Inertial platform stable loop disturbance estimation method based on hybrid network

Abstract

The invention particularly relates to an interference estimation method of an inertial platform stable loop based on an RNN and GRU hybrid neural network (RGNN), and belongs to the field of automatic control. According to the method, a model training set is made by collecting data information from an inertial platform; the data are input into an RGNN model which is composed of RNN and GRU and is connected in parallel for training. The RNN and the GRU respectively capture time sequence information in the data and extract advanced features through a unique mechanism. Next, designing a feature fusion layer, performing feature fusion on advanced features extracted by RNN and GRU, and taking the feature fusion as input of a full connection layer; and finally, the model performs interference estimation on the stable loop of the inertial platform by using the fused advanced features, and efficient interference estimation of the stable loop is realized through the steps. Simulation and tests show that the method effectively improves the stability and the anti-interference capability of the system.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and more specifically to a method for estimating the stability loop disturbance of an inertial platform based on a hybrid network. Background Technology

[0002] In related technologies, the inertial platform system is the core component of the entire inertial navigation system, characterized by high precision and performance, and has been widely used in aviation, navigation, and other fields. The stabilization loop is the main structure of the inertial platform in stable inertial space, enabling the platform to normally sense and measure various navigation parameters, providing an accurate and rapid navigation reference for the inertial navigation system. Today's increasingly complex application scenarios place new demands on the stabilization loop's dynamic accuracy and interference compensation capabilities.

[0003] Disturbances in the stabilization loop of an inertial platform significantly impact the system's accuracy and performance. Traditional disturbance identification and compensation methods exhibit limitations when dealing with complex external disturbances. For example, factors such as unbalanced torque, cable deflection torque, foundation coupling torque, and frictional torque can affect the stability of the inertial platform's stabilization loop.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method for estimating the stability loop disturbance of an inertial platform based on a hybrid network, a computer program product, and an electronic device, which can effectively overcome the defects existing in the prior art.

[0006] Other features and advantages of the invention will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0007] According to a first aspect of the present invention, a method for estimating the stabilization loop disturbance of an inertial platform based on a hybrid network is provided, the method comprising: Collect the platform parameters corresponding to the inertial platform to be processed and configure them as model input parameters. The platform parameters include at least one or any combination of multiple of the following: gyroscope, control circuit signal and motor current signal. The RNN and GRU models in the trained hybrid neural network model are used to extract features from the model input parameters to obtain the corresponding first feature data and second feature data; wherein, the hybrid neural network model is composed of a neural network based on the parallel connection of the RNN model and the GRU model; The first feature data and the second feature data are subjected to feature fusion processing to obtain the corresponding high-level feature data; The high-level feature data is processed using a fully connected layer to determine the perturbation estimation data.

[0008] In some exemplary embodiments, the RNN model includes an input layer, a hidden layer, and an output layer arranged sequentially; Hidden layers consist of a sequence of multiple identical neural network layers, each containing a recurrent neuron for information transfer and state recording between adjacent time steps; The first feature data is short-term feature data based on time series.

[0009] In some exemplary embodiments, the method further includes: an RNN model hidden layer. The output is based on the current model input parameters. and the adjacent previous hidden layer Decisions include:

[0010] Where U represents the connection weight matrix between the input layer and the hidden layer; This represents the connection weight matrix between adjacent hidden layers.

[0011] In some exemplary embodiments, the method further includes: the RNN model influencing the output parameters of the output layer according to each unit in the hidden layer, including:

[0012] in, Indicates the output value of the output layer; This represents the connection weight matrix between the hidden layer and the output layer; This indicates a hidden layer.

[0013] In some exemplary embodiments, the first feature data is long-term feature data; The GRU model extracts features from its input parameters, including: The GRU model utilizes the reset gate, update gate, retention of control information, and forgetting, including:

[0014]

[0015]

[0016]

[0017] in, Indicates the reset gate; Indicates an update to the door; Indicates the current input; This represents a summary of the current input and previous hidden layer states; This represents the output of the hidden layer after the memory has been updated; , and These represent the weight matrices for the reset gate, hidden layer, and update gate, respectively. and These are the Sigmoid and hyperbolic tangent activation functions, respectively.

[0018] In some exemplary embodiments, the method further includes: pre-training a hybrid neural network model, including: Construct hybrid neural network models, including parallel-connected RNN models and GRU models; Gaussian noise interference is injected into the inertial platform at the positions of the platform, gyroscope, and motor. Sample data is collected and a sample dataset is constructed. The sample data includes at least one or any combination of gyroscope, control circuit signals, and motor current signals. The initial hybrid neural network model was trained by backpropagation using a sample dataset; and the Adam optimizer was used to adaptively learn the model's hyperparameters. The model is evaluated using preset model evaluation metrics, and the hybrid neural network model is iteratively trained until the model converges. The preset model evaluation metrics include: root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE).

[0019] According to a second aspect of the present invention, a computer program product is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the above-described method for estimating the inertial platform stability loop disturbance based on a hybrid network is implemented.

[0020] According to a third aspect of the present invention, an electronic device is provided, comprising: Processor; and Memory for storing the executable instructions of the processor; The processor is configured to implement the above-described hybrid network-based inertial platform stability loop perturbation estimation method by executing the executable instructions.

[0021] According to a fourth aspect of the present invention, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for estimating the stability loop disturbance of an inertial platform based on a hybrid network.

[0022] The inertial platform stabilization loop perturbation estimation method based on a hybrid network provided in this invention pre-trains a hybrid neural network based on RNN and GRU, extracts high-level features from the time series using the RNN and GRU parts of the hybrid neural network respectively, and fuses these features in a feature fusion layer, thereby more accurately detecting and diagnosing inertial platform interference. The use of the hybrid neural network significantly improves the robustness and stability of the system, enabling it to exhibit higher accuracy and reliability in complex application scenarios. The data-driven approach reduces reliance on domain-specific knowledge, simplifies model design and debugging, and lowers the application threshold. Through end-to-end interference estimation, this method can directly extract high-level features from the data, improving the model's adaptability and robustness, and further ensuring the stable operation of the system in various complex environments.

[0023] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0025] Figure 1 This diagram illustrates an exemplary embodiment of the present invention: a method for estimating the stabilization loop disturbance of an inertial platform based on a hybrid network. Figure 2 The diagram illustrates an inertial platform stabilization loop according to an exemplary embodiment of the present invention. Figure 3 The diagram illustrates an exemplary embodiment of the RGNN model architecture of the present invention. Figure 4 The diagram illustrates an exemplary embodiment of an RNN network architecture according to the present invention. Figure 5 The diagram illustrates a GRU network architecture according to an exemplary embodiment of the present invention. Figure 6 This diagram illustrates a comparison between predicted and actual values ​​of an RGNN model, as shown in an exemplary embodiment of the present invention. Figure 7 The diagram illustrates the composition of an electronic device according to an exemplary embodiment of the present invention. Detailed Implementation

[0026] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the invention will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0027] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0028] In related technologies, some solutions compensate for unbalanced disturbance torque by offline identification of relevant model parameters and calculation of the unbalanced torque. However, such methods are limited to compensating for disturbance torque along the z-axis. Other solutions use a general-purpose reduced-order observer to estimate and compensate for disturbance torque using the angular rate signal output from the gyroscope and the current signal output from the motor. However, such methods have only been analyzed in simple systems and fail to effectively address the various disturbance factors in complex real-world scenarios.

[0029] To address the shortcomings and deficiencies of existing technologies, this example embodiment provides a method for estimating the disturbance of the stabilization loop of an inertial platform based on a hybrid network, which can be applied to the disturbance estimation of the stabilization loop of an inertial platform system. (Reference) Figure 1 As shown, the method for estimating the stabilization loop disturbance of an inertial platform based on a hybrid network may specifically include the following steps: Step S11: Collect the platform parameters corresponding to the inertial platform to be processed and configure them as model input parameters. The platform parameters include at least one or any combination of multiple of the following: gyroscope, control circuit signal and motor current signal. Step S12: Use the RNN model and GRU model in the trained hybrid neural network model to extract features from the model input parameters to obtain the corresponding first feature data and second feature data; wherein, the hybrid neural network model is composed of a neural network based on the parallel connection of the RNN model and the GRU model; Step S13: Perform feature fusion processing on the first feature data and the second feature data to obtain the corresponding high-level feature data; Step S14: Perform fully connected processing on the high-level feature data using a fully connected layer to determine the perturbation estimation data.

[0030] The following will describe in more detail each step of the inertial platform stabilization loop disturbance estimation method based on hybrid networks in this exemplary embodiment, with reference to the accompanying drawings and embodiments.

[0031] For example, refer to Figure 2 The stabilization loop of the inertial platform system shown in the figure first amplifies the gyroscope data through a preamplifier, and then demodulates it using a demodulator to convert the signal. The digital controller determines the corresponding control signal based on the demodulated signal, and then amplifies the control signal using an amplifier. The amplified digital control signal is then input to the DC torque motor to control the motor's output.

[0032] For example, the method includes: pre-training a hybrid neural network model, including: Step S21: Construct a hybrid neural network model, including a parallel-connected RNN model and a GRU model; Step S22: Inject Gaussian noise interference into the inertial platform at the positions of the platform, gyroscope, and motor, collect sample data, and construct a sample dataset; wherein, the sample data includes at least one or any combination of multiple of the following: gyroscope, control circuit signals, and motor current signals; Step S23: Use the sample dataset to perform backpropagation training on the initial hybrid neural network model; and use the Adam optimizer to adaptively learn the model hyperparameters. Step S24: Evaluate the model using preset model evaluation metrics and control the hybrid neural network model to perform iterative training until the model converges; wherein the preset model evaluation metrics include: root mean square error RMSE, mean square error MSE, and mean absolute error MAE.

[0033] Specifically, refer to Figure 3 The RGNN model architecture is a hybrid neural network model composed of parallel connections between an RNN model and a GRU model. During model training, platform parameters, including gyroscope output values, control circuit signals, and motor current signals, can be collected as input data for the inertial platform stabilization loop and fed into the RGNN model. The control circuit signals can be the values ​​of the input and / or output signals of a digital controller, i.e., the values ​​of PID control signals. Additionally, refer to... Figure 2 As shown, Gaussian noise interference can be injected at the locations of the platform, gyroscope, and motor to simulate external interference on the platform. The added Gaussian noise can be a random value.

[0034] refer to Figure 4As shown, RNN is a neural network structure used for feature extraction from time-series data, widely applied in fields such as acoustic signal processing, natural language recognition, and time-series prediction. Unlike feedforward neural networks, RNNs have time-series-related connections between neurons. Its basic structure is a sequence of multiple identical neural network layers, each with a recurrent neuron representing information transfer and state recording between adjacent time steps. In an RNN, the output at the current time step depends on the current input and all previous inputs. Therefore, RNNs can effectively extract high-level features from time-series data. The forward connection of an RNN can be understood as the propagation of a recurrent neural network, where each time-series data point is interconnected, meaning that the hidden information at the current time can be transmitted to the next time point. RNN hidden layers... From the current input and the previous hidden layer Decision. At time 1, the output value of RNN and hidden layers The definition is as follows: (1) (2) in, This represents the connection weight matrix between the hidden layer and the output layer of an RNN. This represents the connection weight matrix between the input layer and the hidden layer of an RNN. This represents the connection weight matrix between hidden layers in an RNN.

[0035] Each unit in the hidden layer of an RNN can affect the value of the output layer, similar to a fully connected layer, where the value of the output layer can be calculated from all the hidden units.

[0036] refer to Figure 5 As shown, the GRU network can enhance the network's ability to extract global features and historical information. Choosing the complementary approach of GRU and RNN ensures both long-term and short-term high-dimensional features of the time series data. The calculation formula for the GRU network is as follows: (3) (4) (5) (6) in, It's a reset door. It's an update door. This is the current input. This represents a summary of the current input and previous hidden layer states. It is the output of the hidden layer after the memory is updated. , and These are the weight matrices for the reset gate, hidden layer, and update gate, respectively. and These are the Sigmoid and hyperbolic tangent activation functions, respectively.

[0037] In the forward propagation process, at each time step t, the reset gate determines which parts of the previous state should be ignored. The update gate determines which information from the new input and the old state will be retained. Based on the reset gate and the current input, candidate hidden states are computed and then combined with the current input. The final hidden state is a linear combination of the old state and candidate states controlled by the update gate.

[0038] For the GRU backpropagation process, we can first calculate the gradient of the loss function with respect to the hidden state at the current time step; then, we use the chain rule to calculate the gradients with respect to the update gate and the reset gate. Next, we calculate the gradient with respect to the current candidate state and use the derivative of the tanh function. Then, we need to calculate the gradient with respect to the previous hidden state and combine it with the gradients from the backpropagation process to update the weights. Based on the gradients of the hidden state and the current candidate state at the current time step, we calculate the gradient with respect to the input and use the Adam optimizer to update the weights of the GRU model.

[0039] This method selects complementary approaches of GRU and RNN to ensure high-level features in both the long and short term of time series data.

[0040] In the feature fusion layer, a direct concatenation method is used to fuse the high-level features extracted by the RNN and GRU. Specifically, the feature fusion layer concatenates the high-level features output by the RNN and the GRU along the feature dimension, thus forming a comprehensive feature representation. This method can fully utilize the feature information extracted by each neural network, enhancing the model's expressive power and enabling it to have higher accuracy and robustness when dealing with complex application scenarios. The concatenated feature representation will contain the diverse information from both the RNN and GRU, helping to improve the accuracy of interference detection and diagnosis.

[0041] The fused high-level features are used as input to the fully connected layer. The fully connected layer further processes and integrates these fused features to generate the model's final output, which is the estimate of the perturbation at the current time step. The purpose of the fully connected layer is to integrate the cross-attention features trained in parallel and the current state information, and input them into the classification layer. The output of a fully connected network is a neuron corresponding to the perturbation label in the dataset. Sigmoid is chosen as the activation function for the fully connected layer, and Adam is chosen as the optimizer. The evaluation metrics used for training are RMSE, MSE, and MAE. Their calculation formulas are as follows:

[0042]

[0043]

[0044] For example, after the model training is completed, the perturbation estimation of the stable loop of the current inertial platform can be performed in real time during the operation of the inertial platform.

[0045] In step S11, the platform parameters corresponding to the inertial platform to be processed are collected and configured as model input parameters. The platform parameters include at least one or any combination of multiple of the following: gyroscope, control circuit signal and motor current signal.

[0046] In this embodiment, during the operation of the inertial platform, the platform parameters corresponding to the current inertial platform to be processed can be collected in real time and used as input parameters for the model. (Reference) Figure 2 The architecture shown allows multiple parameters, including those from the gyroscope, control circuit signals, and motor current signals, to be used as model input parameters.

[0047] In step S12, the RNN model and GRU model in the trained hybrid neural network model are used to extract features from the model input parameters to obtain the corresponding first feature data and second feature data; wherein, the hybrid neural network model is composed of a neural network based on the parallel connection of the RNN model and the GRU model.

[0048] For example, the RNN model includes an input layer, a hidden layer, and an output layer arranged in sequence; the hidden layer includes a sequence of multiple identical neural network layers, each containing a recurrent neuron for information transmission and state recording between adjacent time steps; wherein, the first feature data is short-term feature data based on time series.

[0049] Specifically, the forward connection of an RNN network can be understood as the propagation of a recurrent neural network, where the data of each time series is interconnected, meaning that the hidden information of the current time can be transmitted to the next time point.

[0050] For example, the first feature data is long-term feature data; the GRU model extracts features from the model input parameters, including: the GRU model utilizes the retention and forgetting of reset gate, update gate control information.

[0051] Specifically, the GRU network can enhance the network's ability to extract global features and historical information. By choosing the complementary approach of GRU and RNN, the high-dimensional features of both the long-term and short-term time series data can be ensured.

[0052] In step S13, feature fusion processing is performed on the first feature data and the second feature data to obtain the corresponding high-level feature data.

[0053] For example, the feature fusion layer directly concatenates the feature vectors extracted by RNN and GRU to form a new feature vector.

[0054] In step S14, the high-level feature data is processed using a fully connected layer to determine the perturbation estimation data.

[0055] For example, the output of a fully connected network is a neuron corresponding to a perturbation label in the dataset. Sigmoid can be chosen as the activation function for the fully connected layer.

[0056] For example, the following experimental data illustrates this method. The computer configuration is: CPU i9-13900, graphics processor NVIDIA GeForce RTX 4090. (Reference) Figure 6 The comparison between the predicted and actual values ​​of the RGNN model shown in the figure demonstrates that the method has a good prediction effect on the disturbances of the stabilization loop of the inertial platform, and the error between the predicted and actual values ​​is small.

[0057] The prediction results of this model with those of the KNN and SVM methods for Gaussian interference at different locations were compared. The results are summarized in Table 1 below: Table 1

[0058] As shown in the table above, for the RGNN method, we observed that the mean square error of the interference prediction for motor input, gyroscope output, and platform output gradually decreases, respectively. , and In contrast, the KNN method can only predict disturbances at the motor input and platform output, with mean square errors at these two points being... The mean square error (MSE) is 0.397, but it cannot provide interference prediction at the gyroscope output. The SVM method can only predict interference at the motor input, and its MSE at the motor input is... However, it cannot provide interference prediction for gyroscope output and platform output. The results show that the RGNN method has advantages over the KNN and SVM methods in interference prediction at different locations. The RGNN method not only has good prediction performance for motor input but also effectively predicts interference for gyroscope output and platform output, demonstrating superiority in interference estimation tasks for inertial platform stabilization loops. The method provided in this invention constructs a parallel-connected RGNN model composed of an RNN and a GRU. The RNN and GRU respectively capture time-series information from the data and extract high-level features through their unique mechanisms. Next, a feature fusion layer is designed to fuse the high-level features extracted by the RNN and GRU, and this fusion is used as input to the fully connected layer. Finally, the model uses the fused high-level features to estimate the disturbance in the stable loop of the inertial platform. Through the above steps, efficient disturbance estimation of the stable loop is achieved. Simulation and testing show that this method effectively improves the stability and disturbance rejection capability of the system.

[0059] It should be noted that the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may, for example, be executed synchronously or asynchronously in multiple modules.

[0060] It should be noted that although several modules or units of the device for performing actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of the present invention, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0061] Figure 7 A schematic diagram of an electronic device suitable for implementing embodiments of the present invention is shown.

[0062] It should be noted that, Figure 7 The electronic device 1000 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0063] like Figure 7As shown, the electronic device 1000 includes a Central Processing Unit (CPU) 1001, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 1002 or programs loaded from storage section 1008 into Random Access Memory (RAM) 1003. The RAM 1003 also stores various programs and data required for system operation. The CPU 1001, ROM 1002, and RAM 1003 are interconnected via a bus 1004. An Input / Output (I / O) interface 1005 is also connected to the bus 1004. Furthermore, the electronic device 1000 also includes an FPGA device and a System-on-a-Chip (SoC) device.

[0064] The following components are connected to I / O interface 1005: an input section 1006 including a keyboard, mouse, etc.; an output section 1007 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to I / O interface 1005 as needed. Removable media 1011, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1010 as needed so that computer programs read from them can be installed into storage section 1008 as needed.

[0065] In particular, according to embodiments of the present invention, the processes described below with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1009, and / or installed from removable medium 1011. When the computer program is executed by central processing unit (CPU) 1001, it performs various functions defined in the system of this application.

[0066] Specifically, the aforementioned electronic devices can be airborne intelligent electronic devices, such as airborne video processing equipment.

[0067] It should be noted that the storage medium shown in the embodiments of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, wherein computer-readable program code is carried. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any storage medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0068] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0069] The units described in the embodiments of the present invention can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0070] It should be noted that, as another aspect, this application also provides a storage medium, which may be included in an electronic device or may exist independently without being assembled into the electronic device. The aforementioned storage medium carries one or more programs, which, when executed by an electronic device, cause the electronic device to perform the methods described in the following embodiments. For example, the electronic device may perform... Figure 1 The steps of the method shown.

[0071] In one embodiment, this application provides a computer program product including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0072] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0073] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims.

[0074] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for estimating the stability loop perturbation of an inertial platform based on a hybrid network, characterized in that, The method includes: Collect the platform parameters corresponding to the inertial platform to be processed and configure them as model input parameters. The platform parameters include at least one or any combination of multiple of the following: gyroscope, control circuit signal and motor current signal. The RNN and GRU models in the trained hybrid neural network model are used to extract features from the model input parameters to obtain the corresponding first feature data and second feature data; wherein, the hybrid neural network model is composed of a neural network based on the parallel connection of the RNN model and the GRU model; The first feature data and the second feature data are subjected to feature fusion processing to obtain the corresponding high-level feature data; The high-level feature data is processed using a fully connected layer to determine the perturbation estimation data.

2. The method according to claim 1, characterized in that, The RNN model includes an input layer, a hidden layer, and an output layer arranged sequentially. Hidden layers consist of a sequence of multiple identical neural network layers, each containing a recurrent neuron for information transfer and state recording between adjacent time steps; The first feature data is short-term feature data based on time series.

3. The method according to claim 2, characterized in that, The method also includes: RNN model hidden layers. The output is based on the current model input parameters. and the adjacent previous hidden layer Decisions include: Where U represents the connection weight matrix between the input layer and the hidden layer; This represents the connection weight matrix between adjacent hidden layers.

4. The method according to claim 2, characterized in that, The method further includes: the RNN model influencing the output parameters of the output layer based on each unit in the hidden layer, including: in, Indicates the output value of the output layer; This represents the connection weight matrix between the hidden layer and the output layer; This indicates a hidden layer.

5. The method according to claim 1, characterized in that, The first feature data is long-term feature data; The GRU model extracts features from its input parameters, including: The GRU model utilizes the reset gate, update gate, retention of control information, and forgetting, including: in, Indicates the reset gate; Indicates an update to the door; Indicates the current input; This represents a summary of the current input and previous hidden layer states; This represents the output of the hidden layer after the memory has been updated; , and These represent the weight matrices for the reset gate, hidden layer, and update gate, respectively. and These are the Sigmoid and hyperbolic tangent activation functions, respectively.

6. The method according to claim 1, characterized in that, The method further includes: pre-training a hybrid neural network model, including: Construct hybrid neural network models, including parallel-connected RNN models and GRU models; Gaussian noise interference is injected into the inertial platform at the positions of the platform, gyroscope, and motor. Sample data is collected and a sample dataset is constructed. The sample data includes at least one or any combination of gyroscope, control circuit signals, and motor current signals. The initial hybrid neural network model was trained by backpropagation using a sample dataset; and the Adam optimizer was used to adaptively learn the model's hyperparameters. The model is evaluated using preset model evaluation metrics, and the hybrid neural network model is iteratively trained until the model converges. The preset model evaluation metrics include: root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE).

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for estimating the stability loop disturbance of an inertial platform based on a hybrid network, as described in any one of claims 1 to 6.

8. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the hybrid network-based inertial platform stability loop perturbation estimation method according to any one of claims 1 to 6 by executing the executable instructions.

9. A storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the method for estimating the stability loop disturbance of an inertial platform based on a hybrid network, as described in any one of claims 1 to 6.

Images