A variable address offset management method for a flight simulator

Abstract

This invention belongs to the technical field of flight simulators, specifically to a method for managing variable address offsets in flight simulators. This method aims to address the problems of existing address offset management, namely, insufficient real-time performance, weak fault tolerance, and severe memory fragmentation. This method uses a hybrid prediction model (SDE + LSTM) to predict variable changes, thereby triggering address offset calculation. Specifically, it optimizes offset calculation using a memory topology map and the Dijkstra algorithm. Combined with BCH + quantum layered verification, when verification fails, overflow variables are processed using the Sinkhorn migration and PPO algorithms to reduce memory fragmentation. This method significantly improves the real-time performance, memory utilization, and system reliability of flight simulations, making it suitable for high-precision aviation simulation scenarios.

Description

Technical Field

[0001] The invention belongs to the technical field of flight simulators, and in particular relates to a variable address offset management method of a flight simulator. Background Art

[0002] Flight simulators are essential equipment for pilot training and flight system testing, so accurate management of internal variable address offsets is crucial. Flight simulations involve a large amount of variable data, such as flight attitude, engine parameters, and avionics system data. The storage and address management of these variables in memory directly impact simulation accuracy and system stability.

[0003] Traditional memory management often uses static allocation or polling monitoring mechanisms. The former determines the memory space of variables at the initial stage, while the latter checks the variable status at fixed time intervals. When the variable address needs to be adjusted, the response is delayed, affecting the real-time performance and accuracy of the simulation.

[0004] In addition, in terms of address offset verification, traditional methods mainly rely on a single CRC (cyclic redundancy check) or ECC, which has weak fault tolerance. When verification fails and migration is required, the migration strategy is based on a greedy algorithm, causing the memory fragmentation rate to increase linearly over time. In actual tests, after 8 hours of operation, the utilization rate dropped from 85% to 62%, and frequent memory compression (Compaction) increased additional energy consumption. Summary of the Invention

[0005] In order to solve the above-mentioned problems in the prior art, namely, the problems of insufficient real-time performance, weak fault tolerance and severe memory fragmentation of the existing address offset management, the first aspect of the present invention provides a variable address offset management method for a flight simulator, the method comprising the following steps:

[0006] Get the flight simulator's variable list, shared memory state matrix, and variable parameters;

[0007] A variable monitoring queue of a sliding window is used to process the variable parameters to obtain a real-time data stream, and a hybrid prediction model is used to predict the real-time data stream to obtain an initial variable prediction value; the initial variable prediction value is combined with a gating weight to obtain a variable prediction value, and an early warning signal is obtained based on the variable prediction value;

[0008] After the early warning signal is triggered, a memory topology map is constructed based on the shared memory state matrix, and the Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology map, thereby obtaining a new variable address offset table;

[0009] Verifying the variable address new offset table by using layered BCH and quantum to obtain a verification result;

[0010] When the verification result is successful, updating the variable list and shared memory state matrix of the flight simulator based on the variable address new offset table;

[0011] When the verification result is failure, based on the new variable address offset table and the set of available memory blocks, the overflow variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal strategy optimization PPO algorithm, and the variable address offset is updated, thereby updating the variable list and shared memory status matrix of the flight simulator.

[0012] In some preferred embodiments, the hybrid prediction model predicts the real-time data stream by:

[0013] The jump-diffusion SDE model is used to calculate small changes, discretize the small changes into time intervals, and approximate the solution in each time interval to obtain a short-term prediction value. The jump-diffusion SDE model is:

[0014] ;

[0015] Where, is the long-term average trend of the variable over time; For the moment state variables; For small time intervals; is the degree of variable fluctuation; is the standard Brownian motion increment; is the jump amplitude; is the Poisson process increment;

[0016] The attention multi-scale residual long short-term memory network is used for long-term prediction, and the long-term prediction value is obtained through multi-scale parallelism, double-layer attention, and residual fusion.

[0017] In some preferred embodiments, the gating weight is calculated as follows:

[0018] Calculate the short-term prediction value error and the long-term prediction value error based on the short-term prediction value and the long-term prediction value respectively;

[0019] splicing the short-term prediction value error and the long-term prediction value error to obtain a combined vector;

[0020] Performing a matrix multiplication operation on the combined vector and the trainable parameter, and adding a bias term after the matrix multiplication operation to obtain an intermediate variable;

[0021] The intermediate variable is input into the Sigmoid activation function to obtain the gating weight.

[0022] In some preferred embodiments, the method for constructing a memory topology map based on a shared memory state matrix is ​​as follows:

[0023] Vertices are defined as variable addresses, and edge weights are constructed by reflecting the variable migration costs. The edge weight calculation formula is:

[0024] ;

[0025] Where, For variables size; is the data transmission delay between memory blocks; The remaining space for the target block; is the block size; is the dependency relationship between variables; are all weight coefficients.

[0026] In some preferred implementations, the Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology graph, and the calculation method is:

[0027] ;

[0028] Where, is the calculated new offset; is the distance between memory blocks; is the degree of memory fragmentation; is the fragmentation penalty coefficient.

[0029] In some preferred embodiments, quantum measurements are made using Pauli operators:

[0030] ;

[0031] Where, is the expected value; is a quantum state; is the Pauli operator; N is the number of measurements.

[0032] In some preferred embodiments, the overflow variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal strategy optimization (PPO) algorithm, as follows:

[0033] Obtain the attributes of the overflow variables and available memory blocks, and construct the migration cost matrix. The migration cost measurement formula is:

[0034] ;

[0035] Where, is the variable migration strategy function; is the initial state of the variable; For variables The target state after mapping; The cost of position change before and after the variable migration; is the weight coefficient; For variables from Migrate to Energy consumption during the process; The variable is in its initial state The probability density function of Initial state of the variable The differential of

[0036] Based on the migration cost measurement formula, the Sinkhorn algorithm is used to solve and determine the optimal transmission plan for migrating overflow variables to adjacent low-fragmentation blocks. The solution goal is: ;

[0037] Where, To find M that minimizes the value of the following expression; for Entropy; is the entropy regularization coefficient;

[0038] The proximal strategy optimization PPO algorithm is used to dynamically adjust the parameters during the migration process.

[0039] In some preferred embodiments, the entropy regularization coefficient Adaptively adjust according to the memory fragmentation rate.

[0040] In some preferred embodiments, the The proximal strategy is used to optimize the PPO algorithm for dynamic adjustment.

[0041] In some preferred embodiments, the N Dynamically adjust according to the error probability.

[0042] Beneficial effects of the present invention:

[0043] A hybrid prediction model (SDE+LSTM) is used to achieve millisecond-level response (≤5ms). Memory topology and Dijkstra algorithm are used to optimize offset calculation (time consumption <10ms). Combined with BCH+quantum layered verification, the fault tolerance is improved by 3 orders of magnitude (missed detection rate <10⁻ 9 ), and used Sinkhorn migration and PPO algorithms to reduce memory fragmentation to ≤10% and energy consumption by 35%; significantly improved the real-time performance, memory utilization, and system reliability of flight simulation, and solved the problems of traditional technologies such as long response delay (>50ms), high fragmentation rate (≥30%), and weak verification capability (missed detection rate 10⁻ 6) and other key issues, and is applicable to high-precision aviation simulation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0045] Figure 1 The present invention is a flowchart of the steps of a variable address offset management method for a flight simulator.

[0046] Figure 2 The present invention is a flowchart of layered BCH and quantum verification of a variable address offset management method of a flight simulator. DETAILED DESCRIPTION

[0047] The present application will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the relevant invention are shown in the accompanying drawings.

[0048] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0049] In order to more clearly illustrate the variable address offset management method of a flight simulator of the present invention, the following is combined with Figures 1 to 2 Each step in the embodiment of the present invention is described in detail.

[0050] The present invention proposes a variable address offset management method for a flight simulator, see Figure 1 , the method comprises the following steps:

[0051] Get the flight simulator's variable list, shared memory state matrix, and variable parameters;

[0052] In this embodiment, the variable list includes the variable ID (used to uniquely identify the variable), the initial size of the storage space occupied by the variable) and the data type (such as integer, floating point, etc.); the shared memory status matrix includes the block size B, the starting address A, the total capacity C, the available memory block D, the block size B Specifies the size of each storage block in the shared memory; the starting address A Indicates the starting position of the shared memory partition in memory; total capacity C Indicates the total amount of data that the shared memory partition can accommodate;

[0053] A variable monitoring queue using a sliding window is used to process the variable parameters to obtain a real-time data stream, and a hybrid prediction model is used to predict the real-time data stream to obtain an initial variable prediction value; the initial variable prediction value is combined with a gating weight to obtain a variable prediction value, and an early warning signal is obtained based on the variable prediction value. Specifically, when the difference between the variable prediction value and the actual value of the variable is greater than a threshold or a confidence interval, a warning signal is triggered;

[0054] In this embodiment, the hybrid prediction model predicts the real-time data stream in the following manner:

[0055] The jump-diffusion SDE model is used to calculate small changes, discretize the small changes into time intervals, and approximate the solution in each time interval to obtain a short-term prediction value. The jump-diffusion SDE model is:

[0056] ;

[0057] Where, is the long-term average trend of the variable over time; is a state variable; For small time intervals; is the degree of variable fluctuation; is the standard Brownian motion increment; is the jump amplitude; is the Poisson process increment;

[0058] The attention multi-scale residual long short-term memory network is used for long-term prediction. The long-term prediction value is obtained through multi-scale parallelism, double-layer attention, and residual fusion. The specific steps include the following:

[0059] Step 1. Get real-time data stream and process it:

[0060] Step 1.1: Get real-time data stream:

[0061] ,in, is the time step Observed values ​​of variables;

[0062] Step 1.2: Normalize and map to interval , and obtain the normalized ;

[0063] Step 2. Parallel extraction of multi-scale features:

[0064] Step 2.1: Normalize the Simultaneously feed into 3 parallel LSTM layers (short-term, medium-term, long-term):

[0065] Short-term LSTM: captures high-frequency changes (such as time step = 1), outputs features ;

[0066] Medium-term LSTM: captures medium-frequency changes (such as time step = 5), outputs features ;

[0067] Long-term LSTM: captures low-frequency changes (such as time step = 10), outputs features ;

[0068] Step 2.2: Concatenate the outputs of the 3-layer LSTM to generate a multi-scale feature vector , the dimension is 1×(D1+D2+D3);

[0069] Step 3. Temporal attention weight calculation:

[0070] Step 3.1: Multi-scale features Perform time step importance weighting:

[0071] The intermediate features are calculated through the first linear layer W1 and the Tanh activation function: ;

[0072] Generate temporal weights through the second linear layer W2 and the Softmax function: ;

[0073] Step 3.2: Time-based weighting Weighting multi-scale features to obtain temporal attention features ;

[0074] Step 4. Sequence Attention Dependency Modeling:

[0075] Step 4.1: Place the Mapped into query (Q), key (K), value (V) vectors:

[0076] ;

[0077] Step 4.2: Calculate the attention score for each time step in the sequence , represents the time step and Dependencies:

[0078] ;

[0079] Step 4.3: Generate context-aware features by weighted summation ;

[0080] Step 5. Residual fusion and feature enhancement:

[0081] Step 5.1: Concatenate multi-scale features and context-aware features ,get:

[0082] ;

[0083] Step 5.2: Use the fusion network (with linear layer and ReLU activation function) to Perform a nonlinear transformation:

[0084] ;

[0085] Step 5.3: Add residual connections to generate fusion features and retain the original feature information:

[0086] ;

[0087] Step 6. Generate predicted values:

[0088] Step 6.1: Extract the last time step Fusion features ;

[0089] Step 6.2: Map the output linear layer to the prediction space and obtain Predicted value at time:

[0090] ;

[0091] Where, is the weight matrix of the output linear layer; is the bias vector of the output linear layer;

[0092] Step 6.3: Forecasting Perform denormalization to restore the original data scale and obtain the final prediction result;

[0093] The calculation method of the gating weight is:

[0094] Calculate the short-term prediction value error and the long-term prediction value error based on the short-term prediction value and the long-term prediction value respectively;

[0095] splicing the short-term prediction value error and the long-term prediction value error to obtain a combined vector;

[0096] Performing a matrix multiplication operation on the combined vector and the trainable parameter, and adding a bias term after the matrix multiplication operation to obtain an intermediate variable;

[0097] The intermediate variable is input into the Sigmoid activation function to obtain the gating weight; the calculation formula of the gating weight is:

[0098] ;

[0099] Where, is the Sigmoid function; is a trainable parameter; is the bias term; They are the short-term forecast value error and the long-term forecast value error, respectively. The forecast value error is the difference between the forecast value and the actual value;

[0100] After the early warning signal is triggered, a memory topology map is constructed based on the shared memory state matrix, and the Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology map, thereby obtaining a new variable address offset table;

[0101] In this embodiment, the method for constructing a memory topology map based on a shared memory state matrix is ​​as follows:

[0102] Vertices are defined as variable addresses, and edge weights are constructed by reflecting the variable migration costs. The edge weight calculation formula is:

[0103] ;

[0104] Where, For variables size; is the data transmission delay between memory blocks; The remaining space for the target block; is the block size; Dependencies between variables (e.g. parent and child variables must be stored adjacently); All are weight coefficients;

[0105] Using proximal strategy to optimize PPO algorithm Perform dynamic adjustment; the specific principle and process are the same as the following: dynamically adjust the parameters during the migration process using the proximal strategy optimization PPO algorithm;

[0106] The Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology map, and the calculation method is:

[0107] ;

[0108] Where, is the calculated new offset; is the distance between memory blocks; is the degree of memory fragmentation; is the fragmentation penalty coefficient;

[0109] See also Figure 2 , verifying the variable address new offset table through layered BCH and quantum to obtain a verification result;

[0110] BCH performs the first level verification;

[0111] The probability of an error being detected in the first level of verification When it is greater than the set value (for example, 0.98), quantum verification is started, and quantum verification uses Pauli operator measurement:

[0112] ;

[0113] Where, is the expected value; is a quantum state; is the Pauli operator; N is the number of measurements;

[0114] The number of measurements of the quantum check N Dynamic adjustment based on error probability, the formula is: N = , (0.05) is the measurement accuracy parameter, ln is the natural logarithm function, and ⌈⌉ indicates rounding up;

[0115] When the verification result is successful, updating the variable list and shared memory state matrix of the flight simulator based on the variable address new offset table;

[0116] When the check result is failure, based on the variable address new offset table and the set of available memory blocks, the overflowed variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal strategy optimization (PPO) algorithm, and the variable address offset is updated, thereby updating the variable list and shared memory state matrix of the flight simulator.

[0117] In this embodiment, the overflow variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal policy optimization (PPO) algorithm. The method is as follows:

[0118] Obtain various attributes of overflow variables and available memory blocks (such as variable size, remaining space in memory blocks, time required for migration, etc.), and construct a migration cost matrix. The migration cost measurement formula is:

[0119] ;

[0120] Where, is the variable migration strategy function; is the initial state of the variable; For variables The target state after mapping; The cost of position change before and after the variable migration; is the weight coefficient; For variables from Migrate to Energy consumption during the process; The variable is in its initial state The probability density function of Initial state of the variable The differential of

[0121] Based on the migration cost measurement formula, the Sinkhorn algorithm is used to solve and determine the optimal transmission plan for migrating overflow variables to adjacent low-fragmentation blocks. The solution goal is: ;

[0122] Where, To find M that minimizes the value of the following expression; for Entropy; is the entropy regularization coefficient;

[0123] The proximal strategy optimization PPO algorithm is used to dynamically adjust the parameters in the migration process to minimize the long-term migration cost and fragmentation rate. Specifically, by defining the state space ( , respectively, resource utilization, fragmentation, energy consumption), action space ( ), reward function ( , is the resource utilization indicator, is the energy consumption indicator, To consume fragmentation indicators (including memory fragmentation rate FragRate), this reward function is used to guide the PPO algorithm to learn a better parameter adjustment strategy to optimize the variable migration process.

[0124] The Sinkhorn algorithm is used to solve the optimal transmission plan, and the entropy regularization coefficient Adaptive adjustment with memory fragmentation rate enables the Sinkhorn algorithm to better balance the optimization of migration cost and memory fragmentation under different memory fragmentation levels. When the fragmentation level is low, it focuses on reducing migration cost; when the fragmentation level is high, it focuses more on reducing fragmentation through migration operations, thereby improving memory utilization. The memory fragmentation rate FragRate is obtained by calculating the ratio of the total size of the fragment blocks in the memory to the total memory space. For example, when FragRate < 0.1, the memory fragmentation level is low, and it is appropriate to reduce it at this time. η The value of η =0.1−0.05×FragRate;

[0125] When 0.1≤FragRate≤0.3, keep η The value of is relatively stable, for example η =0.1. The degree of memory fragmentation is within an acceptable range and there is no need to significantly adjust the entropy regularization coefficient;

[0126] When FragRate>0.3, the memory fragmentation is high, increasing η The value of η =0.1+0.05×(FragRate−0.3);

[0127] Although the various steps in the above embodiment are described in the above-mentioned order, those skilled in the art will understand that in order to achieve the effect of this embodiment, different steps do not have to be executed in such an order. They can be executed simultaneously (in parallel) or in a reverse order. These simple changes are within the scope of protection of the present invention.

[0128] Those skilled in the art will appreciate that the modules and method steps described in the various examples described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of both. The programs corresponding to the software modules and method steps can be stored in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. To clearly illustrate the interchangeability of electronic hardware and software, the components and steps of each example have been generally described in terms of function in the above description. Whether these functions are implemented using electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the present invention. Terms such as "first" and "second" are used to distinguish similar objects and are not used to describe or indicate a specific order or precedence. The term "comprise" or any other similar term is intended to cover non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0129] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A method for managing variable address offsets of a flight simulator, characterized in that: The method comprises the following steps: Get the flight simulator's variable list, shared memory state matrix, and variable parameters; A variable monitoring queue of a sliding window is used to process the variable parameters to obtain a real-time data stream, and a hybrid prediction model is used to predict the real-time data stream to obtain an initial variable prediction value; the initial variable prediction value is combined with a gating weight to obtain a variable prediction value, and an early warning signal is obtained based on the variable prediction value; After the early warning signal is triggered, a memory topology map is constructed based on the shared memory state matrix, and the Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology map, thereby obtaining a new variable address offset table; Verifying the variable address new offset table by using layered BCH and quantum to obtain a verification result; When the verification result is successful, updating the variable list and shared memory state matrix of the flight simulator based on the variable address new offset table; When the verification result is failure, based on the new variable address offset table and the set of available memory blocks, the overflow variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal strategy optimization PPO algorithm, and the variable address offset is updated, thereby updating the variable list and shared memory status matrix of the flight simulator.

2. The variable address offset management method of a flight simulator according to claim 1, characterized in that: The hybrid prediction model predicts the real-time data stream in the following manner: The jump-diffusion SDE model is used to calculate small changes, discretize the small changes into time intervals, and approximate the solution in each time interval to obtain a short-term prediction value. The jump-diffusion SDE model is: ； Where, is the long-term average trend of the variable over time; For the moment state variables; For small time intervals; is the degree of variable fluctuation; is the standard Brownian motion increment; is the jump amplitude; is the Poisson process increment; The attention multi-scale residual long short-term memory network is used for long-term prediction, and the long-term prediction value is obtained through multi-scale parallelism, double-layer attention, and residual fusion.

3. The variable address offset management method of a flight simulator according to claim 2, characterized in that: The calculation method of the gating weight is: Calculate the short-term prediction value error and the long-term prediction value error based on the short-term prediction value and the long-term prediction value respectively; splicing the short-term prediction value error and the long-term prediction value error to obtain a combined vector; Performing a matrix multiplication operation on the combined vector and the trainable parameter, and adding a bias term after the matrix multiplication operation to obtain an intermediate variable; The intermediate variable is input into the Sigmoid activation function to obtain the gating weight.

4. The variable address offset management method of a flight simulator according to claim 1, characterized in that: The method for constructing a memory topology map based on a shared memory state matrix is ​​as follows: Vertices are defined as variable addresses, and edge weights are constructed by reflecting the variable migration costs. The edge weight calculation formula is: ； Where, For variables size; is the data transmission delay between memory blocks; The remaining space for the target block; is the block size; is the dependency relationship between variables; are all weight coefficients.

5. The variable address offset management method of a flight simulator according to claim 1, characterized in that: The Dijkstra algorithm is used to solve the optimal variable address offset on the memory topology map, and the calculation method is: ； Where, is the calculated new offset; is the distance between memory blocks; is the degree of memory fragmentation; is the fragmentation penalty coefficient.

6. The variable address offset management method of a flight simulator according to claim 1, characterized in that: Quantum verification uses Pauli operator measurement: ； Where, is the expected value; is a quantum state; is the Pauli operator; N is the number of measurements.

7. The variable address offset management method of a flight simulator according to claim 1, characterized in that: The overflow variables are migrated to adjacent low-fragmentation blocks using the Sinkhorn algorithm and the proximal strategy optimization PPO algorithm. The method is as follows: Obtain the attributes of the overflow variables and available memory blocks, and construct the migration cost matrix. The migration cost measurement formula is: ； Where, is the variable migration strategy function; is the initial state of the variable; For variables The target state after mapping; The cost of position change before and after the variable migration; is the weight coefficient; For variables from Migrate to Energy consumption during the process; Initial state of the variable The probability density function of Initial state of the variable The differential of Based on the migration cost measurement formula, the Sinkhorn algorithm is used to solve and determine the optimal transmission plan for migrating overflow variables to adjacent low-fragmentation blocks. The solution goal is: ； Where, To find M that minimizes the value of the following expression; for Entropy; is the entropy regularization coefficient; The proximal strategy optimization PPO algorithm is used to dynamically adjust the parameters during the migration process.

8. The variable address offset management method for a flight simulator according to claim 7, characterized in that: The entropy regularization coefficient Adaptively adjust according to the memory fragmentation rate.

9. The variable address offset management method for a flight simulator according to claim 4, characterized in that: described The proximal strategy is used to optimize the PPO algorithm for dynamic adjustment.

10. The variable address offset management method of a flight simulator according to claim 6, characterized in that: described N Dynamically adjust according to the error probability.

Images