Error prediction method for inertial navigation equipment

Abstract

The invention discloses an error prediction method of inertial navigation equipment, which belongs to the technical field of inertial localization, and comprises the steps of data acquisition and preprocessing, error modeling and preliminary fusion, physical constraint-based LSTM modeling, LSTM optimization, and error correction and feedback optimization, and solves the technical problem of improving the real-time performance and accuracy of a navigation system. According to the method, an IMU drift formula and an environment temperature error are fused for modeling, the prediction precision is enhanced, a physical interpretation force is given to the model, the adaptability of a system to a dynamic environment is improved, a real-time error is used as a reward signal, an LSTM weight and a covariance matrix are dynamically optimized, the prediction capability under long-time work is improved, feedback optimization is performed on corrected navigation data, and the prediction precision is improved. And model self-learning and performance iteration are realized.

Description

Technical Field

[0001] The invention belongs to the technical field of inertial positioning, and in particular relates to a method for error prediction of an inertial navigation device. Background Art

[0002] Inertial Navigation System (INS) combines IMU sensors with GNSS receivers to achieve precise navigation. However, IMU sensors are susceptible to drift errors, GNSS signals are unstable in occluded environments, and ambient temperature changes can also affect navigation accuracy. In existing technologies, inertial navigation often uses Kalman filtering to fuse data, or relies on data-driven deep learning models, but both have shortcomings:

[0003] 1. Traditional Kalman filters cannot adapt to complex dynamic environments in real time.

[0004] 2. Purely data-driven models lack physical constraints, making it difficult to explain the results and prone to failure under external disturbances. Summary of the invention

[0005] The purpose of the present invention is to provide a method for error prediction of an inertial navigation device, which solves the technical problem of improving the real-time and accuracy of a navigation system by directly embedding an IMU physical drift formula and environmental error into an LSTM model, combining a dynamic adaptive mechanism and reinforcement learning optimization.

[0006] To achieve the above purpose, the present invention adopts the following technical inventions:

[0007] A method for error prediction of an inertial navigation device comprises the following steps:

[0008] Step 1: The data acquisition module obtains IMU data P from the IMU sensor, GNSS receiver and environmental sensor respectively IMU (t), GNSS positioning data P GNSS (t) and ambient temperature data T env (t), and preprocessing; IMU data PIMU (t) includes acceleration a (t) and angular velocity ω (t);

[0009] Preprocessing includes IMU data P IMU (t) Noise filtering is performed on the ambient temperature data T env (t) is smoothed and the IMU data P is aligned using the timestamp IMU (t) and GNSS positioning data P GNSS (t), and interpolate and extend the GNSS positioning data P GNSS (t) to IMU data P IMU (t) time frequency; time interval error E ΔtPerform mathematical modeling; the preprocessed data and time interval error E Δt Combine and obtain the feature vector X(t);

[0010] Step 2: Data fusion module for IMU data P IMU (t) drift error E IMU (t) and the temperature error E of the ambient temperature data T Mathematical modeling is performed respectively according to the drift error E IMU (t), temperature error E T and time interval error E Δt , generate the initial fusion error E 融合 (t); the drift error E IMU (t) and temperature error E T Add to the feature vector X(t);

[0011] Step 3: The LSTM calculation module retrieves the feature vector X(t) and the fusion error E 融合 (t), using the LSTM model and physical constraints to predict sensor errors, we get the error prediction result E LSTM (t);

[0012] Step 4: Optimize the module to obtain real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the error prediction result E LSTM (t) Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes, use reinforcement learning optimization method to establish a reward function, dynamically adjust the LSTM weights and noise covariance matrix, obtain the optimized LSTM model weights and dynamically adjusted noise covariance matrix, and improve the real-time performance of error prediction and correction;

[0013] Step 5: Use the prediction error to correct the inertial data and continuously optimize the model through a feedback mechanism to output high-precision navigation data.

[0014] Preferably, when executing step 1, the feature vector X(t) is represented by the following formula:

[0015] X(t)=[a(t),ω(t),P GNSS (t),T env (t),E Δt ];

[0016] Time interval error E Δt Calculated by the following formula:

[0017] E Δt =β×Δt;

[0018] Among them, Δt is the time interval, which represents the IMU data P IMU (t) and GNSS positioning data P GNSS (t) is the time synchronization error at the same time; β is the time error influence coefficient, which represents the influence of the time delay between sensors.

[0019] Preferably, when executing step 2, the specific steps are as follows:

[0020] Step 2-1: IMU data P IMU (t) drift error E IMU (t) Modeling, as shown in the following formula:

[0021] E IMU (t) = E IMU (t-1)+α×Δt;

[0022] Among them, E IMU (t) is the IMU error at the current time point t, E IMU (t-1) is the IMU error at the previous time point t-1, α is the drift rate, and Δt is the time interval;

[0023] Step 2-2: Temperature error E of ambient temperature data T Modeling, as shown in the following formula:

[0024] E T =γ×(T env -T0);

[0025] Among them, T env is the ambient temperature, T0 is the calibration temperature of the ambient sensor, γ is the temperature sensitivity coefficient; the ambient sensor is a temperature sensor;

[0026] Step 2-3: Convert the drift error E IMU (t), temperature error E T and time interval error E Δt Fusion is performed to obtain the fusion error E 融合 (t):

[0027] E 融合 (t) = E IMU (t)+E T +E Δt ;

[0028] Step 2-4: Convert the drift error E IMU (t) and temperature error E T After being added to the feature vector X(t), the feature vector X(t) is expressed as follows:

[0029] X(t)=[a(t),ω(t),P GNSS(t),T env (t),E IMU (t),E T ,E Δt ].

[0030] Preferably, when executing step 3, the specific steps are as follows:

[0031] Step 3-1: Feature input, specifically, taking the feature vector X(t) obtained in step 2-4 as the feature vector input into the input layer of the LSTM model;

[0032] Step 3-2: Embed physical constraints in the LSTM model, specifically by transforming the drift error formula E IMU (t) = E IMU (t-1)+α×Δt and temperature error formula E T =γ×(T env -T0) is directly embedded in the input and gating mechanism of the LSTM model;

[0033] Introduce drift rate α and time interval β in the forget gate;

[0034] LSTM model fitting prediction result E LSTM1 (t);

[0035] Step 3-3: Physical constraint regularization, specifically introducing the regularization term L phy Constrain the LSTM loss function:

[0036] L phy =λ phy ×||E L (t)-(E IMU (t)+E T +E Δt )|| 2 ;

[0037] Among them, E L (t) is the error calculation of the intermediate process, which is used to constrain the degree of compliance of the LSTM model with physical laws; λ phy is the regularization coefficient, which adjusts the strength of the physical error constraint;

[0038] Physical constraint regularization is used to encourage the prediction results E LSTM1 (t) satisfy physical constraints;

[0039] Step 3-4: Combine the error loss and physical regularization loss to optimize the LSTM model weights and generate the final prediction result E LSTM (t), the total loss function L of LSTM is:

[0040] L total =Ldata +λL phy ;

[0041] L data =||E true (t)-E LSTM1 (t)|| 2 ;

[0042] Among them, λ is the weight factor, L data is the error loss based on the data.

[0043] Preferably, when executing step 4, the following steps are specifically included:

[0044] Step 4-1: Get real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the prediction result E LSTM (t);

[0045] Step 4-2: Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes:

[0046] Q adaptive =Q0+k1×|T env -T0|;

[0047] R adaptive =R0+k2×SNR IMU ;

[0048] Among them, Q adaptive is the process noise covariance matrix, which is affected by temperature changes; R adaptive is the measurement noise covariance matrix, which is affected by the signal-to-noise ratio; k1 and k2 are adjustment coefficients; Q0 is the initial value of the process noise covariance matrix; R0 is the initial value of the measurement noise covariance matrix;

[0049] Step 4-3: Use reinforcement learning optimization method to establish reward function Reward:

[0050] Reward=-||E true (t)-E LSTM (t)|| 2 ;

[0051] The LSTM weights and Kalman filter covariance matrix are dynamically adjusted through the reward function Reward to optimize the LSTM model and noise covariance matrix.

[0052] Preferably, when executing step 5, the IMU data P is calculated according to the following formula: IMU (t) Make corrections to obtain the corrected navigation data P corrected (t):

[0053] P corrected (t) = P IMU (t)-E LSTM (t);

[0054] The corrected data is used to update the parameters of the LSTM model and the Kalman filter to continuously optimize the model through the feedback mechanism, and the navigation data P is output. corrected (t).

[0055] The error prediction method of an inertial navigation device disclosed in the present invention solves the technical problem of improving the real-time performance and accuracy of a navigation system by directly embedding an IMU physical drift formula and an environmental error into an LSTM model, combining a dynamic adaptive mechanism and reinforcement learning optimization. The present invention integrates the IMU drift formula and the environmental temperature error modeling, enhances the prediction accuracy and gives the model a physical interpretation power, adjusts the noise covariance matrix according to the environmental temperature change and the IMU signal-to-noise ratio, improves the adaptability of the system to a dynamic environment, uses the real-time error as a reward signal, dynamically optimizes the LSTM weight and the covariance matrix, improves the prediction capability under long-term operation, performs feedback optimization on the corrected navigation data, and realizes model self-learning and performance iteration. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is the main flow chart of the present invention;

[0057] Figure 2 is a flow chart of step 1 of the present invention;

[0058] Figure 3 is a flow chart of step 2 of the present invention;

[0059] Figure 4 is a flow chart of step 3 of the present invention;

[0060] Figure 5 is a flow chart of step 4 of the present invention;

[0061] Figure 6 It is a flow chart of step 5 of the present invention. DETAILED DESCRIPTION

[0062] Depend on Figure 1-Figure 6 A method for error prediction of an inertial navigation device is shown, comprising the following steps:

[0063] Step 1: The data acquisition module obtains IMU data P from the IMU sensor, GNSS receiver and environmental sensor respectively IMU (t), GNSS positioning data P GNSS (t) and ambient temperature data T env (t), and preprocessing; IMU data P IMU(t) includes acceleration a(t) and angular velocity ω(t);

[0064] Preprocessing includes IMU data P IMU (t) Noise filtering is performed on the ambient temperature data T env (t) is smoothed and the IMU data P is aligned using the timestamp IMU (t) and GNSS positioning data P GNSS (t), and interpolate and extend the GNSS positioning data P GNSS (t) to IMU data P IMU (t) time frequency; time interval error E Δt Perform mathematical modeling; the preprocessed data and time interval error E Δt Combine and obtain the feature vector X(t);

[0065] When executing step 1, the feature vector X(t) is expressed by the following formula:

[0066] X(t)=[a(t),ω(t),P GNSS (t),T env (t),E Δt ];

[0067] Time interval error E Δt Calculated by the following formula:

[0068] E Δt =β×Δt;

[0069] Among them, Δt is the time interval, which represents the IMU data P IMU (t) and GNSS positioning data P GNSS (t) is the time synchronization error at the same time; β is the time error influence coefficient, which represents the influence of the time delay between sensors.

[0070] In the application process, the data sampling frequencies of IMU sensors and GNSS receivers are usually different. The sampling frequency of IMU sensors is higher (such as 100Hz or higher), while the sampling frequency of GNSS receivers is lower (such as 1Hz); since the time bases of data collection of the two are not completely consistent, time deviation will occur.

[0071] The internal clocks of the IMU sensor and GNSS receiver may drift or be out of sync, which can also cause inconsistent timestamps.

[0072] Environmental factors such as temperature can also affect the internal clock, causing drift.

[0073] When interpolating GNSS data to the IMU frequency, additional time errors may also be introduced due to the imperfections of the interpolation algorithm.

[0074] The present invention corrects the above errors and reduces the time synchronization error Δt, thereby obtaining more accurate positioning data.

[0075] In this embodiment, the timestamp t of each IMU data is IMU and the corresponding GNSS data timestamp t GNSS Make a comparison.

[0076] Since the frequency of GNSS data is low, each data point of the IMU may correspond to multiple GNSS data points. In this case, the present embodiment uses interpolated GNSS data to align to the sampling time of the IMU.

[0077] The time interval Δt is the result of each comparison, which is the difference between the timestamp of the IMU sampling time and the corresponding GNSS data, that is:

[0078] Δt=t IMU -t GNSS ;

[0079] If the GNSS data does not exactly match the IMU timestamp, the GNSS data can be interpolated to get the closest timestamp t GNSS .

[0080] In this embodiment, various sensor data are standardized and aligned through preprocessing to provide data input in a unified format for subsequent modeling. Noise filtering and data smoothing improve data quality and reduce the interference of random errors on subsequent modeling. The GNSS data frequency is expanded by interpolation to be consistent with the IMU frequency, thereby reducing the problem of time resolution mismatch. The modeling of time interval error directly targets the problem of time asynchrony between sensors, thereby improving the accuracy of error prediction.

[0081] Step 2: Data fusion module for IMU data P IMU (t) drift error E IMU (t) and the temperature error E of the ambient temperature data T Mathematical modeling is performed respectively according to the drift error E IMU (t), temperature error E T and time interval error E Δt , generate the initial fusion error E 融合 (t); the drift error E IMU (t) and temperature error E T Add to the feature vector X(t);

[0082] When executing step 2, the specific steps are as follows:

[0083] Step 2-1: IMU data P IMU (t) drift error E IMU (t) Modeling, as shown in the following formula:

[0084] E IMU (t) = E IMU (t-1)+α×Δt;

[0085] Among them, E IMU (t) is the IMU error at the current time point t, E IMU (t-1) is the IMU error at the previous time point t-1, α is the drift rate, and Δt is the time interval;

[0086] Step 2-2: Temperature error E of ambient temperature data T Modeling, as shown in the following formula:

[0087] E T =γ×(T env -T0);

[0088] Among them, T env is the ambient temperature, T0 is the calibration temperature of the ambient sensor, γ is the temperature sensitivity coefficient; the ambient sensor is a temperature sensor;

[0089] Step 2-3: Convert the drift error E IMU (t), temperature error E T and time interval error E Δt Fusion is performed to obtain the fusion error E 融合 (t):

[0090] E 融合 (t) = E IMU (t)+E T +E Δt ;

[0091] Step 2-4: Convert the drift error E IMU (t) and temperature error E T After being added to the feature vector X(t), the feature vector X(t) is expressed as follows:

[0092] X(t)=[a(t),ω(t),P GNSS (t),T env (t),E IMU (t),E T ,E Δt ].

[0093] In this embodiment, the physical characteristics and environmental factors are decomposed into independent models, which helps to analyze the error source more intuitively. The ambient temperature error model (such as temperature sensitivity) allows the system to adapt to different working environments. 融合 (t) It provides a means of quickly estimating errors, providing a basic reference for subsequent complex models, adding physical errors to the feature vector, and enhancing the predictive ability of subsequent LSTM models.

[0094] Step 3: The LSTM calculation module retrieves the feature vector X(t) and the fusion error E 融合 (t), using the LSTM model and physical constraints to predict sensor errors, we get the error prediction result E LSTM (t);

[0095] When executing step 3, the specific steps are as follows:

[0096] Step 3-1: Feature input, specifically, taking the feature vector X(t) obtained in step 2-4 as the feature vector input into the input layer of the LSTM model;

[0097] Step 3-2: Embed physical constraints in the LSTM model, specifically by transforming the drift error formula E IMU (t) = E IMU (t-1)+α×Δt and temperature error formula E T =γ×(T env -T0) is directly embedded in the input and gating mechanism of the LSTM model;

[0098] Introduce drift rate α and time interval β in the forget gate;

[0099] LSTM model fitting prediction result E LSTM1 (t);

[0100] Step 3-3: Physical constraint regularization, specifically introducing the regularization term L phy Constrain the LSTM loss function:

[0101] L phy =λ phy ×||E L (t)-(E IMU (t)+E T +E Δt )|| 2 ;

[0102] Among them, E L (t) is the error calculation of the intermediate process, which is used to constrain the degree of compliance of the LSTM model with physical laws; λ phy is the regularization coefficient, which adjusts the strength of the physical error constraint;

[0103] Physical constraint regularization is used to encourage the prediction results E LSTM1 (t) satisfy physical constraints;

[0104] Step 3-4: Combine the error loss and physical regularization loss to optimize the LSTM model weights and generate the final prediction result E LSTM (t), the total loss function L of LSTM is:

[0105] L total =L data +λL phy ;

[0106] L data =||E true (t)-E LSTM1 (t)|| 2 ;

[0107] Among them, λ is the weight factor, L data is the error loss based on the data. In this embodiment, L data Represents the model prediction error E LSTM1 (t) and the actual error E true The goal is to minimize the loss function and optimize the parameters of the LSTM model so that the prediction result E LSTM1 (t) is closer to the true error E true (t);

[0108] E true (t) is the true error, which represents the deviation between the IMU sensor error in the actual measurement and the ground truth value. When training the LSTM model, E true (t) Pre-input precise position information P true (t) and IMU sensor measurement value P IMU (t); precise location information P true (t) obtained through a high-precision reference system, such as a high-precision GNSS system;

[0109] In this embodiment, when optimizing the LSTM model, the time synchronization error is embedded in the model as a feature, and the model automatically optimizes its influence on the error through learning; the physical constraint regularization term L phy Including E Δt , which encourages the prediction results of model fitting to satisfy the physical laws of time synchronization, thereby reducing the impact of time synchronization errors on the final results.

[0110] In this embodiment, data-driven and physical modeling are combined to make up for the lack of physical interpretability of traditional deep learning models. By regularizing physical constraints, the overfitting risk of the LSTM model during training is reduced and the accuracy of error prediction is improved. The LSTM model can learn dynamic patterns in time series data and can adapt to a variety of navigation scenarios. The total loss function combines physical errors and data errors to effectively balance the model's predictive ability and physical consistency.

[0111] Step 4: Optimize the module to obtain real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the error prediction result E LSTM (t) Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes, use reinforcement learning optimization method to establish a reward function, dynamically adjust the LSTM weights and noise covariance matrix, obtain the optimized LSTM model weights and dynamically adjusted noise covariance matrix, and improve the real-time performance of error prediction and correction;

[0112] When executing step 4, the specific steps include:

[0113] Step 4-1: Get real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the prediction result E LSTM (t);

[0114] Step 4-2: Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes:

[0115] Q adaptive =Q0+k1×|T env -T0|;

[0116] R adaptive =R0+k2×SNR IMU ;

[0117] Among them, Q adaptive is the process noise covariance matrix, which is affected by temperature changes; R adaptive is the measurement noise covariance matrix, which is affected by the signal-to-noise ratio; k1 and k2 are adjustment coefficients; Q0 is the initial value of the process noise covariance matrix; R0 is the initial value of the measurement noise covariance matrix;

[0118] Step 4-3: Use reinforcement learning optimization method to establish reward function Reward:

[0119] Reward=-||E true (t)-E LSTM (t)|| 2 ;

[0120] The LSTM weights and Kalman filter covariance matrix are dynamically adjusted through the reward function Reward to optimize the LSTM model and noise covariance matrix.

[0121] IMU signal-to-noise ratio (SNR) IMU It is the typical value of the signal-to-noise ratio of the IMU sensor, provided by the manufacturer.

[0122] In this embodiment, by dynamically adjusting the noise covariance matrix, the system can continuously optimize performance during real-time operation, and reinforcement learning dynamically adjusts the model parameters and the filter covariance matrix so that the model can adapt to the complex changes in the real-time environment. Through the multi-layer optimization mechanism (LSTM optimization + noise adjustment), error accumulation is further reduced, and the noise parameters can be automatically adjusted according to the real-time ambient temperature and IMU signal-to-noise ratio, thereby improving the robustness of the system.

[0123] Step 5: Use the prediction error to correct the inertial data and continuously optimize the model through a feedback mechanism to output high-precision navigation data.

[0124] When executing step 5, the IMU data P is calculated according to the following formula IMU (t) Make corrections to obtain the corrected navigation data P corrected (t):

[0125] P corrected (t) = P IMU (t)-E LSTM (t);

[0126] The corrected data is used to update the parameters of the LSTM model and the Kalman filter to continuously optimize the model through the feedback mechanism, and the navigation data P is output. corrected (t).

[0127] The following is an application scenario in this embodiment:

[0128] A drone was flying in a complex valley terrain. Due to terrain obstruction, the GNSS signal was intermittently lost, and the system needed to rely on inertial navigation equipment for high-precision positioning.

[0129] Input data:

[0130] IMU data: acceleration a(t) = [0.1, 0.2, -9.8] (m / s 2 ), angular velocity ω(t)=[0.01,-0.02,0.03].

[0131] GNSS positioning data:

[0132] P GNSS(t) = [39.123456, -76.543210, 100] (latitude, longitude and altitude, units: degrees and meters).

[0133] Ambient temperature:

[0134] T env (t) = 35°C.

[0135] Preprocessing:

[0136] Time interval error: E Δt =0.005×Δt,Δt=0.1s.

[0137] Fusion error calculation:

[0138] E IMU (t)=0.01+0.005×0.1=0.0105.

[0139] E T =0.002×(35-25)=0.02.

[0140] E 融合 (t)=0.0105+0.02+0.0005=0.031.

[0141] Error correction:

[0142] Corrected inertial data:

[0143] P corrected (t) = P IMU (t)-E LSTM (t),E LSTM (t) = 0.03.

[0144] Output: P corrected (t)=[39.123455,-76.543209,99.97].

[0145] The error prediction method of an inertial navigation device disclosed in the present invention solves the technical problem of improving the real-time performance and accuracy of a navigation system by directly embedding an IMU physical drift formula and an environmental error into an LSTM model, combining a dynamic adaptive mechanism and reinforcement learning optimization. The present invention integrates the IMU drift formula and the environmental temperature error modeling, enhances the prediction accuracy and gives the model a physical interpretation power, adjusts the noise covariance matrix according to the environmental temperature change and the IMU signal-to-noise ratio, improves the adaptability of the system to a dynamic environment, uses the real-time error as a reward signal, dynamically optimizes the LSTM weight and the covariance matrix, improves the prediction capability under long-term operation, performs feedback optimization on the corrected navigation data, and realizes model self-learning and performance iteration.

Claims

1. A method for error prediction of an inertial navigation device, characterized in that: The steps include: Step 1: The data acquisition module obtains IMU data P from the IMU sensor, GNSS receiver and environmental sensor respectively IMU (t), GNSS positioning data P GNSS (t) and ambient temperature data T env (t), and preprocessing; IMU data PIMU (t) includes acceleration a (t) and angular velocity ω (t); Preprocessing includes IMU data P IMU (t) Noise filtering is performed on the ambient temperature data T env (t) is smoothed and the IMU data P is aligned using the timestamp IMU (t) and GNSS positioning data P GNSS (t), and interpolate and extend the GNSS positioning data P GNSS (t) to IMU data P IMU (t) time frequency; time interval error E Δt Conduct mathematical modeling; The preprocessed data and time interval error E Δt Combine and obtain the feature vector X(t); Step 2: Data fusion module for IMU data P IMU (t) drift error E IMU (t) and the temperature error E of the ambient temperature data T Mathematical modeling is performed respectively according to the drift error E IMU (t), temperature error E T and time interval error E Δt , generate the initial fusion error E 融合 (t); the drift error E IMU (t) and temperature error E T Add to the feature vector X(t); Step 3: The LSTM calculation module retrieves the feature vector X(t) and the fusion error E 融合 (t), using the LSTM model and physical constraints to predict sensor errors, we get the error prediction result E LSTM (t); Step 4: Optimize the module to obtain real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the error prediction result E LSTM (t) Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes, use reinforcement learning optimization method to establish a reward function, dynamically adjust the LSTM weights and noise covariance matrix, obtain the optimized LSTM model weights and dynamically adjusted noise covariance matrix, and improve the real-time performance of error prediction and correction; Step 5: Use the prediction error to correct the inertial data and continuously optimize the model through a feedback mechanism to output high-precision navigation data.

2. The method for error prediction of an inertial navigation device according to claim 1, characterized in that: When executing step 1, the feature vector X(t) is expressed by the following formula: X(t)=[a(t),ω(t),P GNSS (t),T env (t),E Δt ]; Time interval error E Δt Calculated by the following formula: AND Δt =β×Δt; Among them, Δt is the time interval, which represents the IMU data P IMU (t) and GNSS positioning data P GNSS (t) is the time synchronization error at the same time; β is the time error influence coefficient, which represents the influence of the time delay between sensors.

3. The method for error prediction of an inertial navigation device as claimed in claim 2, characterized in that: When executing step 2, the specific steps are as follows: Step 2-1: IMU data P IMU (t) drift error E IMU (t) Modeling, as shown in the following formula: E IMU (t)=E IMU (t-1)+α×Δt; Among them, E IMU (t) is the IMU error at the current time point t, E IMU (t-1) is the IMU error at the previous time point t-1, α is the drift rate, and Δt is the time interval; Step 2-2: Temperature error E of ambient temperature data T Modeling, as shown in the following formula: E T =γ×(T env -T0); Among them, T env is the ambient temperature, T0 is the calibration temperature of the ambient sensor, γ is the temperature sensitivity coefficient; the ambient sensor is a temperature sensor; Step 2-3: Convert the drift error E IMU (t), temperature error E T and time interval error E Δt Fusion is performed to obtain the fusion error E 融合 (t): AND 融合 (t)=And IMU (t)+E T +E Δt ; Step 2-4: Convert the drift error E IMU (t) and temperature error E T After being added to the feature vector X(t), the feature vector X(t) is expressed as follows: X(t)=[a(t),ω(t),P GNSS (t),T env (t),E IMU (t),E T ,E Δt ]。 4. The method for error prediction of an inertial navigation device as claimed in claim 3, characterized in that: When executing step 3, the specific steps are as follows: Step 3-1: Feature input, specifically, taking the feature vector X(t) obtained in step 2-4 as the feature vector input into the input layer of the LSTM model; Step 3-2: Embed physical constraints in the LSTM model, specifically by transforming the drift error formula E IMU (t) = E IMU (t-1)+α×Δt and temperature error formula E T =γ×(T env -T0) is directly embedded in the input and gating mechanism of the LSTM model; Introduce drift rate α and time interval β in the forget gate; LSTM model fitting prediction result E LSTM1 (t); Step 3-3: Physical constraint regularization, specifically introducing the regularization term L phy Constrain the LSTM loss function: L phy =λ phy ×||E L (are you IMU (t)+E T +E Δt )|| 2 4 Among them, E L (t) is the error calculation of the intermediate process, which is used to constrain the degree of compliance of the LSTM model with physical laws; λ phy is the regularization coefficient, which adjusts the strength of the physical error constraint; Physical constraint regularization is used to encourage the prediction results E LSTM1 (t) satisfy physical constraints; Step 3-4: Combine the error loss and physical regularization loss to optimize the LSTM model weights and generate the final prediction result E LSTM (t), the total loss function L of LSTM is: THE total =L data +λL phy ; L data =||E true (t)-E LSTM1 (t)|| 2 ; Among them, λ is the weight factor, L data is the error loss based on the data.

5. The method for error prediction of an inertial navigation device as claimed in claim 4, characterized in that: When executing step 4, the specific steps include: Step 4-1: Get real-time ambient temperature data T env (t), IMU signal-to-noise ratio SNR IMU And the prediction result E LSTM (t); Step 4-2: Establish a dynamic adaptive mechanism to adjust the noise covariance matrix according to environmental changes: Q adaptive =Q0+k1×∣T env -T0∣; R adaptive =R0+k2×SNR IMU ; Among them, Q adaptive is the process noise covariance matrix, which is affected by temperature changes; R adaptive is the measurement noise covariance matrix, which is affected by the signal-to-noise ratio; k1 and k2 are adjustment coefficients; Q0 is the initial value of the process noise covariance matrix; R0 is the initial value of the measurement noise covariance matrix; Step 4-3: Use reinforcement learning optimization method to establish reward function Reward: Reward=-||E true (t)-E LSTM (t)|| 2 ; Among them, E true (t) is the true error; The LSTM weights and Kalman filter covariance matrix are dynamically adjusted through the reward function Reward to optimize the LSTM model and noise covariance matrix.

6. The method for error prediction of an inertial navigation device as claimed in claim 4, characterized in that: When executing step 5, the IMU data P is calculated according to the following formula IMU (t) Make corrections to obtain the corrected navigation data P corrected (t): P corrected (t)=P IMU (t)-E LSTM (t); The corrected data is used to update the parameters of the LSTM model and the Kalman filter to continuously optimize the model through the feedback mechanism, and the navigation data P is output. corrected (t).

Images