Lateral velocity prediction method and device of carrier, computer equipment and readable storage medium

Abstract

The invention relates to a lateral velocity prediction method and device of a carrier, computer equipment and a computer readable storage medium. The method comprises the following steps: acquiring to-be-processed motion characteristic parameters and to-be-processed sensor characteristic parameters of a carrier acquired by the integrated navigation system; inputting the to-be-processed motion characteristic parameters and the to-be-processed sensor characteristic parameters into a pre-lateral velocity prediction model to obtain a lateral velocity prediction value of the carrier; wherein the lateral velocity prediction model is obtained based on sample data training, and the sample data at least comprises one of the following: a sample motion characteristic parameter output by the integrated navigation system, a sample sensor characteristic parameter output by the integrated navigation system, and a sample lateral velocity time sequence synchronously acquired by the lateral velocity acquisition equipment. By adopting the method, the lateral speed prediction accuracy can be improved.

Description

Technical Field

[0001] This application relates to the fields of navigation and artificial intelligence, and in particular to a method and apparatus for predicting the lateral velocity of a carrier, a computer device, and a readable storage medium. Background Technology

[0002] Integrated navigation systems, by fusing information from multiple navigation sensors, effectively compensate for the shortcomings of single navigation methods, providing continuous and reliable navigation parameters for the vehicle in complex scenarios. Lateral velocity refers to the velocity component of the vehicle perpendicular to its direction of travel. Lateral velocity is one of the core motion parameters of an integrated navigation system, directly determining the accuracy of navigation and positioning. The accuracy of its measurement and calculation is crucial for the vehicle's path planning, attitude control, and safe operation.

[0003] In related technologies, lateral velocity can be obtained through data calculation from navigation sensors, such as integrating accelerometer data from an inertial measurement unit. However, this method is susceptible to sensor drift and noise accumulation, and the error increases significantly after prolonged use. Alternatively, it can be measured directly using high-precision reference devices, such as GPS differential sensors or laser trackers. However, these devices are expensive and difficult to operate stably in complex environments. Summary of the Invention

[0004] Therefore, it is necessary to provide a method and apparatus for predicting the lateral velocity of a carrier, a computer device, and a readable storage medium that can improve the accuracy of lateral velocity prediction, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for predicting the lateral velocity of a carrier, comprising:

[0006] Acquire the motion characteristic parameters of the carrier and the sensor characteristic parameters to be processed from the integrated navigation system;

[0007] The motion characteristic parameters and sensor characteristic parameters to be processed are input into the lateral velocity prediction model to obtain the predicted value of the lateral velocity of the carrier. The lateral velocity prediction model is trained based on sample data, which includes at least one of the following: sample motion characteristic parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and sample lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

[0008] In one embodiment, the integrated navigation system includes an accelerometer and a positioning module, and the motion characteristic parameters to be processed include at least one of the following: a steering angle time sequence, a centripetal acceleration time sequence, and a lateral displacement time sequence; wherein, the steering angle time sequence is acquired by the steering angle acquisition module, the centripetal acceleration time sequence is acquired by the accelerometer, and the lateral displacement time sequence is determined by the positioning module.

[0009] In one embodiment, the sensor characteristic parameters to be processed include at least one of the following: angle random walk of the inertial measurement unit, gyroscope white noise sequence, and accelerometer white noise sequence.

[0010] In an optional embodiment, the lateral velocity prediction method for the carrier further includes: obtaining a nominal angle random walk value and a measured angle random walk value, determining a correction coefficient based on the nominal angle random walk value and the measured angle random walk value, correcting the nominal angle random walk value based on the correction coefficient, and generating the angle random walk; obtaining an initial angular velocity value and an initial acceleration value, and performing DC component removal processing on the initial angular velocity value and the initial acceleration value to obtain processed angular velocity value and acceleration value; performing Allen variance calculation on the processed angular velocity and acceleration value to generate the gyroscope white noise sequence and the accelerometer white noise sequence.

[0011] In an optional embodiment, after obtaining the predicted lateral velocity value of the carrier, the lateral velocity prediction method of the carrier further includes: smoothing the predicted lateral velocity value.

[0012] In one embodiment, the lateral velocity prediction model is trained based on the following steps: for multiple historical scenarios, sample motion feature parameters and sample sensor characteristic parameters output by the integrated navigation system in each historical scenario are obtained; the lateral velocity time series in the historical scenario is obtained; the sample lateral velocity time series is used as supervision data, and the lateral velocity prediction model is trained based on the sample motion feature parameters and sample sensor characteristic parameters to obtain the lateral velocity prediction model.

[0013] In an optional embodiment, the sample motion characteristic parameters include samples of at least one of the following parameters: steering angle, centripetal acceleration, lateral acceleration, real-time gyroscope z-axis angular velocity, average gyroscope z-axis angular velocity, change in gyroscope z-axis angular velocity, change in lateral acceleration, vehicle speed, and position increments (including forward position increment, right position increment, and downward position increment) mechanically arranged by the Inertial Navigation System (INS). The position increments in the INS mechanical arrangement are the displacement changes of the vehicle in its own coordinate system along three orthogonal directions: forward / backward, left / right, and up / down. These are the results obtained by the inertial navigation system through two integrations (acceleration → velocity → position).

[0014] In one embodiment, the lateral velocity prediction method further includes: acquiring environmental parameters in each historical scene, determining the attention weight factor corresponding to each historical scene based on the environmental parameters, and updating the lateral velocity prediction model based on the attention weight factor.

[0015] Secondly, this application also provides a lateral velocity prediction device for a carrier, comprising:

[0016] The data acquisition module is used to acquire the motion characteristic parameters of the carrier and the sensor characteristic parameters to be processed collected by the integrated navigation system.

[0017] The lateral velocity prediction module is used to input the motion feature parameters to be processed and the sensor characteristic parameters to be processed into the lateral velocity prediction model to obtain the predicted value of the lateral velocity of the carrier. The lateral velocity prediction model is trained based on sample data, which includes at least one of the following: sample motion feature parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and sample lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method embodiments.

[0019] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method embodiments.

[0020] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method embodiments.

[0021] The lateral velocity prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product of the aforementioned carrier, wherein the lateral velocity prediction method utilizes the fusion of motion characteristic parameters output by the integrated navigation system and sensor characteristic parameters as input to the lateral velocity prediction model, and considers the influence of the sensor's own characteristics on the lateral velocity when performing lateral velocity prediction, thereby improving the accuracy of lateral velocity prediction. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a method for predicting the lateral velocity of a carrier in one embodiment;

[0024] Figure 2 This is a flowchart illustrating the lateral velocity prediction method for a carrier in another embodiment;

[0025] Figure 3 This is a structural block diagram of a lateral velocity prediction device for a carrier in one embodiment;

[0026] Figure 4 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0028] In one embodiment, such as Figure 1 As shown, a method for predicting the lateral velocity of a carrier is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0029] Step 102: Obtain the motion characteristic parameters and sensor characteristic parameters of the carrier to be processed from the integrated navigation system.

[0030] Among them, the integrated navigation system is a system used to deeply fuse data from two or more navigation devices with different principles and complementary error characteristics through information fusion algorithms (such as Kalman filters) to output navigation information that is more accurate, reliable and continuous than any single device.

[0031] In one embodiment, the integrated navigation system can be applied to fields such as autonomous vehicles, drones, robots, aerospace, surveying and exploration, and other equipment such as autonomous vehicles, drones, and robots can be identified as carriers. For different types of carriers, the factors generating lateral velocity differ, but all provide important information for the carrier's navigation. For example, for land-based carriers such as cars or robots, significant lateral velocity is generated when sideslip, drift, or understeer / oversteer occurs; for carriers such as ships / underwater vehicles, lateral velocity (drift) is generated by water flow, lateral thrust, and steering, affecting heading control and positioning accuracy; for carriers such as aircraft, lateral motion is key to describing their lateral movement and is directly related to the sideslip angle.

[0032] In one embodiment, the integrated navigation system may include, but is not limited to, sensors, fusion algorithms, and processing units.

[0033] Optionally, the sensors may include, but are not limited to, Global Navigation Satellite System (GNSS), Inertial Measurement Unit (IMU), odometer, and magnetometer.

[0034] In one specific embodiment, the sensors of the integrated navigation system include GNSS and IMU, and the sensor characteristic parameters include GNSS parameters and IMU parameters.

[0035] Alternatively, the fusion algorithm can be an extended Kalman filter or an unscented Kalman filter.

[0036] Optionally, the processing unit is used to run the fusion algorithm. For example, the processing unit can be a microcontroller unit (MCU) or the like.

[0037] Optionally, the motion characteristic parameters to be processed may include, but are not limited to, the steering angle time series, the centripetal acceleration time series, and the lateral displacement time series.

[0038] Among them, the steering angle time sequence is used to reflect the change in the vehicle's driving direction.

[0039] In one embodiment, the timing sequence of the steering angle is acquired by the steering angle acquisition module. Optionally, the steering angle acquisition module can be a steering sensor built into the integrated navigation system, or an external steering angle acquisition module.

[0040] Among them, the centripetal acceleration time sequence is used to reflect the centrifugal force characteristics during the curvilinear motion of the carrier.

[0041] In one embodiment, the timing sequence of centripetal acceleration is acquired by the accelerometer of the integrated navigation system.

[0042] The lateral displacement time sequence is used to reflect the lateral offset of the carrier relative to the preset trajectory.

[0043] In one embodiment, the timing sequence of lateral displacement is calculated and determined by the positioning module of the integrated navigation system.

[0044] In one embodiment, the sensor characteristic parameters to be processed include at least one of the IMU angular random walk (ARW), gyroscope white noise sequence, and accelerometer white noise sequence.

[0045] In one embodiment, the step of obtaining the angle random walk may include: obtaining the angle random walk nominal value, obtaining the angle random walk measurement value, determining the correction coefficient based on the angle random walk nominal value and the angle random walk measurement value, correcting the random walk nominal value based on the correction coefficient, and generating the angle random walk.

[0046] The "scene" is used to characterize the application scenario in which the carrier is located. Optionally, the scene may include, but is not limited to, urban roads, highways, indoor venues, off-road mountain roads, maritime navigation, and aerial patrols.

[0047] Alternatively, the angle random walk rating can be obtained from the technical manual provided by the IMU manufacturer.

[0048] Optionally, the angle random walk measurement can be the output data of a static test IMU for a continuously preset duration (e.g., 1 hour). Optionally, the angle random walk measurement can be a time series.

[0049] In an optional embodiment, after obtaining the rated value and measured value of the random walk angle, a correction coefficient can be determined based on the rated value and measured value of the random walk angle. The rated value of the random walk angle is then corrected using the correction coefficient, and the rated value of the random walk angle obtained after correction is the random walk angle.

[0050] For example, the formula for determining the correction coefficient can be:

[0051]

[0052] For example, the formula for determining the angle random walk can be:

[0053]

[0054] In one embodiment, the steps of generating the gyroscope white noise sequence and the accelerometer white noise sequence may include: obtaining initial angular velocity values ​​and initial acceleration values, and performing DC component removal processing on the initial angular velocity values ​​and initial acceleration values ​​to obtain processed angular velocity values ​​and acceleration values; performing Allen variance calculation on the processed angular velocity values ​​and acceleration values ​​to generate the gyroscope white noise sequence and the accelerometer white noise sequence.

[0055] For example, the carrier (such as a vehicle) on which the IMU is installed can be parked on a level, solid ground (such as a concrete parking lot) to ensure that the vehicle is in a parked state with the engine off, eliminating interference from engine vibration and vehicle body shaking; at the same time, electromagnetic interference sources around the vehicle should be cleared and the ambient temperature and humidity controlled to be stable (temperature 25±2℃, humidity 45%~65%).

[0056] Then, the IMU and vehicle-mounted data acquisition system are activated, and the initial angular velocity and initial acceleration values ​​output by the IMU are continuously acquired at a sampling frequency of 1000Hz for at least 2 hours. During the acquisition process, contact with the vehicle and acquisition equipment should be avoided to ensure data stability. The data must be correlated with microsecond-level timestamps. Optionally, the initial angular velocity value can be the angular velocity of a three-axis gyroscope, and the initial acceleration value can be the acceleration of a three-axis accelerometer.

[0057] Optionally, after obtaining the initial angular velocity and initial acceleration values, the DC component of the initial angular velocity and initial acceleration values ​​is removed to eliminate abnormal values ​​caused by accidental vibration.

[0058] Optionally, after removing the DC component from the initial angular velocity and initial acceleration values, the processed angular velocity and acceleration values ​​are obtained. The Galen variance can be calculated on the processed angular velocity and acceleration values, and the gyroscope white noise sequence and accelerometer white noise sequence can be determined based on the Allen variance calculation results.

[0059] For example, processing the angular velocity values ​​by removing the DC component and calculating the Allen variance yields a gyroscope white noise sequence, while processing the acceleration values ​​by removing the DC component and calculating the Allen variance yields an accelerometer white noise sequence.

[0060] Step 104: Input the motion characteristic parameters to be processed and the sensor characteristic parameters to be processed into the lateral velocity prediction model to obtain the predicted value of the lateral velocity of the carrier.

[0061] In one embodiment, after obtaining the lateral velocity prediction value, the lateral velocity prediction value can be smoothed.

[0062] Optionally, a prediction result smoothing module can be pre-generated, and the lateral velocity prediction value can be input into the prediction result smoothing module to achieve smoothing of the lateral velocity prediction value. Optionally, the prediction result smoothing module uses the exponential moving average (EMA) algorithm to smooth the lateral velocity prediction value, reducing prediction fluctuations.

[0063] The lateral velocity prediction model is trained based on sample data.

[0064] In one embodiment, after obtaining the motion characteristic parameters and sensor characteristic parameters of the carrier to be processed, the motion characteristic parameters and sensor characteristic parameters to be processed can be preprocessed, and then the preprocessed motion characteristic parameters and sensor characteristic parameters to be processed can be input into the lateral velocity prediction model, and the lateral velocity prediction value of the carrier output by the lateral velocity prediction model can be received.

[0065] Optionally, preprocessing may include time-series alignment of the motion feature parameters and sensor characteristic parameters to be processed based on timestamps.

[0066] In an exemplary embodiment, the timing alignment operation may include: acquiring a first timestamp of the motion feature parameters to be processed and a second timestamp of the sensor characteristic parameters to be processed, and aligning the motion feature parameters to be processed and the sensor characteristic parameters to be processed based on the first timestamp and the second timestamp. Using the first timestamp and the second timestamp to align the data can control the time synchronization error between the motion feature parameters to be processed and the sensor characteristic parameters to be processed within 1ms, ensuring data accuracy and thus improving the accuracy of lateral velocity prediction.

[0067] Optionally, preprocessing may include outlier removal using the 3σ criterion and sliding window detection method.

[0068] Optionally, preprocessing may include standardizing the data using Z-score.

[0069] Optionally, preprocessing may include time series alignment based on timestamps, outlier removal, standardization, and then constructing time series samples from the obtained data within a fixed time window. Optionally, the window length can be adjusted according to the scenario, ranging from 50ms to 200ms.

[0070] Optionally, the sample data includes sample motion characteristic parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and sample lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

[0071] In one embodiment, the lateral velocity acquisition device can be a high-precision velocity acquisition device attached to the carrier, used to acquire the lateral velocity time series of the carrier during its movement, as part of the sample data for generating a lateral velocity prediction model.

[0072] In one embodiment, the lateral velocity prediction model may include: for multiple historical scenarios, obtaining sample motion feature parameters and sample sensor characteristic parameters output by the integrated navigation system in each historical scenario; obtaining the sample lateral velocity time series in the historical scenarios; using the sample lateral velocity time series as supervision data, and training the lateral velocity prediction model based on the sample motion feature parameters and sample sensor characteristic parameters.

[0073] In an optional embodiment, the sample motion characteristic parameters include at least one of the time sequence of sample steering angle, sample centripetal acceleration, and sample lateral displacement.

[0074] In an exemplary embodiment, the sample motion characteristic parameters include samples of at least one of the following parameters: steering angle, centripetal acceleration, lateral acceleration, real-time gyroscope z-axis angular velocity, average gyroscope z-axis angular velocity, change in gyroscope z-axis angular velocity, change in lateral acceleration, vehicle speed, and position increments (including forward, right, and downward position increments) mechanically arranged by the Inertial Navigation System (INS). The position increments in the INS mechanical arrangement represent the displacement changes of the vehicle in its own coordinate system along three orthogonal directions: forward / backward, left / right, and up / down. These increments are obtained by the inertial navigation system through two integrations (acceleration → velocity → position).

[0075] Among them, the time sequence of the sample steering angle is used to reflect the change in the vehicle's driving direction.

[0076] In one embodiment, the time sequence of the sample steering angles is acquired by the steering angle acquisition module. Optionally, the steering angle acquisition module can be a steering sensor built into the integrated navigation system, or an external steering angle acquisition module.

[0077] Among them, the centripetal acceleration time series of the sample is used to reflect the centrifugal force characteristics during the carrier's curvilinear motion.

[0078] In one embodiment, the sample centripetal acceleration time sequence is acquired by the accelerometer of the integrated navigation system.

[0079] Among them, the sample lateral displacement time sequence is used to reflect the lateral offset of the carrier relative to the preset trajectory.

[0080] In one embodiment, the lateral displacement time sequence of the sample is determined by the positioning module of the integrated navigation system.

[0081] In one embodiment, the sample sensor characteristic parameters include at least one of the following: sample angle random walk of the inertial measurement unit, sample gyroscope white noise sequence, and sample accelerometer white noise sequence.

[0082] In one embodiment, the step of obtaining the sample angle random walk may include: obtaining the sample angle random walk nominal value, obtaining the sample angle random walk measured value, determining the correction coefficient based on the sample angle random walk nominal value and the sample angle random walk measured value, correcting the angle random walk nominal value based on the correction coefficient, and generating the sample angle random walk.

[0083] The "scene" is used to characterize the application scenario in which the carrier is located. Optionally, the scene may include, but is not limited to, urban roads, highways, indoor venues, off-road mountain roads, maritime navigation, and aerial patrols.

[0084] Optionally, the historical sample data can be sample motion characteristic parameters and sample sensor characteristic parameters from at least three different application scenarios.

[0085] Alternatively, the sample angle random walk ratings can be obtained from the technical manual provided by the inertial measurement unit (IMU) manufacturer.

[0086] Optionally, the sample angle random walk measurements can be the output data of a static test IMU for a continuously preset duration (e.g., 1 hour).

[0087] In an optional embodiment, after obtaining the sample angle random walk nominal value and the sample angle random walk measurement value, the sample angle random walk nominal value is corrected using the sample correction coefficient, and the corrected sample angle random walk nominal value is the sample angle random walk.

[0088] For example, the formula for determining the sample correction coefficient can be:

[0089]

[0090] For example, the formula for determining the sample angle random walk can be:

[0091]

[0092] In one embodiment, the generation of the sample gyroscope white noise sequence and the sample accelerometer white noise sequence may include: obtaining initial sample angular velocity values ​​and initial sample acceleration values, and performing DC component removal processing on the initial sample angular velocity values ​​and initial sample acceleration values ​​to obtain processed sample angular velocity values ​​and sample acceleration values; performing Allen variance calculation on the processed sample angular velocity and sample acceleration values ​​to generate the sample gyroscope white noise sequence and the sample accelerometer white noise sequence.

[0093] For example, the carrier (such as a vehicle) on which the IMU is installed can be parked on a level, solid ground (such as a concrete parking lot) to ensure that the vehicle is in a parked state with the engine off, eliminating interference from engine vibration and vehicle body shaking; at the same time, electromagnetic interference sources around the vehicle should be cleared and the ambient temperature and humidity controlled to be stable (temperature 25±2℃, humidity 45%~65%).

[0094] Then, the IMU and vehicle-mounted data acquisition system are activated, and the initial sample angular velocity and initial sample acceleration values ​​output by the IMU are continuously collected for at least 2 hours at a sampling frequency of 1000Hz. During the acquisition process, contact with the vehicle and acquisition equipment should be avoided to ensure data stability. The data must be correlated with microsecond-level timestamps. Optionally, the initial sample angular velocity value can be a three-axis gyroscope angular velocity, and optionally, the initial sample acceleration value can be a three-axis accelerometer acceleration.

[0095] Optionally, after obtaining the initial sample angular velocity value and the initial sample acceleration value, the DC component of the initial sample angular velocity value and the initial sample acceleration value are removed to eliminate abnormal values ​​caused by accidental vibration.

[0096] Optionally, after removing the DC component from the initial sample angular velocity and initial sample acceleration values, processed sample angular velocity and sample acceleration values ​​are obtained. Allen variance can be calculated on the processed sample angular velocity and sample acceleration values, and the sample gyroscope white noise sequence and sample accelerometer white noise sequence can be determined based on the Allen variance calculation results. For example, the sample angular velocity value corresponds to the sample gyroscope white noise sequence, and the sample acceleration value corresponds to the sample accelerometer white noise sequence.

[0097] For example, processing the sample angular velocity values ​​by removing the DC component and calculating the Allen variance yields a sample gyroscope white noise sequence, while processing the sample acceleration values ​​by removing the DC component and calculating the Allen variance yields a sample accelerometer white noise sequence.

[0098] In one embodiment, after obtaining the historical motion feature parameters and historical sensor characteristic parameters of the samples, training samples are generated based on the historical motion feature parameters and historical sensor characteristic parameters of the samples, and the training samples are used to generate a lateral velocity prediction model.

[0099] In one embodiment, during the generation of the lateral velocity prediction model, a long short-term memory (LSTM) model can be trained using the generated training samples to generate the lateral velocity prediction model.

[0100] In an exemplary embodiment, when generating a lateral velocity prediction model for training based on training samples, the training samples can be divided according to a preset ratio. For example, the training samples can be divided into a training set, a validation set, and a test set in a ratio of 7:2:1, and the model can be trained using the divided training samples.

[0101] In one embodiment, a sliding window method can be used to process the training samples to obtain a training set, a validation set, and a test set.

[0102] Optionally, the input window length and prediction step size can be set, and the training samples can be processed according to the input window length and prediction step size.

[0103] In an alternative embodiment, single-step prediction or multi-step prediction can be used to process the training samples.

[0104] Optionally, for single-step prediction, it may include:

[0105] (1) Determine the length of the historical window and select a fixed time step T as the number of past observations for the model input.

[0106] (2) Construct input and output samples sequentially for the original time series. For example, the original time series is... The following sample is constructed: Input X is The output y is Continue sliding the window and enter the following: The output is And so on, until the end of the sequence.

[0107] (3) Generate a complete training sample set. Based on the input and output samples, NT samples can be constructed, for example: , , , , ..., , .

[0108] (4) Divide the training set, validation set and test set, and divide the complete training sample set in chronological order to obtain the training set, validation set and test set.

[0109] For example, the complete training sample set is divided into a fixed proportion, such as the first 70% of the training sample set as the training set, the middle 20% of the training sample set as the validation set, and the last 10% as the test set.

[0110] For example, the complete training sample set can also be divided using a rolling window, such as by using a loop or slicing method to extract time series subsequences, and then dividing the samples based on the extracted subsequences.

[0111] For example, time-series cross-validation can also be used to generate training, validation, and test sets, such as dividing the data into K folds in chronological order, with the training set being the first i segments and the test set being the subsequent segments.

[0112] Optionally, for multi-step prediction, it may include:

[0113] (1) Define parameters.

[0114] Let the time series length be N, meaning there are N observations.

[0115] Let the historical window length be T, and use the past T time steps as input.

[0116] Let the prediction compensation (H) predict the value for the next H time steps.

[0117] (2) Generate sliding window samples

[0118] Starting from the beginning of the time series, slide one time step at a time to extract T consecutive values ​​as input X. Correspondingly, take the next H consecutive values ​​as output y. The sample generation cutoff condition is to ensure that the future H does not exceed the sequence boundary, that is, the last valid starting index is NTH.

[0119] (3) Mathematical expression

[0120] Given a time series

[0121] The t-th training sample can be:

[0122] a. Input

[0123] b. Output

[0124] The value of t is t=1, 2, ..., NTH.

[0125] (4) Sample size

[0126] The total number of samples that can be generated is .

[0127] (5) Divide the dataset into training set, validation set and test set.

[0128] See step (4) of the single-step prediction; it will not be repeated here.

[0129] In one embodiment, during model training, the training strategy can be dynamically adjusted using the learning rate. This may include: if the validation set loss decreases for five consecutive training epochs, then the initial learning rate is decayed, for example, decayed to 1 / 10 of the initial learning rate.

[0130] In an optional embodiment, during the training of the lateral velocity prediction model using training samples, an attention mechanism can be employed for training. Optionally, the attention mechanism can be embodied in attention weight factors, which can be determined based on the environmental parameters of the scene. Optionally, the process of determining the attention weight factors may include: acquiring environmental parameters from each historical scene; determining the attention weight factors corresponding to each historical scene based on the environmental parameters; and updating the lateral velocity prediction model based on the attention weight factors.

[0131] In one embodiment, the model structure of the lateral velocity prediction model trained using training samples includes: an input embedding layer, a multi-head attention mechanism layer, a temporal prediction layer, and a fully connected output layer.

[0132] The input embedding layer uses independent feature embedding modules to embed and encode motion parameter features (steering angle, centripetal acceleration, lateral displacement) and sensor characteristic features (ARW, white noise sequence) respectively, and then uses a feature fusion gating unit to achieve adaptive fusion of the two types of features.

[0133] The multi-head attention mechanism layer employs a spatiotemporal attention mechanism to simultaneously capture the temporal correlation and spatial coupling between features. This attention weight is then introduced into the LSTM module to enhance the model's ability to capture long-term temporal dependencies. Finally, a fully connected layer outputs the prediction lateral velocity. By combining attention patterns during model training, the model effectively captures the long- and short-term dependencies and feature coupling relationships in temporal data, significantly improving prediction accuracy in complex scenarios.

[0134] Optionally, a loss function can be set, and the lateral velocity prediction model can be trained based on the loss function.

[0135] In one embodiment, the loss function can be a hybrid loss function. For example, the hybrid loss function includes mean squared error (MSE) loss and cosine similarity loss, wherein the MSE loss is used to constrain the numerical error between the predicted value and the reference value, and the cosine similarity loss is used to constrain the consistency of the temporal change trend of the two.

[0136] In one embodiment, the lateral velocity of the sample can be obtained while acquiring the sample motion feature parameters and sample sensor characteristic parameters in each historical scenario. The acquired lateral velocity of the sample is used as reference data, and the lateral velocity of the sample is marked with a supervision label.

[0137] Optionally, the lateral velocity of the sample can be obtained using a high-precision reference device. For example, the high-precision reference device can be a differential GPS positioning system, a laser Doppler velocimeter, or a fiber optic gyroscope inertial navigation system, etc.

[0138] In an optional embodiment, after obtaining the sample motion characteristic parameters, sample sensor characteristic parameters, and sample lateral velocity, data preprocessing can be performed on the sample motion characteristic parameters, sample sensor characteristic parameters, and sample lateral velocity. Optionally, when there are multiple sample lateral velocities, a sample lateral velocity time series can be formed.

[0139] For example, preprocessing may include time-series alignment, outlier removal, feature normalization, and sequence construction.

[0140] Optionally, time alignment may include: obtaining a third timestamp of the sample motion feature parameters, a fourth timestamp of the sample sensor characteristic parameters, and a fifth timestamp of the sample lateral velocity, and aligning the sample motion feature parameters, sample sensor characteristic parameters, and sample lateral velocity based on the third, fourth, and fifth timestamps. Using the third, fourth, and fifth timestamps to align the data can control the time synchronization error between the sample motion feature parameters, sample sensor characteristic parameters, and sample lateral velocity within 1ms, ensuring the accuracy of the training samples and thus improving the accuracy of the lateral velocity prediction model.

[0141] Optionally, outlier removal can be achieved by combining the 3σ criterion with the sliding window detection method to remove abnormal data points caused by sensor malfunction or environmental interference. The 3σ criterion and sliding window detection method are existing technologies and will not be elaborated upon here.

[0142] Optionally, feature standardization can be achieved by using Z-score to standardize the sample motion feature parameters, sample sensor characteristic parameters, and sample lateral velocity, thereby eliminating the impact of dimensional differences on model training.

[0143] Optionally, sequence construction can involve constructing time-series samples by combining preprocessed sample motion feature parameters, sample sensor characteristic parameters, and sample lateral velocities within a fixed time window, with each sample corresponding to a reference lateral velocity label. Optionally, the window length of the time window can be dynamically adjusted according to the scenario, ranging from 50ms to 200ms.

[0144] In one embodiment, training samples are generated based on preprocessed sample motion feature parameters, sample sensor characteristic parameters, and sample lateral velocity, and a lateral velocity prediction model is generated based on the training samples.

[0145] In the aforementioned lateral velocity prediction method, the motion characteristic parameters output by the integrated navigation system and the sensor characteristic parameters are fused as input to the lateral velocity prediction model. This approach considers the influence of sensor characteristics on lateral velocity during prediction, thus improving the accuracy of lateral velocity prediction. Furthermore, the lateral velocity prediction model also considers sensor characteristics during training, ensuring the accuracy of the generated model and further enhancing the overall accuracy of lateral velocity prediction.

[0146] In one embodiment, such as Figure 2 As shown, a method for predicting the lateral velocity of a carrier is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0147] Step 202: For multiple historical scenarios, obtain the sample motion feature parameters and sample sensor characteristic parameters output by the integrated navigation system in each historical scenario.

[0148] Optionally, the sample motion characteristic parameters include at least one of the time series of sample steering angle, sample centripetal acceleration, and sample lateral displacement; the method further includes:

[0149] In one embodiment, the time sequence of the sample steering angle is obtained by the steering angle acquisition module;

[0150] In one embodiment, the time sequence of the centripetal acceleration of the sample is obtained by the accelerometer of the integrated navigation system;

[0151] In one embodiment, the temporal sequence of the sample's lateral displacement is determined by the positioning module of the integrated navigation system.

[0152] Optionally, the sample sensor characteristic parameters include at least one of the following: sample angle random walk of the inertial measurement unit, sample gyroscope white noise sequence, and sample accelerometer white noise sequence.

[0153] In one embodiment, the sample angle random walk generation step may include: obtaining a sample angle random walk nominal value, obtaining a sample angle random walk measurement value, determining a correction coefficient based on the sample angle random walk nominal value and the sample angle random walk measurement value, correcting the sample angle random walk nominal value based on the correction coefficient, and generating a sample angle random walk.

[0154] In one embodiment, the steps for generating the sample gyroscope white noise sequence and the sample accelerometer white noise sequence may include: obtaining initial sample angular velocity values ​​and initial sample acceleration values, and performing DC component removal processing on the initial sample angular velocity values ​​and initial sample acceleration values ​​to obtain processed sample angular velocity values ​​and sample acceleration values; performing Allen variance calculation on the processed sample angular velocity values ​​and sample acceleration values ​​to generate the sample gyroscope white noise sequence and the sample accelerometer white noise sequence.

[0155] Step 204: Obtain the sample lateral velocity time series in historical scenarios.

[0156] Step 206: Using the sample lateral velocity time series as supervision data, a lateral velocity prediction model is trained based on the sample motion feature parameters and sample sensor characteristic parameters.

[0157] Step 208: Obtain environmental parameters in each historical scene, and determine the attention weight factor corresponding to each historical scene based on the environmental parameters.

[0158] Step 210: Update the lateral velocity prediction model based on the attention weight factor to obtain the updated lateral velocity prediction model.

[0159] Step 212: Obtain the motion characteristic parameters and sensor characteristic parameters of the carrier to be processed from the integrated navigation system.

[0160] Step 214: Input the motion feature parameters to be processed and the sensor characteristic parameters to be processed into the updated lateral velocity prediction model to obtain the lateral velocity prediction value.

[0161] Step 216: Smooth the predicted lateral velocity values ​​to obtain smoothed predicted lateral velocity values.

[0162] In the aforementioned lateral velocity prediction method, the motion characteristic parameters output by the integrated navigation system and the sensor characteristic parameters are fused as input to the lateral velocity prediction model. This approach considers the influence of sensor characteristics on lateral velocity during prediction, thus improving the accuracy of lateral velocity prediction. Furthermore, the lateral velocity prediction model also considers sensor characteristics during training, ensuring the accuracy of the generated model and further enhancing the overall accuracy of lateral velocity prediction.

[0163] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0164] Based on the same inventive concept, this application also provides a lateral velocity prediction device for a carrier to implement the lateral velocity prediction method for the carrier described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the lateral velocity prediction device for a carrier provided below can be found in the limitations of the lateral velocity prediction method for a carrier described above, and will not be repeated here.

[0165] In one exemplary embodiment, such as Figure 3 As shown, a lateral velocity prediction device 300 for a carrier is provided, comprising: a data acquisition module 302 and a lateral velocity prediction module 304, wherein:

[0166] Data acquisition module 302 is used to acquire the motion characteristic parameters to be processed and the sensor characteristic parameters to be processed output by the integrated navigation system;

[0167] The lateral velocity prediction module 304 is used to input the motion feature parameters to be processed and the sensor characteristic parameters into the lateral velocity prediction model to obtain the lateral velocity prediction value of the carrier. The lateral velocity prediction model is trained based on sample data, and the sample data includes at least one of the following: sample motion feature parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

[0168] In one embodiment, the integrated navigation system includes an accelerometer and a positioning module. The motion characteristic parameters to be processed include at least one of the following: a steering angle time sequence, a centripetal acceleration time sequence, and a lateral displacement time sequence; wherein the steering angle time sequence is acquired by the steering angle acquisition module, the centripetal acceleration time sequence is acquired by the accelerometer, and the lateral displacement time sequence is determined by the positioning module.

[0169] In an optional embodiment, the sensor characteristic parameters to be processed include at least one of the following: angle random walk of the inertial measurement unit, gyroscope white noise sequence, and accelerometer white noise sequence.

[0170] In an exemplary embodiment, the lateral velocity prediction device for the carrier further includes a data determination module, configured to acquire a nominal angle random walk value and a measured angle random walk value, determine a correction coefficient based on the nominal angle random walk value and the measured angle random walk value, correct the nominal angle random walk value based on the correction coefficient, and generate the angle random walk; acquire an initial angular velocity value and an initial acceleration value, and perform DC component removal processing on the initial angular velocity value and the initial acceleration value to obtain processed angular velocity value and acceleration value; perform Allen variance calculation on the processed angular velocity value and acceleration value to generate the gyroscope white noise sequence and the accelerometer white noise sequence.

[0171] In one embodiment, the lateral velocity prediction device for the carrier further includes a smoothing module for smoothing the lateral velocity prediction value after obtaining it.

[0172] In one embodiment, the lateral velocity prediction device of the carrier further includes a model generation module, which is used to acquire sample motion feature parameters and sample sensor characteristic parameters output by the integrated navigation system in each historical scenario for multiple historical scenarios; acquire sample lateral velocity time series in the historical scenarios; use the sample lateral velocity time series as supervision data, and train a lateral velocity prediction model based on the sample motion feature parameters and sample sensor characteristic parameters.

[0173] In one embodiment, the model generation module is further configured to acquire environmental parameters in each historical scene, determine the attention weight factor corresponding to each historical scene based on the environmental parameters, and update the lateral velocity prediction model based on the attention weight factor.

[0174] Each module in the lateral velocity prediction device of the aforementioned carrier can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0175] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for predicting the lateral velocity of a carrier. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0176] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0177] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method embodiments.

[0178] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method embodiments.

[0179] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the above-described method embodiments.

[0180] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0181] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0182] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0183] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting the lateral velocity of a carrier, characterized in that, The method includes: Acquire the motion characteristic parameters of the carrier and the sensor characteristic parameters to be processed from the integrated navigation system; The motion characteristic parameters to be processed and the sensor characteristic parameters to be processed are input into the lateral velocity prediction model to obtain the lateral velocity prediction value of the carrier. The lateral velocity prediction model is trained based on sample data, which includes at least one of the following: sample motion feature parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and sample lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

2. The method according to claim 1, characterized in that, The integrated navigation system includes an accelerometer and a positioning module. The motion characteristic parameters to be processed include at least one of the following: steering angle time sequence, centripetal acceleration time sequence, and lateral displacement time sequence. The steering angle timing sequence is acquired by the steering angle acquisition module, the centripetal acceleration timing sequence is acquired by the accelerometer, and the lateral displacement timing sequence is determined by the positioning module.

3. The method according to claim 1, characterized in that, The sensor characteristic parameters to be processed include at least one of the following: angle random walk of the inertial measurement unit, gyroscope white noise sequence, and accelerometer white noise sequence.

4. The method according to claim 3, characterized in that, The method further includes: Obtain the rated value and measured value of the random walk angle, determine the correction coefficient based on the rated value and measured value of the random walk angle, correct the rated value of the random walk angle based on the correction coefficient, and generate the random walk angle. The initial angular velocity and initial acceleration values ​​are obtained, and the DC component of the initial angular velocity and initial acceleration values ​​are removed to obtain the processed angular velocity and acceleration values. The processed angular velocity and acceleration values ​​are subjected to Allen variance calculation to generate the gyroscope white noise sequence and the accelerometer white noise sequence.

5. The method according to claim 1, characterized in that, After obtaining the predicted lateral velocity value of the carrier, the method further includes: The predicted lateral velocity values ​​are smoothed.

6. The method according to any one of claims 1 to 4, characterized in that, The lateral velocity prediction model is trained based on the following steps: For multiple historical scenarios, sample motion feature parameters and sample sensor characteristic parameters output by the integrated navigation system in each of the historical scenarios are obtained respectively; Obtain the time-series sequence of lateral velocity samples in the historical scenario; The lateral velocity time series of the samples is used as supervision data, and the lateral velocity prediction model is trained based on the sample motion feature parameters and sample sensor characteristic parameters.

7. The method according to claim 6, characterized in that, The method further includes: Obtain environmental parameters from each of the historical scenarios, and determine the attention weight factor corresponding to each of the historical scenarios based on the environmental parameters; The lateral velocity prediction model is updated based on the attention weight factor.

8. A lateral velocity prediction device for a carrier, characterized in that, The device includes: The data acquisition module is used to acquire the motion characteristic parameters of the carrier and the sensor characteristic parameters to be processed collected by the integrated navigation system. The lateral velocity prediction module is used to input the motion feature parameters to be processed and the sensor characteristic parameters to be processed into the lateral velocity prediction model to obtain the lateral velocity prediction value of the carrier; wherein, the lateral velocity prediction model is trained based on sample data, and the sample data includes at least one of the following: sample motion feature parameters output by the integrated navigation system, sample sensor characteristic parameters output by the integrated navigation system, and sample lateral velocity time series synchronously acquired by the lateral velocity acquisition device.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images