Optimized multi-preamble satellite random access deep segmentation collision detection method

Abstract

The application belongs to the technical field of satellite communication, and discloses a satellite random access deep segmentation conflict detection method and system with optimized multiple roots, which screens important load sample features in a correlation result according to a cyclic shift offset value of an MRP under different conflict conditions, and performs conflict detection and load estimation on the correlation result of the current MRP in a segmented iteration manner according to the extracted features by embedding a pre-trained neural network; as a neural network in the embedded algorithm, a lightweight Transform architecture is further designed, which only integrates an attention module and a multi-layer perception module, so as to analyze the correlation of complex RA data samples with low overhead. Finally, an improved satellite random access deep segmentation conflict detection model based on multiple roots is constructed, which can be used for accurate conflict load prediction in high-dynamic large-scale satellite random access.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of satellite communication, and particularly relates to a satellite random access deep segmentation collision detection method and system for optimizing multiple preambles. BACKGROUND

[0002] In recent years, with the deployment of the fifth generation network, the sixth generation network will push wireless communication technology to a new height. Due to the explosive growth of intelligent devices and data traffic, the current terrestrial networks (TNs) far cannot meet the extensive access demand of global Internet of Things devices. On the contrary, satellite communication has been used to deploy non-terrestrial networks (NTN) to make up for the deficiency of TNs due to its ubiquitous service availability. In this regard, one of the main challenges is how to solve the influence of communication characteristics in high dynamic satellite scenarios on air transmission. Especially in the NTN scenario with complex channel environment, when a large number of devices access the satellite system concurrently, serious overload problems will occur, and serious preamble collisions will inevitably occur, resulting in that the devices causing these collisions cannot successfully establish a connection with the satellite network. Therefore, it is essential to design an efficient collision detection scheme for high dynamic satellite communication scenarios.

[0003] For implementing large-scale random access collision detection, the traditional method is to combine the preamble characteristics to design a threshold-based detection method. For example, the collision detection scheme based on the label preamble (T-eCD) uses the correlation results of the label preamble. This scheme can realize collision detection and load estimation in the first step of random access. Or the collision detection scheme based on a single preamble (S-eCD) first designs a single preamble sequence that is robust to noise, and based on this, uses a threshold to capture the preamble correlation results exceeding the threshold for collision detection. However, these methods face significant performance degradation in large-scale concurrent access scenarios or in high dynamic scenarios with complex channel environments. Therefore, it is necessary to improve the existing methods to realize efficient collision detection in high dynamic satellite scenarios.

[0004] Through the above analysis, the problems and defects of the prior art are:

[0005] In the NTN scenario with complex channel environment, when a large number of devices access the satellite system concurrently, serious overload problems will occur, and serious preamble collisions will inevitably occur, resulting in that the devices causing these collisions cannot successfully establish a connection with the satellite network. Especially in large-scale concurrent access scenarios or in high dynamic scenarios with complex channel environments, the correlation peak value corresponding to the time delay path will also exceed the threshold, resulting in that the existing threshold-based collision detection method is easily affected by the multipath effect and faces significant performance degradation. SUMMARY

[0006] In view of the problems existing in the prior art, the application provides a satellite random access deep segmentation collision detection method for optimizing multiple root preambles.

[0007] The application is implemented as follows: a satellite random access deep segmentation collision detection method for optimizing multiple root preambles comprises the following steps:

[0008] Step 1: Based on the characteristics of a high-dynamic satellite scene, a multiple root preamble (MRP) optimization method is designed to obtain an optimized MRP set capable of reducing the interference of multipath effects.

[0009] Step 2: Based on each optimized MRP, a related result data set containing different collision conditions and complex channel characteristics is constructed.

[0010] Step 3: For any MRP, a lightweight Transformer architecture integrating only one attention module and one multi-layer perception module is used to preliminarily train the entire data set of the MRP.

[0011] Step 4: The MRP related results are windowed according to the cyclic shift characteristic, and the cyclic shift value and the multipath time delay are combined to extract features from the related results in each window. A segmentation detection algorithm based on feature extraction is designed. The pre-trained Transformer network is embedded into the detection algorithm to obtain a satellite random access deep segmentation collision detection model based on optimized multiple root preambles.

[0012] Step 5: The satellite random access deep segmentation collision detection model based on optimized multiple root preambles is used to perform signal collision detection and load estimation on a large-scale access device in a high-dynamic satellite scene.

[0013] Further, the specific method in step 1 is as follows: the influence of the characteristics of a high-dynamic satellite scene on the MRP related results is analyzed, a cyclic shift configuration set of the MRP is designed in combination with the multipath time delay law, and the related results of the MRP are optimized to reduce the interference of multipath effects.

[0014] Further, the specific method for constructing the data set in step 2 is as follows: different collision conditions and channel models are set, and related results for each optimized MRP are generated as a data set based thereon.

[0015] Further, based on the constructed data set, the lightweight Transformer network architecture integrating one attention module and one multi-layer perception module in step 3 is used to preliminarily train the data set in step 2 for different collision conditions of the optimized MRP, so as to sufficiently learn the correlation between random access samples.

[0016] Further, the step 4 needs to window the optimized MRP correlation results according to the cyclic shift characteristic, and combine the correlation results in each window with the cyclic shift value and the multipath time delay for feature extraction, construct a segmented detection algorithm module based on feature extraction, and embed a pre-trained Transformer network into the detection algorithm.

[0017] Another object of the present application is to provide an optimized multi-root preamble satellite random access deep segmentation conflict detection system comprising:

[0018] An optimization module is used to design an MRP optimization method based on the characteristics of high dynamic satellite scenes, so as to obtain an optimized MRP set capable of reducing multipath effect interference;

[0019] A construction module is used to construct a correlation result data set containing different conflict situations and complex channel characteristics based on each optimized MRP;

[0020] A training module is used to preliminarily train the entire data set of the MRP by using a lightweight Transformer architecture integrating only one attention module and one multi-layer perception module for any MRP;

[0021] A feature extraction module is used to window the MRP correlation results according to the cyclic shift characteristic, and combine the correlation results in each window with the cyclic shift value and the multipath time delay for feature extraction, design a segmented detection algorithm based on feature extraction, and embed a pre-trained Transformer network into the detection algorithm to obtain an optimized multi-root preamble satellite random access deep segmentation conflict detection model;

[0022] A detection module is used to perform signal conflict detection and load estimation on a large-scale access device in a high dynamic satellite scene by using the optimized multi-root preamble satellite random access deep segmentation conflict detection model.

[0023] Another object of the present application is to provide a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to enable the processor to execute the steps of the optimized multi-root preamble satellite random access deep segmentation conflict detection method.

[0024] Another object of the present application is to provide a computer readable storage medium storing a computer program, and the computer program is executed by a processor to enable the processor to execute the steps of the optimized multi-root preamble satellite random access deep segmentation conflict detection method.

[0025] Another object of the present application is to provide an information data processing terminal for implementing the optimized multi-preamble satellite random access deep segmentation collision detection system.

[0026] In combination with the above technical solutions and the technical problems solved, the technical solutions of the present application have the following advantages and positive effects:

[0027] First, the present application aims to provide an effective solution that can achieve accurate collision detection in high dynamic satellite random access scenarios. Based on the optimized MRP, a collision detection method embedded with a lightweight Transformer network is designed. By inputting the relevant results of MRP into the network of the algorithm, accurate collision detection and load estimation can be achieved. This method first optimizes the MRP preamble, designs a cyclic shift configuration set combined with high dynamic characteristics, and reduces the interference of multipath effects on MRP. Then, based on each improved MRP, a correlation result dataset containing complex channel features is constructed under different collision conditions for network training. Finally, a lightweight Transformer architecture is constructed by integrating an attention module and a multi-layer perception module to embed the designed feature extraction-based segmentation collision detection method. A satellite random access deep segmentation collision detection model based on optimized multi-preambles is finally constructed, which is trained offline and used for efficient collision detection and load estimation in the online phase.

[0028] (1) The MRP optimized by high dynamic scene characteristics reduces the interference of multipath effects, avoids introducing a large amount of interference data into the dataset, and enables the neural network to fully establish the mapping relationship between the optimized MRP correlation results and the collision load.

[0029] (2) The constructed lightweight Transformer network architecture can fully learn the long-range dependencies in the dataset with lower overhead compared to traditional complex Transformer networks.

[0030] (3) The MRP correlation result window is used to fully exploit the internal regularity of the correlation result in each window under different collision conditions, and the cyclic shift position and time delay position are deduced for feature extraction. A segmentation collision detection method embedded with a pre-trained network is designed to improve the detection accuracy of the model in high dynamic satellite scenarios.

[0031] Second, as the creative auxiliary evidence of the claims of the present application, it is also reflected in the following important aspects:

[0032] (1) The expected revenue and commercial value of the technical solutions of the present application after transformation are:

[0033] Due to the limited coverage and services of the current TNs, which cannot meet the extensive access requirements of 6G global Internet of Things devices, deploying NTN is considered an effective solution. Considering the existence of a large number of devices distributed in the area covered by a satellite in the future, in the random access process, these devices will access the satellite base station with limited payload in a high dynamic NTN scenario, resulting in a serious decline in satellite access efficiency. The present application mainly tends to enable the satellite base station to perform accurate conflict detection for a large number of accessing devices in a complex channel environment of high dynamic scene, and provide accurate load information for subsequent conflict resolution steps, thereby improving the satellite access efficiency. The present application can ensure the accuracy of the satellite base station's conflict detection for a large number of users in a complex channel scenario, improve the communication quality in the NTN environment, enhance the resource utilization rate of the satellite, and help optimize the performance and user experience of the wireless communication system.

[0034] (2) The technical solution of the present application fills the gap in the domestic and foreign industry:

[0035] The satellite random access deep segmentation conflict detection method and system based on optimized multiple pilots designed by the present application can solve the gap that the existing traditional threshold-based detection method cannot accurately detect conflicts and estimate loads for a large number of devices in a high dynamic complex channel environment. Specifically, the existing traditional threshold-based conflict detection algorithm only relies on a predefined threshold to capture the relevant peaks of the conflict for conflict detection, and records the number of captured peaks as the current conflict load, resulting in a sharp deterioration of the detection performance of the traditional method in a high dynamic NTN scenario with a complex channel model. On the one hand, the relevant peaks will exceed the predefined threshold at the position of the time delay spread due to the influence of the multipath effect, and on the other hand, it is difficult to distinguish noise from actual relevant peaks under high noise conditions when a large number of users access. In summary, the technical solution of the present application starts from the characteristics of the pilot, optimizes the MRP to reduce the interference of the multipath effect, and combines the conflict detection technology with the deep learning technology based on this, designs a lightweight Transformer architecture with strong data correlation analysis capability and feature selection technology to avoid the above problems, realizes efficient conflict detection and load estimation for a large number of accessing devices in a high dynamic complex channel model, and thus fills the gap in the domestic and foreign industry.

[0036] (3) The technical solution of the present application solves the technical problems that people have been eager to solve but have always failed to succeed:

[0037] The technical scheme of the present application solves the technical problem that the existing satellite conflict detection and load estimation methods based on threshold at home and abroad cannot perform accurate conflict detection under the condition of massive device access in a high dynamic scene with a complex channel environment affected by multipath effect and high noise. Specifically, the traditional conflict detection method based on threshold is usually designed based on the preamble and relies on a predefined threshold to capture the correlation peak as the identification of the conflict device, but the correlation result as the basis for the above detection will be affected by multipath effect and high noise, resulting in the phenomenon of conflict peak interference with each other. On the contrary, the present application first optimizes the MRP based on the characteristics of the high dynamic scene and the cyclic shift configuration set of the MRP to reduce the influence of multipath effect. Secondly, the present application combines the cyclic shift characteristics of the optimized MRP to window the correlation result and designs a segmented detection algorithm based on feature extraction to avoid the data redundancy phenomenon introduced by conflict peak interference with each other. In summary, the present application solves the technical problem that the satellite base station cannot perform accurate conflict detection when massive devices access in the scene of multipath effect and high noise.

[0038] (4) The technical scheme of the present application overcomes the technical bias:

[0039] The present application overcomes the technical bias. Specifically, for the traditional conflict detection algorithm based on threshold, the received signal is correlated with the local preamble sequence, and the correlation peak is captured according to the threshold to perform conflict detection and obtain the load value. However, when using the traditional conflict detection algorithm based on threshold in a high dynamic satellite scene, due to the excessive dependence on the threshold setting and the data interference of the correlation peak caused by multipath effect and high noise, efficient and accurate conflict detection and load prediction cannot be performed. In order to overcome the problems existing in the above method, the present application is based on the design of optimized MRP, and uses a data-driven deep segmented conflict detection method based on a lightweight Transformer network architecture. By optimizing the cyclic shift configuration set of the MRP, and based on this, the correlation result is windowed and segmented feature extraction is performed, the Transformer network fully establishes the correlation between the conflict load and the target of each detection window, and finally the satellite random access deep segmented conflict detection method and system based on the optimized multiple preambles can perform efficient conflict detection and load estimation in the high dynamic satellite access scene and the high noise scene.

[0040] Thirdly, the present application solves the problem of inaccurate signal conflict detection in the high dynamic satellite scene in the prior art. Due to multipath effect and channel complexity, the existing technology often leads to unreliable signal conflict detection results, thereby affecting the overall performance of satellite communication. The present application designs an optimized multiple preamble (MRP) set, effectively reduces the multipath effect interference, and improves the accuracy and reliability of detection.

[0041] The present application has made significant progress in constructing a dataset containing different conflict situations and complex channel characteristics. By constructing a relevant result dataset based on an optimized MRP, various conflict situations and channel characteristics are covered, making the model training data more comprehensive and representative. This method greatly enhances the generalization ability of the model in complex satellite communication environments.

[0042] The introduction of a lightweight Transformer architecture combined with a feature extraction segmentation detection algorithm is a major innovation of the present application. The lightweight design using only one attention module and one multi-layer perception module not only reduces the demand for computing resources, but also improves the running efficiency of the model. At the same time, through feature extraction and segmentation detection algorithm, the detection result is more fine and accurate, effectively improving the detection performance.

[0043] Finally, the present application realizes efficient signal collision detection and load estimation for large-scale access devices in high dynamic satellite scenarios. Based on the optimized multi-root preamble and deep segmentation collision detection model, the present application not only improves the accuracy of signal collision detection, but also realizes accurate load estimation. These technical advances significantly improve the overall performance and reliability of satellite communication systems, providing strong technical support for future high dynamic satellite communication scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 is a flow chart of the satellite random access deep segmentation collision detection method provided by the embodiment of the present application.

[0045] Figure 2 is a block diagram of the satellite random access deep segmentation collision detection system structure provided by the embodiment of the present application.

[0046] Figure 3 is a satellite random access deep segmentation collision detection scheme diagram based on the optimized multi-root preamble provided by the embodiment of the present application.

[0047] Figure 4 is a design architecture diagram of the lightweight Transformer neural network provided by the embodiment of the present application.

[0048] Figure 5 is a segmentation detection algorithm flow chart based on feature extraction provided by the embodiment of the present application.

[0049] Figure 6 is a flow chart of online collision detection and load estimation provided by the embodiment of the present application.

[0050] Figure 7 is a conflict detection performance diagram in a high dynamic satellite scenario provided by the method of the embodiment of the present application.

[0051] Figure 8 is the load estimation performance chart of the method provided by the embodiment of the application in a high dynamic satellite scenario.

[0052] Figure 9 is the performance comparison chart of the method provided by the embodiment of the application with a traditional collision detection method under an AWGN channel model.

[0053] Figure 10 Load estimation accuracy in different collision situations in a high-density device deployment area in a city. DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of the application clearer, the application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application.

[0055] Embodiment one: application in a low-orbit satellite communication system

[0056] In a low-orbit satellite communication system, due to the high-speed movement of satellites and the multipath effect, signal collision and interference problems are particularly prominent. In order to improve the system performance, the following steps can be taken to implement the satellite random access deep segmentation collision detection method with optimized multiple pilots:

[0057] 1) MRP optimization: first, in view of the high dynamic characteristics of low-orbit satellites, an MRP optimization method is designed and implemented. This includes analyzing the influence of satellite orbit, speed, direction and other factors on the multipath effect, so as to select an MRP set that can reduce these interferences to the greatest extent.

[0058] 2) Construction of data set: then, based on each optimized MRP, data is collected in the actual low-orbit satellite communication environment, which contains various collision situations and complex channel characteristics. The data set can include signal strength, time delay, Doppler shift and other parameters.

[0059] 3) Preliminary training: a lightweight Transformer architecture is adopted, which only integrates one attention module and one multi-layer perception module, to preliminarily train the entire data set of each MRP. In this way, important features can be retained while reducing computational complexity and improving training efficiency.

[0060] 4) Feature extraction and segmentation detection algorithm design: the MRP correlation results are windowed using the cyclic shift property, and the correlation results in each window are extracted based on the cyclic shift value and the multipath time delay. Based on these features, a segmentation detection algorithm is designed, and the pre-trained Transformer network is embedded in the detection algorithm.

[0061] 5) Conflict detection and load estimation: Finally, the constructed satellite random access deep segmentation conflict detection model based on optimized multiple-preamble is utilized to perform signal conflict detection and load estimation for large-scale access devices in low-orbit satellite scenarios. This helps the system to adjust resource allocation in time and optimize communication performance.

[0062] Example Two: Application in High Dynamic Ocean Satellite Communication

[0063] In the ocean environment, satellite communication often faces high dynamic, variable climate conditions and complex electromagnetic environment. In order to achieve reliable communication in such an environment, the following steps can be taken to implement the satellite random access deep segmentation conflict detection method based on optimized multiple-preamble:

[0064] 1) MRP optimization: Considering the variability of the ocean environment, the first step is to design an MRP optimization method for this environmental characteristic. This includes comprehensive analysis of ocean climate, sea waves, electromagnetic interference and other factors to determine the optimal MRP set.

[0065] 2) Data collection: In the ocean environment, data sets containing different conflict situations and channel characteristics are collected through actual communication or simulation experiments. These data can reflect the signal propagation characteristics in the ocean environment.

[0066] 3) Model training: The data set is preliminarily trained using a lightweight Transformer architecture to extract key features and build a preliminary model.

[0067] 4) Feature extraction and algorithm embedding: Feature extraction is performed on the relevant results through cyclic shift characteristics and multi-path time delay, and a segmentation detection algorithm is designed. The pre-trained Transformer network is embedded into the detection algorithm to improve the accuracy of conflict detection.

[0068] 5) Application and evaluation: Finally, the constructed model is applied to the high dynamic ocean satellite communication system for signal conflict detection and load estimation. By monitoring and evaluating the system performance in real time, communication strategies can be adjusted in time to ensure the stability and reliability of communication.

[0069] These two examples demonstrate how to apply the satellite random access deep segmentation conflict detection method based on optimized multiple-preamble in different high dynamic satellite scenarios to improve the performance and stability of the communication system.

[0070] As shown in Figure 1 , 3 , a satellite random access deep segmentation conflict detection method based on optimized multiple-preamble provided by an embodiment of the present application includes the following steps:

[0071] S101, based on the high dynamic satellite scene features, a MRP optimization method is designed to obtain an optimized MRP set capable of reducing multipath effect interference;

[0072] S102, based on each optimized MRP, a related result data set containing different conflict situations and complex channel features is constructed;

[0073] S103, for any MRP, a lightweight Transformer architecture integrating only one attention module and one multi-layer perception module is used to preliminarily train the entire data set of the MRP;

[0074] S104, the MRP related results are windowed according to the cyclic shift characteristic, and the cyclic shift value and the multipath time delay are combined to extract features of the related results in each window, a segmented detection algorithm based on feature extraction is designed, and a satellite random access deep segmented conflict detection model based on an optimized multiple root preamble is obtained by embedding the pre-trained Transformer network into the detection algorithm;

[0075] S105, the satellite random access deep segmented conflict detection model based on the optimized multiple root preamble is used to perform signal conflict detection and load estimation on a large-scale access device in a high dynamic satellite scene.

[0076] The specific method in S101 provided by the embodiment of the application is that the influence of high dynamic satellite scene features on MRP related results is analyzed, a cyclic shift configuration set of MRP is designed in combination with a multipath time delay, and the related results of MRP are optimized to reduce the interference of the multipath effect.

[0077] The specific method for constructing the data set in S102 provided by the embodiment of the application is that different conflict situations and channel models are set, and related results of each optimized MRP are generated as a data set based thereon.

[0078] According to the constructed data set, the lightweight Transformer network architecture integrating one attention module and a multi-layer perception module in S103 is used to preliminarily train the data set in S102 for different conflict situations of the optimized MRP, and the correlation between random access samples is fully learned.

[0079] In S104 provided by the embodiment of the application, the optimized MRP related results need to be windowed according to the cyclic shift characteristic, and the cyclic shift value and the multipath time delay are combined to extract features of the related results in each window, a segmented detection algorithm module based on feature extraction is constructed, and the pre-trained Transformer network is embedded into the detection algorithm; finally, a satellite random access deep segmented conflict detection model based on an optimized multiple root preamble is obtained.

[0080] The method aims to solve the signal collision problem caused by multipath effects and channel complexity in high dynamic satellite scenarios. Here is the detailed working principle of the method:

[0081] Step 1: Design MRP optimization method

[0082] - Goal: Reduce multipath effect interference.

[0083] - Method: Based on the characteristics of high dynamic satellite scenarios, design an optimized multiple root preamble (MRP) preamble cyclic shift configuration set.

[0084] - Result: Get an optimized MRP set that can effectively reduce multipath effect interference.

[0085] Step 2: Build relevant result dataset

[0086] - Goal: Obtain data containing different collision situations and complex channel characteristics.

[0087] - Method: Based on each optimized MRP, build relevant result datasets containing various collision situations and complex channel characteristics.

[0088] - Result: Form a detailed dataset for subsequent model training and feature extraction.

[0089] Step 3: Lightweight Transformer preliminary training

[0090] - Goal: Preliminary training of MRP dataset.

[0091] - Method: Use a lightweight Transformer architecture that integrates only one attention module and one multi-layer perception module to preliminarily train the entire dataset of optimized MRP.

[0092] - Result: Train a preliminary Transformer model for subsequent feature extraction and collision detection.

[0093] Step 4: Segmented detection algorithm based on feature extraction

[0094] - Goal: Feature extraction of MRP related results and design segmented detection algorithm.

[0095] - Method:

[0096] 1) Windowing: Windowing processing of MRP related results according to cyclic shift characteristics.

[0097] 2) Feature extraction: Combine cyclic shift values and multipath time delay to extract features from the correlation results in each window.

[0098] 3) Detection algorithm: design a segmented detection algorithm based on feature extraction, and embed a pre-trained Transformer network into the detection algorithm.

[0099] - Result: form a satellite random access deep segmentation conflict detection model based on optimized multiple pilots.

[0100] Step 5: Perform signal conflict detection and load estimation

[0101] Objective: Detect signal conflict and estimate load in high dynamic satellite scenarios.

[0102] Method: Use the satellite random access deep segmentation conflict detection model based on optimized multiple pilots to detect signal conflict and estimate load for large-scale access devices.

[0103] Result: Achieve efficient signal conflict detection and load estimation, improve communication performance and reliability in high dynamic satellite scenarios.

[0104] Through the above five steps, this method can use optimized multiple pilots and lightweight Transformer network to achieve efficient and accurate signal conflict detection and load estimation in complex satellite communication environment, thereby improving the overall performance of satellite communication system.

[0105] As shown in Figure 2 , the satellite random access deep segmentation conflict detection system based on optimized multiple pilots provided by the embodiment of the application comprises:

[0106] An optimization module is configured to design an MRP optimization method based on the characteristics of high dynamic satellite scenarios, and obtain an optimized MRP set capable of reducing multipath effect interference.

[0107] A construction module is configured to construct a related result dataset containing different conflict situations and complex channel characteristics based on each optimized MRP.

[0108] A training module is configured to use a lightweight Transformer architecture integrating only one attention module and one multi-layer perception module to preliminarily train the entire dataset of the MRP for any MRP.

[0109] A feature extraction module is configured to window the MRP related results according to the cyclic shift characteristic, extract features of the related results in each window in combination with the cyclic shift value and the multipath time delay, design a segmented detection algorithm based on feature extraction, and embed a pre-trained Transformer network into the detection algorithm to obtain a satellite random access deep segmentation conflict detection model based on optimized multiple pilots.

[0110] The detection module is used for performing signal collision detection and load estimation on a large number of access devices in a high-dynamic satellite scene by using the satellite random access deep segmentation collision detection model based on the optimized multiple roots of the preamble.

[0111] Another object of the present application is to provide a computer device comprising a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to make the processor execute the steps of the satellite random access deep segmentation collision detection method based on the optimized multiple roots of the preamble.

[0112] Another object of the present application is to provide a computer-readable storage medium storing a computer program, the computer program being executed by a processor to make the processor execute the steps of the satellite random access deep segmentation collision detection method based on the optimized multiple roots of the preamble.

[0113] Another object of the present application is to provide an information data processing terminal for implementing the satellite random access deep segmentation collision detection system based on the optimized multiple roots of the preamble.

[0114] First step: study the influence of high-dynamic scene characteristics on MRP, and design an MRP optimization method

[0115] The preamble in random access is generated by a ZC sequence with good correlation and low mean-to-peak ratio, and the generation mode of the ZC sequence is: z r (n)=exp[-jπrn(n+1) / N ZC ],0≤N≤N ZC -1, wherein N ZC is the length of the ZC sequence, r is the root index and r∈{1,...,N ZC -1}. Considering the balance between collision avoidance capability and noise suppression capability, MRP sequences can be used to construct all random access preambles, denoted as:

[0116]

[0117] where i∈{1,...,N P} represents the preamble index, N P is the number of available preambles. l={1,...,N SS} represents the cyclic shift offset multiple, wherein N CS is the cyclic shift configuration of the preamble. In order to reduce the interference of multipath effect in NTN scene on MRP, which affects the detection performance of the network. The designed cyclic shift configuration optimization criterion is:

[0118] :

[0119] :

[0120] where T ZC is the sequence duration, τ′ max is the maximum multipath delay spread value. According to the design of the cyclic shift configuration optimization criterion, all available MRP that meets the optimization criterion can be generated, and the user randomly selects the ith preamble to be transmitted on the physical random access channel. After passing through the typical satellite line-of-sight channel, the preamble sequence received at the receiving end is:

[0121]

[0122] where I = {i1,...,i U} represents the preamble index set of U access devices, where d∈{1,...,u i} is the access user index, E d , and represent the number of arrival paths, path gain and round-trip delay of the dth user, respectively. ω(n) is the noise. The correlation result of the correlation operation between the received signal and the local root sequence is as follows:

[0123]

[0124] where τ∈{0,...,(N ZC -1)}, the correlation result of the optimized MRP will not be affected by the influence of multipath effect, so that there is a large amount of interference data in the correlation result based on formula (3), and therefore the neural network can fully establish the mapping relationship between the correlation result of the optimized MRP and the conflict load.

[0125] Second step: constructing a data set based on each optimized MRP

[0126] For all available optimized MRPs, the correlation result data set of non-repeated samples is generated in the preset conflict load range and under different channel models, where the data set of the ith preamble is:

[0127] H i = {(c i,1 ,u i,1 ),(c i,2 ,u i,2 ),...,(c i,G ,u i,G )} (4)

[0128] where G represents the size of the data set, c i,G is the correlation result of the optimized MRP, and the channel model is set to AWGN, NTN-B and NTN-D respectively. The finally constructed data set can be represented as Hi = 3 - H i .

[0129] Third step: Constructing the lightweight Transformer and using the dataset of optimized MRP for initial training Figure 4

[0130] The whole lightweight Transformer is integrated by attention module and multi-layer perception module, the attention module is composed of full connection layer, layer normalization operation, ReLu activation function and a double-headed attention mechanism, the output of the double-headed attention mechanism is connected with the full connection layer in residual connection; the multi-layer perception module is stacked by two full connection layers for data fitting. The input sample can be fully extracted after entering the attention module, and the long-range correlation between the data is improved, and the representation ability of the model is improved; then, the important sample data after correlation scoring is fitted by using the multi-layer perception module, the mapping function between input and output is established, and the initial training of the lightweight Transformer neural network is completed.

[0131] Fourth step: Embedding the pre-trained Transformer neural network into the feature extraction-based segment detection algorithm

[0132] The flow of the feature extraction-based segment detection algorithm is shown in Figure 5

[0133] First, the related results of the optimized MRP are windowed, and the lth detection window can be expressed as:

[0134]

[0135] Since the N CS of the MRP can be obtained in advance, and the value of l affects the division of the detection window, when multiple users select the same MRP for access at the same time, the significant position of the conflict feature can be determined in advance at lN CS . Secondly, due to the influence of multipath effect in high dynamic satellite scene, the correlation peak at lN CS will be expanded to the multipath delay. Therefore, it is concluded that no matter how many users select the same MRP to generate conflicts, the position of the correlation peak generated by the conflict can be obtained in advance in the corresponding detection window, and the size of the peak at these positions will be more greatly affected by the number of access conflict loads. Therefore, the important feature position in each window can be extracted according to the N CS obtained in advance and the multipath delay τ, the features extracted in the lth detection window of the ith MRP can be expressed as: ​​

[0136] c i,l =[C(lN CS ),C(lN CS +τ1)...,C(lN CS +τ max )] (6)

[0137] Based on the feature extraction results in each window, the pre-trained lightweight Transformer network embedding detection algorithm is used to perform detection in a segmented iterative manner.

[0138] Step 5: Detect and estimate the conflict load according to the satellite random access deep segment conflict detection method model based on optimized multiple preambles.

[0139] First, the correlation operation result is obtained based on the correlation operation of the target received signal and the local sequence based on the optimized MRP. The obtained correlation result is input into the segmented detection algorithm based on feature extraction for windowing and feature extraction. Then, the pre-trained lightweight Transformer performs regression prediction on the extracted features in each window in a segmented iterative manner to obtain the accurate conflict situation of the current load and the load occupancy situation. The exact value of is estimated as follows:

[0140]

[0141] Where f(·) is the mapping function between the correlation result after feature extraction and the real conflict load, is the estimated value of the target. The online conflict detection and load estimation process is as follows Figure 6 As shown:

[0142] Simulation Results

[0143] The performance of the invention is analyzed below in combination with simulation. In the simulation, the maximum label of the i-th leading data set is u i,g =100, and select N that meets the cyclic shift configuration optimization criteria CS The simulation is carried out with an MRP of 10. The AWGN, NTN-B, and NTN-D channel models are used to simulate the satellite random access scenario based on high dynamic NTN.

[0144] Figure 7 The collision detection probability performance curves of the present invention under the NTN-B, NTN-D, and AWGN channel models are given, where the collision situation on the i-th preamble is set as u i =100,u i= 300. As can be seen from the figure, the application has conflict detection capability under different signal-to-noise ratios in the high-dynamic satellite random access scene, and the conflict detection probability under different channel models can reach more than 90% when the signal-to-noise ratio is greater than -10dB.

[0145] Figure 8 The load estimation accuracy performance curves of the application under the NTN-B, NTN-D, and AWGN channel models are given, and the conflict on the ith preamble is also set to u i = 100, u i = 300. As can be seen from the figure, the application has excellent load estimation capability under different signal-to-noise ratios in the high-dynamic satellite random access scene, and the load estimation accuracy under different channel models can reach more than 94% when the signal-to-noise ratio is greater than -18dB. Figure 9 The conflict detection probability performance of the application under different signal-to-noise ratios is compared with that of the traditional detection method, which is the conflict detection scheme based on the label preamble (T-eCD) and the conflict detection method based on a single preamble (S-eCD), and the conflict scale on the ith preamble in the simulation is set to u i = 2. As can be seen from the figure, compared with the traditional detection method, the application can obtain better conflict detection performance than the traditional method under any signal-to-noise ratio.

[0146] Specific application field: conflict detection and load estimation in high-density device deployment areas in urban areas based on high-dynamic NTN scenarios.

[0147] 1. Optimization of MRP set generation:

[0148] Based on the designed MRP optimization method, a preamble set is generated for random access of a large number of devices in a high-density device deployment area.

[0149] 2. Data collection and preprocessing:

[0150] In a high-dynamic NTN scenario in a high-density device deployment area in an urban area, communication data between a satellite base station and a large number of devices, especially received signals in the random access process, are collected. The collected data is preprocessed, including compensation for timing advance based on satellite ephemeris and user location. And the received signal is correlated with the local sequence to construct multiple correlation result data sets based on the optimized MRP.

[0151] 3. Network design and pre-training

[0152] A lightweight Transformer network architecture is constructed, including an attention mechanism module and a multi-layer perception module. And the dataset made in a high-density device deployment area in an urban area in a high-dynamic NTN scenario is used for initial training.

[0153] 4. Network embedding feature extraction based segmentation detection method

[0154] According to the cyclic shift characteristics of the MRP, the correlation results are windowed, and the feature extraction is performed on the correlation results in each window in combination with the cyclic shift offset value of the massive devices initiating access and the multipath delay. A pre-trained Transformer network architecture is embedded into the segmentation detection method to construct a deep segmentation conflict detection model.

[0155] 5. Conflict detection and load estimation

[0156] In the urban high-density device deployment area of the high-dynamic NTN scene, the constructed deep segmentation conflict detection model is used. The received signal of the massive devices initiating access is correlated with the local sequence, and is input into the model to perform windowing and feature extraction. The model outputs the conflict situation of the current load, which is used for subsequent resource allocation.

[0157] Figure 10 Load estimation accuracy of different conflict situations in urban high-density device deployment areas. Figure 10 For the load estimation accuracy of different conflict situations in urban high-density device deployment areas, it can be seen that when the number of devices in conflict during access is 100 and 300 respectively, the proposed scheme can maintain an estimation accuracy of more than 0.9 under the condition of higher than -18dB.

[0158] It should be noted that the embodiments of the present application can be realized by hardware, software or a combination of software and hardware. The hardware part can be realized by special logic; the software part can be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or a specially designed hardware. Those skilled in the art can understand that the above-mentioned devices and methods can be realized by computer executable instructions and / or included in processor control code, such as provided on a carrier medium, such as a magnetic disk, CD or DVD-ROM, a programmable memory, such as a read-only memory (firmware), or a data carrier, such as an optical or electronic signal carrier. The devices and modules of the present application can be realized by hardware circuit, such as ultra-large-scale integrated circuit or gate array, semiconductor, such as logic chip, transistor, etc., or programmable hardware device, such as field programmable gate array, programmable logic device, etc., can also be realized by software executed by various types of processors, or by a combination of the above-mentioned hardware circuit and software, such as firmware.

[0159] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any modification, equivalent replacement and improvement within the technical range disclosed by the present application and within the spirit and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for optimizing satellite random access deep segment conflict detection for multiple preambles, characterized in that: The following steps are involved: Step 1: Based on the characteristics of highly dynamic satellite scenes, an MRP optimization method is designed to obtain an optimized MRP set that can reduce multipath interference. Step 2: Based on each optimized MRP, a relevant result dataset containing different conflict situations and complex channel characteristics is constructed; Step 3: For any MRP, a lightweight Transformer architecture that only integrates one attention module and one multi-layer perception module is used to perform preliminary training on the entire dataset of the MRP. Step 4: Window the MRP correlation results according to the cyclic shift characteristics, and extract features from the correlation results in each window by combining the cyclic shift value and multipath delay. A segmented detection algorithm based on feature extraction is designed. The pre-trained Transformer network is embedded in the detection algorithm to obtain a satellite random access deep segmented conflict detection model based on optimized multiple preambles. Step 5: Utilize the satellite random access deep segment conflict detection model based on optimized multiple preambles to perform signal conflict detection and load estimation on large-scale access devices in a highly dynamic satellite scenario.

2. The method for detecting satellite random access depth segment conflicts by optimizing multiple preambles according to claim 1, wherein: The specific approach in step 1 is: analyzing the impact of high-dynamic satellite scene characteristics on MRP correlation results, designing a cyclic shift configuration set of MRP in combination with multipath delay, and optimizing the MRP correlation results to reduce the interference of multipath effects.

3. The method for detecting satellite random access depth segment conflicts by optimizing multiple preambles according to claim 1, wherein: The specific method of constructing the data set in step 2 is: setting different conflict situations and channel models, and based on these, generating relevant results as a data set for each optimized MRP.

4. The method for detecting satellite random access depth segment conflicts by optimizing multiple preambles according to claim 1, wherein: According to the constructed data set, in step 3, an attention module and a multi-layer perception module are integrated to construct a lightweight Transformer network architecture. For the data set in step 2, preliminary training is performed on the conflict situations of different optimized MRPs to fully learn the correlation between randomly accessed samples.

5. The method for detecting satellite random access depth segment conflicts by optimizing multiple preambles according to claim 1, wherein: In step 4, the optimized MRP correlation results need to be windowed according to the cyclic shift characteristics, and the correlation results in each window need to be feature extracted in combination with the cyclic shift value and multipath delay. A segmented detection algorithm module based on feature extraction is constructed, and the pre-trained Transformer network is embedded in the detection algorithm. Finally, a satellite random access deep segment conflict detection model based on optimized multiple preambles is obtained.

6. A system for optimizing satellite random access deep segment conflict detection for multiple preambles, which implements the method for optimizing satellite random access deep segment conflict detection for multiple preambles as claimed in any one of claims 1 to 5, characterized in that: The satellite random access deep segment conflict detection system for optimizing multiple preambles includes: The optimization module is used to design an MRP optimization method based on the characteristics of high-dynamic satellite scenes to obtain an optimized MRP set that can reduce multipath interference; A construction module is used to construct a relevant result dataset containing different conflict situations and complex channel characteristics based on each optimized MRP; The training module is used to perform preliminary training on the entire dataset of any MRP using a lightweight Transformer architecture that only integrates one attention module and one multi-layer perception module; The feature extraction module is used to window the MRP correlation results according to the cyclic shift characteristics, and extract features from the correlation results in each window by combining the cyclic shift value and multipath delay. This module then designs a segmentation detection algorithm based on feature extraction. The pre-trained Transformer network is then embedded into the detection algorithm to obtain a deep segmentation conflict detection model for satellite random access based on optimized multiple preambles. The detection module is used to perform signal conflict detection and load estimation on large-scale access devices in a high-dynamic satellite scenario by using the satellite random access deep segment conflict detection model based on optimized multiple preambles.

7. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method for optimizing satellite random access deep segment conflict detection for multiple preambles as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor performs the steps of the method for optimizing deep segment conflict detection of satellite random access with multiple preambles according to any one of claims 1 to 5.

9. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the satellite random access deep segment conflict detection system for optimizing multiple preambles as described in claim 6.

Images