Satellite selection method for advanced receiver autonomous integrity monitoring

Abstract

The application discloses a satellite selection method for advanced receiver autonomous integrity monitoring, which comprises the following steps: obtaining a multi-level feature vector of a current effective satellite set, including satellite-level features, constellation-level features and environment features, inputting the multi-level feature vector into a pre-trained constellation distribution prediction model, and outputting a current optimal constellation proportion vector; generating a plurality of candidate satellite groups based on the optimal constellation proportion vector and the effective satellite set, each candidate satellite group comprising at least one satellite and at least one constellation; generating a comprehensive score and a failure trend factor of each candidate satellite group by using a preset multi-level cooperative scoring mechanism; and generating a main combination and a backup strategy thereof based on the comprehensive scores and the failure trend factors of all candidate satellite groups by using a preset main-backup combination satellite selection decision. Thus, the decision efficiency is improved, and the positioning accuracy and the reliability of integrity monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of satellite navigation technology, and in particular to a satellite selection method for advanced receiver autonomous integrity monitoring. Background Technology

[0002] In the field of satellite navigation and positioning, with the continuous evolution and improvement of the four major global satellite navigation systems—BeiDou, GPS, GLONASS, and the European Galileo system—multi-constellation navigation technology is becoming increasingly widespread. Users can now choose from and use a wider range of satellites, which should greatly enhance the flexibility and potential accuracy of positioning. However, how to accurately select the optimal satellite combination from this vast pool of satellites to effectively improve positioning performance has become a critical challenge that urgently needs to be overcome.

[0003] In real-world applications, satellite navigation systems face complex and ever-changing external environmental challenges. Ionospheric disturbances interfere with the propagation path of satellite signals, and various weather conditions such as heavy rain and dense fog adversely affect signal strength and quality. These factors combined result in inconsistent satellite signal quality, severely impacting positioning accuracy. Traditional satellite selection methods often have limitations in addressing these complex environmental factors, failing to comprehensively and accurately consider their influence on satellite signals. This makes it difficult to achieve ideal satellite selection quality, ultimately leading to decreased positioning performance and an inability to meet the ever-increasing demand for high-precision positioning. Especially in applications such as aviation, maritime navigation, and autonomous driving, where system reliability and safety are paramount, the stability and accuracy of satellite navigation systems are crucial. Even a minor deviation in a satellite navigation system can trigger serious safety incidents and cause incalculable losses.

[0004] To effectively address fault issues in satellite navigation systems, Advanced Receiver Autonomous Integrity Monitoring (ARAIM) technology has emerged. Compared to traditional Receiver Autonomous Integrity Monitoring (RAIM), ARAIM technology possesses more powerful fault monitoring capabilities, simultaneously monitoring various complex fault modes such as multi-satellite faults and constellation faults, providing strong assurance for the safe and reliable operation of satellite navigation systems. Currently, multi-constellation ARAIM algorithms have been widely applied in numerous fields. However, in a multi-constellation ARAIM environment, while the significant increase in the number of satellites makes global availability potentially reach or approach 100%, it also brings new challenges. With the increase in the number of satellites, the potential fault modes grow exponentially, and the number and complexity of the fault subsets that need to be monitored increase dramatically. At this point, simply selecting satellites based on their geometric configuration is no longer sufficient to meet the high requirements for satellite selection quality of terminal algorithms such as ARAIM, making it difficult to fully leverage the advantages of multi-constellation ARAIM technology. Therefore, researching a satellite selection optimization method adapted to the multi-constellation ARAIM environment has significant practical significance and application value. Summary of the Invention

[0005] This application provides a satellite selection method for autonomous integrity monitoring of advanced receivers, which improves satellite selection efficiency and enhances the positioning reliability of ARAM monitoring.

[0006] This application provides a satellite selection method for advanced receiver autonomous integrity monitoring, including:

[0007] S101: Obtain the multi-level feature vector of the current effective satellite set, including satellite-level features, constellation-level features and environmental features, input it into the pre-trained constellation distribution prediction model, and output the current optimal constellation ratio vector;

[0008] S102, based on the optimal constellation ratio vector and the effective satellite set, generate several candidate satellite groups, each candidate satellite group including at least one satellite and at least one constellation;

[0009] S103 utilizes a pre-set multi-level collaborative scoring mechanism to generate a comprehensive score and fault tendency factor for each candidate satellite group;

[0010] S104, based on the comprehensive scores and fault tendency factors of all candidate satellite groups, uses the preset primary and backup combination satellite selection decision to generate the primary combination and its backup strategy.

[0011] Preferably, the multi-level feature vector is obtained as follows:

[0012] A1. Generate satellite-level and constellation-level features based on the performance attributes of each valid satellite. The performance attributes include user ranging error, satellite elevation angle, and signal-to-noise ratio.

[0013] A2. Environmental characteristics consist of the ionospheric disturbance index and weather codes;

[0014] A3. Concatenate satellite-level features, constellation-level features, and environmental features sequentially to form a multi-level feature vector.

[0015] Preferably, the step of generating satellite-level features and constellation-level features based on the performance attributes of each valid satellite includes:

[0016] a1. Based on the performance attributes of each valid satellite, group them according to their constellation affiliation and classify them into the corresponding constellation. Calculate the statistical characteristics of the performance attributes within each constellation, including the mean, standard deviation, range, and proportion of high elevation angle satellites.

[0017] a2. Combine the mean, standard deviation, and range of each performance attribute within each constellation, along with the proportion of high-elevation satellites in that constellation, to form constellation-level characteristics;

[0018] a3. Calculate the global characteristics of the performance attributes of all valid satellites, including the global mean, global standard deviation, and global range, and combine the global characteristics of each attribute of the performance attributes of all valid satellites into satellite-level characteristics.

[0019] Preferably, the training method for the constellation distribution prediction model is as follows:

[0020] B1. Collect historical GNSS observation data that meet the preset conditions. The preset conditions are the scenarios in which ARAIM is successfully running. Each scenario consists of all observation values ​​at a specific time and location. The continuous data is divided into multiple historical scenario slices according to the preset time step.

[0021] B2. Based on each historical scene slice, generate the corresponding multi-level feature vector and the constellation ratio vector of the satellites actually selected for the corresponding historical scene slice in ARAIM. Use the constellation ratio vector to label the multi-level feature vector.

[0022] B3. Using all labeled multi-level feature vectors as the training set, train the pre-selected neural network structure, continuously optimize the model parameters, and generate the final constellation distribution prediction model.

[0023] Preferably, S102 specifically includes:

[0024] S201, classify the valid satellite set into the corresponding constellation, and obtain the satellite subset [S1, S2, ..., S] of each constellation. i ,...,S n ], S i For a subset of satellites in the i-th constellation, count the number of satellites s in that subset. i n is the total number of constellation types;

[0025] S202, Calculate the number of selected stars for each constellation based on the optimal constellation ratio vector and the preset number of selected stars:

[0026] k i =[p i ×s i ]

[0027] Where, k i p represents the number of stars selected for the i-th constellation. i Let s represent the proportion of the i-th constellation in the candidate satellite group. i This represents the number of satellites in a subset of the constellation's satellites, with [] indicating rounding down;

[0028] S203, If the total number of selected stars for all constellations is less than the preset number of selected stars, the preset replacement mechanism is used to update the number of selected stars for each constellation so that the total number of selected stars for all constellations is equal to the preset number of selected stars.

[0029] S204: Based on the satellite subsets of each constellation, generate all candidate satellite subsets for that constellation according to the number of selected satellites, and obtain candidate satellite groups based on the combination of candidate satellite subsets of all constellations.

[0030] Preferably, the preset replacement mechanism is set as follows:

[0031] Calculate the number of satellites to be replaced: L = NK, where N is the preset number of satellites selected, K is the total number of satellites selected for all constellations, and L is the number of satellites to be replaced.

[0032] Arrange the constellations according to their corresponding (p) i ×s i -k i Sort the constellations in descending order of size, select the top L constellations and increase the number of selected constellations by one, then update the number of selected constellations for each constellation.

[0033] Preferably, the preset multi-level collaborative scoring mechanism specifically includes:

[0034] S301, input the performance attributes of each satellite in the candidate satellite group into the preset satellite individual performance comprehensive scoring model, and output the first score value of the candidate satellite group;

[0035] S302, input the candidate satellite group into the preset satellite synergy comprehensive scoring model, and output the second score value of the candidate satellite group;

[0036] S303, input the candidate satellite group into the preset fault tendency prediction model, and output the fault tendency factor of the candidate satellite group;

[0037] S304. Based on the first score, the second score, and the fault tendency factor, the comprehensive score of the candidate satellite group is obtained.

[0038] Preferably, the preset satellite individual performance comprehensive scoring model is as follows:

[0039]

[0040] w1 + w2 + w3 = 1

[0041] Where Score1 is the first score of the candidate satellite group, and L is the total number of satellites in the candidate satellite group. Let be the normalized value of the user ranging error for the k-th satellite. The normalized value of the elevation angle of the k-th satellite. Let P′ be the normalized signal-to-noise ratio of the k-th satellite. k The degree of fit of the constellation ratio vector of this candidate satellite group is represented by w1, w2, and w3, which are the influence values ​​of user ranging error, elevation angle, and signal-to-noise ratio on the first score, respectively. These values ​​are set based on expert experience and actual conditions. P′ k The degree of fit between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector is set as the similarity value between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector.

[0042] The preset satellite synergy comprehensive scoring model is as follows:

[0043] S401, calculate the spatial geometric accuracy factor of the candidate satellite group according to the preset satellite distribution convex hull volume algorithm;

[0044] S402, normalize the spatial ensemble precision factor of the candidate satellite group to obtain the second score value Score2.

[0045] Preferably, S303 specifically includes: obtaining fault trend feature information of candidate satellite groups within a preset time window, inputting it into a preset fault trend prediction model, and outputting a fault trend factor; the fault trend feature information is obtained by: obtaining the change range of satellite-level features, constellation-level features, and environmental features of candidate satellite groups within a preset time window, determining them as fault trend feature information, and determining the change range based on the difference between the maximum slope value and the minimum slope value corresponding to the change curve of each parameter within the preset time window;

[0046] The fault trend prediction model includes an input terminal, several fault trend prediction sub-models, a probability statistics module, and an output terminal, specifically used for:

[0047] The input end receives the actual constellation proportion vector and fault trend feature information of the input candidate satellite group, calculates the similarity between the actual constellation proportion vector of the input candidate satellite group and the central constellation proportion vector of each sub-model, and selects the sub-model with the highest similarity as the prediction path.

[0048] The input end inputs the fault trend characteristics of the candidate satellite group into the corresponding fault trend prediction sub-model according to the prediction path, and outputs the prediction results to the probability statistics module; the prediction results include 0 and 1, where 0 represents normal and 1 represents fault.

[0049] The probability and statistics module is used to output the fault tendency factor based on the prediction result. If the prediction result is 0, the fault tendency factor is determined to be 0. If the prediction result is 1, the similarity between the fault trend feature information and the sub-training set used for training the corresponding sub-model is obtained, and the maximum similarity value is determined as the fault tendency factor.

[0050] Preferably, S104 specifically includes:

[0051] S501: Based on the comprehensive score of all candidate satellite groups, the candidate satellite group with the highest comprehensive score is selected as the primary combination;

[0052] S502, sort the remaining candidate satellite groups in descending order of their comprehensive score, and select the candidate satellite groups in the top third as alternative satellite groups;

[0053] S503, sort the candidate satellite groups in descending order of their failure tendency factors, and select the candidate satellite group in the top third as the backup combination;

[0054] S504, according to a preset time interval, steps S101 to S104 are executed periodically. In each cycle, the fault tendency factor of the primary combination is monitored and updated in real time, and the change curve of the fault tendency factor is generated in real time. If the slope value of the change curve of the fault tendency factor is detected to be greater than the preset change rate threshold, the switching mechanism is triggered to switch the primary combination to the standby combination.

[0055] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0056] By optimizing the constellation ratio, geometric accuracy, failure probability, and monitoring thresholds can be balanced, thereby improving system safety and reliability. The hierarchical strategy significantly reduces algorithm complexity; by controlling the constellation ratio, it reduces the number of combinations and improves star selection efficiency, making it suitable for scenarios with high real-time requirements.

[0057] By introducing an optimal constellation proportion vector to accurately characterize system operation features, and combining it with a multi-level collaborative scoring mechanism to comprehensively evaluate the system status from multiple dimensions, a solid basis for subsequent decision-making is provided. Specifically, clustering the training data of the fault tendency prediction model based on the constellation proportion vector enables the rational allocation and efficient utilization of model resources, enhancing the model's adaptability to different scenarios. Simultaneously, the fault tendency prediction model proactively identifies potential system failure risks, allowing for preventative measures, while the switching mechanism ensures rapid and accurate switching between operating modes when system status changes or failures occur, guaranteeing stable system operation. The synergistic effect of these key technologies significantly improves system operational stability, fault response capabilities, and resource utilization efficiency, substantially reducing system failure rates and maintenance costs, and effectively ensuring the efficient and reliable operation of business.

[0058] This technology effectively solves three core challenges in multi-constellation satellite navigation environments: satellite selection combination explosion, sensitivity to environmental interference, and lagging fault monitoring. First, it uses a constellation distribution prediction model to generate the optimal constellation ratio, compressing the exponential combination search into a finite candidate set under proportional constraints, thus improving decision-making efficiency. Second, it simultaneously optimizes individual satellite accuracy (URE / elevation angle / SNR), spatial geometric distribution (convex hull volume factor), and fault tendency (real-time risk prediction) through a multi-level scoring mechanism, thereby improving positioning accuracy. Finally, it achieves seamless switching based on primary / backup combination decision-making and fault factor slope monitoring, ensuring continuous and reliable positioning in complex environments (such as ionospheric disturbances and extreme weather), especially meeting the integrity requirements in safety-critical scenarios such as aviation and autonomous driving. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating the satellite selection method for advanced receiver autonomous integrity monitoring according to an embodiment of the present invention. Detailed Implementation

[0060] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0061] It should be noted that the terms "vertical," "horizontal," "up," "down," "left," "right," and similar expressions used in this article are for illustrative purposes only and do not represent the only possible implementation.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0063] Example 1: Figure 1 This is a schematic flowchart of a satellite selection method for advanced receiver autonomous integrity monitoring according to an embodiment of the present invention.

[0064] like Figure 1 As shown, a satellite selection method for advanced receiver autonomous integrity monitoring includes the following steps:

[0065] S101: Obtain the multi-level feature vector of the current effective satellite set, including satellite-level features, constellation-level features and environmental features, input it into the pre-trained constellation distribution prediction model, and output the current optimal constellation ratio vector.

[0066] Among them, satellites with an elevation angle (the angle of the satellite relative to the receiver) greater than 5° are defined as the valid satellite set.

[0067] In some embodiments, the multi-level feature vectors are obtained as follows:

[0068] A1. Generate satellite-level features and constellation-level features based on the performance attributes of each valid satellite. Performance attributes include, but are not limited to, user ranging error, satellite elevation angle, and signal-to-noise ratio.

[0069] Specifically, satellite-level features and constellation-level features are generated based on the performance attributes of each valid satellite, including:

[0070] a1. Based on the performance attributes of each valid satellite, group them according to their constellation affiliation and classify them into the corresponding constellation. Calculate the statistical characteristics of the performance attributes within each constellation, including the mean, standard deviation, range, and the proportion of high-elevation satellites.

[0071] The performance attributes are represented as [URE, el, SNR]. URE is the user ranging error, i.e., the satellite ranging accuracy error, reflecting the satellite's ranging accuracy; el is the satellite elevation angle, i.e., the elevation angle of the satellite relative to the receiver. Low-elevation satellite signals are more susceptible to atmospheric and multipath effects; SNR is the signal-to-noise ratio, i.e., signal strength, reflecting anti-interference capability; the proportion of high-elevation satellites is set as the ratio of the number of high-elevation satellites to the total number of satellites in the constellation. When the elevation angle is greater than a preset elevation angle threshold, it is determined to be a high-elevation satellite. It should be noted that if a certain constellation type of satellite does not exist in the current valid satellite set, then all statistical characteristics of that constellation can be assigned 0.

[0072] a2. Combine the mean, standard deviation, and range of each performance attribute within each constellation, along with the proportion of high-elevation satellites in that constellation, to form constellation-level characteristics.

[0073] Constellation-level characteristics reflect the group performance, internal consistency, and available satellite distribution structure of each individual constellation at the current moment and location, capturing the key differences between different constellations.

[0074] a3. Calculate the global characteristics of the performance attributes of all valid satellites, including the global mean, global standard deviation, and global range, and combine the global characteristics of each attribute of the performance attributes of all valid satellites into satellite-level characteristics.

[0075] Satellite-level characteristics represent the global statistical properties of all currently effective satellites, reflecting the overall performance and distribution of all available satellites.

[0076] A2. Environmental characteristics are determined by the ionospheric disturbance index TEC and weather codes W. code composition.

[0077] The ionospheric disturbance index is obtained by real-time download of the Global Ionospheric Map (GIM), and the weather code is set to a digital code. For example, 0 = sunny, 1 = rain, and 2 = snow.

[0078] A3. Concatenate satellite-level features, constellation-level features, and environmental features sequentially to form a multi-level feature vector.

[0079] In some embodiments, the constellation distribution prediction model is trained as follows:

[0080] B1. Collect historical GNSS observation data that meet the preset conditions. The preset conditions are scenarios where ARAIM is running successfully. Each scenario consists of all observations at a specific time and location. Divide the continuous data into multiple historical scenario slices according to the preset time step (e.g., 2 minutes). The actual star selection group corresponding to the continuous data in each historical scenario slice must be the same.

[0081] Among them, the scenarios in which ARAIM runs successfully are those that meet the integrity conditions (positioning error is within the protection level range, horizontal protection level HPL < horizontal alarm limit HAL, vertical protection level VPL < vertical alarm limit VAL), and these scenarios represent reliable positioning results.

[0082] Specifically, historical GNSS observation data comes from actual receiver records at different locations, times, and under different environmental conditions (such as cities, suburbs, canyons, open areas, etc.). It includes valid satellite observations (including satellite identifier, constellation type, user ranging error URE, elevation angle, signal-to-noise ratio SNR) and corresponding environmental information (location, time, ionospheric index, sky obscuration, etc.). These historical scenarios must be highly intact, i.e., scenarios verified by ARAIM (positioning error within the protection level, HPL HAL, VPL). <VAL)。

[0083] B2. Based on each historical scene slice, generate corresponding multi-level feature vectors and ARAIM constellation ratio vectors of the satellites actually selected for the corresponding historical scene slice, and use the constellation ratio vectors to label the multi-level feature vectors.

[0084] The constellation ratio vector includes the ratio of the number of selected satellites to the actual number of selected satellites in each constellation. If a constellation has no selected satellites, the value is set to 0.

[0085] B3. Using all labeled multi-level feature vectors as the training set, train the pre-selected neural network structure, continuously optimize the model parameters, and generate the final constellation distribution prediction model.

[0086] S102, based on the optimal constellation ratio vector and the effective satellite set, generates several candidate satellite groups, each of which includes at least one satellite and at least one constellation.

[0087] For example, the optimal constellation ratio vector is represented as P = [p1, p2, ..., p...]. i ,...,p n ], p i Let be the proportion that the i-th constellation should occupy in the candidate satellite group, and n be the total number of constellation types.

[0088] Specifically, step S102 includes:

[0089] S201, classify the valid satellite set into the corresponding constellation, and obtain the satellite subset [S1, S2, ..., S] of each constellation. i ,...,S n ], S i For a subset of satellites in the i-th constellation, count the number of satellites s in that subset. i .

[0090] S202, based on the optimal constellation ratio vector and the preset number of selected stars (a preliminary standard is set based on actual requirements and experience, for example, 8 stars), calculate the number of selected stars for each constellation:

[0091] k i =[p i ×s i ]

[0092] Where, k i p represents the number of stars selected for the i-th constellation. i Let s represent the proportion of the i-th constellation in the candidate satellite group. i This represents the number of satellites in the subset of satellites in this constellation, with [] indicating rounding down.

[0093] S203, if the total number of selected stars for all constellations is less than the preset number of selected stars, use the preset replacement mechanism to update the number of selected stars for each constellation so that the total number of selected stars for all constellations equals the preset number of selected stars.

[0094] Specifically, the preset replacement mechanism is set as follows:

[0095] Calculate the number of satellites to be replaced: L = NK, where N is the preset number of satellites selected, K is the total number of satellites selected for all constellations, and L is the number of satellites to be replaced.

[0096] Arrange the constellations according to their corresponding (p) i ×s i -k i Sort the constellations in descending order of size, select the top L constellations and increase the number of selected constellations by one, then update the number of selected constellations for each constellation.

[0097] S204: Based on the satellite subsets of each constellation, generate all candidate satellite subsets for that constellation according to the number of selected satellites, and obtain candidate satellite groups based on the combination of candidate satellite subsets of all constellations.

[0098] Specifically, the number of candidate satellite subsets for the i-th constellation is For each candidate satellite subset, it is combined with candidate satellite subsets from other constellations to form a candidate satellite group. For example, if there are n constellations, each constellation has C... i A subset of candidate satellites can then generate a total of A group of candidate satellites.

[0099] S103 utilizes a pre-defined multi-level collaborative scoring mechanism to generate a comprehensive score and fault tendency factor for each candidate satellite group.

[0100] In some embodiments, the preset multi-level collaborative scoring mechanism specifically includes:

[0101] S301: Input the performance attributes of each satellite in the candidate satellite group into the preset satellite individual performance comprehensive scoring model, and output the first score value of the candidate satellite group.

[0102] Specifically, the pre-defined comprehensive performance evaluation model for individual satellites is as follows:

[0103]

[0104] w1 + w2 + w3 = 1

[0105] Where Score1 is the first score of the candidate satellite group, and L is the total number of satellites in the candidate satellite group. Let be the normalized value of the user ranging error for the k-th satellite. The normalized value of the elevation angle of the k-th satellite. Let P′ be the normalized signal-to-noise ratio of the k-th satellite. k The degree of fit of the constellation ratio vector of the candidate satellite group is represented by w1, w2, and w3, which are the influence values ​​of user ranging error, elevation angle, and signal-to-noise ratio on the first score value, respectively, and are set according to expert experience and actual conditions.

[0106] Among them, P′ k The degree of fit between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector is set as the similarity value between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector.

[0107] It should be noted that:

[0108] The normalized value for user ranging error is set as follows: URE k For the user ranging error of the k-th satellite in the candidate constellation group, URE max The maximum user ranging error in the candidate constellation group, URE min This represents the minimum user ranging error in the candidate constellation group. URE reflects the accuracy error of satellite ranging; the smaller the value, the higher the satellite ranging accuracy. This formula maps the URE value to the interval [0, 1]. The smaller the URE, the larger the normalized value, indicating better satellite ranging performance.

[0109] The normalized elevation angle value is set as follows: el k Let el be the elevation angle of the k-th satellite in the candidate constellation group. max The maximum elevation angle among the candidate constellation groups, el min This represents the minimum elevation angle among the candidate constellation groups. The larger the satellite elevation angle, the less the signal is affected by atmospheric and multipath effects, resulting in better positioning performance. A larger el value indicates a higher normalized value and better satellite positioning performance.

[0110] The signal-to-noise ratio normalization value is set as follows: SNR k Let SNR be the signal-to-noise ratio of the k-th satellite in the candidate constellation group. maxThe maximum signal-to-noise ratio (SNR) in the candidate constellation group. min This represents the minimum signal-to-noise ratio (SNR) among the candidate constellation groups. SNR reflects the signal strength; a higher SNR indicates stronger anti-interference capability and a larger normalized value.

[0111] S302, input the candidate satellite group into the preset satellite synergy comprehensive scoring model, and output the second score value of the candidate satellite group.

[0112] Specifically, the pre-set comprehensive satellite synergy scoring model is as follows:

[0113] S401, based on the preset satellite distribution convex hull volume algorithm, calculates the spatial geometric accuracy factor of the candidate satellite group.

[0114] The preset satellite distribution convex hull volume algorithm specifically includes:

[0115] C1. With the receiver position as the origin, the position vector of the i-th satellite in the candidate satellite group relative to the receiver is r. i =(x i ,y i ,z i Let i = 1, 2, ..., L. Using convex hull algorithms in computational geometry (such as Graham scan algorithm, QuickHull algorithm, etc.), we can obtain the convex hull containing all satellite points in the candidate constellation group. The convex hull is the smallest convex polyhedron containing all satellite points. In three-dimensional space, the convex hull is composed of at least one tetrahedron, and the vertices of the tetrahedron are the points corresponding to the satellite position vectors in the candidate satellite group.

[0116] C3. Calculate the convex hull volume using the following formula:

[0117]

[0118] Where V is the volume of the convex hull, M is the number of tetrahedrons in the convex hull, and V m The volume of the m-th tetrahedron is obtained as follows:

[0119] The indices of the satellites corresponding to the four vertices of the m-th tetrahedron in the candidate satellite group are obtained as o1, o2, o3, and o4 (o1, o2, o3, and o4 are in the range of 1 to L), and their corresponding position vectors are respectively...

[0120] i, j, and k are the unit vectors in the x, y, and z directions, respectively.

[0121] C4. Use the convex hull volume as the spatial geometric accuracy factor for this candidate satellite group.

[0122] The larger the convex hull volume, the better the satellite geometry distribution (lower DOP value), which directly improves positioning accuracy and avoids satellites being concentrated in a single azimuth angle in obstructed environments such as urban areas and canyons.

[0123] S402, normalize the spatial ensemble precision factor of the candidate satellite group to obtain the second score value Score2.

[0124] S303 inputs the candidate satellite group into the preset fault tendency prediction model and outputs the fault tendency factor of the candidate satellite group.

[0125] Specifically, step S303 includes: obtaining fault trend characteristic information of candidate satellite groups within a preset time window, inputting it into a preset fault trend prediction model, and outputting fault trend factors.

[0126] The fault trend characteristic information is obtained by obtaining the change range of satellite-level characteristics, constellation-level characteristics and environmental characteristics of the candidate satellite group within a preset time window, which is determined as fault trend characteristic information. The change range can be determined based on the difference between the maximum slope value and the minimum slope value corresponding to the change curve of each parameter within the preset time window. The change range of satellite-level characteristics is composed of the difference between the maximum slope value and the minimum slope value corresponding to the change curve of all parameters within the preset time window. The same applies to constellation-level characteristics and environmental characteristics, which will not be elaborated further.

[0127] In some embodiments, the training method for the preset fault tendency prediction model is as follows:

[0128] D1. Collect historical GNSS observation data under historical fault scenarios and normal operation, ensuring that the data covers performance attributes such as satellite identification, constellation type, user ranging error (URE), elevation angle, signal-to-noise ratio (SNR), as well as environmental information and fault records, to obtain several historical scene slices.

[0129] D2. Obtain the fault trend feature information of each historical scene slice within a preset time window (e.g., set to the past 5 minutes), and based on its constellation proportion vector, use a preset clustering algorithm (e.g., K-means) to cluster the fault trend feature information of the historical scene slices, classify scenes with similar constellation distributions into the same category, and obtain several clusters. Each cluster includes at least one fault trend feature information and corresponds to a sub-training set.

[0130] Specifically, the fault trend feature information is obtained by obtaining the change range of satellite-level features, constellation-level features, and environmental features of the corresponding historical scene slice within a preset time window, which is then determined as fault trend feature information. The change range can be determined based on the difference between the maximum slope value and the minimum slope value corresponding to the change curve of each parameter within the preset time window. The change range of satellite-level features is composed of the difference between the maximum slope value and the minimum slope value corresponding to the change curve of all parameters within the preset time window. The same applies to constellation-level features and environmental features, which will not be elaborated further.

[0131] Latent failure modes (such as single-constellation collective bias) are identified by clustering sub-models (grouped by constellation distribution).

[0132] D3. Based on each sub-training set, label each fault trend feature information in it, and set the label content to 0 (normal) and 1 (fault).

[0133] D4. Using each sub-training set, train the pre-selected neural network structure, continuously optimize the model parameters, and obtain the fault trend prediction sub-model. Each sub-model is assigned a guide label, which is set as the central constellation proportion vector of the corresponding cluster, that is, the average value of the constellation proportion vector in the cluster.

[0134] D5. Combine several fault tendency prediction sub-models into a fault tendency prediction model, which is used to determine the prediction path based on the actual constellation proportion vector of the input candidate satellite group and output the corresponding fault tendency factor.

[0135] Specifically, the fault tendency prediction model includes an input, several fault tendency prediction sub-models, a probability and statistics module, and an output, which is used for:

[0136] The input end receives the actual constellation proportion vector and fault trend feature information of the input candidate satellite group, calculates the similarity between the actual constellation proportion vector of the input candidate satellite group and the central constellation proportion vector of each sub-model, and selects the sub-model with the highest similarity as the prediction path.

[0137] The input end inputs the fault trend feature information of the candidate satellite group into the corresponding fault trend prediction sub-model according to the prediction path, and outputs the prediction results to the probability statistics module.

[0138] The probability and statistics module is used to output the fault tendency factor based on the prediction result. If the prediction result is 0, the fault tendency factor is determined to be 0. If the prediction result is 1, the similarity between the fault trend feature information and the sub-training set of the corresponding sub-model is obtained, and the maximum similarity value is determined as the fault tendency factor.

[0139] S304. Based on the first score, the second score, and the fault tendency factor, the comprehensive score of the candidate satellite group is obtained.

[0140] Specifically, the overall score of the candidate satellite group is calculated according to the following formula:

[0141]

[0142] Wherein, Score is the comprehensive score of the candidate satellite group, Score1 is the first score, Score2 is the second score, F is the fault tendency factor, and e1, e2, and e3 are preset weight factors, which are used to represent the degree of influence of the first score, the second score, and the fault tendency factor on the comprehensive score, respectively. They are set according to expert experience or actual needs. The greater the degree of influence, the larger the weight factor, and e1+e2+e3=1.

[0143] The first score represents the overall performance score of the individual satellite, reflecting the performance of each satellite in the candidate satellite group. The second score represents the overall performance score of satellite synergy, reflecting the spatial geometric distribution quality of the candidate satellite group. The fault tendency factor reflects the probability of the candidate satellite group failing in the future; the smaller the value, the lower the probability of failure.

[0144] S104, based on the comprehensive scores and fault tendency factors of all candidate satellite groups, uses the preset primary and backup combination satellite selection decision to generate the primary combination and its backup strategy.

[0145] In some embodiments, step S104 specifically includes:

[0146] S501 selects the candidate satellite group with the highest comprehensive score as the primary combination based on the comprehensive score of all candidate satellite groups.

[0147] S502, sort the remaining candidate satellite groups in descending order of their comprehensive scores, and select the top third of the candidate satellite groups as alternative satellite groups.

[0148] S503, sort the candidate satellite groups in descending order of their failure tendency factors, and select the candidate satellite group in the top third as the backup combination.

[0149] S504, according to a preset time interval, steps S101 to S104 are executed periodically. In each cycle, the fault tendency factor of the primary combination is monitored and updated in real time, and the change curve of the fault tendency factor is generated in real time. If the slope value of the change curve of the fault tendency factor is detected to be greater than the preset change rate threshold (set according to expert experience, used to measure the degree of change of the fault tendency factor), the switching mechanism is triggered to switch the primary combination to the standby combination.

[0150] In another embodiment, within each cycle, the vertical and horizontal protection levels under the current selection combination conditions can be calculated in real time and compared with preset standard thresholds (such as the LPV-200 standard) to determine whether the current selection combination meets the safety and availability requirements of the navigation system, i.e., whether ARAIM is available. If it does not meet the requirements, a switching mechanism is triggered. If the backup combination still does not meet the requirements, the process jumps directly to a new cycle to reselect the primary combination. It should be noted that the determination of the availability of the current ARAIM can be referred to the descriptions in relevant prior art, and this invention will not elaborate on them.

[0151] In summary, satellite performance (URE, elevation angle, SNR), constellation statistical characteristics (mean / standard deviation / percentage of high-elevation satellites), and environmental factors (ionospheric TEC, weather) are integrated into a unified feature vector. This vector encompasses satellite-level features (reflecting individual satellite performance), constellation-level features (reflecting overall constellation performance), and environmental features (describing external interference factors). This provides a rich data foundation for subsequent model analysis. Utilizing a pre-trained constellation distribution prediction model (trained from historical observation data validated by ARAIM, ensuring the proportion vector implicitly contains integrity conditions), the optimal constellation proportion for the current moment is predicted based on multi-level feature vectors. This determines the proportion of each constellation in the selected satellite combination, making the prediction results more closely aligned with real-world scenarios and avoiding the static limitations of traditional geometric configuration-based satellite selection. Therefore, by clearly defining the proportion of each constellation in the selected satellite combination, blind satellite selection is avoided. The exponentially growing satellite combination problem is transformed into a finite combination search constrained by proportions, improving satellite selection efficiency. Furthermore, considering the impact of environmental factors on satellite signals makes the selection results more robust, maintaining good positioning performance under different environments. The optimal constellation ratio vector is an important basis for generating candidate satellite groups, ensuring that the generated candidate satellite groups conform to the optimal constellation distribution under the current environment.

[0152] The effective satellite set is categorized by constellation, resulting in satellite subsets for each constellation. This facilitates subsequent satellite selection based on the optimal constellation ratio. The number of satellites to be selected for each constellation is calculated based on the optimal constellation ratio vector and a preset selection quantity. When the total number of selected satellites for all constellations is less than the preset quantity, a replacement mechanism is used to update the selection quantity for each constellation, ensuring the required number of satellites is met. Based on the satellite subsets and selection quantity for each constellation, all candidate satellite subsets for that constellation are generated, and these subsets are combined to obtain candidate satellite groups. Thus, by rationally allocating the number of selected satellites for each constellation according to the optimal constellation ratio and the preset selection quantity, the diversity of satellite selection combinations is ensured. The replacement mechanism can handle situations where the number of selected satellites is insufficient, ensuring that the number of generated candidate satellite groups meets the subsequent screening requirements. By combining candidate satellite subsets from different constellations, a large number of candidate satellite groups are generated, providing more choices for subsequent screening. A rich selection of candidate satellite groups increases the probability of finding the optimal satellite combination, thereby improving positioning performance.

[0153] The performance attributes of each satellite in the candidate satellite group are input into the satellite individual performance comprehensive scoring model, outputting a first score value for the candidate satellite group, reflecting the individual satellite performance. The candidate satellite group is input into the satellite synergy comprehensive scoring model, and a second score value is obtained by calculating and normalizing the spatial geometric precision factor, reflecting satellite synergy. Fault trend characteristics of the candidate satellite group within a preset time window are obtained and input into the fault trend prediction model, outputting a fault trend factor, reflecting the probability of the candidate satellite group failing in the future. Based on the first score value, the second score value, and the fault trend factor, a comprehensive score for the candidate satellite group is calculated according to preset weights. Thus, the multi-level collaborative scoring mechanism evaluates the candidate satellite group from three aspects: individual satellite performance, satellite synergy, and fault trend. The evaluation results are more comprehensive and accurate. By calculating the score value and the fault trend factor, the performance and fault risk of the candidate satellite group are quantified, facilitating comparison and screening. It can screen out candidate satellite groups with excellent performance and low fault risk, providing a high-quality selection for the final satellite selection.

[0154] The candidate satellite group with the highest comprehensive score is selected as the primary combination to ensure its performance advantage. By selecting alternative and backup satellite groups, a robust backup mechanism is established to improve system reliability and security. Real-time monitoring of the primary combination's failure tendency factors allows for timely detection of potential failure risks, and a dynamic switching mechanism ensures the system always uses the highest-performing satellite combination. The backup mechanism and real-time monitoring and switching functions effectively address failures in the primary combination, ensuring the stable operation of the satellite navigation system. Continuously using a high-performing satellite combination improves positioning accuracy and reliability, meeting the needs of various application scenarios. The primary / backup switching mechanism ensures that a single point of failure does not affect positioning continuity (crucially applicable to autopilot and aircraft landing).

[0155] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0156] By optimizing the constellation ratio, geometric accuracy, failure probability, and monitoring thresholds can be balanced, thereby improving system safety and reliability. The hierarchical strategy significantly reduces algorithm complexity; by controlling the constellation ratio, it reduces the number of combinations and improves star selection efficiency, making it suitable for scenarios with high real-time requirements.

[0157] By introducing an optimal constellation proportion vector to accurately characterize system operation features, and combining it with a multi-level collaborative scoring mechanism to comprehensively evaluate the system status from multiple dimensions, a solid basis for subsequent decision-making is provided. Specifically, clustering the training data of the fault tendency prediction model based on the constellation proportion vector enables the rational allocation and efficient utilization of model resources, enhancing the model's adaptability to different scenarios. Simultaneously, the fault tendency prediction model proactively identifies potential system failure risks, allowing for preventative measures, while the switching mechanism ensures rapid and accurate switching between operating modes when system status changes or failures occur, guaranteeing stable system operation. The synergistic effect of these key technologies significantly improves system operational stability, fault response capabilities, and resource utilization efficiency, substantially reducing system failure rates and maintenance costs, and effectively ensuring the efficient and reliable operation of business.

[0158] This technology effectively solves three core challenges in multi-constellation satellite navigation environments: satellite selection combination explosion, sensitivity to environmental interference, and lagging fault monitoring. First, it uses a constellation distribution prediction model to generate the optimal constellation ratio, compressing the exponential combination search into a finite candidate set under proportional constraints, thus improving decision-making efficiency. Second, it simultaneously optimizes individual satellite accuracy (URE / elevation angle / SNR), spatial geometric distribution (convex hull volume factor), and fault tendency (real-time risk prediction) through a multi-level scoring mechanism, thereby improving positioning accuracy. Finally, it achieves seamless switching based on primary / backup combination decision-making and fault factor slope monitoring, ensuring continuous and reliable positioning in complex environments (such as ionospheric disturbances and extreme weather), especially meeting the integrity requirements in safety-critical scenarios such as aviation and autonomous driving.

[0159] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A satellite selection method for advanced receiver autonomous integrity monitoring, characterized in that, include: S101: Obtain the multi-level feature vector of the current effective satellite set, including satellite-level features, constellation-level features and environmental features, input it into the pre-trained constellation distribution prediction model, and output the current optimal constellation ratio vector; S102, based on the optimal constellation ratio vector and the effective satellite set, generate several candidate satellite groups, each candidate satellite group including at least one satellite and at least one constellation; S103, using a pre-set multi-level collaborative scoring mechanism, generate a comprehensive score and fault tendency factor for each candidate satellite group; S301, input the performance attributes of each satellite in the candidate satellite group into a pre-set satellite individual performance comprehensive scoring model, and output the first score value of the candidate satellite group. The performance attributes include user ranging error, satellite elevation angle, and signal-to-noise ratio; S302, input the candidate satellite group into a pre-set satellite collaborative comprehensive scoring model, and output the second score value of the candidate satellite group; S303, input the candidate satellite group into a pre-set fault tendency prediction model, and output the fault tendency factor of the candidate satellite group; S304, based on the first score value, the second score value, and the fault tendency factor, obtain the comprehensive score of the candidate satellite group; S104, based on the comprehensive scores and fault tendency factors of all candidate satellite groups, uses the preset primary and backup combination satellite selection decision to generate the primary combination and its backup strategy.

2. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 1, characterized in that, The method for obtaining the multi-level feature vectors is as follows: A1. Generate satellite-level and constellation-level features based on the performance attributes of each valid satellite; A2. Environmental characteristics consist of the ionospheric disturbance index and weather codes; A3. Concatenate satellite-level features, constellation-level features, and environmental features sequentially to form a multi-level feature vector.

3. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 2, characterized in that, The generation of satellite-level and constellation-level features based on the performance attributes of each valid satellite includes: a1. Based on the performance attributes of each valid satellite, group them according to their constellation affiliation and classify them into the corresponding constellation. Calculate the statistical characteristics of the performance attributes within each constellation, including the mean, standard deviation, range, and proportion of high elevation angle satellites. a2. Combine the mean, standard deviation, and range of each performance attribute within each constellation, along with the proportion of high-elevation satellites in that constellation, to form constellation-level characteristics; a3. Calculate the global characteristics of the performance attributes of all valid satellites, including the global mean, global standard deviation, and global range, and combine the global characteristics of each attribute of the performance attributes of all valid satellites into satellite-level characteristics.

4. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 1, characterized in that, The training method for the constellation distribution prediction model is as follows: B1. Collect historical GNSS observation data that meet the preset conditions. The preset conditions are the scenarios in which ARAIM is successfully running. Each scenario consists of all observation values ​​at a specific time and location. The continuous data is divided into multiple historical scenario slices according to the preset time step. B2. Based on each historical scene slice, generate the corresponding multi-level feature vector and the constellation ratio vector of the satellites actually selected for the corresponding historical scene slice in ARAIM. Use the constellation ratio vector to label the multi-level feature vector. B3. Using all labeled multi-level feature vectors as the training set, train the pre-selected neural network structure, continuously optimize the model parameters, and generate the final constellation distribution prediction model.

5. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 1, characterized in that, S102 specifically includes: S201, classify the valid satellite set into the corresponding constellation, and obtain the satellite subset of each constellation. , For a subset of satellites in the i-th constellation, count the number of satellites in that subset. n is the total number of constellation types; S202, Calculate the number of selected stars for each constellation based on the optimal constellation ratio vector and the preset number of selected stars: ； in, Let be the number of stars selected for the i-th constellation. Let i be the proportion that the i-th constellation should occupy in the candidate satellite group. This represents the number of satellites in a subset of the constellation's satellites, with [] indicating rounding down; S203, If the total number of selected stars for all constellations is less than the preset number of selected stars, the preset replacement mechanism is used to update the number of selected stars for each constellation so that the total number of selected stars for all constellations is equal to the preset number of selected stars. S204: Based on the satellite subsets of each constellation, generate all candidate satellite subsets for that constellation according to the number of selected satellites, and obtain candidate satellite groups based on the combination of candidate satellite subsets of all constellations.

6. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 5, characterized in that, The preset replacement mechanism is set as follows: Calculate the number of satellites to be replaced: L=NK, N is the preset number of satellites selected, K is the total number of satellites selected for all constellations, and L is the number of satellites to be replaced. Arrange the constellations according to their corresponding Sort the constellations in descending order of size, select the top L constellations and increase their selection count by one, then update the selection count for each constellation.

7. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 3, characterized in that, The preset satellite individual performance comprehensive scoring model is as follows: ； ； in, The first score for the candidate satellite group is L, where L is the total number of satellites in the candidate satellite group. Let be the normalized value of the user ranging error for the k-th satellite. The normalized value of the elevation angle of the k-th satellite. Let be the normalized signal-to-noise ratio value of the k-th satellite. The degree of fit of the constellation scale vector for this candidate satellite group. These are the influence values ​​of user ranging error, elevation angle, and signal-to-noise ratio on the first score, set based on expert experience and actual conditions. The degree of fit between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector is set as the similarity value between the constellation ratio vector of the candidate satellite group and the optimal constellation ratio vector. The preset satellite synergy comprehensive scoring model is as follows: S401, calculate the spatial geometric accuracy factor of the candidate satellite group according to the preset satellite distribution convex hull volume algorithm; S402, normalize the spatial ensemble precision factor of the candidate satellite group to obtain the second score value. .

8. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 3, characterized in that, S303 specifically includes: obtaining fault trend feature information of candidate satellite groups within a preset time window, inputting it into a preset fault trend prediction model, and outputting fault trend factors; the fault trend feature information is obtained by: obtaining the change range of satellite-level features, constellation-level features and environmental features of candidate satellite groups within a preset time window, and determining it as fault trend feature information, the change range is determined according to the difference between the maximum slope value and the minimum slope value corresponding to the change curve of each parameter within the preset time window; The fault trend prediction model includes an input terminal, several fault trend prediction sub-models, a probability statistics module, and an output terminal, specifically used for: The input end receives the actual constellation proportion vector and fault trend feature information of the input candidate satellite group, calculates the similarity between the actual constellation proportion vector of the input candidate satellite group and the central constellation proportion vector of each sub-model, and selects the sub-model with the highest similarity as the prediction path. The input end inputs the fault trend characteristics of the candidate satellite group into the corresponding fault trend prediction sub-model according to the prediction path, and outputs the prediction results to the probability and statistics module; the prediction results include 0 and 1, where 0 represents normal and 1 represents fault. The probability and statistics module is used to output the fault tendency factor based on the prediction result. If the prediction result is 0, the fault tendency factor is determined to be 0. If the prediction result is 1, the similarity between the fault trend feature information and the sub-training set used for training the corresponding sub-model is obtained, and the maximum similarity value is determined as the fault tendency factor.

9. The satellite selection method for advanced receiver autonomous integrity monitoring as described in claim 7, characterized in that, S104 specifically includes: S501: Based on the comprehensive score of all candidate satellite groups, the candidate satellite group with the highest comprehensive score is selected as the primary combination; S502, sort the remaining candidate satellite groups in descending order of their comprehensive score, and select the candidate satellite groups in the top third as alternative satellite groups; S503, sort the candidate satellite groups in descending order of their failure tendency factors, and select the candidate satellite group in the top third as the backup combination; S504, according to a preset time interval, steps S101 to S104 are executed periodically. In each cycle, the fault tendency factor of the primary combination is monitored and updated in real time, and the change curve of the fault tendency factor is generated in real time. If the slope value of the change curve of the fault tendency factor is detected to be greater than the preset change rate threshold, the switching mechanism is triggered to switch the primary combination to the standby combination.

Images