A data synchronization method for a flight simulator configuration database

Abstract

The present application belongs to the technical field of flight simulator database management, and relates to a data synchronization method for a flight simulator configuration database, aiming to solve the problems of low synchronization efficiency and difficult guarantee of data consistency. The present application generates / calibrates synchronization parameters through graph neural operators, optimizes the optimal synchronization sequence containing execution timing through manifold, generates a semantic consistency mapping table through a parameterized quantum model in combination with type constraint Hamiltonian and semantic matching baseline; then, based on the resource allocation benchmark in the synchronization parameters, the subsystem state is modeled as a stochastic partial differential equation, and the dynamic priority strategy is obtained by solving the HJB equation; after executing the synchronization task according to the optimal synchronization sequence and the mapping table, the consistency is verified by the directional slice distance, and if it exceeds the standard, rollback is triggered. The present application constructs a closed-loop system of parameter generation, sequence planning, field mapping, dynamic control and consistency verification, improves the synchronization adaptability and accuracy, and guarantees the data consistency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of flight simulator database management, and particularly relates to a data synchronization method for a flight simulator configuration database. BACKGROUND

[0002] In the running process of a flight simulator, its configuration database needs to store a large amount of flight simulation related data, including core information such as configuration parameters of different aircraft models, flight training trajectory data, and subsystem running state data. With the iterative upgrade of flight simulation technology and the continuous emergence of new aircraft models, higher requirements are put forward for the data synchronization mechanism of the configuration database, which is specifically reflected in the following four interrelated aspects:

[0003] 1. New aircraft model adaptation level: The introduction of new aircraft models significantly increases the complexity of obtaining and calibrating synchronization parameters. Since there are differences in the configuration of each aircraft model, traditional synchronization methods cannot meet the diversified configuration needs of new aircraft models, resulting in insufficient adaptability of data synchronization.

[0004] 2. Inter-table dependency management level: There are complex dependency relationships between configuration data tables in the configuration database. If the data synchronization order cannot be scientifically determined, it is easy to cause data inconsistency or integrity loss, which poses a challenge to the logic and planning of data synchronization.

[0005] 3. Field mapping accuracy level: When mapping fields between different configuration tables, accurately identifying semantic consistency is the key to avoiding data conversion errors. Once the semantic identification deviates, it will directly affect the accuracy and availability of data synchronization.

[0006] 4. Subsystem dynamic regulation level: The synchronization state of each subsystem of the flight simulator is affected by various factors such as hardware performance and network environment. This requires the synchronization process to have the ability to dynamically adjust the priority control strategy to optimize synchronization efficiency.

[0007] In summary, the existing data synchronization method has the problems of low synchronization efficiency and difficulty in guaranteeing data consistency when dealing with complex scenarios such as flight simulator configuration databases, and cannot meet the high-precision and high-reliability operation requirements of flight simulators. SUMMARY

[0008] In order to solve the above-mentioned problems in the prior art, i.e. the problems of difficult synchronization parameter acquisition and calibration, unreasonable synchronization order, inaccurate field mapping, and lack of dynamic optimization of subsystem synchronization strategy in the synchronization process of the flight simulator configuration database, resulting in low synchronization efficiency and difficulty in guaranteeing data consistency, the present application proposes, in a first aspect, a data synchronization method for a flight simulator configuration database, which comprises the following steps:

[0009] The new model synchronization parameters are obtained by using a graph domain neural operator architecture, and the conventional model synchronization parameters are calibrated based on historical synchronization performance data; the synchronization parameters include table dependency weights, field mapping initial threshold values, subsystem resource allocation benchmarks, and consistency qualified threshold values;

[0010] A plurality of configuration data tables in a configuration database are obtained, a dependency graph is constructed based on inter-table dependency relationships, synchronization sequences are encoded as orthogonal matrices on a Stiefel manifold, the graph structure constraints are adjusted in combination with table dependency weights, and an optimal synchronization sequence and its probability distribution are generated through Riemann optimization; the synchronization sequence is an ordered task queue, including a source configuration table, a target configuration table, field mapping priorities, and execution timing;

[0011] Based on the source configuration table and the target configuration table, to-be-mapped fields are identified, a parameterized quantum model is constructed to jointly encode semantic features of the fields into quantum states, a type constraint Hamiltonian is introduced, the field mapping initial threshold values are taken as semantic matching baselines, the optimal synchronization sequence and its probability distribution are combined as mapping score prior weights, and a semantic consistency mapping table is obtained;

[0012] The subsystem synchronization state is modeled as a multi-population coupled stochastic partial differential equation of a space-time density field, the execution timing is taken as a time benchmark, and the subsystem resource allocation benchmark is taken as an initial value of a control variable to solve a corresponding optimal control equation and obtain a dynamic priority control strategy;

[0013] Based on the optimal synchronization sequence, a configuration data update task queue is generated, the semantic consistency mapping table is applied to each pair of source configuration table and target configuration table for data conversion, and a converted task queue is obtained; the dynamic priority control strategy is applied to the converted task queue for dynamic scheduling, a priority sorting result is obtained, the priority sorting result is executed through a database interface, and data synchronization of the flight simulator configuration database is completed.

[0014] In some preferred embodiments, after completing the data synchronization of the flight simulator configuration database, the method further includes:

[0015] The source configuration table and the target configuration table are mapped to a diffusion geometry space, a direction slice distance is used for cross-structure semantic consistency measurement, a consistency verification value is obtained, if the consistency verification value exceeds a preset threshold value, rollback is triggered and data synchronization is re-executed until the consistency verification value is lower than the preset threshold value.

[0016] In some preferred embodiments, the method for obtaining new model synchronization parameters by using a graph domain neural operator architecture includes:

[0017] extracting a meta-feature vector of a new model configuration database, the meta-feature vector including table structure complexity, field correlation, and operation and maintenance resource preference;

[0018] constructing a parameter dependency graph, nodes of the parameter dependency graph including table dependency weight, field mapping initial threshold, subsystem resource allocation benchmark, and consistency eligibility threshold, and edges of the parameter dependency graph representing cooperative relationships between the nodes;

[0019] using a graph attention network as a core of a graph domain neural operator, initializing initial features of the nodes in the parameter dependency graph based on the meta-feature vector, calculating attention weights between the nodes through a kernel function, and obtaining aggregated neighbor node features of the attention weights; wherein the kernel function is based on node feature vectors, edge features, attention weight vectors, and linear transformation matrices, and after being processed by a LeakyReLU activation function, the normalized attention weights between the nodes are obtained through normalization;

[0020] based on the parameter dependency graph and the aggregated neighbor node features of the attention weights, updating feature representations of the nodes through feature aggregation, and after iterative optimization of a multi-layer network, mapping the node features to specific numerical values to obtain the synchronization parameters.

[0021] In some preferred embodiments, the method for generating the optimal synchronization order and its probability distribution is as follows:

[0022] obtaining a plurality of configuration data tables in a flight simulator configuration database, constructing a dependency graph with data tables as nodes and dependency relationships between tables as edges based on foreign key associations and trigger dependencies between the data tables, and assigning table dependency weights to the edges of the dependency graph to obtain a weighted dependency graph;

[0023] encoding the synchronization order into an orthogonal matrix on a Stiefel manifold;

[0024] constructing an optimization objective function based on the weighted dependency graph, the optimization objective function combining synchronization cost, graph structure constraints, and regularization terms;

[0025] updating the orthogonal matrix using a Riemannian gradient descent method with curvature compensation based on the optimization objective function;

[0026] using a temperature parameter controlled Gumbel-Sinkhorn operator sampling to generate the optimal synchronization order and its probability distribution for the updated orthogonal matrix.

[0027] In some preferred embodiments, the temperature parameter controlled Gumbel-Sinkhorn operator sampling to generate the optimal synchronization order and its probability distribution includes:

[0028] Gumbel noise is injected into each element of the updated orthogonal matrix to obtain a noisy matrix;

[0029] The Gumbel-Sinkhorn operator controlled by a temperature parameter is applied to the noisy matrix to generate a probability distribution of the synchronous sequence;

[0030] An optimal synchronous sequence is sampled from the probability distribution.

[0031] In some preferred embodiments, the semantic consistency mapping table is obtained by:

[0032] Based on the source configuration table and the target configuration table, the source fields and the target fields to be mapped are identified by matching the field name similarity, data type compatibility and business description keywords to form a candidate field pair set;

[0033] A parameterized quantum model is constructed to jointly encode the semantic features of the source fields and the target fields into a quantum state, and the measurement probability representing the semantic matching degree of the fields is obtained by measuring the quantum state;

[0034] A penalty result is obtained based on a type constraint Hamiltonian, which includes multiple penalty terms that impose energy penalties on candidate field pairs with incompatible data types, and the penalty term weight is positively correlated with the initial threshold value of field mapping;

[0035] The optimal synchronous sequence and its probability distribution are used as the prior weight of the mapping score, and the comprehensive matching score of the candidate field pairs is calculated by combining the measurement probability and the penalty result. Field pairs with scores higher than the initial threshold value of field mapping are selected to form a semantic consistency mapping table from the source fields to the target fields.

[0036] In some preferred embodiments, the parameterized quantum model is constructed to jointly encode the semantic features of the source fields and the target fields into a quantum state, and the measurement probability representing the semantic matching degree of the fields is obtained by measuring the quantum state, by:

[0037] The semantic features of the fields are extracted and quantized into classical feature vectors; the semantic features include text description vectors, data type identifiers and business attribute labels;

[0038] The classical feature vectors are mapped to the initial ground state of the quantum bits by an initialization circuit;

[0039] The initial ground state of the quantum bits is subjected to unitary evolution by a parameterized quantum circuit to obtain a quantum state;

[0040] And the measurement probability representing the semantic matching degree of the fields is obtained by measuring the quantum state.

[0041] In some preferred embodiments, the random partial differential equation is:

[0042] ;

[0043] wherein, is a synchronization density field of the i-th subsystem, is a control velocity field, is a diffusion coefficient; is a Laplace operator; is a divergence operator; is a noise intensity of the i-th subsystem, dynamically adjusted according to a consistency verification value, is a differential form of a Wiener process, is a noise term; is a coupling coefficient between subsystems. In some preferred embodiments, the optimal control equation is obtained by solving the HJB equation, and the synchronization total cost minimization is taken as the target in the solving process.

[0044] In some preferred embodiments, the trigger rollback is realized based on a proximal policy optimization algorithm, and the rollback path that minimizes the total recovery cost is selected to re-execute synchronization by constructing a rollback cost function.

[0045] In some preferred embodiments, the trigger rollback is realized based on a proximal policy optimization algorithm, and the rollback path that minimizes the total recovery cost is selected to re-execute synchronization by constructing a rollback cost function.

[0046] Advantages of the present application:

[0047] By combining the graph domain neural operator architecture with the new model meta-feature vector to obtain synchronization parameters, and using historical synchronization performance data to calibrate conventional model synchronization parameters, the configuration requirements of different models can be more accurately adapted, and the rationality and accuracy of synchronization parameters can be improved. On this basis, multi-dimensional optimization means are deeply coordinated to realize exponential improvement of data synchronization performance:

[0048] Driver-dependent intelligent collaboration: a dependency graph is constructed based on inter-table dependencies, and an optimization algorithm on the Stiefel manifold is used to generate an optimal synchronization order and its probability distribution, providing a data flow framework for parameterized quantum models; under the joint action of the type-constrained Hamiltonian and the prior weight of the optimal synchronization order, a semantic consistency mapping table can be accurately constructed, not only avoiding data inconsistency caused by improper order, but also reducing data conversion errors through accurate field mapping, forming a dual protection mechanism of "sequence planning - semantic alignment";

[0049] ​Dynamic control intelligent response: model the subsystem synchronization state as a stochastic partial differential equation, combine the dynamic priority control strategy obtained from the optimal control equation to form a dynamic response closed loop with the synchronization sequence optimization; according to the real-time state and resource allocation of the subsystem, the priority strategy dynamically adjusts the synchronization sequence execution rhythm, so that the cross-structure semantic consistency measure takes effect in real time in the data synchronization process; when the consistency verification value is not up to standard, immediately trigger the rollback mechanism to ensure the full-link consistency from data conversion to system cooperation, and finally realize the high-performance stable operation of the flight simulator configuration database in complex scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0050] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the drawings:

[0051] Figure 1 is a step flow chart of a data synchronization method of a flight simulator configuration database according to the present application. DETAILED DESCRIPTION

[0052] The application will be further described in detail below with reference to the drawings and examples. It should be understood that the specific examples described herein are intended to explain the relevant application, but not to limit the application. In addition, it should be noted that only the parts related to the application are shown in the drawings for ease of description.

[0053] It should be noted that the embodiments and features in the embodiments in the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0054] In order to more clearly describe the data synchronization method of the flight simulator configuration database according to the present application, the following will be described in combination with Figure 1 The steps in the embodiments of the present application will be described in detail.

[0055] The present application proposes a data synchronization method of a flight simulator configuration database, referring to Figure 1 The method comprises the following steps:

[0056] The graph neural operator architecture is used to obtain the synchronization parameters of the new model, and the historical synchronization performance data is used to calibrate the synchronization parameters of the conventional model; the synchronization parameters include table dependency weight, field mapping initial threshold, subsystem resource allocation benchmark and consistency qualified threshold;

[0057] In the present embodiment, the method for obtaining the synchronization parameters of the new model by using the graph neural operator architecture is as follows:

[0058] The meta-feature vector of the new model configuration database is extracted, and the meta-feature vector includes table structure complexity (including the number of data tables, the average number of fields, the hierarchical depth of primary and foreign keys, forming a 48-dimensional feature sub-vector), field correlation (calculated by field co-occurrence frequency and business logic correlation degree, such as the correlation degree of “rotational speed” and “fuel consumption” is set to 0.8, forming a 40-dimensional feature sub-vector), and operation and maintenance resource preference (resource allocation tendency extracted based on historical operation and maintenance logs, such as CPU / memory proportion preference, forming a 40-dimensional feature sub-vector);

[0059] A parameter dependency graph is constructed with nodes and edges ; The node includes table weight (a weight parameter representing the synchronization priority between data tables), field mapping initial threshold (a baseline threshold for filtering valid field pairs), subsystem resource allocation baseline (resource proportion control of each subsystem, such as flight control and avionics), and consistency qualified threshold (qualified standard for determining data synchronization consistency); The edge represents the cooperative relationship between nodes, for example, positive cooperative edge (positive correlation between table dependency weight and field mapping threshold), resource constraint edge (resource allocation baseline constraint on consistency threshold, the more sufficient the resource, the more relaxed the threshold), and each edge is assigned a cooperative strength weight in the interval [0, 1];

[0060] A graph attention network is used as the core of the graph domain neural operator, based on the meta-feature vector, the initial features of each node in the parameter dependency graph are initialized, the attention weight between nodes is calculated through a kernel function, the attention weight is obtained, and the neighbor node features are aggregated to update the current node representation; The kernel function is based on the node feature vector, the edge feature, the attention weight vector, and the linear transformation matrix, processed by the LeakyReLU activation function, and the normalized attention weight between nodes is obtained by normalization; Specifically, the kernel function expression is:

[0061] ;

[0062] In the formula, is the current feature vector of the node , (the first layer GAT is 128-dimensional, and the second layer is 64-dimensional);

[0063] W is a linear transformation matrix (the first layer , and the second layer );

[0064] is the feature vector of the edge (the edge weight 0-1 is quantized into a vector with the same dimension as , such as a 64-dimensional edge feature vector with all elements being the edge weight value 0.85);

[0065] For attention weight vector (first layer , corresponding to 64+64+64 splicing dimension; second layer , corresponding to 32+32+32 splicing dimension; For node The neighbor node set of; For activation function;

[0066] After the multi-layer (usually 2 layers) network iteration optimization, the optimized node feature is mapped to a specific numerical value through a linear mapping layer to obtain the synchronization parameter; the expression of the optimized node feature is: , wherein For activation function;

[0067] The linear mapping layer is mapped to a specific numerical value, which is: the node feature (32 dimensions) of the table dependent weight is mapped to the weight value (such as 0.75) in the interval of 0-1;

[0068] The field mapping initial threshold node feature is mapped to the threshold value (such as 0.65) in the interval of 0.5-0.8;

[0069] The subsystem resource allocation benchmark node feature is mapped to the resource proportion of each subsystem (such as 0.35 for flight control system and 0.25 for avionics system);

[0070] The consistency qualified threshold node feature is mapped to the qualified rate threshold (such as 95%) in the interval of 0-100%;

[0071] The specific numerical value is the synchronization parameter of the new model, which is the core parameter basis for subsequent dependency graph construction and field mapping;

[0072] The historical synchronization performance data includes the success rate and time consumption parameters in the last three synchronization processes, and the dynamic calibration method is to perform weighted fusion on the historical optimal synchronization parameter And the initial synchronization parameter To generate the final synchronization parameter ;

[0073] The weighted fusion method includes:

[0074] The historical synchronization performance data is constructed into an effectiveness vector , wherein Is the kth synchronization success rate, Is the time consumption; the last The historical parameters Are weighted and fused:

[0075] ;

[0076] ;

[0077] in, These are performance weights used to adjust the initial synchronization parameters. Perform dynamic calibration. These are the weighting coefficients;

[0078] Multiple configuration data tables in the configuration database are obtained, a dependency graph is constructed based on the inter-table dependencies, the synchronization order is encoded as an orthogonal matrix on the Stiefel manifold, and the optimal synchronization order and its probability distribution are generated by Riemann optimization in combination with the table dependency weight adjustment graph structure constraints. The synchronization order is an ordered task queue, which includes the source configuration table, the target configuration table, the field mapping priority, and the execution time sequence.

[0079] In this embodiment, the optimal synchronization sequence and its probability distribution are obtained as follows:

[0080] Obtain multiple configuration data tables from the flight simulator configuration database. Based on the foreign key relationships and trigger dependencies between the data tables, construct a dependency graph with data tables as nodes and inter-table dependencies as edges, and assign table dependency weights to the edges of the dependency graph.

[0081] Encode the synchronization sequence into a Stiefel manifold On Orthogonal matrix Y, Where n is the number of configuration data tables, and k is the latent semantic dimension, typically taken as n. To ensure that the matrix size matches the complexity of the dependency graph, encoding can be achieved using orthogonal encoding rules. accomplish;

[0082] Based on the weighted dependency graph, an optimization objective function is constructed that incorporates the fusion synchronization cost, graph structure constraints, and regularization term. The optimization objective function is as follows:

[0083] ;in, For the expected cost of synchronization, Temperature parameter Controlled Gumbel-Sinkhorn operator, For synchronization order, As a cost of synchronization, The Laplace matrix of the dependency graph. D is the degree matrix of the dependency graph (the diagonal elements are the weighted degree of the node, that is, the sum of the table dependency weights of all associated edges of the node), and W is the adjacency matrix of the dependency graph (the elements are the table dependency weights of the corresponding edges). For matrix trace operations, The term is a penalty term for graph structure constraints. For regularization terms, For gradient operators, It is the Frobenius norm. These are the weighting coefficients;

[0084] Based on the aforementioned objective function, the orthogonal matrix is ​​updated using the Riemann gradient descent method with curvature compensation. The update formula is as follows:

[0085] ;in, For exponential mapping, For tangent space projection, For the objective function in Riemann gradient at that point Let be the local curvature tensor of the manifold. For learning rate, The curvature compensation coefficient is used; when the difference in the Frobenius norm between adjacent matrices is less than the threshold, the iteration stops, and the updated orthogonal matrix Y is obtained.

[0086] For the updated orthogonal matrix Y, temperature parameters are used. The controlled Gumbel-Sinkhorn operator samples to generate the optimal synchronization order. Its probability distribution includes:

[0087] For each element in the updated orthogonal matrix Y Injecting Gumbel noise yields a noisy matrix. ; ;

[0088] in, Temperature is used as a parameter to control noise intensity; the lower the value, the weaker the noise impact and the more concentrated the sampling results.

[0089] The noisy matrix is ​​processed by temperature parameters. The controlled Gumbel-Sinkhorn operator processes the data to generate a probability distribution of the synchronization sequence. , Represents a given orthogonal matrix Time synchronization sequence The probability of occurrence; specifically:

[0090] For noisy matrices Perform row standardization and calculate the softmax probability;

[0091] Perform Sinkhorn iterations (usually 3-5 rounds) to make the matrix satisfy the double random constraint. ; Obtain the probability distribution of the synchronization sequence , among which, element denotes the th data table is assigned to the th synchronization timing;

[0092] sampling the optimal synchronization order from the probability distribution , i.e. ; specifically, an argmax sampling strategy can be adopted: ;

[0093] the optimal synchronization order is an ordered queue of length k, where denotes the data table index of the th timing node performing synchronization;

[0094] the optimal synchronization order is an ordered task queue, each task containing a source configuration table, a target configuration table, a field mapping priority, and an execution timing;

[0095] based on the source configuration table and the target configuration table, identify the fields to be mapped, construct a parameterized quantum model to jointly encode the semantic features of the fields into a quantum state, introduce a type constraint Hamiltonian, take the initial threshold value of the field mapping as the semantic matching baseline, combine the optimal synchronization order and its probability distribution as the mapping score prior weight, and obtain a semantic consistency mapping table;

[0096] In this embodiment, the semantic consistency mapping table is obtained by:

[0097] Based on the source configuration table and the target configuration table, identify the source fields and target fields to be mapped by matching field name similarity, data type compatibility, and business description keywords, and form a candidate field pair set;

[0098] Construct a parameterized quantum model to jointly encode the semantic features of the source fields and the target fields into a quantum state, and measure the quantum state to obtain a measurement probability representing the semantic matching degree of the fields, the method being:

[0099] Extract the semantic features of the fields and quantize them into a classical feature vector; the semantic features include a text description vector, a data type identifier, and a business attribute label;

[0100] Map the classical feature vector to a quantum bit initial ground state (such as ) through an initialization circuit;

[0101] The quantum bit initial ground state is subjected to unitary evolution by a parameterized quantum circuit to obtain a quantum state;

[0102] The unitary evolution of the parameterized quantum circuit is:

[0103] ;

[0104] wherein, is the number of circuit layers, is the number of qubits (adapted to the feature dimension), , are the Y-axis, Z-axis rotation gates of the layer th qubit, are the learnable rotation angle parameters, is the set of qubit connection edges, is the controlled-Z gate between the , th qubits; the quantum state output by the unitary evolution encodes the field semantic features and the correlation information between the features;

[0105] and measuring the quantum state to obtain the measurement probability representing the field semantic matching degree;

[0106] obtaining a penalty result based on the type constraint Hamiltonian , the type constraint Hamiltonian includes multiple penalty terms, and energy penalty is applied to candidate fields (such as integer and string, date and Boolean) that are incompatible in data types, and the penalty term weight is positively correlated with the field mapping initial threshold value, that is, the higher the threshold value, the greater the penalty strength for type deviation; the expression is:

[0107] ;

[0108] wherein, is the summation symbol, and all field pairs belonging to the “mismatch set ” are traversed , that is, the subsequent penalty calculation is performed for each group of mismatched and accumulated; is the penalty coefficient; is the field mapping initial threshold value; is the projection operator of the quantum state, is the state vector of the th qubit (or a certain state vector of a quantum system), is its dual left vector; indicates the operation of projecting the quantum state onto the subspace spanned by ; similarly, is the projection operation on the th qubit (or the corresponding state vector); is the tensor product symbol;

[0109] The optimal synchronization order and its probability distribution are used as prior weights for the mapping score. That is, the field pairs contained in the table with high priority (high probability) in the synchronization order are given a higher weight in the mapping score. The measurement probability of the quantum state output by the parameterized quantum model (representing the semantic matching degree of the field) and the penalty result of the type constraint Hamiltonian are combined to calculate the comprehensive matching score of the candidate field pairs. Field pairs with scores higher than the initial threshold of field mapping are selected to form a semantic consistency mapping table from the source field to the target field.

[0110] The comprehensive matching score is calculated as follows:

[0111] ;

[0112] in, Represents field pairs The overall matching score; Fields representing the output of the quantum model The probability of measurement; This indicates the penalty result for type-constrained Hamiltonians; Represents field pair The probability distribution corresponding to the optimal synchronization order, also known as the prior weights;

[0113] The synchronization state of the subsystem is modeled as a multi-population coupled stochastic partial differential equation of a space-time density field. Using the execution timing as the time base and the subsystem resource allocation base as the initial value of the control variable, the corresponding optimal control equation is solved to obtain the dynamic priority control strategy.

[0114] In this embodiment, the stochastic partial differential equation is:

[0115] ;

[0116] in, For the first The synchronization density field of each subsystem (characterizing the distribution of synchronization completion degree of the subsystem at spatial location x and time t, i.e.) The time variable t is based on the execution time sequence and is strictly aligned with the execution time sequence of the synchronous tasks. To control the velocity field (characterizing the intensity of resource allocation's regulation of the synchronization state, its initial distribution is set based on the subsystem resource allocation reference vector). The diffusion coefficient (describes the spatial propagation characteristics of a synchronized state). For the Laplace operator; For divergence operators; For the first Noise intensity of each subsystem This is the differential form of the Wiener process (characterizing the time evolution of random perturbations). is the noise term (characterizing random disturbance in the synchronization process); is the coupling coefficient between subsystems (quantifying the mutual influence of the synchronization states of different subsystems);

[0117] the noise intensity is dynamically adjusted according to the consistency verification value (specifically, dynamically adjusted according to the consistency verification value of the historical synchronization period), which satisfies:

[0118]

[0119] wherein, is the basic noise intensity, is the adjustment coefficient, and the greater the consistency verification value DCVC is, the higher the noise intensity to enhance the robustness of the system;

[0120] The consistency verification value is defined as the directional slice Gromov-Wasserstein distance DCVC, which is calculated after the data synchronization of the flight simulator configuration database is completed, and then the noise intensity is optimized through feedback:

[0121]

[0122] wherein, is a d-dimensional unit sphere, is the square of the second-order Gromov-Wasserstein distance (measuring the structural similarity of two metric spaces), and are the edge distributions of the source configuration table and the target configuration table in the unit vector direction, is the Haar measure on the unit sphere; and the greater the consistency verification value DCVC is, the higher the noise intensity to enhance the robustness of the system;

[0123] The optimal control equation is obtained by solving the Hamilton-Jacobi-Bellman (HJB) equation, and in the solving process, the minimum total synchronization cost (including resource consumption, time delay, etc.) is taken as the target, and the output dynamic priority control strategy quantifies the resource allocation weight and task priority sorting rule of each subsystem at different time nodes;

[0124] The target expression for minimizing the total synchronization cost is:

[0125]

[0126] wherein, represents the total cost of the synchronization process;​​​ denotes the total length of the synchronization process; denotes the summation index; denotes the cost coefficient associated with the i-th term; denotes the velocity variable (or other dynamic variable) associated with the i-th term; denotes the regularization parameter used to balance the weight of different parts in the total cost; denotes the density variable associated with the i-th term, denotes the square norm of the target value 1; denotes the density variable associated with the i-th term, denotes the square norm of the target value 1;

[0127] The HJB equation is:

[0128] ;

[0129] where, denotes the partial derivative of the value function with respect to time t; denotes the optimization of the control variable ; denotes the gradient vector of the value function ; denotes the dynamic vector field; denotes the instantaneous cost function, which is the integrand of the value function ;

[0130] Based on the optimal synchronization order, a configuration data update task queue is generated, the semantic consistency mapping table is applied for data conversion for each pair of source configuration table and target configuration table to obtain a converted task queue, the dynamic priority control strategy is used for dynamic scheduling of the converted task queue to obtain a priority sorting result, and the priority sorting result is executed through a database interface to complete data synchronization of the flight simulator configuration database;

[0131] After completing the data synchronization of the flight simulator configuration database, the method further includes:

[0132] The source configuration table and the target configuration table are mapped to a diffusion geometry space, a directional slice distance is used for cross-structure semantic consistency measurement to obtain a consistency verification value, if the consistency verification value exceeds a preset threshold, rollback is triggered and data synchronization is re-executed until the consistency verification value is lower than the preset threshold, a plurality of groups of DCVCs are calculated, and when the upper quartile of these values is 0.25 (for example, the maximum DCVC of a historical qualified sample is 0.25, and a redundancy is reserved, and the preset threshold is set to 0.3, that is, when DCVC>0.3), it is determined that the post-synchronization structure semantic consistency is not up to standard, and rollback is triggered); ​

[0133] The trigger rollback is implemented based on a proximal policy optimization (PPO) algorithm, a rollback cost function is constructed, and a rollback path that minimizes the total recovery cost is selected to re-execute synchronization.

[0134] Rollback cost function:

[0135] ;

[0136] wherein, represents the cost (or cost) of the rollback operation; represents the time cost weight coefficient; represents the weight coefficient of the data loss cost caused by the rollback operation;

[0137] In this way, through the graph domain neural operator architecture, the automatic generation of the synchronization parameters of the new model is realized, the dynamic calibration of the historical synchronization performance data is used for the conventional model, and the accuracy and adaptability of the parameter configuration are improved; based on the Stiefel manifold, the synchronization order is encoded as an orthogonal matrix, the Gumbel-Sinkhorn operator and the Riemannian gradient descent method with curvature compensation are introduced, the optimal synchronization order and its execution timing are solved under the premise of meeting the inter-table dependency constraint, and the global optimality of the scheduling plan is ensured; further, a parameterized quantum model is constructed to jointly encode the semantic features of the source table and the target table fields, a type constraint Hamiltonian is introduced and combined with the synchronization order probability prior, and a high-precision semantic consistency mapping table is generated; the subsystem synchronization state is modeled as a multi-population coupled stochastic partial differential equation of the space-time density field, the execution timing is taken as the time reference, and the resource allocation benchmark accuracy is taken as the control initial value, the optimal control strategy is solved, and the dynamic allocation of resources and intelligent regulation and control of task priority are realized; after synchronization is completed, the cross-structure consistency is verified by using the direction slice Gromov-Wasserstein distance, when the verification value exceeds the standard, the rollback mechanism based on the proximal policy optimization (PPO) is triggered, and the recovery path with the minimum cost is selected to re-execute; the scheme integrates the graph neural network, differential geometry, quantum computing and stochastic control, forms a closed-loop system of "perception-planning-execution-verification-feedback", and significantly improves the automation level, semantic accuracy and system robustness of the configuration data synchronization of the flight simulator.

[0138] Although the steps in the above embodiments are described in the above order, those skilled in the art can understand that, in order to achieve the effect of the embodiments, the steps do not have to be executed in such an order, and they can be executed simultaneously (in parallel) or in a reversed order, and these simple changes are within the protection scope of the present application.

[0139] Those skilled in the art should clearly understand that the modules and method steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware, computer software or a combination of both. The programs corresponding to the software modules and method steps can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. In order to clearly illustrate the interchangeability of electronic hardware and software, the components and steps of each example have been described in the above description in general terms. Whether the functions are performed by electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0140] The terms "first", "second", and the like are used to distinguish similar objects, rather than to describe or indicate a particular order or sequence.

[0141] The term "comprising" or any other similar term is intended to encompass a non-exclusive inclusion, such that a process, method, article, or apparatus / device including a series of elements includes not only those elements expressly listed, but also other elements inherent in such process, method, article, or apparatus / device or latent in such process, method, article, or apparatus / device.

[0142] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after such changes or replacements will fall within the protection scope of the present application.

Claims

1. A data synchronization method of a flight simulator configuration database, characterized by, The method comprises the following steps: The method comprises the following steps: Synchronization parameters include table dependency weights, field mapping initial threshold values, subsystem resource allocation benchmarks, and consistency qualified threshold values. Obtain a plurality of configuration data tables in a configuration database, construct a dependency graph based on inter-table dependency relationships, encode synchronization sequences into orthogonal matrices on a Stiefel manifold, adjust graph structure constraints based on table dependency weights, and generate optimal synchronization sequences and their probability distributions through Riemann optimization; the synchronization sequences are ordered task queues, including source configuration tables, target configuration tables, field mapping priorities, and execution timing sequences. Identify fields to be mapped based on the source configuration tables and the target configuration tables, construct a parameterized quantum model to jointly encode semantic features of the fields into quantum states, introduce a type constraint Hamiltonian, use the field mapping initial threshold values as a semantic matching baseline, combine the optimal synchronization sequences and their probability distributions as mapping score prior weights, and obtain a semantic consistency mapping table; Model the subsystem synchronization state as a multi-population coupled stochastic partial differential equation of a space-time density field, use the execution timing sequences as a time reference and the subsystem resource allocation benchmarks as initial values of control variables, solve the corresponding optimal control equation to obtain a dynamic priority control strategy; 2. The data synchronization method for a flight simulator configuration database according to claim 1, wherein, Generate a configuration data update task queue based on the optimal synchronization sequences, apply the semantic consistency mapping table to each pair of source configuration table and target configuration table for data conversion, and obtain a converted task queue; dynamically schedule the converted task queue according to the dynamic priority control strategy to obtain a priority ordering result, execute the priority ordering result through a database interface, and complete data synchronization of the flight simulator configuration database. After completing data synchronization of the flight simulator configuration database, the method further comprises:

3. The data synchronization method of claim 1, wherein, Map the source configuration tables and the target configuration tables to a diffusion geometry space, use directional slice distance to measure cross-structure semantic consistency, obtain a consistency verification value, and if the consistency verification value exceeds a preset threshold value, trigger rollback and re-execute data synchronization until the consistency verification value is lower than the preset threshold value. The method for obtaining synchronization parameters of a new model using a graph domain neural operator architecture comprises the following steps: Extract meta-feature vectors of the configuration database of the new model, wherein the meta-feature vectors include table structure complexity, field correlation, and operation and maintenance resource preferences; Construct a parameter dependency graph; the nodes of the parameter dependency graph include table dependency weights, field mapping initial threshold values, subsystem resource allocation benchmarks, and consistency qualified threshold values; the edges of the parameter dependency graph represent the cooperative relationships between the nodes; Use a graph attention network as the core of the graph domain neural operator, initialize the initial features of each node in the parameter dependency graph based on the meta-feature vectors, calculate the attention weights between the nodes through a kernel function, and obtain aggregated neighbor node features based on the attention weights; wherein the kernel function is based on node feature vectors, edge features, attention weight vectors, and linear transformation matrices, is processed through a LeakyReLU activation function, and then normalized to obtain normalized attention weights between the nodes; Based on the parameter-dependent graph, the neighbor node features are aggregated according to the attention weights, the node feature representations are updated, and after iterative optimization of a multi-layer network, the node features are mapped to specific numerical values to obtain the synchronization parameters.

4. The data synchronization method of a flight simulator configuration database according to claim 1, wherein, The method for generating the optimal synchronization sequence and its probability distribution is as follows: A plurality of configuration data tables in a flight simulator configuration database are obtained, a dependency graph with data tables as nodes and dependency relationships between tables as edges is constructed based on foreign key associations and trigger dependencies between the data tables, and table dependency weights are assigned to the edges of the dependency graph to obtain a weighted dependency graph. The synchronization sequence is encoded as an orthogonal matrix on a Stiefel manifold. An optimization objective function is constructed based on the weighted dependency graph, which integrates synchronization cost, graph structure constraints and regularization terms. The orthogonal matrix is updated using a Riemannian gradient descent method with curvature compensation based on the optimization objective function. The updated orthogonal matrix is sampled using a temperature parameter controlled Gumbel-Sinkhorn operator to generate the optimal synchronization sequence and its probability distribution.

5. The data synchronization method of claim 4, wherein, The temperature parameter controlled Gumbel-Sinkhorn operator sampling method for generating the optimal synchronization sequence and its probability distribution includes: Gumbel noise is injected into each element of the updated orthogonal matrix to obtain a noisy matrix. The noisy matrix is processed by a temperature parameter controlled Gumbel-Sinkhorn operator to generate a probability distribution of the synchronization sequence. An optimal synchronization sequence is sampled from the probability distribution.

6. The data synchronization method of a flight simulator configuration database according to claim 1, wherein, The method for obtaining the semantic consistency mapping table is as follows: Based on the source configuration table and the target configuration table, the source fields and the target fields to be mapped are identified by matching field name similarity, data type compatibility and business description keywords to form a candidate field pair set. A parameterized quantum model is constructed to jointly encode the semantic features of the source fields and the target fields into a quantum state, and the quantum state is measured to obtain a measurement probability representing the semantic matching degree of the fields. A penalty result is obtained based on a type constraint Hamiltonian, which includes a plurality of penalty terms that apply energy penalties to candidate field pairs with incompatible data types, and the penalty term weights are positively correlated with the initial threshold value of field mapping. The optimal synchronization sequence and its probability distribution are used as prior weights of the mapping score, and the measurement probability and the penalty result are combined to calculate the comprehensive matching score of the candidate field pairs. Field pairs with scores higher than the initial threshold value of field mapping are selected to form a semantic consistency mapping table from the source fields to the target fields.

7. The data synchronization method for a flight simulator configuration database according to claim 6, wherein, The method for constructing a parameterized quantum model to jointly encode the semantic features of the source fields and the target fields into a quantum state and measuring the quantum state to obtain a measurement probability representing the semantic matching degree of the fields is as follows: The semantic features of the fields are extracted and quantized into classical feature vectors; the semantic features include text description vectors, data type identifiers and business attribute labels. The classical feature vectors are mapped to a quantum bit initial ground state through an initialization circuit. The quantum bit initial ground state is subjected to unitary evolution by a parameterized quantum circuit to obtain a quantum state. And measure the quantum state to obtain a measurement probability of semantic matching degree of the characterization field.

8. The data synchronization method of a flight simulator configuration database according to claim 1, wherein, The random partial differential equation is: ； wherein, is the synchronization density field of the th subsystem, is the control velocity field, is the diffusion coefficient; is the Laplacian operator; is the divergence operator; is the noise intensity of the th subsystem, dynamically adjusted according to the consistency verification value, is the differential form of the Wiener process, is the noise term; is the coupling coefficient between subsystems.

9. The data synchronization method of claim 1, wherein, The optimal control equation is obtained by solving the HJB equation, and the synchronous total cost minimization is taken as the target in the solving process.

10. The data synchronization method of claim 2, wherein, The trigger rollback is realized based on a proximal policy optimization algorithm, a rollback cost function is constructed, and a rollback path that minimizes the total recovery cost is selected to re-execute synchronization.

Images