Vehicle-mounted inertial positioning method and device based on data and model combined driving

Abstract

The invention relates to the technical field of navigation positioning, in particular to a vehicle-mounted inertial positioning method and device based on data and model combined driving, and the method comprises the steps: carrying out the data preprocessing of the acceleration data and gyroscope data of a vehicle-mounted IMU, inputting a three-dimensional speed estimation network to obtain a pseudo measurement constraint and a measurement noise scale factor, and carrying out the calculation of the pseudo measurement constraint and the measurement noise scale factor; calculating a measurement noise covariance according to the two; inputting the preprocessed data into a process noise estimation network to obtain a process noise scale factor and process noise uncertainty, and calculating a process noise covariance according to the process noise scale factor and the process noise uncertainty; constructing a loss function of uncertainty estimation, and adjusting the error penalty weight of the three-dimensional speed estimation network and the error penalty weight of the noise estimation network; and carrying out mechanical arrangement on the data, carrying out extended Kalman filtering in combination with pseudo measurement constraint and adaptive noise covariance, and carrying out vehicle-mounted inertial positioning according to a filtering result. Therefore, the problem that the vehicle positioning precision is reduced in a complex environment in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of navigation and positioning technology, and in particular to a vehicle-mounted inertial positioning method and device based on joint drive of data and model. Background Art

[0002] High-precision vehicle positioning is a basic requirement for the development of smart driving and IoT technologies. GNSS (Global Navigation Satellite System), LiDAR, cameras, and inertial sensors have been widely used for vehicle navigation, control, and decision-making. However, in challenging and complex urban environments, such as urban canyons, thin textures, and severe weather conditions, the positioning accuracy of GNSS and vision-based technologies may drop significantly, and the application of most devices is usually limited by complex installation requirements and high costs. INS (Inertial Navigation System) can perform continuous autonomous positioning without any external infrastructure, making it a key component of smart vehicle navigation. However, INS errors accumulate over time and the estimated position drifts rapidly. In order for vehicles to accurately perform dynamic movements at a high update rate in all scenarios, ensure safe driving, and avoid extreme risks, it is crucial to address the limitations of INS.

[0003] Noise in traditional Kalman filters for vehicle positioning is usually based on fixed empirical assumptions. However, noise characteristics often vary with the environment and motion state, and it is challenging to manually construct a suitable noise model, which limits the accuracy of Kalman filter state estimation and leads to a decrease in vehicle positioning accuracy in complex environments. In addition, velocity pseudo-measurements such as odometers and NHC (Non-Holonomic Constraint) can provide additional constraints to the Kalman filter under GNSS rejection conditions, thereby improving the accuracy of state estimation. However, obtaining odometer data requires access to the vehicle's controller area network, which limits its application in some cases. The NHC assumption also becomes inaccurate when measuring complex motions such as sideslips and jumps. Summary of the invention

[0004] The present application provides a vehicle-mounted inertial positioning method and device based on the joint drive of data and model to solve the problem of reduced vehicle positioning accuracy in complex environments in related technologies.

[0005] The first aspect of the present application provides a vehicle-mounted inertial positioning method based on joint data and model driving, comprising the following steps: preprocessing the acceleration data and gyroscope data of the vehicle-mounted IMU; inputting the preprocessed data into a three-dimensional velocity estimation network, the three-dimensional velocity estimation network outputs the vehicle's three-dimensional velocity and measurement noise uncertainty, taking the vehicle's three-dimensional velocity as a pseudo-measurement constraint, taking the measurement noise uncertainty as a measurement noise scale factor, and calculating the measurement noise covariance according to the measurement noise scale factor; inputting the preprocessed data into a process noise estimation network, the process noise estimation network outputs a process noise scale factor and process noise uncertainty, and calculating the process noise covariance according to the process noise scale factor; constructing a loss function for uncertainty estimation, adjusting the error penalty weight of the three-dimensional velocity estimation network according to the loss function and the measurement noise uncertainty, and adjusting the error penalty weight of the process noise estimation network according to the loss function and the process noise uncertainty; mechanically arranging the acceleration data and gyroscope data of the vehicle-mounted IMU, performing extended Kalman filtering according to the mechanically arranged data, pseudo-measurement constraints, zero-speed detection results, measurement noise covariance, and process noise covariance, and performing vehicle-mounted inertial positioning according to the filtering results.

[0006] Optionally, the acceleration data and gyroscope data of the vehicle-mounted IMU are preprocessed, including: adding random Gaussian noise to the acceleration data and gyroscope data; and segmenting the acceleration data and gyroscope data using a sliding window.

[0007] Optionally, the three-dimensional speed estimation network includes an input block, a residual block and an output block, wherein the input block includes a convolution layer, a batch normalization layer and a maximum pooling layer, the convolution layer is used for feature extraction, the batch normalization layer normalizes the output of the convolution layer, and the maximum pooling layer selects the feature of the maximum value in the pooling window; the residual block controls two convolution layers, the two convolution layers are connected through batch normalization and a residual, and a self-attention module is fused after each residual block; the output block outputs the three-dimensional speed of the vehicle and the measurement noise uncertainty, the output block includes a fully connected layer and a batch normalization layer, and multiple Dropout layers are set after the fully connected layer and the batch normalization layer.

[0008] Optionally, the output block of the three-dimensional speed estimation network is provided with a Tanh function, and the Tanh function is replaced with a Sigmoid function, and the vehicle state is binary classified to obtain a zero-speed detection result.

[0009] Optionally, the process noise estimation network includes two feature extraction modules, wherein the two feature extraction modules have the same structure, the feature extraction module includes a convolution layer, a ReLU function and a compression excitation module, and after the convolution layer extracts features and is activated using the ReLU function, the compression excitation module is used to dynamically adjust the weight of each channel.

[0010] Optionally, the compression excitation module includes a global average pooling layer, a fully connected layer, and a Sigmoid function activation, wherein the global average pooling layer globally aggregates the spatial information of the channels of each original feature map, and the globally aggregated information is activated by the Sigmoid function through the fully connected layer, and the channels of the original feature map are reweighted and output according to the activated result.

[0011] Optionally, the loss function for uncertainty estimation is:

[0012]

[0013] Where N is the sliding window length; y k is the reference true value; f(x k ) is the input x k The estimated value of the corresponding output; σ(x k ) 2 is the model uncertainty.

[0014] Optionally, the state vector of the extended Kalman filter is defined as follows:

[0015]

[0016] Among them, p n is the position in the navigation system; v n is the speed under the navigation system; For posture; ω is the gyroscope bias under the load system; b a is the accelerometer bias under the load system; is the IMU installation angle between the carrier system and the vehicle system; It is the IMU arm between the carrier system and the vehicle system;

[0017] The covariance matrix is ​​updated as follows:

[0018]

[0019] in, is the prior state covariance at time k; is the posterior state covariance at time k-1; F k-1 and G k-1 are the Jacobian matrices of the nonlinear functions of state propagation; Q is the process noise covariance; v is the coordinate transformation matrix from the carrier system to the navigation system at time k-1; k-1 is the speed of the navigation system at time k-1; p k-1 is the position in the navigation system at time k-1; g is gravity; dt is the time interval; (·×) represents the antisymmetric matrix.

[0020] The second aspect of the present application provides a vehicle-mounted inertial positioning device based on joint data and model driving, including: a processing module, which is used to preprocess the acceleration data and gyroscope data of the vehicle-mounted IMU; a first calculation module, which is used to input the preprocessed data into a three-dimensional velocity estimation network, the three-dimensional velocity estimation network outputs the vehicle's three-dimensional velocity and measurement noise uncertainty, uses the vehicle's three-dimensional velocity as a pseudo-measurement constraint, uses the measurement noise uncertainty as a measurement noise scaling factor, and calculates the measurement noise covariance according to the measurement noise scaling factor; a second calculation module, which is used to input the preprocessed data into a process noise estimation network, the process noise estimation network outputs a process noise covariance. Noise scaling factor and process noise uncertainty, calculate the process noise covariance according to the process noise scaling factor; an adjustment module, used to construct a loss function for uncertainty estimation, adjust the error penalty weight of the three-dimensional velocity estimation network according to the loss function and measurement noise uncertainty, and adjust the error penalty weight of the process noise estimation network according to the loss function and process noise uncertainty; a filtering module, used to mechanically arrange the acceleration data and gyroscope data of the on-board IMU, perform extended Kalman filtering according to the mechanically arranged data, pseudo-measurement constraints, zero-speed detection results, measurement noise covariance and process noise covariance, and perform on-board inertial positioning according to the filtering results.

[0021] A third aspect of the present application provides a vehicle, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle-mounted inertial positioning method based on joint driving of data and models as in the first aspect.

[0022] Therefore, this application includes the following beneficial effects:

[0023] The embodiment of the present application preprocesses the acceleration data and gyroscope data of the vehicle-mounted IMU, inputs them into the three-dimensional velocity estimation network, the network outputs pseudo-measurement constraints and measurement noise scale factors, and calculates the measurement noise covariance according to the measurement noise scale factors; inputs the preprocessed data into the process noise estimation network, the network outputs the process noise scale factor and process noise uncertainty, and calculates the process noise covariance according to the process noise scale factor; constructs a loss function for uncertainty estimation, adjusts the error penalty weight of the three-dimensional velocity estimation network and the error penalty weight of the noise estimation network; mechanically arranges the acceleration data and gyroscope data of the vehicle-mounted IMU, performs extended Kalman filtering, performs vehicle-mounted inertial positioning according to the filtering results, estimates the three-dimensional velocity of the vehicle and the corresponding uncertainty based on deep learning, introduces a data-driven noise covariance adapter to dynamically adjust the process noise covariance, and achieves accurate and robust position estimation. Thus, the problem of reduced vehicle positioning accuracy in complex environments in related technologies is solved.

[0024] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0026] Figure 1 A flowchart of a vehicle-mounted inertial positioning method based on joint driving of data and models provided according to an embodiment of the present application;

[0027] Figure 2 A schematic diagram of a vehicle-mounted inertial positioning method based on joint driving of data and models provided according to an embodiment of the present application;

[0028] Figure 3 A schematic diagram of the structure of a residual network based on an attention mechanism provided according to an embodiment of the present application;

[0029] Figure 4 A schematic diagram of a process noise estimation network structure provided according to an embodiment of the present application;

[0030] Figure 5 This is an example diagram of a vehicle-mounted inertial positioning device based on joint driving of data and model according to an embodiment of the present application;

[0031] Figure 6 It is a schematic diagram of the structure of a vehicle provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0033] The following describes the vehicle-mounted inertial positioning method and device based on data and model joint drive of the embodiment of the present application with reference to the accompanying drawings. In view of the problem that the vehicle positioning accuracy of the related technology mentioned in the above background technology decreases in complex environments, the present application provides a vehicle-mounted inertial positioning method based on data and model joint drive, in which the acceleration data and gyroscope data of the vehicle-mounted IMU are preprocessed and input into the three-dimensional velocity estimation network, the network outputs pseudo-measurement constraints and measurement noise scale factors, and the measurement noise covariance is calculated according to the measurement noise scale factor; the preprocessed data is input into the process noise estimation network, the network outputs the process noise scale factor and the process noise uncertainty, and the process noise covariance is calculated according to the process noise scale factor; the loss function of uncertainty estimation is constructed, and the error penalty weight of the three-dimensional velocity estimation network and the error penalty weight of the noise estimation network are adjusted; the acceleration data and gyroscope data of the vehicle-mounted IMU are mechanically arranged, and extended Kalman filtering is performed, and vehicle-mounted inertial positioning is performed according to the filtering results, and the three-dimensional velocity and corresponding uncertainty of the vehicle are estimated based on deep learning, and a data-driven noise covariance adapter is introduced to dynamically adjust the process noise covariance, so as to achieve accurate and robust position estimation. Thereby, the problem of decreased vehicle positioning accuracy in complex environments in related technologies is solved.

[0034] Specifically, Figure 1 A flow chart of a vehicle-mounted inertial positioning method based on joint driving of data and model provided in an embodiment of the present application.

[0035] like Figure 1 As shown, the vehicle-mounted inertial positioning method based on joint driving of data and model includes the following steps:

[0036] In step S101, the acceleration data and gyroscope data of the vehicle-mounted IMU are preprocessed.

[0037] It can be understood that the embodiments of the present application can collect acceleration data and gyroscope data through the vehicle-mounted IMU, and perform data preprocessing on the acceleration data and gyroscope data. The preprocessing method will be described in detail below and will not be repeated here.

[0038] In an embodiment of the present application, the acceleration data and gyroscope data of the vehicle-mounted IMU are preprocessed, including: adding random Gaussian noise to the acceleration data and gyroscope data; and using a sliding window to segment the acceleration data and gyroscope data.

[0039] Among them, the random Gaussian noise is generally set according to the data provided by the IMU manufacturer, and is not specifically limited here; the formula for segmenting the original data sequence using a sliding window is: ω and a are the original 3D gyroscope data and 3D accelerometer data of IMU respectively; L is the length of the original data sequence; and is the processed data with random Gaussian noise added; N is the sliding window length; Num seg The number of slices to be divided.

[0040] It can be understood that the method for preprocessing the acceleration data and gyroscope data of the vehicle-mounted IMU in the embodiment of the present application is to add random Gaussian noise to the acceleration data and gyroscope data, and use a sliding window to calculate the data using the formula Split the acceleration data and gyroscope data.

[0041] In step S102, the preprocessed data is input into a three-dimensional velocity estimation network, which outputs the vehicle's three-dimensional velocity and measurement noise uncertainty. The vehicle's three-dimensional velocity is used as a pseudo-measurement constraint, and the measurement noise uncertainty is used as a measurement noise scaling factor. The measurement noise covariance is calculated based on the measurement noise scaling factor.

[0042] Among them, the 3D speed estimation network is a residual network designed based on the attention mechanism. The preprocessed data is input into the 3D speed estimation network. The 3D speed estimation network outputs the vehicle 3D speed and the measurement noise uncertainty formula as follows: and is the gyroscope data and accelerometer data with Gaussian noise added, v c =[v for ,v lat ,v up ] T is the estimated 3D velocity of the vehicle; γ is the uncertainty; F 1 (θ) is the designed residual network based on the attention mechanism; the uncertainty is defined as the measurement noise scaling factor that adjusts the measurement noise covariance of the pseudo measurement, and the formula is: R k is the measurement noise covariance at time k; is the initial value; k is the diagonal matrix of uncertainty transformation of the velocity network output; 10 λ is a constant parameter.

[0043] It can be understood that the embodiment of the present application can input the preprocessed acceleration and gyroscope data into the three-dimensional velocity estimation network, output the three-dimensional velocity of the vehicle and the measurement noise uncertainty through the three-dimensional velocity estimation network, use the three-dimensional velocity of the vehicle as a pseudo-measurement constraint, and use the obtained measurement noise uncertainty as a scaling factor. The measurement noise covariance can be calculated according to the measurement noise scaling factor. Rk is the measurement noise covariance at time k.

[0044] In an embodiment of the present application, a three-dimensional speed estimation network includes an input block, a residual block and an output block, wherein the input block includes a convolution layer, a batch normalization layer and a maximum pooling layer, the convolution layer is used for feature extraction, the batch normalization layer normalizes the output of the convolution layer, and the maximum pooling layer selects the feature of the maximum value in the pooling window; the residual block controls two convolution layers, the two convolution layers are connected through batch normalization and a residual, and a self-attention module is fused after each residual block; the output block outputs the three-dimensional speed of the vehicle and the measurement noise uncertainty, the output block includes a fully connected layer and a batch normalization layer, and multiple Dropout layers are set after the fully connected layer and the batch normalization layer.

[0045] Among them, the Dropout layer can randomly discard a part of neurons, which can prevent the model from overfitting and improve the generalization ability.

[0046] It can be understood that the three-dimensional speed estimation network of the embodiment of the present application includes an input block, a residual block and an output block. The input block includes a convolution layer, a batch normalization layer and a maximum pooling layer. Feature extraction can be achieved through the convolution layer, and the output of the convolution layer is normalized by the batch normalization layer, and the maximum pooling layer is used to select the feature of the maximum value in the pooling window; the residual block controls two convolution layers, the two convolution layers have batch normalization operations, and a residual connection is added to ensure smooth information flow, and a self-attention module is fused after each residual block. The output block outputs the three-dimensional speed of the vehicle and the measurement noise uncertainty; the output block includes a fully connected layer and a batch normalization layer. Multiple Dropout layers are set after the fully connected layer and the batch normalization layer. By randomly discarding a part of the neurons, the model can be prevented from overfitting and the generalization ability can be improved.

[0047] In an embodiment of the present application, the output block of the three-dimensional speed estimation network is provided with a Tanh function, and the Tanh function is replaced with a Sigmoid function, and the vehicle state is binary classified to obtain a zero-speed detection result.

[0048] Among them, binary classification refers to the task of assigning a state to one of two mutually exclusive categories, which can distinguish between "vehicle moving" and "vehicle stationary".

[0049] It can be understood that the embodiment of the present application can replace the Tanh function set in the output block with the Sigmoid function, use binary classification to classify the vehicle state, and perform zero-speed detection on the vehicle to obtain a zero-speed detection result.

[0050] In step S103, the preprocessed data is input into a process noise estimation network, the process noise estimation network outputs a process noise proportional factor and a process noise uncertainty, and the process noise covariance is calculated according to the process noise proportional factor.

[0051] Among them, the process of inputting the preprocessed data into the process noise estimation network and the process noise estimation network outputting the process noise proportional factor and process noise uncertainty is: ρ is the process noise proportional factor; μ is the process noise uncertainty; F 2 (θ) is the designed process noise estimation network; the process noise estimation network structure will be described in detail below and will not be repeated here; the process noise covariance is calculated according to the process noise scale factor as shown below: Q k =Q·ρ k , Q k is the adaptive process noise covariance at time k; is the initial matrix; ρ k is the diagonal matrix converted from the process noise scale factor at time k.

[0052] It can be understood that the embodiment of the present application designs a process noise estimation network, and the preprocessed data can be input into the process noise estimation network, and the process noise estimation network outputs the process noise proportional factor and the process noise uncertainty. k =Q·ρ k , calculate the adaptive process noise covariance at time k.

[0053] In an embodiment of the present application, the process noise estimation network includes two feature extraction modules, wherein the structures of the two feature extraction modules are the same, the feature extraction modules include a convolutional layer, a ReLU function and a compression excitation module, and after the convolutional layer extracts features and is activated using the ReLU function, the compression excitation module is used to dynamically adjust the weight of each channel.

[0054] It can be understood that the process noise estimation network of the embodiment of the present application includes two feature extraction modules with the same structure. The feature extraction module includes a convolution layer, a ReLU function and a compression excitation module. In the convolution layer, it is responsible for extracting local features in the preprocessed data, and then activating these features using the ReLU function. Finally, the compression excitation module is used to dynamically adjust the weights of each channel according to the global information.

[0055] In an embodiment of the present application, the compression excitation module includes a global average pooling layer, a fully connected layer, and a sigmoid function activation, wherein the global average pooling layer globally aggregates the spatial information of the channels of each original feature map, and the globally aggregated information is activated by the sigmoid function through the fully connected layer, and the channels of the original feature map are reweighted and output according to the activated result.

[0056] It can be understood that the compression excitation module of the embodiment of the present application includes a global average pooling layer, a fully connected layer, and a sigmoid function activation. The spatial information of each channel is first globally aggregated through the global average pooling layer, and then passed through two fully connected layers and activated by the sigmoid function. The results are used to reweight the channels of the original feature map and then output.

[0057] In step S104, a loss function for uncertainty estimation is constructed, and the error penalty weight of the three-dimensional velocity estimation network is adjusted according to the loss function and the measurement noise uncertainty, and the error penalty weight of the process noise estimation network is adjusted according to the loss function and the process noise uncertainty.

[0058] Among them, the loss function of uncertainty estimation will be described in detail below and will not be repeated here.

[0059] It can be understood that the embodiments of the present application can construct a loss function for uncertainty estimation, use the loss function for uncertainty estimation and measurement noise uncertainty, process noise uncertainty to estimate the error penalty weight of the network, and dynamically adjust the error penalty weight to improve the accuracy and robustness of the network output in various motion states.

[0060] In the embodiment of the present application, the loss function of uncertainty estimation is:

[0061]

[0062] Where N is the sliding window length; y k is the reference true value; f(x k ) is the input x k The estimated value of the corresponding output; σ(x k ) 2 is the model uncertainty.

[0063] In step S105, the acceleration data and gyroscope data of the vehicle-mounted IMU are mechanically arranged, and extended Kalman filtering is performed based on the mechanically arranged data, pseudo-measurement constraints, zero-speed detection results, measurement noise covariance and process noise covariance, and vehicle-mounted inertial positioning is performed based on the filtering results.

[0064] Among them, mechanical choreography refers to the process of converting the raw output of the accelerometer and gyroscope into position, velocity and attitude; the extended Kalman filter will be described in detail below and will not be repeated here; the pseudo-measurement constrains the three-dimensional velocity of the vehicle and is defined as follows: [v for ,v lat ,v up ] T is the estimated three-dimensional vehicle speed, n c is the measurement noise, and its covariance is the measurement noise covariance estimated above.

[0065] It can be understood that the embodiments of the present application can perform INS mechanical arrangement of the IMU raw data, perform extended Kalman filtering in combination with the estimated pseudo-measurement constraints, zero-speed detection results, measurement noise covariance and process noise covariance, and obtain the filtering results for vehicle inertial positioning.

[0066] In the embodiment of the present application, the state vector of the extended Kalman filter is defined as follows:

[0067]

[0068] Among them, p n is the position in the navigation system; v n is the speed under the navigation system; For posture; ω is the gyroscope bias under the load system; b a is the accelerometer bias under the load system; is the IMU installation angle between the carrier system and the vehicle system; It is the IMU arm between the carrier system and the vehicle system;

[0069] The covariance matrix is ​​updated as follows:

[0070]

[0071] in, is the prior state covariance at time k; is the posterior state covariance at time k-1; F k-1 and G k-1 are the Jacobian matrices of the nonlinear functions of state propagation; Q is the process noise covariance; v is the coordinate transformation matrix from the carrier system to the navigation system at time k-1; k-1 is the speed of the navigation system at time k-1; p k-1 is the position in the navigation system at time k-1; g is gravity; dt is the time interval; (·×) represents the antisymmetric matrix.

[0072] It should be noted that in the update step, the vehicle speed calculated by INS is: in, is the coordinate transformation matrix between the carrier system and the vehicle system, i.e., the IMU installation angle; for the lever arm; and is zero-mean Gaussian noise, and the vehicle 3D velocity is used as a pseudo-measurement constraint, defined as [v for ,v lat ,v up ] T is the three-dimensional speed of the vehicle estimated in step S102; n c is the measurement noise, whose covariance is the R estimated in step S102 k , based on the above formula, the measurement equation when the vehicle is moving is: n c is the noise of the pseudo-measurement of the vehicle's three-dimensional velocity, with a covariance of R; H 1 is the Jacobian matrix of the nonlinear function of the measured value in the moving state; based on zero-speed detection, when the vehicle is stationary, the measurement equation is: H 2 =[θ 3×3 ,I 3×3 ,0 15×3 ], where n c0 is the noise of pseudo-measurement of vehicle 3D velocity; H 2 is the Jacobian matrix of the nonlinear function of the measurements at rest.

[0073] It can be understood that the embodiment of the present application can expand the state vector of the Kalman filter and update the covariance matrix through the above formula, calculate the vehicle measurement equation when the vehicle is moving or stationary during the update process, and obtain the filtering result for vehicle inertial positioning.

[0074] According to the vehicle-mounted inertial positioning method based on joint driving of data and model proposed in the embodiment of the present application, the acceleration data and gyroscope data of the vehicle-mounted IMU are preprocessed and input into the three-dimensional velocity estimation network, the network outputs pseudo-measurement constraints and measurement noise scaling factors, and the measurement noise covariance is calculated according to the measurement noise scaling factors; the preprocessed data are input into the process noise estimation network, the network outputs the process noise scaling factor and process noise uncertainty, and the process noise covariance is calculated according to the process noise scaling factor; a loss function for uncertainty estimation is constructed, and the error penalty weight of the three-dimensional velocity estimation network and the error penalty weight of the noise estimation network are adjusted; the acceleration data and gyroscope data of the vehicle-mounted IMU are mechanically arranged, and extended Kalman filtering is performed, and vehicle-mounted inertial positioning is performed according to the filtering results, the three-dimensional velocity of the vehicle and the corresponding uncertainty are estimated based on deep learning, and a data-driven noise covariance adapter is introduced to dynamically adjust the process noise covariance, thereby achieving accurate and robust position estimation.

[0075] The vehicle-mounted inertial positioning method based on the joint drive of data and model is further described below through a specific embodiment:

[0076] like Figure 2 As shown, the vehicle-mounted inertial positioning method based on joint driving of data and model provided in this embodiment specifically includes the following steps:

[0077] Step 1: preprocess the original acceleration data and gyroscope data of the vehicle-mounted IMU, including adding Gaussian zero-mean noise and segmenting the data sequence using a sliding window, as follows:

[0078] In order to improve the robustness of the neural network training phase to the noise of the input data, random Gaussian noise is added to the original observation data. The random Gaussian noise is generally set according to the data provided by the IMU manufacturer. In addition, since the IMU data at a single moment cannot be mined, a sliding window is used to segment the original data sequence so as to serve as the input of the network in steps 2 and 3, as shown below:

[0079]

[0080] Wherein, ω and a are the original 3D gyroscope data and 3D accelerometer data of IMU respectively; L is the length of the original data sequence; and is the processed data with random Gaussian noise added; N is the sliding window length; Num seg The number of slices to be divided.

[0081] Step 2: Based on the preprocessed data in step 1, a residual network combined with an attention mechanism is used to estimate the vehicle's 3D speed and uncertainty through multi-task learning. The vehicle's 3D speed is used as a pseudo-measurement constraint, and the uncertainty of the accompanying speed output is used as the measurement noise covariance, as follows:

[0082] The input and output of the 3D velocity estimation network driven by IMU data are as follows:

[0083]

[0084] in, and is the gyroscope data and accelerometer data with Gaussian noise added; v c =[v for ,v lat ,v up ] T is the predicted three-dimensional speed of the vehicle; γ is the uncertainty N is the sliding window length; F 1 (θ) is the designed residual network based on the attention mechanism.

[0085] like Figure 3 As shown in the figure, the residual network based on the attention mechanism consists of three parts: input block, residual block and output block. The specific structure design is as follows:

[0086] Input block: It consists of a convolutional layer, a batch normalization layer, and a maximum pooling layer. The convolutional layer is used for initial local feature extraction, the batch normalization layer stabilizes the process by normalizing the output of the convolutional layer, and the maximum pooling layer retains the most important features by selecting the maximum value in the pooling window, which helps reduce irrelevant features and noise.

[0087] Residual Block: The core of the network is the residual block consisting of four residual blocks. The classic residual block consists of two convolutional layers with batch normalization and a residual connection. To optimize the structure, a self-attention module is fused after each residual block to avoid a significant increase in computational overhead. This part of the network structure first extracts local features and then captures long-term dependencies in the global feature space based on the self-attention mechanism.

[0088] Output block: The high-dimensional features are flattened and fed to the fully connected layer and passed through the Tanh function for final prediction. Multiple dropout layers are set after the fully connected layer and batch normalization layer to prevent the model from overfitting. The output of the network contains two vectors: the estimated 3D vehicle velocity and its uncertainty.

[0089] The uncertainty is defined as the measurement noise scaling factor that adjusts the measurement noise covariance of the pseudo-measurement:

[0090] Rk =R·γ k (3)

[0091]

[0092] Among them, R k is the measurement noise covariance at time k; is the initial value; k is the diagonal matrix of uncertainty transformation of the velocity network output; 10 λ is a constant parameter.

[0093] Based on the speed estimation network structure, the Tanh function in the output block is replaced with the Sigmoid function to perform binary classification on the vehicle state, thereby realizing zero-speed detection. Based on zero-speed detection, additional constraints can be further used in the subsequent adaptive Kalman filter when the vehicle is stationary.

[0094] Step 3: Based on the preprocessed data in step 1, a convolutional network with a compression-excitation mechanism is used to estimate the process noise covariance and its uncertainty through multi-task learning, as follows:

[0095] The input of the process noise estimation network driven by IMU data is the gyroscope data and accelerometer data after data preprocessing, and the network output is the process noise scaling factor:

[0096]

[0097] in, and are the gyroscope data and accelerometer data with Gaussian noise added; ρ is the process noise scale factor; μ is the process noise uncertainty; N is the sliding window length; F 2 (θ) is the designed process noise estimation network.

[0098] like Figure 4 As shown, the process noise estimation network structure is as follows:

[0099] Two identical modules are used to extract features from the input data. In each module, a one-dimensional convolutional layer is first used to extract the initial features and activated by the ReLU function, and then the weight of each channel is dynamically adjusted by the compression excitation module. In the compression excitation module, the spatial information of each channel is first globally aggregated by GAP, and then passed through two fully connected layers and activated by the sigmoid function. The result is re-weighted and output for the channels of the original feature map, and the process noise scale factor output by the noise estimation network is used to dynamically adjust the process noise covariance, as shown below:

[0100] Q k =Q·ρk (6)

[0101]

[0102] Among them, Q k is the adaptive process noise covariance at time k; is the initial matrix; ρ k is the diagonal matrix converted from the process noise scale factor at time k.

[0103] Step 4: For the network in steps 2 and 3, a loss function based on uncertainty estimation of sensor data noise is constructed, and the error penalty weight is dynamically adjusted to improve the accuracy and robustness of the network output under various motion states;

[0104] Furthermore, step 4 is as follows: By using multi-task training in the data-driven module, the network estimates the uncertainty associated with each sensor and incorporates it into the training, thereby improving the estimation performance of the model. Both the proposed speed estimation network and the process noise estimation network are optimized using the negative log-likelihood loss function. In the case of incorporating uncertainty, maximizing the likelihood of the predicted distribution can effectively capture the random uncertainty present in the data:

[0105]

[0106] y k is the reference true value; f(x k ) is the input x k The estimated value of the corresponding output; σ(x k ) 2 is the model uncertainty.

[0107] Specifically, the velocity estimation network y in step 2 k is the real three-dimensional velocity, f(x k ) is the output three-dimensional velocity. Since the sensor data exhibits inconsistent noise under different motion states, the accuracy of velocity estimation can be improved by implementing uncertainty estimation. For the process noise covariance estimation network in step 3, since the process noise covariance reflects the state of the filter, when constructing the loss function of the process noise network, y k is the real displacement, f(x k ) is the displacement output by the adaptive Kalman filter.

[0108] By introducing uncertainty in the loss function, the proposed network can more effectively capture and quantify the noise characteristics in the real-world environment, thereby enabling the neural Kalman filter system to adapt to state changes and improve the localization performance.

[0109] In step 5, INS mechanical arrangement is performed based on the IMU raw data. Combining the pseudo-measurement constraints and measurement noise estimated in step 2 and the process noise covariance estimated in step 3, an adaptive Kalman filter is constructed for vehicle inertial positioning to improve the positioning accuracy and robustness in multi-motion states in a GNSS-denied environment.

[0110] Furthermore, step 5 is specifically as follows: using extended Kalman filtering for inertial positioning, the state vector is defined as follows:

[0111]

[0112] Among them, p n is the position in the navigation system; v n is the speed under the navigation system; For posture; ω is the gyroscope bias under the load system; b a is the accelerometer bias under the load system; is the IMU installation angle between the carrier system and the vehicle system; It is the IMU arm between the carrier system and the vehicle system.

[0113] In the prediction step, the covariance matrix is ​​updated as follows:

[0114]

[0115] in, is the prior state covariance at time k; is the posterior state covariance at time k-1; F k-1 and G k-1 are the Jacobian matrices of the nonlinear functions of state propagation respectively; Q is the process noise covariance, which is dynamically estimated by step 3; v is the coordinate transformation matrix from the carrier system to the navigation system at time k-1; k-1 is the speed of the navigation system at time k-1; p k-1 is the position in the navigation system at time k-1; g is gravity; dt is the time interval; (·×) represents the antisymmetric matrix.

[0116] In the update step, the vehicle speed calculated by the INS is:

[0117]

[0118] in Mounting angle for the IMU; for the lever arm; and is zero-mean Gaussian noise.

[0119] The vehicle 3D velocity is used as a pseudo-measurement constraint and is defined as follows:

[0120]

[0121] Among them, [v for ,v lat ,v up ] T is the three-dimensional velocity of the vehicle estimated in step 2; n c is the measurement noise, whose covariance is the R estimated in step 2 k .

[0122] Based on the above two formulas, the measurement equation when the vehicle is moving is:

[0123]

[0124] Among them, n c is the noise of the pseudo-measurement of the vehicle's three-dimensional velocity, with a covariance of R; H 1 is the Jacobian matrix of the nonlinear function of the measurements under moving conditions.

[0125] Based on the zero-speed detection in step 2, when the vehicle is stationary, the measurement equation is as follows:

[0126]

[0127] H 2 =[0 3×3 ,I 3×3 ,0 15×3 ] (18)

[0128] Among them, n c0 is the noise of pseudo-measurement of vehicle 3D velocity; H 2 is the Jacobian matrix of the nonlinear function of the measurements at rest.

[0129] Next, a vehicle-mounted inertial positioning device based on joint drive of data and model proposed in accordance with an embodiment of the present application is described with reference to the accompanying drawings.

[0130] Figure 5 It is a block diagram of a vehicle-mounted inertial positioning device based on joint drive of data and model according to an embodiment of the present application.

[0131] like Figure 5 As shown, the vehicle-mounted inertial positioning device 10 based on the joint driving of data and model includes: a processing module 201 , a first calculation module 202 , a second calculation module 203 , an adjustment module 204 and a filtering module 205 .

[0132] Among them, the processing module 201 is used to preprocess the acceleration data and gyroscope data of the vehicle-mounted IMU; the first calculation module 202 is used to input the preprocessed data into the three-dimensional velocity estimation network, the three-dimensional velocity estimation network outputs the vehicle's three-dimensional velocity and measurement noise uncertainty, uses the vehicle's three-dimensional velocity as a pseudo-measurement constraint, uses the measurement noise uncertainty as a measurement noise scaling factor, and calculates the measurement noise covariance based on the measurement noise scaling factor; the second calculation module 203 is used to input the preprocessed data into the process noise estimation network, and the process noise estimation network outputs the process noise scaling factor and the process noise uncertainty , calculate the process noise covariance according to the process noise scale factor; the adjustment module 204 is used to construct a loss function for uncertainty estimation, adjust the error penalty weight of the three-dimensional velocity estimation network according to the loss function and the measurement noise uncertainty, and adjust the error penalty weight of the process noise estimation network according to the loss function and the process noise uncertainty; the filtering module 205 is used to mechanically arrange the acceleration data and gyroscope data of the vehicle-mounted IMU, perform extended Kalman filtering according to the mechanically arranged data, pseudo-measurement constraints, zero-speed detection results, measurement noise covariance and process noise covariance, and perform vehicle-mounted inertial positioning according to the filtering results.

[0133] In the embodiment of the present application, the processing module 201 is further used to: add random Gaussian noise to the acceleration data and the gyroscope data; and segment the acceleration data and the gyroscope data using a sliding window.

[0134] In an embodiment of the present application, a three-dimensional speed estimation network includes an input block, a residual block and an output block, wherein the input block includes a convolution layer, a batch normalization layer and a maximum pooling layer, the convolution layer is used for feature extraction, the batch normalization layer normalizes the output of the convolution layer, and the maximum pooling layer selects the feature of the maximum value in the pooling window; the residual block controls two convolution layers, the two convolution layers are connected through batch normalization and a residual, and a self-attention module is fused after each residual block; the output block outputs the three-dimensional speed of the vehicle and the measurement noise uncertainty, the output block includes a fully connected layer and a batch normalization layer, and multiple Dropout layers are set after the fully connected layer and the batch normalization layer.

[0135] In an embodiment of the present application, the output block of the three-dimensional speed estimation network is provided with a Tanh function, and the Tanh function is replaced with a Sigmoid function, and the vehicle state is binary classified to obtain a zero-speed detection result.

[0136] In an embodiment of the present application, the process noise estimation network includes two feature extraction modules, wherein the structures of the two feature extraction modules are the same, the feature extraction modules include a convolutional layer, a ReLU function and a compression excitation module, and after the convolutional layer extracts features and is activated using the ReLU function, the compression excitation module is used to dynamically adjust the weight of each channel.

[0137] In an embodiment of the present application, the compression excitation module includes a global average pooling layer, a fully connected layer, and a sigmoid function activation, wherein the global average pooling layer globally aggregates the spatial information of the channels of each original feature map, and the globally aggregated information is activated by the sigmoid function through the fully connected layer, and the channels of the original feature map are reweighted and output according to the activated result.

[0138] In the embodiment of the present application, the loss function of uncertainty estimation is:

[0139]

[0140] Where N is the sliding window length; y k is the reference true value; f(x k ) is the input x k The estimated value of the corresponding output; σ(x k ) 2 is the model uncertainty.

[0141] In the embodiment of the present application, the state vector of the extended Kalman filter is defined as follows:

[0142]

[0143] Among them, p n is the position in the navigation system; v n is the speed under the navigation system; For posture; ω is the gyroscope bias under the load system; b a is the accelerometer bias under the load system; is the IMU installation angle between the carrier system and the vehicle system; It is the IMU arm between the carrier system and the vehicle system.

[0144] The covariance matrix is ​​updated as follows:

[0145]

[0146]

[0147] in, is the prior state covariance at time k; is the posterior state covariance at time k-1; F k-1 and G k-1 are the Jacobian matrices of the nonlinear functions of state propagation; Q is the process noise covariance; v is the coordinate transformation matrix from the carrier system to the navigation system at time k-1; k-1 is the speed of the navigation system at time k-1; p k-1is the position in the navigation system at time k-1; g is gravity; dt is the time interval; (·×) represents the antisymmetric matrix.

[0148] It should be noted that the above explanation of the embodiment of the vehicle-mounted inertial positioning method based on the joint driving of data and model is also applicable to the vehicle-mounted inertial positioning device based on the joint driving of data and model in this embodiment, and will not be repeated here.

[0149] According to the vehicle-mounted inertial positioning device based on joint data and model driving proposed in the embodiment of the present application, the acceleration data and gyroscope data of the vehicle-mounted IMU can be preprocessed and input into the three-dimensional velocity estimation network, the network outputs pseudo-measurement constraints and measurement noise scaling factors, and the measurement noise covariance is calculated according to the measurement noise scaling factors; the preprocessed data is input into the process noise estimation network, the network outputs the process noise scaling factor and process noise uncertainty, and the process noise covariance is calculated according to the process noise scaling factor; a loss function for uncertainty estimation is constructed, and the error penalty weight of the three-dimensional velocity estimation network and the error penalty weight of the noise estimation network are adjusted; the acceleration data and gyroscope data of the vehicle-mounted IMU are mechanically arranged, and extended Kalman filtering is performed, and vehicle-mounted inertial positioning is performed according to the filtering results, the three-dimensional velocity of the vehicle and the corresponding uncertainty are estimated based on deep learning, and a data-driven noise covariance adapter is introduced to dynamically adjust the process noise covariance, thereby achieving accurate and robust position estimation.

[0150] Figure 6 A schematic diagram of the structure of a vehicle provided in an embodiment of the present application. The vehicle may include:

[0151] A memory 301 , a processor 302 , and a computer program stored in the memory 301 and executable on the processor 302 .

[0152] When the processor 302 executes the program, the vehicle-mounted inertial positioning method based on joint driving of data and model provided in the above embodiment is implemented.

[0153] Furthermore, the vehicle also includes:

[0154] The communication interface 303 is used for communication between the memory 301 and the processor 302 .

[0155] The memory 301 is used to store computer programs that can be run on the processor 302 .

[0156] The memory 301 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0157] If the memory 301, the processor 302 and the communication interface 303 are implemented independently, the communication interface 303, the memory 301 and the processor 302 can be connected to each other through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0158] Optionally, in a specific implementation, if the memory 301, the processor 302 and the communication interface 303 are integrated on a chip, the memory 301, the processor 302 and the communication interface 303 can communicate with each other through an internal interface.

[0159] The processor 302 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0160] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0161] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0162] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0163] It should be understood that the various parts of the present application can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, the steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array, a field programmable gate array, etc.

[0164] A person of ordinary skill in the art may understand that all or part of the steps carried by the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the above-mentioned program may be stored in a computer-readable storage medium, which, when executed, includes one of the steps of the method embodiment or a combination thereof.

[0165] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A vehicle-mounted inertial positioning method based on joint driving of data and model, characterized in that: The following steps are involved: Preprocess the acceleration data and gyroscope data of the vehicle-mounted IMU; Inputting the preprocessed data into a three-dimensional velocity estimation network, the three-dimensional velocity estimation network outputs a three-dimensional velocity of the vehicle and a measurement noise uncertainty, taking the three-dimensional velocity of the vehicle as a pseudo-measurement constraint, taking the measurement noise uncertainty as a measurement noise scaling factor, and calculating the measurement noise covariance according to the measurement noise scaling factor; Inputting the preprocessed data into a process noise estimation network, the process noise estimation network outputs a process noise scale factor and a process noise uncertainty, and calculating the process noise covariance according to the process noise scale factor; Constructing a loss function for uncertainty estimation, adjusting the error penalty weight of the three-dimensional velocity estimation network according to the loss function and the measurement noise uncertainty, and adjusting the error penalty weight of the process noise estimation network according to the loss function and the process noise uncertainty; The acceleration data and gyroscope data of the vehicle-mounted IMU are mechanically arranged, and extended Kalman filtering is performed according to the mechanically arranged data, the pseudo-measurement constraint, the zero-speed detection result, the measurement noise covariance and the process noise covariance, and vehicle-mounted inertial positioning is performed according to the filtering result.

2. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 1 is characterized in that: The data preprocessing of the acceleration data and gyroscope data of the vehicle-mounted IMU includes: Adding random Gaussian noise to the acceleration data and the gyroscope data; The acceleration data and the gyroscope data are segmented using a sliding window.

3. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 1 is characterized in that: The three-dimensional velocity estimation network includes an input block, a residual block and an output block, wherein: The input block includes a convolution layer, a batch normalization layer and a maximum pooling layer. The convolution layer is used for feature extraction. The batch normalization layer normalizes the output of the convolution layer. The maximum pooling layer selects the feature of the maximum value in the pooling window. The residual block controls two convolutional layers, which are connected through batch normalization and a residual connection, and a self-attention module is fused after each residual block; The output block outputs the three-dimensional speed of the vehicle and the measurement noise uncertainty. The output block includes a fully connected layer and a batch normalization layer. A plurality of Dropout layers are arranged after the fully connected layer and the batch normalization layer.

4. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 3 is characterized in that: The output block of the three-dimensional speed estimation network is provided with a Tanh function, and the Tanh function is replaced with a Sigmoid function, and the vehicle state is binary classified to obtain a zero-speed detection result.

5. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 1 is characterized in that: The process noise estimation network includes two feature extraction modules, wherein the structures of the two feature extraction modules are the same, the feature extraction modules include a convolution layer, a ReLU function and a compression excitation module, and after the convolution layer extracts features and is activated using the ReLU function, the compression excitation module is used to dynamically adjust the weight of each channel.

6. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 5 is characterized in that: The compression excitation module includes a global average pooling layer, a fully connected layer, and a Sigmoid function activation, wherein the global average pooling layer globally aggregates the spatial information of the channels of each original feature map, and the globally aggregated information is activated by the Sigmoid function through the fully connected layer, and the channels of the original feature map are reweighted and output according to the activated results.

7. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 1 is characterized in that: The loss function of the uncertainty estimation is: Where N is the sliding window length; y k is the reference true value; f(x k ) is the input x k The estimated value of the corresponding output; σ(x k ) 2 is the model uncertainty.

8. The vehicle-mounted inertial positioning method based on joint driving of data and model according to claim 1 is characterized in that: The state vector of the extended Kalman filter is defined as follows: Among them, p n is the position in the navigation system; v n is the speed under the navigation system; For posture; ω is the gyroscope bias under the load system; b a is the accelerometer bias under the load system; is the IMU installation angle between the carrier system and the vehicle system; It is the IMU arm between the carrier system and the vehicle system; The covariance matrix is ​​updated as follows: in, is the prior state covariance at time k; is the posterior state covariance at time k-1; F k-1 and G k-1 are the Jacobian matrices of the nonlinear functions of state propagation; Q is the process noise covariance; v is the coordinate transformation matrix from the carrier system to the navigation system at time k-1; k-1 is the speed of the navigation system at time k-1; p k-1 is the position in the navigation system at time k-1; g is gravity; dt is the time interval; (·×) represents the antisymmetric matrix.

9. A vehicle-mounted inertial positioning device based on joint drive of data and model, characterized in that: include: A processing module, used for preprocessing the acceleration data and gyroscope data of the vehicle-mounted IMU; A first calculation module is used to input the preprocessed data into a three-dimensional speed estimation network, the three-dimensional speed estimation network outputs a three-dimensional speed of the vehicle and a measurement noise uncertainty, the three-dimensional speed of the vehicle is used as a pseudo measurement constraint, the measurement noise uncertainty is used as a measurement noise scaling factor, and the measurement noise covariance is calculated according to the measurement noise scaling factor; A second calculation module is used to input the preprocessed data into a process noise estimation network, the process noise estimation network outputs a process noise scale factor and a process noise uncertainty, and calculates the process noise covariance according to the process noise scale factor; An adjustment module, configured to construct a loss function for uncertainty estimation, adjust the error penalty weight of the three-dimensional velocity estimation network according to the loss function and the measurement noise uncertainty, and adjust the error penalty weight of the process noise estimation network according to the loss function and the process noise uncertainty; The filtering module is used to mechanically arrange the acceleration data and gyroscope data of the vehicle-mounted IMU, perform extended Kalman filtering according to the mechanically arranged data, the pseudo-measurement constraint, the zero-speed detection result, the measurement noise covariance and the process noise covariance, and perform vehicle-mounted inertial positioning according to the filtering result.

10. A vehicle, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the vehicle-mounted inertial positioning method based on joint driving of data and model as described in any one of claims 1 to 8.

Images