An Error Compensation Method for Hemispherical Resonant Gyroscopes Based on LSTM
By using an LSTM-based error compensation method, the problem of complex temperature error models in hemispherical resonant gyroscopes was solved, achieving high-precision and highly adaptable error compensation, thus improving the measurement accuracy and real-time performance of the gyroscope.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING AUTOMATION CONTROL EQUIP INST
- Filing Date
- 2024-12-27
- Publication Date
- 2026-06-30
AI Technical Summary
In existing technologies, the temperature error model of hemispherical resonant gyroscopes is complex and difficult to effectively compensate for, affecting the accuracy and real-time performance of the gyroscope.
An LSTM-based error compensation method is adopted. By acquiring the gyroscope output at multiple temperature points, spline interpolation is performed for noise reduction preprocessing to establish an LSTM model, eliminating the influence of temperature on the gyroscope output. The compensation model is then programmed into the control circuit to achieve real-time compensation.
It achieves high-precision, high-adaptability, and low-complexity error compensation, improving the angular velocity measurement accuracy and overall performance of hemispherical resonant gyroscopes.
Smart Images
Figure CN122306036A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of hemispherical resonator gyroscope technology, and particularly relates to an error compensation method for hemispherical resonator gyroscopes based on LSTM. Background Technology
[0002] The physical field factors affecting the performance of a hemispherical resonator gyroscope mainly include temperature, stress, acoustic, and electromagnetic fields, with temperature playing a dominant role. Temperature error is a key factor limiting the accuracy of the gyroscope. Establishing an accurate temperature compensation model for the hemispherical resonator gyroscope is crucial for its application.
[0003] In a hemispherical resonant gyroscope, the temperature-sensitive components are the gyroscope head and the control circuit. Temperature error includes the temperature error of the gyroscope head and the temperature error of the gyroscope control circuit.
[0004] When a hemispherical resonator gyroscope is operating normally, changes in the external environment and the heat generated by the vibration of the gyroscope's resonator will cause changes in the gyroscope's temperature. On the one hand, when the resonator's temperature changes, parameters such as the resonator's elastic modulus, Poisson's ratio, material density, and membrane stress will be affected, thus altering the gyroscope's mathematical model and causing significant errors in the drive control and detection processes. On the other hand, changes in the gyroscope's temperature will also alter the resonator's stiffness coefficient and the damping coefficient within the gyroscope's resonant cavity, causing changes in the resonant frequency and consequently altering the gyroscope's quality factor (Q value), further affecting the gyroscope's accuracy.
[0005] After the gyroscope is started, the digital control circuit inevitably generates heat. Since the circuit board is installed in a relatively enclosed and small housing, the heat is not easily dissipated in time, causing the circuit temperature to rise and resulting in temperature errors in the circuit's measurement output. From cold start to thermal equilibrium, the circuit's temperature changes with the ambient temperature, altering the performance parameters of the components used in the circuit and introducing measurement errors.
[0006] Because the temperature drift affecting the accuracy of a hemispherical resonator (HNR) is complex, the temperature error model of the HNR can be viewed as a gray box model. This involves selecting a suitable model from a set of models and, according to certain criteria, ensuring it optimally describes the nature of the HNR's temperature error. The established HNR temperature error model should meet the following requirements: 1. Accuracy: The established mathematical model should optimally describe the temperature error characteristics of the HNR according to predetermined target criteria. 2. Applicability: The mathematical model of the HNR's temperature error characteristics should be effectively applicable to the HNR under different environments and conditions. Generally, it is advisable to avoid establishing multiple different mathematical models for the same HNR, as this reduces the model's versatility. 3. Complexity: Overly complex models are difficult to implement in engineering applications. Compensation requires a significant amount of computation time, compromising real-time performance and negating the fundamental purpose of modeling and compensation. Summary of the Invention
[0007] The present invention aims to solve at least one of the technical problems existing in the prior art.
[0008] This invention provides an LSTM-based method for compensating errors in a hemispherical resonator gyroscope. The method includes: acquiring the gyroscope output at multiple temperature points; performing spline interpolation noise reduction preprocessing on the acquired gyroscope output; calculating the output drift of the hemispherical resonator gyroscope at each temperature point as a reference value for evaluating the temperature error compensation effect; establishing a temperature error compensation model for the hemispherical resonator gyroscope based on an LSTM model; verifying the temperature error compensation effect of the hemispherical resonator gyroscope by combining the output drift of the hemispherical resonator gyroscope at each temperature point; and programming the verified temperature error compensation model into the hemispherical resonator gyroscope control circuit to achieve real-time error compensation.
[0009] Furthermore, the LSTM model consists of, in sequence: a linear layer, a Dropout layer, an LSTM layer, and a fully connected layer.
[0010] Furthermore, the LSTM layer includes several LSTM neural network units.
[0011] Furthermore, the input to any LSTM neural network unit is the data x at the current time. t The output h of the LSTM neural network unit at the previous time step t-1 and the memory state c hidden in the previous LSTM neural network unit t-1 The corresponding output is the output value h at the current time. t and the memory state c of the LSTM neural network unit at the current moment. t .
[0012] Furthermore, any LSTM neural network unit has three gate states: the forget gate f t Input gate i t and output gate o t Forgotten Gate f t The memory state from the previous moment is filtered and retained, and the input gate i t The input and output gates determine the current moment's data input to be stored in the memory unit. t The memory unit that controls the output at the current moment.
[0013] Furthermore, the forgetting gate f t =σ(W f ·[h t-1 ,x t ]+b f ), where W f The weight matrix of the forget gate, b f Let represent the bias vector of the forget gate, σ(·) represent the sigmoid activation function, and [·,·] represent vector concatenation.
[0014] Furthermore, input gate i t =σ(W i ·[h t-1 ,x t ]+b i ), where W i Let b represent the weight matrix of the input gate. i This represents the bias vector of the input gate.
[0015] Furthermore, output gate o t =σ(W o ·[h t-1 ,x t ]+b o ), where W o Let b represent the weight matrix of the output gate. o This represents the bias vector of the output gate.
[0016] Furthermore, the memory state of the LSTM neural network unit at the current moment. in, tanh(·) denotes the hyperbolic tangent activation function, W c The weight matrix representing the cell state, b c A bias vector representing the cell state.
[0017] Furthermore, the output value h at the current moment t =o t *tanh(c t ).
[0018] This invention provides an LSTM-based error compensation method for hemispherical resonant gyroscopes. This method performs spline interpolation noise reduction preprocessing on the acquired gyroscope output. Based on a deep learning LSTM model, it establishes the relationship between temperature and hemispherical resonant gyroscope error, eliminating the influence of temperature on the output angular rate of the hemispherical resonant gyroscope. This achieves high-precision, highly adaptable, and low-complexity error compensation for hemispherical resonant gyroscopes, improving the accuracy of hemispherical resonant gyroscope angular velocity measurement and enhancing the overall performance of hemispherical resonant gyroscopes in practical applications. Compared with existing technologies, this invention solves the technical problem of complex hemispherical resonant gyroscope error models and difficulties in effective compensation in existing technologies. Attached Figure Description
[0019] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, 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 creative effort.
[0020] Figure 1 A flowchart illustrating an LSTM-based hemispherical resonator gyroscope error compensation method according to a specific embodiment of the present invention is shown.
[0021] Figure 2 A schematic diagram of the structure of an LSTM model provided according to a specific embodiment of the present invention is shown;
[0022] Figure 3 A schematic diagram of the structure of an LSTM neural network unit provided according to a specific embodiment of the present invention is shown. Detailed Implementation
[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0025] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0026] like Figure 1 As shown in the figure, a method for compensating for errors in a hemispherical resonant gyroscope based on LSTM is provided according to a specific embodiment of the present invention. The method includes:
[0027] Obtain the gyroscope output of a hemispherical resonant gyroscope at multiple temperature points;
[0028] The acquired gyroscope output is preprocessed using spline interpolation for noise reduction.
[0029] The output drift of the hemispherical resonant gyroscope at various temperature points is calculated as a reference value for evaluating the temperature error compensation effect;
[0030] A temperature error compensation model for a hemispherical resonant gyroscope is established based on the LSTM model.
[0031] The temperature error compensation effect of the hemispherical resonant gyroscope is verified by combining the output drift of the hemispherical resonant gyroscope at various temperature points;
[0032] The verified temperature error compensation model for the hemispherical resonator gyroscope is programmed into the hemispherical resonator gyroscope control circuit to achieve real-time error compensation.
[0033] This configuration method provides an LSTM-based error compensation method for hemispherical resonant gyroscopes. The method preprocesses the acquired gyroscope output with spline interpolation for noise reduction. Based on a deep learning LSTM model, it establishes the relationship between temperature and hemispherical resonant gyroscope error, eliminating the influence of temperature on the output angular rate of the hemispherical resonant gyroscope. This achieves high-precision, highly adaptable, and low-complexity error compensation for hemispherical resonant gyroscopes, improving the accuracy of hemispherical resonant gyroscope angular velocity measurement and enhancing the overall performance of hemispherical resonant gyroscopes in practical applications.
[0034] First, in this invention, the gyroscope output of the hemispherical resonator gyroscope at multiple temperature points is obtained; then, spline interpolation noise reduction preprocessing is performed on the obtained gyroscope output.
[0035] The high-frequency output angular rate of a hemispherical resonant gyroscope contains high-frequency random noise, which reduces the fitting accuracy of the model during model building. Therefore, it is necessary to first perform noise reduction preprocessing on the angular velocity rate output by the hemispherical resonant gyroscope to ensure the accuracy of subsequent modeling. This invention proposes to use spline interpolation to perform noise reduction processing on the high-frequency output angular rate of the hemispherical resonant gyroscope.
[0036] For the interpolation function f(x), the nodes x0, x1, ..., xn are known. n The function value at a given point is used to divide adjacent nodes into n interpolation intervals. Within each interpolation interval, a polynomial of degree k is used for interpolation to satisfy the interpolation conditions and k-1 order smoothness.
[0037] S i (x i )=f(x i ), S i (x i+1 )=f(x i+1 )
[0038]
[0039] Where, f(x) i ) and f(x i+1 ) respectively represent at node x i and x i+1 The interpolation function to be interpolated, i.e., the original output angular velocity of the hemispherical resonant gyroscope to be interpolated, S i (x i ) and S i (x i+1 ) respectively represent at node x i and x i+1 Polynomial function S constructed from intervals i (x) at node x i and x i+1 The spline interpolation result, i.e., the angular velocity after noise reduction, S i+1(x i+1 ) indicates at node x i+1 and x i+2 Polynomial function S constructed from intervals i+1 (x) at node x i+1 The spline interpolation results, S represents i (x i+1 The j-th derivative of ) S represents i+1 (x i+1 The j-th derivative of ), where n is the number of interpolation subintervals and k is the degree of the polynomial.
[0040] Furthermore, in this invention, the output drift of the hemispherical resonant gyroscope at each temperature point is calculated as a reference value for evaluating the temperature error compensation effect; a temperature error compensation model for the hemispherical resonant gyroscope is established based on the LSTM model.
[0041] Fused silica is highly sensitive to temperature changes and easily affected by base vibrations, resulting in strong nonlinearity in the output of hemispherical resonant gyroscopes. Traditional fitting methods cannot accurately describe this nonlinearity. This invention establishes a model based on LSTM to model the relationship between temperature and the angular rate error of the hemispherical resonant gyroscope.
[0042] The compensation model of this invention adopts a fusion model based on few-shot learning prototype network and LSTM. The LSTM model structure is as follows: Figure 2 As shown, it includes, in sequence: a linear layer, a Dropout layer, an LSTM layer, and a fully connected layer.
[0043] The denoised preprocessed data is input into the LSTM model in batches according to time steps for latent representation extraction; the LSTM layer structure diagram is as follows. Figure 3 As shown, the LSTM layer consists of several LSTM neural network units. For each LSTM neural unit, the input is the data x at the current time. t The output h of the LSTM neural network unit at the previous time step t-1 and the memory state c hidden in the previous LSTM neural network unit t-1 The corresponding output is the output value h at the current time. t and the memory state c of the LSTM neural network unit at the current moment. t .like Figure 3 As shown, in an LSTM neural network unit, rectangles represent different activation functions, and circles represent different vector operations. Together, they constitute the three gating states in the LSTM neuron unit: the forget gate f t Input gate i t and output gate o tThe forget gate filters and retains the memory state from the previous time step. The input gate determines how much of the current data input can be remembered by the memory unit, and the output gate controls how much of the memory unit can output to the current output. The state of each gate is updated according to the following formula:
[0044] Forgotten Gate: f t =σ(W f ·[h t-1 ,x t ]+b f )
[0045] Input gate: i t =σ(W i ·[h t-1 ,x t ]+b i )
[0046] Cell state:
[0047]
[0048] Output gate: o t =σ(W o ·[h t-1 ,x t ]+b o )
[0049] h t =o t *tanh(c t )
[0050] Among them, W f W i W c and W o Let b represent the weight matrices for the forget gate, input gate, cell state, and output gate, respectively. f b i b c and b o Let represent the bias vectors of the forget gate, input gate, cell state, and output gate, respectively; tanh(·) represents the hyperbolic tangent activation function; σ(·) represents the sigmoid activation function; and [·,·] represents vector concatenation.
[0051] Furthermore, in this invention, the temperature error compensation effect of the hemispherical resonant gyroscope is verified by combining the output drift of the hemispherical resonant gyroscope at various temperature points. The verified temperature error compensation model of the hemispherical resonant gyroscope is then burned into the hemispherical resonant gyroscope control circuit to achieve real-time error compensation.
[0052] In summary, this invention proposes an LSTM-based error compensation method for hemispherical resonant gyroscopes. Based on a deep learning LSTM model, the method establishes the relationship between temperature and hemispherical resonant gyroscope error, eliminating the influence of temperature on the output angular rate of the hemispherical resonant gyroscope. This achieves high-precision, highly adaptable, and low-complexity error compensation for hemispherical resonant gyroscopes, thereby improving the performance of the hemispherical resonant gyroscope.
[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for error compensation of a hemispherical resonant gyroscope based on LSTM, characterized in that, The LSTM-based hemispherical resonator gyroscope error compensation method includes: Obtain the gyroscope output of a hemispherical resonant gyroscope at multiple temperature points; The acquired gyroscope output is preprocessed using spline interpolation for noise reduction. The output drift of the hemispherical resonant gyroscope at various temperature points is calculated as a reference value for evaluating the temperature error compensation effect; A temperature error compensation model for a hemispherical resonant gyroscope is established based on the LSTM model. The temperature error compensation effect of the hemispherical resonant gyroscope is verified by combining the output drift of the hemispherical resonant gyroscope at various temperature points; The verified temperature error compensation model for the hemispherical resonator gyroscope is programmed into the hemispherical resonator gyroscope control circuit to achieve real-time error compensation.
2. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 1, characterized in that, The LSTM model consists of, in sequence: a linear layer, a Dropout layer, an LSTM layer, and a fully connected layer.
3. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 1, characterized in that, The LSTM layer includes several LSTM neural network units.
4. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 3, characterized in that, The input of any LSTM neural network unit is the data x at the current time t , the output h of the previous time LSTM neural network unit t-1 , and the hidden memory state c of the previous LSTM neural network unit t-1 The corresponding output is the output value h at the current time t and the memory state c of the current LSTM neural network unit t .
5. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 4, characterized in that, Any LSTM neural network unit has three gate states: the forget gate f t Input gate i t and output gate o t Forgotten Gate f t The memory state from the previous moment is filtered and retained, and the input gate i t The input and output gates determine the current moment's data input to be stored in the memory unit. t The memory unit that controls the output at the current moment.
6. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 5, characterized in that, Forgotten Gate f t =σ(W f ·[h t-1 ,x t ]+b f ), where W f The weight matrix of the forget gate, b f Let represent the bias vector of the forget gate, σ(·) represent the sigmoid activation function, and [·,·] represent vector concatenation.
7. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 5, characterized in that, Input gate i t =σ(W i ·[h t-1 ,x t ]+b i ), where W i Let b represent the weight matrix of the input gate. i This represents the bias vector of the input gate.
8. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 5, characterized in that, Output gate o t =σ(W o ·[h t-1 ,x t ]+b o ), where W o Let b represent the weight matrix of the output gate. o This represents the bias vector of the output gate.
9. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 5, characterized in that, The memory state of the LSTM neural network unit at the current moment in, tanh(·) denotes the hyperbolic tangent activation function, W c The weight matrix representing the cell state, b c A bias vector representing the cell state.
10. The LSTM-based hemispherical resonator gyroscope error compensation method according to claim 5, characterized in that, The output value h at the current moment t =o t *tanh(c t ).