of the handover frame is used as the service transmission phase.

Images