A satellite portable station automatic star alignment system

Abstract

The invention discloses an automatic satellite alignment system for a portable satellite station, which belongs to the technical field of automatic satellite alignment and comprises a single chip microcomputer, a geographic positioning module, a posture sensing module, a stepping motor control module, a learning optimization module, a scanning path optimization module, a signal strength detection module, an antenna azimuth search module, a signal analysis module, a learning correction module and a data storage module. The invention can avoid falling into a local optimum too early, does not require manual intervention, reduces the complexity of manual debugging, can find more complex and diverse optimization paths, can provide accurate control paths more quickly, and ensure that accurate alignment is completed in a short time; improves the accuracy and efficiency of antenna alignment, reduces the time of the fine-tuning stage, can adapt to different signal environments, quickly finds the maximum value of the signal strength, ensures the communication quality, and reduces the amount of calculation and resource consumption.

Description

Technical Field

[0001] The invention relates to the technical field of automatic star alignment, and in particular to an automatic star alignment system for a portable satellite station. Background Art

[0002] At present, the portable satellite station usually calculates the theoretical values ​​of azimuth and elevation angles through formulas based on its own positioning longitude and latitude and the longitude of the satellite, uses a mechanical compass to measure the actual values ​​of the azimuth and elevation angles of the satellite antenna, and manually adjusts the antenna to achieve the alignment. However, there are difficulties such as deviations between theoretical values ​​and actual conditions, errors in manual readings of mechanical compasses, and inaccurate manual direction adjustment, which lead to problems such as slow alignment speed, poor accuracy, and low satellite signals. In view of the above problems, an automatic alignment system for a portable satellite station is proposed to achieve the purpose of simple operation, improved accuracy, and rapid positioning.

[0003] The existing automatic alignment system of portable satellite stations is prone to fall into local optimality too early, and the complexity of manual debugging is high. At the same time, it is unable to find more complex and diverse optimization paths, and cannot quickly provide accurate control paths. In addition, the existing automatic alignment system of portable satellite stations has low antenna alignment accuracy and efficiency, and the fine-tuning stage takes a long time, and it cannot adapt to different signal environments. Therefore, we propose an automatic alignment system for a portable satellite station. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose an automatic satellite pointing system for a portable satellite station.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A satellite portable station automatic star pointing system, comprising a single chip microcomputer, a geographic positioning module, a posture sensing module, a stepping motor control module, a learning optimization module, a scanning path optimization module, a signal strength detection module, an antenna azimuth search module, a signal analysis module, a learning correction module and a data storage module;

[0007] The single chip microcomputer is used to collect and process data from each module, and perform rough adjustment and precise fine adjustment of the antenna;

[0008] The geographic positioning module is used to collect the geographic location information of the satellite portable station using the Beidou / GPS dual-mode positioning system;

[0009] The attitude sensing module is used to collect the current azimuth and pitch angle of the antenna in real time;

[0010] The stepper motor control module is used to control the stepper motors in the azimuth and elevation directions of the antenna, so that the antenna angle is adjusted according to the corrected theoretical value to perform preliminary satellite alignment operations;

[0011] The learning optimization module is used to optimize the scanning strategy according to the historical satellite data and predict the best scanning area;

[0012] The scanning path optimization module is used to plan the antenna scanning path and optimize the control path of the stepper motor;

[0013] The signal strength detection module is used to monitor the received satellite signal strength in real time and feed it back to the single chip microcomputer for direction adjustment;

[0014] The antenna orientation search module is used to quickly locate the optimal antenna direction during the signal strength fine-tuning stage;

[0015] The signal analysis module is used to perform spectrum analysis and demodulation on the received satellite signal;

[0016] The learning correction module is used to record the deviation between the theoretical value and the actual value, update the initial parameters of the system, and dynamically correct the deviation between the initial positioning value and the antenna adjustment process;

[0017] The data storage module is used to store all data generated during the satellite alignment process.

[0018] As a further solution of the present invention, the specific steps of the preliminary star alignment operation of the stepper motor control module are as follows:

[0019] S1.1: Based on the geographical location information collected by the geo-positioning module, the theoretical values ​​of the azimuth and elevation angles of the satellite portable station antenna are calculated according to the formula, and the azimuth is corrected using the magnetic declination;

[0020] S1.2: Compare the corrected theoretical value with the current antenna azimuth and pitch angle collected by the attitude sensor module. According to the comparison result, the azimuth and pitch high-precision stepper motors are controlled by the stepper motor control module to drive the gears to rotate for preliminary alignment. After the preliminary alignment process is completed, the stepper motor is controlled to scan in a smaller area.

[0021] As a further solution of the present invention, the specific steps of predicting the optimal scanning area by the learning optimization module are as follows:

[0022] S2.1: extract historical alignment data from the data storage module, including geographic location information, initial antenna angle, signal strength and final alignment angle, clean and extract features from each set of historical alignment data, standardize each set of data to the interval [0,1], divide each set of processed data into a training set, a validation set and a test set, and expand each set of data by adding random noise or generating virtual samples;

[0023] S2.2: Construct input and output features, and create a set of deep learning models based on the RNN network architecture. Randomly initialize the weights and biases of the network, then input the training set data into the deep learning model, calculate layer by layer through forward propagation, obtain the output prediction value, and calculate the deviation between the predicted value and the actual value through the mean square error loss function;

[0024] S2.3: Input the calculated deviation value from the output layer of the deep learning model. Based on the chain rule, calculate the gradient of the deviation value for the weights and biases of each network layer of the model layer by layer, and use the Adam optimizer to update the weights and biases. After each training cycle, calculate the loss value with the test set. If the test loss stops decreasing, stop training in advance. Otherwise, adjust the number of network layers, number of nodes, and learning rate hyperparameters, and continue training.

[0025] S2.4: Use the test set to test the final performance indicators of the deep learning model after training, evaluate the generalization ability of the model, collect and preprocess the real-time data required by the current portable station to the satellite, pass the preprocessed input data to the trained deep learning model, and obtain the predicted optimal scanning area center angle and scanning radius;

[0026] S2.5: Generate a scanning path for the antenna using the predicted optimal scanning area center angle and scanning radius. According to the generated scanning path, gradually adjust the antenna angle through a stepper motor. After the scanning is completed, further fine-tune the antenna direction according to the position of the point with the maximum signal strength.

[0027] As a further solution of the present invention, the specific steps of the scanning path optimization module to optimize the control path of the stepper motor are as follows:

[0028] S3.1: According to the results of the preliminary scanning area generated by the learning optimization module, the center point and radius of the fine-tuning range are determined, the angle interval of the segmentation is determined according to the fine-tuning accuracy requirements and the minimum step angle of the stepper motor, and the area within the fine-tuning range is discretized and segmented to generate a grid structure and corresponding grid points;

[0029] S3.2: Screen the generated grid points, remove invalid points beyond the circular fine-tuning range, establish adjacency relationships for all valid grid points, generate a fine-tuning path grid containing a grid point set and an adjacency matrix, and initialize the number of search bodies, pheromone concentrations of each path, and pheromone volatility;

[0030] S3.3: According to the signal strength change between each grid point and the angular displacement of the stepper motor, the corresponding heuristic factor is calculated, the initial grid point position of each search body is initialized, and the selection probability of each selection node is calculated according to the pheromone concentration of each path and the corresponding heuristic factor, and the direction of the next move is selected according to the probability rule;

[0031] S3.4: When each group of search bodies completes the construction of a complete path, the maximum signal strength of each group of paths is counted, and the pheromone concentration of the path is adjusted through global and local update rules. The node selection, path construction and pheromone update are repeated until the pheromone concentration change of each group of paths converges to the preset threshold and stops;

[0032] S3.5: Select the path with the highest signal strength score as the optimal control path of the stepper motor, and feed back the fine-tuning strategy corresponding to the path to the stepper motor control module. The stepper motor control module controls the antenna to adjust the direction along this path until the maximum signal strength point is reached.

[0033] As a further solution of the present invention, the specific steps of the antenna orientation search module to quickly locate the optimal antenna direction are as follows:

[0034] S4.1: collect the current angle position of the antenna during the fine-tuning process, the angle point from which the antenna moves to the adjacent angle point, and the signal strength gain of the current antenna position, construct a set of search trees, and use each node in the tree to represent the angle state, and the edge to represent the action of the antenna moving from one angle state to another. Use the initial angle position of the fine-tuning area in the optimal control path as the root node of the search tree;

[0035] S4.2: Starting from the root node, use the upper confidence interval formula to select the node with the highest UCB1 value along the search tree to build the current optimal path until reaching the leaf node. When the leaf node is not fully expanded, multiple groups of child nodes are generated based on the actions that can be selected by the current child node, and the generated child nodes are added to the search tree as new leaf nodes. Starting from the new leaf node, randomly select actions to simulate, move the antenna angle, and calculate the corresponding signal strength reward;

[0036] S4.3: The reward value of the simulated path is traced back to each node on the path, the number of visits and the cumulative reward of the node are updated, and the selection, expansion, simulation and backtracking steps are repeated until the fixed number of iterations is reached or the specified search time limit is reached. The node with the most visits is selected from the child nodes of the root node, and its corresponding state is used as the optimal antenna direction.

[0037] As a further solution of the present invention, the specific steps of dynamic correction of the learning correction module are as follows:

[0038] S5.1: Randomly generate a set of antibodies as the initial population, where each antibody consists of a set of parameters, representing the amount of adjustment after correction, simulate the antenna angle and position corrected based on each set of parameters, evaluate each antibody, and calculate its corresponding fitness value based on the signal strength in the corrected antenna direction;

[0039] S5.2: Record the fitness value of each antibody, select antibodies according to the fitness value through the roulette method, clone the selected antibodies, generate a new offspring antibody pool, and then randomly perturb each parameter of the selected antibodies, calculate the fitness value of each antibody after the perturbation, and compare it with the original antibody. If the fitness value of the mutated antibody is greater than the original antibody, replace the original antibody. After the replacement is completed, remove the antibodies with high similarity for antibody inhibition;

[0040] S5.3: Repeat the fitness evaluation, selection, cloning, mutation and inhibition process until the preset number of iterations is reached or the antibody fitness value is improved and converged to a preset range, output the antibody with the highest fitness value, and use the correction parameters corresponding to the antibody as the dynamic correction result to correct the actual initial positioning value and the deviations that occurred during the adjustment process.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. The present invention discretizes and segments the area within the fine-tuning range, generates a grid structure, and generates corresponding grid points. The generated grid points are screened to eliminate invalid points beyond the circular fine-tuning range, and an adjacency relationship is established for all valid grid points to generate a fine-tuning path grid including a grid point set and an adjacency matrix. The number of search bodies, the pheromone concentration of each path, and the pheromone volatility are initialized. According to the change of the signal strength between each grid point and the angular displacement of the stepping motor, the corresponding heuristic factor is calculated, the initial grid point position of each search body is initialized, and the selection probability of each band selection node is calculated according to the pheromone concentration of each path and the corresponding heuristic factor, and the direction of the next move is selected according to the probability rule. When each group of search bodies completes a complete path After the construction, the maximum signal strength of each group of paths is counted, and the pheromone concentration of the path is adjusted through global and local update rules. The node selection, path construction and pheromone update are repeated until the pheromone concentration change of each group of paths converges to the preset threshold and stops. The path with the highest signal strength score is selected as the optimal control path of the stepper motor, and the fine-tuning strategy corresponding to the path is fed back to the stepper motor control module. The antenna is controlled by the stepper motor control module to adjust the direction along this path until the maximum signal strength point is reached. This can avoid falling into the local optimum too early, and does not require manual intervention, reducing the complexity of manual debugging, and can find more complex and diverse optimization paths. It can provide accurate control paths more quickly, ensuring that accurate satellite alignment can be completed in a short time.

[0043] 2. The present invention collects the current angular position of the antenna during the fine-tuning process, the antenna moves from the current position to the adjacent angular point, and the signal strength gain of the current antenna position, constructs a set of search trees, and each node in the tree represents the angular state, and the edge represents the action of the antenna moving from one angular state to another. The initial angular position of the fine-tuning area in the optimal control path is used as the root node of the search tree. Starting from the root node, the upper confidence interval formula is used to select the node with the highest UCB1 value along the search tree to construct the current optimal path until the leaf node is reached. When the leaf node is not fully expanded, multiple groups of child nodes are generated based on the actions that can be selected by the current child node, and the generated child nodes are used as new leaf nodes. The point is added to the search tree, starting from the new leaf node, randomly selecting actions for simulation, moving the antenna angle, calculating the corresponding signal strength reward, tracing the reward value of the simulated path back to each node on the path, updating the number of node visits and cumulative rewards, repeating the selection, expansion, simulation and backtracking steps until the fixed number of iterations is reached or the specified search time limit is reached, and the node with the most visits is selected from the child nodes of the root node, and its corresponding state is used as the optimal antenna direction, which improves the accuracy and efficiency of the antenna to the satellite, reduces the time of the fine-tuning stage, can adapt to different signal environments, quickly find the maximum signal strength, ensure communication quality, and reduce calculation and resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0045] Figure 1 This is a system block diagram of a satellite portable station automatic alignment system proposed by the present invention;

[0046] Figure 2 This is a design model diagram of the mechanical components of a satellite portable station automatic alignment system proposed by the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0048] Example 1

[0049] Reference Figure 1-2 A portable satellite station automatic star pointing system includes a single chip microcomputer, a geographic positioning module, a posture sensing module, a stepping motor control module, a learning optimization module, a scanning path optimization module, a signal strength detection module, an antenna azimuth search module, a signal analysis module, a learning correction module and a data storage module.

[0050] The single-chip microcomputer is used to collect and process data from each module, and to perform rough adjustment and precise fine-tuning of the antenna; the geo-positioning module is used to utilize the Beidou / GPS dual-mode positioning system to collect the geographical location information of the satellite portable station; the attitude sensing module is used to collect the current azimuth and elevation angles of the antenna in real time; the stepper motor control module is used to control the stepper motors in the azimuth and elevation directions of the antenna, so that it adjusts the antenna angle according to the corrected theoretical value to perform preliminary satellite alignment operations.

[0051] In addition, by Figure 2 It can be seen that the motor that controls the direction of the satellite antenna is set on the satellite antenna, and the signal strength received by the feed source is transmitted to the microcontroller together with the data collected by the geolocation module for analysis and calculation to find the best angle for the satellite. In the design of mechanical components, gears, screws and other mechanical components are used to connect the stepper motor to the satellite antenna, thereby achieving the purpose of using a stepper motor to control the azimuth and pitch angle adjustment of the satellite antenna.

[0052] Specifically, based on the geographic location information collected by the geographic positioning module, the theoretical values ​​of the azimuth and elevation angles of the satellite portable station antenna are calculated according to the formula, and the azimuth is corrected using the magnetic declination. The corrected theoretical value is compared with the current antenna azimuth and elevation angle collected by the attitude sensing module. Based on the comparison result, the azimuth and elevation high-precision stepper motors are controlled by the stepper motor control module to drive the gears to rotate for preliminary star alignment. When the preliminary star alignment process is completed, the stepper motor is controlled to scan in a smaller area.

[0053] The learning optimization module is used to optimize the scanning strategy based on historical satellite data and predict the best scanning area.

[0054] Specifically, historical satellite alignment data, including geographic location information, initial antenna angle, signal strength and final alignment angle, are extracted from the data storage module, each set of collected historical satellite alignment data is cleaned and feature extracted, and each set of data is standardized to the interval [0,1], and each set of processed data is divided into a training set, a validation set and a test set. Each set of data is expanded by adding random noise or generating virtual samples, and input and output features are constructed. A set of deep learning models is created based on the RNN network architecture, and the weights and biases of the network are randomly initialized. The training set data is then input into the deep learning model, and the output prediction value is obtained by forward propagation layer by layer. The deviation between the predicted value and the actual value is calculated by the mean square error loss function, and the calculated deviation value is input from the output layer of the deep learning model. Based on the chain rule, the deviation value pair is calculated layer by layer. The gradient of the weights and biases of each network layer of the model is calculated, and the Adam optimizer is used to update the weights and biases. After each training cycle, the test set is used to calculate the loss value. If the test loss stops decreasing, the training is stopped in advance. Otherwise, the number of network layers, the number of nodes, and the learning rate hyperparameters are adjusted, and the training is continued. The final performance indicators of the deep learning model after training are tested by the test set to evaluate the generalization ability of the model. The real-time data required by the current portable station to the satellite is collected and preprocessed, and the preprocessed input data is passed to the trained deep learning model, and the predicted optimal scanning area center angle and scanning radius are obtained. The predicted optimal scanning area center angle and scanning radius are used to generate the scanning path of the antenna. According to the generated scanning path, the antenna angle is gradually adjusted by a stepper motor. After the scanning is completed, the antenna direction is further fine-tuned according to the position of the point with maximum signal strength.

[0055] The scanning path optimization module is used to plan the antenna scanning path and optimize the control path of the stepper motor.

[0056] Specifically, according to the results of the preliminary scanning area generated by the learning optimization module, the center point and radius of the fine-tuning range are determined, the segmentation angle interval is determined according to the fine-tuning accuracy requirements and the minimum step angle of the stepper motor, and the area within the fine-tuning range is discretized and segmented to generate a grid structure and corresponding grid points. The generated grid points are screened to remove invalid points that exceed the circular fine-tuning range, and an adjacency relationship is established for all valid grid points to generate a fine-tuning path grid containing a grid point set and an adjacency matrix. The number of search bodies, the pheromone concentration of each path, and the pheromone volatility rate are initialized. According to the change in signal strength between each grid point and the angular displacement of the stepper motor, the corresponding heuristic factor is calculated to initialize the initial value of each search body. The grid point position is calculated, and the selection probability of each selection node is calculated according to the pheromone concentration of each path and the corresponding heuristic factor, and the direction of the next move is selected according to the probability rule. When each group of search bodies completes the construction of a complete path, the maximum signal strength of each group of paths is counted, and the pheromone concentration of the path is adjusted through global and local update rules. Node selection, path construction and pheromone update are repeated until the pheromone concentration change of each group of paths converges to the preset threshold and then stops. The path with the highest signal strength score is selected as the optimal control path of the stepper motor, and the fine-tuning strategy corresponding to the path is fed back to the stepper motor control module. The stepper motor control module controls the antenna to adjust the direction along this path until the maximum signal strength point is reached.

[0057] Example 2

[0058] Reference Figure 1-2 A portable satellite station automatic star pointing system includes a single chip microcomputer, a geographic positioning module, a posture sensing module, a stepping motor control module, a learning optimization module, a scanning path optimization module, a signal strength detection module, an antenna azimuth search module, a signal analysis module, a learning correction module and a data storage module.

[0059] The signal strength detection module is used to monitor the received satellite signal strength in real time and feed it back to the microcontroller for direction adjustment; the antenna azimuth search module is used to quickly locate the optimal antenna direction during the signal strength fine-tuning stage.

[0060] Specifically, the current angle position of the antenna, the angle point from which the antenna moves from the current position to the adjacent angle point, and the signal strength gain of the current antenna position are collected during the fine-tuning process, a set of search trees are constructed, and each node in the tree represents an angle state, and the edge represents the action of the antenna moving from one angle state to another. The initial angle position of the fine-tuning area in the optimal control path is used as the root node of the search tree. Starting from the root node, the upper confidence interval formula is used to select the node with the highest UCB1 value along the search tree to construct the current optimal path until the leaf node is reached. When the leaf node is not fully expanded, multiple groups of child nodes are generated based on the actions that can be selected by the current child node, and the generated child nodes are added to the search tree as new leaf nodes. Starting from the new leaf node, actions are randomly selected for simulation, the antenna angle is moved, and the corresponding signal strength reward is calculated. The reward value of the simulated path is traced back to each node on the path, the number of visits and the cumulative reward of the node are updated, and the selection, expansion, simulation and backtracking steps are repeated until a fixed number of iterations is reached or the specified search time limit is reached. The node with the most visits is selected from the child nodes of the root node, and its corresponding state is used as the optimal antenna direction.

[0061] The signal analysis module is used to perform spectrum analysis and demodulation on the received satellite signals; the learning correction module is used to record the deviation between the theoretical value and the actual value, update the initial parameters of the system, and dynamically correct the deviation between the initial positioning value and the antenna adjustment process; the data storage module is used to store all data generated during the satellite alignment process.

[0062] Specifically, a set of antibody sets is randomly generated as the initial population, wherein each antibody is composed of a set of parameters, representing the adjustment amount after correction. The antenna angle and position corrected based on each set of parameters are simulated, each antibody is evaluated, and its corresponding fitness value is calculated based on the signal strength in the corrected antenna direction, and the fitness value of each antibody is recorded. The antibody is selected according to the fitness value through the roulette method, and the selected antibody is cloned to generate a new offspring antibody pool. After that, a random perturbation is performed on each parameter of the selected antibody, and the fitness value of each antibody after the perturbation is calculated and compared with the original antibody. If the fitness value of the mutated antibody is greater than the original antibody, the original antibody is replaced. After the replacement is completed, the antibodies with high similarity are removed for antibody suppression, and the fitness evaluation, selection, cloning, mutation and suppression processes are repeated until a preset number of iterations is reached or the antibody fitness value is improved and converged to a preset range, and the antibody with the highest fitness value is output, and the correction parameter corresponding to the antibody is used as a dynamic correction result to correct the actual initial positioning value and the deviation occurring during the adjustment process.

Claims

1. A satellite portable station automatic star alignment system, characterized in that: It includes a single chip microcomputer, a geographic positioning module, a posture sensing module, a stepping motor control module, a learning optimization module, a scanning path optimization module, a signal strength detection module, an antenna orientation search module, a signal analysis module, a learning correction module and a data storage module; The single chip microcomputer is used to collect and process data from each module, and perform rough adjustment and precise fine adjustment of the antenna; The geographic positioning module is used to collect the geographic location information of the satellite portable station using the Beidou / GPS dual-mode positioning system; The attitude sensing module is used to collect the current azimuth and pitch angle of the antenna in real time; The stepper motor control module is used to control the stepper motors in the azimuth and elevation directions of the antenna, so that the antenna angle is adjusted according to the corrected theoretical value to perform preliminary satellite alignment operations; The learning optimization module is used to optimize the scanning strategy according to the historical satellite data and predict the best scanning area; The scanning path optimization module is used to plan the antenna scanning path and optimize the control path of the stepper motor; The signal strength detection module is used to monitor the received satellite signal strength in real time and feed it back to the single chip microcomputer for direction adjustment; The antenna orientation search module is used to quickly locate the optimal antenna direction during the signal strength fine-tuning stage; The signal analysis module is used to perform spectrum analysis and demodulation on the received satellite signal; The learning correction module is used to record the deviation between the theoretical value and the actual value, update the initial parameters of the system, and dynamically correct the deviation between the initial positioning value and the antenna adjustment process; The data storage module is used to store all data generated during the satellite alignment process.

2. The portable satellite station automatic alignment system according to claim 1, characterized in that: The specific steps of the preliminary star alignment operation of the stepper motor control module are as follows: S1.1: Based on the geographical location information collected by the geo-positioning module, the theoretical values ​​of the azimuth and elevation angles of the satellite portable station antenna are calculated according to the formula, and the azimuth is corrected using the magnetic declination; S1.2: Compare the corrected theoretical value with the current antenna azimuth and pitch angle collected by the attitude sensor module. According to the comparison result, the azimuth and pitch high-precision stepper motors are controlled by the stepper motor control module to drive the gears to rotate for preliminary alignment. After the preliminary alignment process is completed, the stepper motor is controlled to scan in a smaller area.

3. The portable satellite station automatic alignment system according to claim 2, characterized in that: The specific steps of the learning optimization module predicting the optimal scanning area are as follows: S2.1: extract historical alignment data from the data storage module, including geographic location information, initial antenna angle, signal strength and final alignment angle, clean and extract features from each set of historical alignment data, standardize each set of data to the interval [0,1], divide each set of processed data into a training set, a validation set and a test set, and expand each set of data by adding random noise or generating virtual samples; S2.2: Construct input and output features, and create a set of deep learning models based on the RNN network architecture. Randomly initialize the weights and biases of the network, then input the training set data into the deep learning model, calculate layer by layer through forward propagation, obtain the output prediction value, and calculate the deviation between the predicted value and the actual value through the mean square error loss function; S2.3: Input the calculated deviation value from the output layer of the deep learning model. Based on the chain rule, calculate the gradient of the deviation value for the weights and biases of each network layer of the model layer by layer, and use the Adam optimizer to update the weights and biases. After each training cycle, calculate the loss value with the test set. If the test loss stops decreasing, stop training in advance. Otherwise, adjust the number of network layers, number of nodes, and learning rate hyperparameters, and continue training. S2.4: Use the test set to test the final performance indicators of the deep learning model after training, evaluate the generalization ability of the model, collect and preprocess the real-time data required by the current portable station to the satellite, pass the preprocessed input data to the trained deep learning model, and obtain the predicted optimal scanning area center angle and scanning radius; S2.5: Generate a scanning path for the antenna using the predicted optimal scanning area center angle and scanning radius. According to the generated scanning path, gradually adjust the antenna angle through a stepper motor. After the scanning is completed, further fine-tune the antenna direction according to the position of the point with the maximum signal strength.

4. The portable satellite station automatic alignment system according to claim 3, characterized in that: The specific steps of the scanning path optimization module to optimize the control path of the stepper motor are as follows: S3.1: According to the results of the preliminary scanning area generated by the learning optimization module, the center point and radius of the fine-tuning range are determined, the angle interval of the segmentation is determined according to the fine-tuning accuracy requirements and the minimum step angle of the stepper motor, and the area within the fine-tuning range is discretized and segmented to generate a grid structure and corresponding grid points; S3.2: Screen the generated grid points, remove invalid points beyond the circular fine-tuning range, establish adjacency relationships for all valid grid points, generate a fine-tuning path grid containing a grid point set and an adjacency matrix, and initialize the number of search bodies, pheromone concentrations of each path, and pheromone volatility; S3.3: According to the signal strength change between each grid point and the angular displacement of the stepper motor, the corresponding heuristic factor is calculated, the initial grid point position of each search body is initialized, and the selection probability of each selection node is calculated according to the pheromone concentration of each path and the corresponding heuristic factor, and the direction of the next move is selected according to the probability rule; S3.4: When each group of search bodies completes the construction of a complete path, the maximum signal strength of each group of paths is counted, and the pheromone concentration of the path is adjusted through global and local update rules. The node selection, path construction and pheromone update are repeated until the pheromone concentration change of each group of paths converges to the preset threshold and stops; S3.5: Select the path with the highest signal strength score as the optimal control path of the stepper motor, and feed back the fine-tuning strategy corresponding to the path to the stepper motor control module. The stepper motor control module controls the antenna to adjust the direction along this path until the maximum signal strength point is reached.

5. The portable satellite station automatic alignment system according to claim 4, characterized in that: The specific steps of the antenna direction search module to quickly locate the best antenna direction are as follows: S4.1: collect the current angle position of the antenna during the fine-tuning process, the angle point from which the antenna moves to the adjacent angle point, and the signal strength gain of the current antenna position, construct a set of search trees, and use each node in the tree to represent the angle state, and the edge to represent the action of the antenna moving from one angle state to another. Use the initial angle position of the fine-tuning area in the optimal control path as the root node of the search tree; S4.2: Starting from the root node, use the upper confidence interval formula to select the node with the highest UCB1 value along the search tree to build the current optimal path until reaching the leaf node. When the leaf node is not fully expanded, multiple groups of child nodes are generated based on the actions that can be selected by the current child node, and the generated child nodes are added to the search tree as new leaf nodes. Starting from the new leaf node, randomly select actions to simulate, move the antenna angle, and calculate the corresponding signal strength reward; S4.3: The reward value of the simulated path is traced back to each node on the path, the number of visits and the cumulative reward of the node are updated, and the selection, expansion, simulation and backtracking steps are repeated until the fixed number of iterations is reached or the specified search time limit is reached. The node with the most visits is selected from the child nodes of the root node, and its corresponding state is used as the optimal antenna direction.

6. The portable satellite station automatic alignment system according to claim 1, characterized in that: The specific steps of dynamic correction of the learning correction module are as follows: S5.1: Randomly generate a set of antibodies as the initial population, where each antibody consists of a set of parameters, representing the amount of adjustment after correction, simulate the antenna angle and position corrected based on each set of parameters, evaluate each antibody, and calculate its corresponding fitness value based on the signal strength in the corrected antenna direction; S5.2: Record the fitness value of each antibody, select antibodies according to the fitness value through the roulette method, clone the selected antibodies, generate a new offspring antibody pool, and then randomly perturb each parameter of the selected antibodies, calculate the fitness value of each antibody after the perturbation, and compare it with the original antibody. If the fitness value of the mutated antibody is greater than the original antibody, replace the original antibody. After the replacement is completed, remove the antibodies with high similarity for antibody inhibition; S5.3: Repeat the fitness evaluation, selection, cloning, mutation and inhibition process until the preset number of iterations is reached or the antibody fitness value is improved and converged to a preset range, output the antibody with the highest fitness value, and use the correction parameters corresponding to the antibody as the dynamic correction result to correct the actual initial positioning value and the deviations that occurred during the adjustment process.

Images