Transverse error calibration method and related device

Abstract

The invention discloses a transverse error calibration method and a related device, and the method comprises the steps: determining a low-frequency error signal and a high-frequency error signal under the condition that the current transverse error of a target vehicle is greater than a preset transverse error, and carrying out the alignment processing of the low-frequency error signal and the high-frequency error signal, and obtaining a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same moment, and further determining a target transverse error for calibrating the target vehicle according to the first signal and the second signal. Thus, according to the embodiment of the invention, the error signals (namely the low-frequency error signal and the high-frequency error signal) of different dimensions are determined, then the error signals of different dimensions are aligned at the same moment, and the target transverse error is determined through the first signal and the second signal at the same moment. The error signals of different dimensions are fused, and the accuracy and reliability of the transverse error are improved.

Description

Technical Field

[0001] This application relates to the field of error calibration technology, and in particular to a lateral error calibration method and related apparatus. Background Technology

[0002] With the rapid development of intelligent and automated technologies, autonomous vehicles possess multiple functions such as real-time positioning, environmental perception, path planning, and motion control to achieve autonomous driving. Among these, the lateral control of the chassis of an autonomous vehicle is one of the key factors in ensuring safety and reliability during driving.

[0003] For example, take the tracked photovoltaic cleaning robot as an example of an autonomous vehicle. Because it is equipped with wide tracks, the tracked photovoltaic cleaning robot can travel in working environments such as mud, uneven ground, potholes, and slopes. However, due to the complexity of the working environment, the tracks of the tracked photovoltaic cleaning robot may slip, slide, or sink between themselves and the ground, which will cause a lateral deviation between the actual running path of the tracked photovoltaic cleaning robot and the set running path.

[0004] Therefore, in related technologies, positioning techniques (such as signal positioning, trajectory estimation positioning, etc.) are used to correct the resulting lateral deviations so that autonomous vehicles can travel along the set operating path. However, due to the limitations of the positioning techniques in these technologies, it is difficult to meet the requirements of high precision and high reliability, resulting in low accuracy and reliability of lateral errors, and consequently, poor accuracy of the lateral error correction results. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a lateral error calibration method and related apparatus, aiming to solve the problems of low accuracy and reliability of lateral errors.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] In a first aspect, embodiments of this application provide a lateral error calibration method, the method comprising:

[0008] If the current lateral error of the target vehicle is greater than the preset lateral error, determine the low-frequency error signal and the high-frequency error signal;

[0009] Alignment processing is performed on the low-frequency error signal and the high-frequency error signal to obtain the first signal corresponding to the low-frequency error signal and the second signal corresponding to the high-frequency error signal at the same time.

[0010] The target lateral error is determined based on the first and second signals; the target lateral error is used to calibrate the target vehicle.

[0011] In conjunction with the first aspect, in one possible implementation, determining the low-frequency error signal and the high-frequency error signal includes:

[0012] Obtain vehicle motion data of the target vehicle;

[0013] Based on vehicle motion data, low-frequency error signals and high-frequency error signals are determined from the low-frequency channel and high-frequency channel, respectively.

[0014] In conjunction with the first aspect, in one possible implementation, the vehicle motion data includes inertial measurement unit (IMU) data of the target vehicle, pressure sensor data, and current data of the target vehicle's motors; it also includes obtaining low-frequency error signals through the following methods:

[0015] The ground pressure distribution and slip angle of the target vehicle are determined by using a pre-trained slip rate-frequency response model based on IMU data, pressure sensor data, and current data. The pre-trained slip rate-frequency response model is obtained by training the slip rate-frequency response model with first historical data, which includes historical IMU data, historical pressure sensor data, and historical current data.

[0016] The low-frequency error signal is determined based on the ground pressure distribution value, slip angle, and current lateral error.

[0017] In conjunction with the first aspect, in one possible implementation, the low-frequency error signal is determined based on the ground pressure distribution value, the slip angle, and the current lateral error, including:

[0018] Based on the ground pressure distribution value and the slip angle, the current lateral error is corrected to obtain the first error signal;

[0019] The first error signal is decomposed into a first preset frequency domain range to obtain the first prediction error signal;

[0020] The first prediction error signal is processed using a first Kalman filter to obtain a low-frequency error signal.

[0021] In conjunction with the first aspect, in one possible implementation, the vehicle motion data includes lidar data; it also includes obtaining high-frequency error signals through the following methods:

[0022] Point cloud matching calculations are performed on the lidar data to obtain the second error signal;

[0023] The second error signal is decomposed into a second preset frequency domain range to obtain the second prediction error signal;

[0024] The second prediction error signal is processed using a second Kalman filter to obtain a high-frequency error signal.

[0025] In conjunction with the first aspect, in one possible implementation, the low-frequency error signal and the high-frequency error signal are aligned to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time, including:

[0026] The low-frequency error signal and the high-frequency error signal are separated by frequency band separation to obtain the first signal corresponding to the low-frequency error signal and the second signal corresponding to the high-frequency error signal at the same time.

[0027] In conjunction with the first aspect, in one possible implementation, frequency band separation processing is performed on the low-frequency error signal and the high-frequency error signal respectively to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal, including:

[0028] By using a low-pass filter to perform frequency band separation on the low-frequency error signal and a band-pass filter to perform frequency band separation on the high-frequency error signal, a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time are obtained.

[0029] In conjunction with the first aspect, in one possible implementation, determining the target lateral error based on the first and second signals includes:

[0030] The first and second signals are weighted according to the terrain stiffness coefficient to obtain the target lateral error; the terrain stiffness coefficient characterizes the stiffness of the terrain traversed by the target vehicle.

[0031] In conjunction with the first aspect, one possible implementation also includes obtaining the terrain stiffness coefficient through the following method:

[0032] Obtain terrain data of the target vehicle's location, and the first stiffness coefficient corresponding to the terrain represented by the terrain data;

[0033] The terrain stiffness coefficient is determined based on the first stiffness coefficient and the preset mapping relationship; wherein, the preset mapping relationship is constructed based on the second historical data, which includes historical terrain data and historical coefficient error data, and the preset mapping relationship characterizes the relationship between terrain type and coefficient error.

[0034] In conjunction with the first aspect, in one possible implementation, the first and second signals are weighted according to the terrain stiffness coefficient to obtain the target lateral error, including:

[0035] The first weighting coefficient is determined based on the terrain stiffness coefficient, stiffness threshold, and adjustment factor.

[0036] The weighted average of the first and second signals is calculated based on the first weighting coefficient and used as the target lateral error.

[0037] In conjunction with the first aspect, one possible implementation also includes:

[0038] The observation vector is determined based on IMU data using the observation model; the observation model is used to observe the velocity and angular velocity of the target vehicle.

[0039] Based on the target's lateral error, the observation vector is updated using a Kalman filter-based algorithm to obtain the state vector;

[0040] The lateral error compensation value is calculated based on the state vector; the lateral error compensation value is used to correct the lateral error of the target vehicle.

[0041] Secondly, embodiments of this application provide a vehicle, which includes a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to execute the lateral error calibration method as described in the first aspect to correct the lateral error of the vehicle.

[0042] Thirdly, embodiments of this application provide a photovoltaic cleaning robot, including: a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to execute the lateral error calibration method as described in the first aspect to correct the lateral error of the photovoltaic cleaning robot.

[0043] In conjunction with the third aspect, one possible implementation also includes: tracks; and a motor connected to the tracks.

[0044] Fourthly, embodiments of this application provide a control device, including a processor and a memory, wherein the processor is connected to the memory, the memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to perform the lateral error calibration method as described in the first aspect.

[0045] Fifthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which is loaded by a processor to execute the lateral error calibration method as described in the first aspect.

[0046] Sixthly, embodiments of this application provide a lateral error calibration device, the device comprising:

[0047] The determination module is used to determine the low-frequency error signal and the high-frequency error signal when the current lateral error of the target vehicle is greater than the preset lateral error.

[0048] The alignment module is used to align the low-frequency error signal and the high-frequency error signal to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

[0049] The lateral error calibration module is used to determine the target lateral error based on the first signal and the second signal; the target lateral error is used to calibrate the target vehicle.

[0050] Beneficial effects:

[0051] The lateral error calibration method provided in this application determines low-frequency and high-frequency error signals when the current lateral error of the target vehicle is greater than a preset lateral error. These signals are then aligned to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time. The target lateral error for calibrating the target vehicle is further determined based on the first and second signals. Thus, this application embodiment achieves the fusion of error signals from different dimensions (i.e., low-frequency and high-frequency error signals), aligns these signals simultaneously, and determines the target lateral error using the first and second signals at the same time. This improves the accuracy and reliability of the lateral error calibration. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A schematic flowchart of a lateral error calibration method provided in an embodiment of this application;

[0054] Figure 2 A schematic diagram of a lateral error calibration method for a tracked photovoltaic cleaning robot provided in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of the structure of a lateral error calibration device provided in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the structure of a control device provided in an embodiment of this application. Detailed Implementation

[0057] As an example, let's take signal positioning technology. Signal positioning technology uses satellites such as GPS and BeiDou for positioning. It locates automated vehicles by measuring the propagation time of satellite signals and further corrects for lateral errors. However, due to problems such as signal blockage, multipath effects, and low update frequency, the accuracy and reliability of lateral errors are low.

[0058] As an example, let's take trajectory-based positioning technology. This technology continuously calculates the current position and lateral error by measuring the vehicle's speed and heading angle in real time, combined with the initial position. However, trajectory-based positioning requires odometers and gyroscopes to calculate the position of automated vehicles, but sensor errors (such as gyroscope bias and odometer slippage) accumulate over time. For example, when the gyroscope bias error is 0.1° / h, the lateral error of an automated driving vehicle can reach 0.17 meters after one hour of driving, and it cannot correct itself, resulting in low accuracy and reliability of the lateral error.

[0059] It is evident that the positioning technology provided by the relevant technologies has limitations, making it difficult to meet the requirements of high precision and high reliability, resulting in low accuracy and reliability of lateral error, and consequently, poor accuracy of the lateral error correction results.

[0060] To improve the accuracy and reliability of lateral error, and thus improve the precision of the lateral error correction result, this application provides a lateral error calibration method and related apparatus. The method includes: determining a low-frequency error signal and a high-frequency error signal when the current lateral error of the target vehicle is greater than a preset lateral error, aligning the low-frequency error signal and the high-frequency error signal to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time, and further determining the target lateral error for calibrating the target vehicle based on the first signal and the second signal.

[0061] Thus, this embodiment of the application determines error signals of different dimensions (i.e., low-frequency error signals and high-frequency error signals), then aligns the error signals of different dimensions at the same time, and then determines the target lateral error through the first signal and the second signal at the same time, thereby realizing the fusion of error signals of different dimensions and improving the accuracy and reliability of lateral error.

[0062] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0063] To facilitate the description of the following embodiments, the terms that may appear in the following embodiments will first be explained:

[0064] LiDAR (Light Detection and Ranging) is a remote sensing technology that uses laser beams to accurately detect the distance, angle, and three-dimensional shape of target objects by measuring the return time or frequency changes of the reflected light. LiDAR data refers to the three-dimensional point cloud data of the surrounding environment of a target vehicle generated by emitting laser pulses and measuring the time difference of the reflected light. This three-dimensional point cloud data includes at least one point cloud coordinate (x, y, z), which represents the position of the obstacle surface relative to the LiDAR.

[0065] An Inertial Measurement Unit (IMU) is an integrated sensor consisting of an accelerometer and a gyroscope. Its core function is to measure the linear acceleration and angular velocity of an object. IMU data refers to the linear acceleration and angular velocity data of a target vehicle collected by the inertial measurement unit.

[0066] A pressure sensor is a sensor that uses the characteristic of pressure devices (such as pressure-sensitive resistors, capacitor plates, and piezoelectric materials) to collect pressure data from the running gear (such as tires and tracks) of a target vehicle, based on the changes in pressure. Pressure sensor data refers to the pressure data of the running gear of the target vehicle collected by the pressure sensor.

[0067] An electric motor is a device that converts electrical energy into mechanical energy, used to drive the wheels or steer the system of a vehicle. Current data refers to the instantaneous current flowing through the windings of the electric motor.

[0068] The Multi-Scale Kalman Filter (MSKF) is an extension of the traditional Kalman filter. Its core idea is to decompose the system state into different frequency bands through multi-resolution analysis, apply Kalman filtering independently in each frequency band, and finally fuse the results of each frequency band to achieve the global optimal estimate.

[0069] The ICP (Iterative Closest Point) algorithm is a classic algorithm for 3D point cloud matching. It achieves point cloud alignment or registration by iteratively optimizing the spatial transformation (rotation and translation) between two frames of point clouds.

[0070] A low-pass filter is a filter that allows low-frequency signals to pass through while suppressing high-frequency signals.

[0071] A bandpass filter is a filter that allows signals in a specific frequency band (passband) to pass through while suppressing low-frequency and high-frequency signals outside the passband.

[0072] See Figure 1 The figure is a flowchart illustrating a lateral error calibration method provided in an embodiment of this application.

[0073] Combination Figure 1 As shown, the lateral error calibration method provided in this application embodiment may include:

[0074] S11: If the current lateral error of the target vehicle is greater than the preset lateral error, determine the low-frequency error signal and the high-frequency error signal.

[0075] The target vehicle refers to the vehicle to be calibrated for lateral error. In one possible implementation, the target vehicle includes, but is not limited to, automated guided vehicles (AGVs), driverless cars, photovoltaic cleaning robots, etc., without specific limitations.

[0076] The current lateral error refers to the lateral error generated by the target vehicle at the current moment. The current lateral error can be determined by methods such as, but not limited to, lidar point cloud matching, camera visual detection, trajectory extrapolation technology, dynamic models, etc., without specific limitations here.

[0077] The preset lateral error refers to the condition used to measure whether low-frequency error signals and high-frequency error signals are determined. The preset lateral error can be 0.1 meters, 0.2 meters, etc., and no specific limitation is made here.

[0078] It should be understood that, in the embodiments of this application, a lateral error (i.e., the current lateral error) can be initially estimated, and then the relationship between the current lateral error and the preset lateral error is judged to determine whether the lateral error of the target vehicle needs to be corrected. If the current lateral error is less than or equal to the preset lateral error, it indicates that the target vehicle can drive normally and no lateral error correction is required; if the current lateral error is greater than the preset lateral error, it indicates that the lateral error of the target vehicle may affect the normal driving of the target vehicle, and in this case, the lateral error needs to be corrected, i.e., the lateral error determination and correction process is executed.

[0079] Low-frequency error signals refer to error signals with a frequency range in the low-frequency range, such as 0.1~1Hz; high-frequency error signals refer to error signals with a frequency range in the high-frequency range, such as 1kHz~10kHz.

[0080] It should be understood that in actual application scenarios, the target vehicle may have error signals in different frequency bands. Error signals in different frequency bands may affect the accuracy and reliability of lateral error correction. Therefore, by determining the low-frequency error signal and the high-frequency error signal, this embodiment of the application is beneficial to use the low-frequency error signal and the high-frequency error signal together to correct the lateral error, so as to achieve error coverage in the full frequency band and improve the accuracy and reliability of lateral error correction.

[0081] In one possible implementation, vehicle motion data of the target vehicle can be acquired, and then low-frequency error signal and high-frequency difference signal can be determined from the low-frequency channel and high-frequency channel respectively based on the vehicle motion data.

[0082] Vehicle motion data refers to the data generated by the target vehicle during its operation. In one possible implementation, vehicle motion data includes, but is not limited to, lidar data, inertial measurement unit (IMU) data, pressure sensor data, and the current data of the target vehicle's motor.

[0083] A low-frequency channel refers to a transmission path in a system that allows low-frequency error signals to pass through while suppressing high-frequency noise or interference. A high-frequency channel refers to a transmission path in a system that allows high-frequency signals to pass through while suppressing low-frequency drift or systematic deviations.

[0084] It should be understood that, in the embodiments of this application, since the actual operating conditions of the target vehicle are diverse and complex, determining the low-frequency error signal from the low-frequency channel and the high-frequency error signal from the high-frequency channel by acquiring vehicle motion data can make the low-frequency error signal and the high-frequency error signal more consistent with the actual operating conditions of the target vehicle at the current moment, thereby improving the accuracy and reliability of the signal.

[0085] It should be noted that, in one possible implementation, the system parameters of the target vehicle can be initialized before executing step S11, for example, by setting the control cycle to... The terrain stiffness parameter is The adjustment factor is k, and the bandwidth of the slip ratio is The bandwidth for matching the lidar point cloud is And the bandwidth of the slip-frequency response model .

[0086] It should be understood that initialization parameters are an optional configuration step before the system, algorithm, or model starts. The purpose of initialization parameters is to set a reasonable initial state for the operating environment. By setting reasonable initialization parameters, the system stability, convergence speed, and final performance can be improved to a certain extent.

[0087] S12: Align the low-frequency error signal and the high-frequency error signal to obtain the first signal corresponding to the low-frequency error signal and the second signal corresponding to the high-frequency error signal at the same time.

[0088] Alignment processing refers to the process of synchronizing or matching low-frequency error signals and high-frequency error signals in time, space, and other dimensions.

[0089] It should be noted that the low-frequency error signal and the high-frequency error signal are out of sync due to their different sampling timings. For example, if the low-frequency error signal is sampled at 1Hz and the high-frequency error signal is sampled at 100Hz, then the timestamps of the low-frequency error signal and the high-frequency error signal will not match.

[0090] This time asynchrony makes it impossible to fuse low-frequency error signals and high-frequency error signals. Therefore, the embodiments of this application can make the time bases of low-frequency error signals and high-frequency error signals consistent through alignment processing, so as to obtain the first signal and the second signal at the same time.

[0091] As an example, assuming the sampling frequency of the low-frequency error signal is twice every 10 seconds and the sampling frequency of the high-frequency error signal is ten times every 10 seconds, then the timestamps of the low-frequency error signal and the high-frequency error signal can be unified to the same reference time through alignment processing, thereby achieving synchronization of the sampling timing.

[0092] Thus, the embodiments of this application achieve spatiotemporal alignment between different error signals (i.e., low-frequency error signals and high-frequency error signals) through alignment processing, thereby improving the accuracy and reliability of lateral error correction.

[0093] S13: Determine the target lateral error based on the first and second signals.

[0094] The target lateral error is used to calibrate the target vehicle.

[0095] It should be understood that the embodiments of this application determine error signals of different dimensions (i.e., low-frequency error signals and high-frequency error signals), and then align the error signals of different dimensions at the same time. Then, the target lateral error is determined by the first signal and the second signal at the same time, thereby realizing the spatiotemporal alignment of heterogeneous error sources, breaking the limitation of the fusion between different error signals in related technologies, and thus improving the accuracy of lateral error.

[0096] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, the low-frequency error signal can also be obtained in the following way:

[0097] A1: Obtain vehicle motion data.

[0098] The vehicle motion data includes inertial measurement unit (IMU) data, pressure sensor data, and motor current data of the target vehicle.

[0099] It should be noted that, in the embodiments of this application, the vehicle motion data can be data collected by the target vehicle's acquisition device (such as an inertial measurement unit, pressure sensor, or ammeter) according to a preset period, or the vehicle motion data can be data collected by the acquisition device when the target vehicle's current lateral error is greater than a preset lateral error. No specific limitation is made here.

[0100] A2: Using a pre-trained slip rate-frequency response model, the ground pressure distribution and slip angle of the target vehicle are determined based on IMU data, pressure sensor data, and current data.

[0101] The slip-rate-frequency response model refers to a model used to calculate the ground pressure distribution and slip angle. A pre-trained slip-rate-frequency response model can be obtained by training the model with first historical data, which includes historical IMU data, historical pressure sensor data, and historical current data.

[0102] It should be noted that since the current lateral error is a preliminary estimate, its accuracy may be low. Therefore, to further improve the accuracy of the lateral error, this embodiment of the application can determine the ground pressure distribution value and slip angle by establishing a pre-trained slip rate-frequency response model (i.e., a kinematic model). Since the ground pressure distribution value can reflect the actual contact state between the running parts (e.g., tracks) and the ground, the tire load distribution can be quantified. At the same time, the slip angle characterizes the degree of lateral deformation of the tracks; that is, the ground pressure distribution value and slip angle can describe the motion characteristics of the target vehicle under different terrain conditions.

[0103] A3: Determine the low-frequency error signal based on the ground pressure distribution value, slip angle, and current lateral error.

[0104] Ground pressure distribution value refers to the vertical pressure (unit: kPa or N / cm²) borne per unit area within the contact zone between the traveling component and the ground. The ground pressure distribution value reflects the mechanical characteristics of the contact between the traveling component and the ground.

[0105] The slip angle (or sideslip angle) is the angle (unit: degrees or radians) between the vehicle's travel direction (the longitudinal axis of the vehicle body) and the actual movement direction of the traveling components (such as the rolling direction of the tracks). The slip angle is a key parameter describing the lateral deformation of the tracks and reflects the lateral motion characteristics of the target vehicle when turning.

[0106] It should be understood that, in the embodiments of this application, the low-frequency error signal is obtained by using the ground pressure distribution value, slip angle and current lateral error. This allows the low-frequency error signal to be integrated with the environmental factors of the target vehicle, improving the adaptability of the control system to complex operating environments and thus improving the accuracy of the lateral error.

[0107] In one possible implementation, step A3 can be achieved in the following way:

[0108] B1: Correct the current lateral error based on the ground pressure distribution value and slip angle to obtain the first error signal.

[0109] It should be understood that since the ground pressure distribution value and slip angle can describe the motion characteristics of the target vehicle under different terrain conditions, correcting the current lateral error by using the ground pressure distribution value and slip angle can make the corrected first error signal integrate the environmental factors of the target vehicle, improve the adaptability of the control system to complex operating environments, and further improve the accuracy of the lateral error.

[0110] B2: Decompose the first error signal into a first preset frequency domain range to obtain the first prediction error signal.

[0111] The first preset frequency range is a low-frequency range, such as 0.1~1Hz. In one possible implementation, the first preset frequency range can be the bandwidth of the slip-frequency response model in the initialization parameters. .

[0112] It should be understood that, since the first error signal is a low-frequency signal, in order to reduce the impact of the high-frequency components in the first error signal on subsequent control, and to facilitate the fusion of the low-frequency error signal and the high-frequency error signal in the future, the first error signal can be decomposed into a first preset frequency domain range to ensure that the signal is concentrated in the responsive frequency band. At the same time, it can enhance the anti-interference ability of the signal and improve its robustness.

[0113] In one possible implementation, wavelet packet decomposition can be used to map the first error signal to a first preset frequency domain range to obtain the first prediction error signal.

[0114] Wavelet packet decomposition is an extension of wavelet analysis, enabling comprehensive multi-scale and multi-resolution analysis of signals through finer frequency domain partitioning and adaptive basis function selection. Wavelet packet decomposition offers advantages such as high-resolution decomposition across the entire frequency band, adaptive basis function selection, and strong noise resistance, making the first prediction error signal more accurate and reliable.

[0115] In this embodiment of the application, wavelet packet decomposition is used to limit the first error signal to the target low-frequency band (i.e., within the first preset frequency domain range), which can reduce high-frequency noise interference.

[0116] B3: The first prediction error signal is processed using the first Kalman filter to obtain the low-frequency error signal.

[0117] The first Kalman filter refers to the filter that performs the optimal error estimation for the first prediction error signal. In one possible implementation, the first Kalman filter can be a multi-scale Kalman filter for the low-frequency channel. The multi-scale Kalman filter for the low-frequency channel is a multi-scale filter optimized for the low-frequency error signal (the first prediction error signal).

[0118] In this embodiment, the first prediction error signal obtained after frequency domain decomposition is processed by a first Kalman filter to capture low-frequency features at different scales. Then, the results from each frequency band are fused to achieve a globally optimal estimate, ultimately yielding the low-frequency error signal. This reduces noise interference, improves computational efficiency, and enhances the accuracy of the low-frequency error signal.

[0119] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, the high-frequency error signal can also be obtained in the following way:

[0120] C1: Acquire vehicle motion data.

[0121] The vehicle motion data includes LiDAR data.

[0122] It should be noted that, in the embodiments of this application, the vehicle motion data can be data collected by the target vehicle's acquisition device according to a preset cycle, or the vehicle motion data can be data collected by the acquisition device when the target vehicle's current lateral error is greater than a preset lateral error. No specific limitation is made here.

[0123] C2: Perform point cloud matching calculations on the lidar data to obtain the second error signal.

[0124] Point cloud matching computation refers to estimating a second error signal by calculating the spatial transformation relationship (such as rotation and translation) between two or more frames of point clouds. In one possible implementation, the ICP (Iterative Closest Point) algorithm can be used to perform point cloud matching computation on lidar data to obtain the second error signal. The ICP algorithm has advantages such as high accuracy, strong versatility, and high scalability, thus making the determined second error signal more accurate and reliable.

[0125] In one possible implementation, the lidar data can be preprocessed before step C2, including noise filtering and landmark removal.

[0126] Noise filtering refers to the noise reduction processing of the LiDAR data (such as noise caused by environmental interference, sensor errors, etc.) to improve the convergence of the subsequent matching algorithm and improve the accuracy of the second error signal.

[0127] Landmarks refer to stable feature points in the environment that are irrelevant to the current task. They are useful for traffic signs, building walls, etc. By removing landmarks, the probability of landmarks being misidentified as dynamic obstacles can be reduced, as well as the computational burden can be reduced and the real-time performance of the system can be improved.

[0128] C3: Decompose the second error signal into a second preset frequency domain range to obtain the second prediction error signal.

[0129] The second preset frequency range is a high-frequency range, such as 10~100Hz. In one possible implementation, the second preset frequency range can be the bandwidth matched to the lidar point cloud in the initialization parameters. .

[0130] It should be understood that, since the second error signal is a high-frequency signal, in order to reduce the impact of the low-frequency components in the first error signal on subsequent control, and to facilitate the fusion of the low-frequency error signal and the high-frequency error signal, the second error signal can be decomposed into a second preset frequency domain range to ensure that the signal is concentrated in the responsive frequency band. At the same time, this can enhance the signal's anti-interference ability and improve its robustness.

[0131] In one possible implementation, wavelet packet decomposition can be used to map the second error signal to a second preset frequency domain range to obtain the second prediction error signal.

[0132] In this embodiment, the second error signal is limited to the target high-frequency band (i.e., the second preset frequency domain range) by wavelet packet decomposition, so as to avoid the interference of low-frequency components on high-frequency estimation, while the local correlation of high-frequency components is used to improve the estimation accuracy.

[0133] C4: The second prediction error signal is processed using the second Kalman filter to obtain the high-frequency error signal.

[0134] The second Kalman filter refers to the signal that processes the second prediction error signal. In one possible implementation, the second Kalman filter can be a multi-scale Kalman filter for a high-frequency channel.

[0135] It should be understood that, since the second error signal typically exhibits local bursts and non-stationarity, multi-scale decomposition can disperse the second error signal across different frequency bands, reducing the estimation complexity at a single scale. For example, in lidar point cloud matching, high-frequency error signals may be caused by sensor jitter or dynamic obstacles; multi-scale decomposition can separate static structures from dynamic noise.

[0136] In this embodiment, a second error signal is obtained by performing point cloud matching calculations on lidar data. This second error signal is then mapped to a second preset frequency domain to obtain a second prediction error signal. A second Kalman filter is then used to process the second prediction error signal to obtain a high-frequency error signal. This reduces noise interference, improves computational efficiency, and enhances the accuracy of the high-frequency error signal.

[0137] It should be noted that in the embodiments of this application, different error signals (i.e., low-frequency error signals and high-frequency error signals) are processed using different algorithms and Kalman filters, which solves the limitations of related technologies in processing heterogeneous error sources.

[0138] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, step S12 can be implemented in the following way: perform frequency band separation processing on the low-frequency error signal and the high-frequency error signal respectively to obtain the first signal corresponding to the low-frequency error signal and the second signal corresponding to the high-frequency error signal at the same time.

[0139] Frequency band separation processing is a technique that decomposes complex signals into different frequency components (frequency bands). Its core purpose is to achieve independent analysis, processing or reconstruction of the characteristics of each frequency band by separating the low-frequency, mid-frequency and high-frequency components in the signal.

[0140] It should be understood that, in the embodiments of this application, frequency band separation processing is performed on different error signals (i.e., low-frequency error signals and high-frequency error signals) respectively. On the one hand, this achieves interference suppression of different error signals and improves signal quality; on the other hand, it facilitates the subsequent fusion of the first signal and the second signal, realizes the spatiotemporal alignment of heterogeneous error sources, breaks the limitation of fusion between different error signals in related technologies, and thus improves the accuracy of lateral error.

[0141] In one possible implementation, the process of performing frequency band separation processing on the low-frequency error signal and the high-frequency error signal can be as follows: using a low-pass filter to perform frequency band separation on the low-frequency error signal, and using a band-pass filter to perform frequency band separation on the high-frequency error signal, to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

[0142] It should be understood that, since the low-frequency error signal is a low-frequency error signal, the embodiments of this application can use a low-pass filter to perform frequency band separation on the low-frequency error signal in order to eliminate high-frequency fluctuations, extract low-frequency trends, and suppress interference. Furthermore, it can be combined with a band-pass filter to achieve full-band separation.

[0143] It should be understood that, since the high-frequency error signal is a high-frequency error signal, the embodiments of this application can use a bandpass filter to perform frequency band separation on the high-frequency error signal, so as to separate a specific frequency band from the complex high-frequency error signal. At the same time, it can suppress interference outside the passband, improve signal quality, and further cooperate with a low-pass filter to achieve full-band separation.

[0144] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, step S13 can be achieved by weighting the first signal and the second signal according to the terrain stiffness coefficient to obtain the target lateral error.

[0145] The terrain stiffness coefficient characterizes the stiffness of the terrain traversed by the target vehicle. Terrain stiffness can include, but is not limited to, grass stiffness, soil stiffness, and rock stiffness, etc., without specific limitations here.

[0146] In one possible implementation, the terrain stiffness coefficient can be obtained in the following way:

[0147] D1: Obtain the terrain data of the target vehicle's location, and the first stiffness coefficient corresponding to the terrain represented by the terrain data.

[0148] Terrain data includes, but is not limited to, elevation information, terrain features, geographic coordinates, texture attributes, and metadata, etc., without specific limitations.

[0149] The first stiffness factor refers to the initial or current stiffness factor. It should be understood that the first stiffness factor may differ for different terrains.

[0150] D2: Determine the terrain stiffness coefficient based on the first stiffness coefficient and the preset mapping relationship.

[0151] The preset mapping relationship is constructed based on the second historical data, which includes historical terrain data and historical coefficient error data. The preset mapping relationship represents the relationship between terrain type and coefficient error.

[0152] The preset mapping relationship can be displayed in tabular form, showing the error coefficients or error values ​​corresponding to different terrains. Furthermore, by combining the error coefficients or error values ​​in the preset mapping relationship, the first stiffness coefficient can be calibrated to obtain the terrain stiffness coefficient. As an example, assuming the first stiffness coefficient of terrain 1 is 2, and the error coefficient corresponding to terrain 1 in the preset mapping relationship is 1.1, then the terrain stiffness coefficient corresponding to terrain 1 is 2 × 1.1 = 2.2.

[0153] It should be understood that due to factors such as sampling, environment, and device accuracy, there may be a deviation between the stiffness coefficient corresponding to the terrain and the actual stiffness coefficient. Therefore, in order to reduce the impact of stiffness coefficient error on the accuracy of lateral error,

[0154] Weighted processing means assigning different weights to the first and second signals to obtain the target lateral error. The target lateral error is used to calibrate the target vehicle.

[0155] In one possible implementation, a first weighting coefficient can be determined based on the terrain stiffness coefficient, stiffness threshold, and adjustment factor. Then, a weighted average of the first and second signals can be calculated based on the first weighting coefficient as the target lateral error.

[0156] In one possible implementation, the first weighting coefficient It can be determined by the following formula (1):

[0157] ;

[0158] in, This represents the terrain stiffness coefficient (typically ranging from 0 to 1). This represents the stiffness threshold (e.g., 0.5), and k represents the adjustment factor (e.g., 10).

[0159] It should be noted that, since high-frequency error signals are more accurate than low-frequency error signals, in one possible implementation, the weight of the second signal can be greater than that of the first signal to ensure that the high-frequency error signal is the main reference signal for the transverse error and improve the reliability of the transverse error.

[0160] This application embodiment determines a first signal and a second signal at the same time, and then performs weighted processing on the first signal and the second signal to obtain the target lateral error. This achieves the fusion of error signals of different dimensions, improving the accuracy and reliability of the lateral error. At the same time, by introducing a terrain stiffness coefficient to perform weighted processing on the first signal and the second signal, that is, by using the terrain stiffness coefficient to reflect the terrain environment in which the target vehicle is located, the perception and evaluation of the terrain environment are realized, further improving the accuracy and reliability of the lateral error.

[0161] Based on the lateral error calibration method provided in the above embodiments, after determining the target lateral error, one possible implementation may further include:

[0162] E1: Determine the observation vector based on IMU data using the observation model.

[0163] The observation model is used to observe the speed and angular velocity of the target vehicle.

[0164] It should be understood that by using an observation model, the observed values ​​of velocity and angular velocity can be unified into the same state space, which is beneficial for fusing multi-source data and improving the robustness of the estimation.

[0165] E2: Based on the target lateral error, the observation vector is updated using a Kalman filter-based algorithm to obtain the state vector.

[0166] The Kalman filter algorithm is an optimal linear dynamic system state estimation method. It combines the system model (prediction) and observation data (update) to recursively estimate the system state in the sense of minimum mean square error (MMSE).

[0167] It should be understood that by fusing external data (i.e., target lateral error) and internal data (i.e., observation vector) to determine the state vector, error accumulation can be suppressed, thereby improving the robustness and accuracy of lateral error correction.

[0168] E3: Calculate the lateral error compensation value based on the state vector.

[0169] Lateral error compensation values ​​are used to correct lateral errors of the target vehicle, such as steering angle correction.

[0170] In one possible implementation, after determining the lateral error compensation value, the lateral error compensation value can be directly input into the kinematic model of the target vehicle, and then the kinematic model can be used to correct the current lateral error. For example, if the current steering angle is A1 and the steering angle compensation value is A2, then the corrected steering angle is A1+A2.

[0171] It should be understood that, in the embodiments of this application, the control input can be determined based on the corrected lateral error to control the target vehicle; furthermore, the updated state is used as the initial value for the next cycle to achieve closed-loop iterative control, and the accuracy and reliability of lateral error calibration are improved by introducing a more accurate target lateral error.

[0172] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, the target vehicle can be a tracked photovoltaic cleaning robot. The inventors discovered in their research on tracked photovoltaic cleaning robots that pollutants such as dust, sand, and bird droppings accumulated on the surface of photovoltaic panels can block sunlight and reduce power generation efficiency. Therefore, a tracked photovoltaic cleaning robot can be used to clean the surface of photovoltaic panels by periodically removing pollutants to restore the light transmittance of the photovoltaic panels, thereby significantly improving power generation.

[0173] After cleaning, the tracked photovoltaic cleaning robot will return to the charging area or sleep area according to a pre-planned path to await the next cleaning instruction. When the tracked photovoltaic cleaning robot travels between the photovoltaic panels and the charging area (or sleep area), the working environment along its path is relatively complex, such as muddy, uneven, potholes, and slopes. Therefore, it needs to use its tracks to travel in the complex working environment.

[0174] However, when the tracked photovoltaic cleaning robot travels along the pre-planned path, its tracks may slip, slide, or sink due to the complex working environment, which may lead to lateral deviation.

[0175] While related technologies can use positioning techniques (such as signal positioning and trajectory estimation positioning) to correct lateral deviations, enabling tracked photovoltaic cleaning robots to travel along a set path, the limitations of these positioning techniques make it difficult to meet the requirements of high precision and high reliability. This results in low accuracy and reliability of lateral errors, leading to poor precision in the correction results.

[0176] Therefore, combining Figure 2 As shown in the embodiments of this application, a lateral error calibration method for a tracked photovoltaic cleaning robot is also provided, which may include:

[0177] Step 21: Initialize the system parameters of the tracked photovoltaic cleaning robot.

[0178] Here, it is assumed that the control cycle is set as follows: The terrain stiffness parameter is The adjustment factor is k, and the bandwidth of the slip ratio is The bandwidth for matching the lidar point cloud is And the bandwidth of the slip-frequency response model .

[0179] Step 22: Collect motion data of the tracked photovoltaic cleaning robot.

[0180] The motion data includes pressure sensor data, motor current data, lidar data, and IMU data. Pressure sensor data is pressure data collected by pressure sensors on the tracks of the tracked photovoltaic cleaning robot. LiDAR data is point cloud data of the surrounding environment of the tracked photovoltaic cleaning robot collected by lidar. IMU data is angular velocity and linear velocity data of the tracked photovoltaic cleaning robot collected using an inertial measurement unit. Current data is the current of the motor connected to the tracks.

[0181] Step 23: Determine whether the current lateral error of the tracked photovoltaic sweeping robot is less than or equal to the preset lateral error. If the current lateral error is less than or equal to the preset lateral error, return to step 1; if the current lateral error is greater than the preset lateral error, proceed to step 4.

[0182] Step 24: Determine the target lateral error of the tracked photovoltaic cleaning robot.

[0183] Step 24 is the same as the process of determining the target lateral error in steps S12-S13 of the above embodiments, so it will not be described again. For the relevant explanation of step 24 in the embodiments of this application, please refer to the explanation of steps S12-S13.

[0184] Step 25: Determine the lateral error correction value based on the IMU data and the target lateral error.

[0185] The lateral error correction value is used to correct the lateral error of the chassis of the tracked photovoltaic sweeping robot.

[0186] This application embodiment determines error signals of different dimensions (i.e., low-frequency error signals and high-frequency error signals), and then performs frequency band separation and weighting processing on the error signals of different dimensions to obtain the target lateral error. This achieves the fusion of error signals of different dimensions, thereby improving the accuracy and reliability of the lateral error.

[0187] In addition, the first and second signals can be weighted by the terrain stiffness coefficient, that is, the terrain stiffness coefficient can be used to reflect the terrain environment in which the tracked photovoltaic sweeping robot is located, so as to realize the perception and assessment of the terrain environment and further improve the accuracy and reliability of the lateral error.

[0188] Based on the lateral error calibration method provided in the above embodiments, see [link to relevant documentation]. Figure 3The figure is a schematic diagram of the structure of a lateral error calibration device provided in an embodiment of this application.

[0189] Combination Figure 3 As shown, the lateral error calibration device 30 provided in this application embodiment may include:

[0190] The determination module 31 is used to determine the low-frequency error signal and the high-frequency error signal when the current lateral error of the target vehicle is greater than the preset lateral error.

[0191] Alignment module 32 is used to align low-frequency error signals and high-frequency error signals to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

[0192] The lateral error calibration module 33 is used to determine the target lateral error based on the first signal and the second signal; the target lateral error is used to calibrate the target vehicle.

[0193] In one possible implementation, the determining module 31 includes: a determining unit, configured to: acquire vehicle motion data of the target vehicle; and determine low-frequency error signals and high-frequency error signals from the low-frequency channel and the high-frequency channel, respectively, based on the vehicle motion data.

[0194] In one possible implementation, the vehicle motion data includes inertial measurement unit (IMU) data of the target vehicle, pressure sensor data, and current data of the target vehicle's motor; it also includes a low-frequency error signal determination unit for obtaining a low-frequency error signal.

[0195] The ground pressure distribution and slip angle of the target vehicle are determined by using a pre-trained slip rate-frequency response model based on IMU data, pressure sensor data, and current data. The pre-trained slip rate-frequency response model is obtained by training the slip rate-frequency response model with first historical data, which includes historical IMU data, historical pressure sensor data, and historical current data.

[0196] The low-frequency error signal is determined based on the ground pressure distribution value, slip angle, and current lateral error.

[0197] In one possible implementation, the low-frequency error signal determination unit is specifically used for:

[0198] The current lateral error is corrected based on the ground pressure distribution value and the slip angle to obtain the first error signal;

[0199] The first error signal is decomposed into a first preset frequency domain range to obtain the first prediction error signal;

[0200] The first prediction error signal is processed using a first Kalman filter to obtain a low-frequency error signal.

[0201] In one possible implementation, the vehicle motion data includes lidar data; it also includes a high-frequency error signal determination unit for obtaining a high-frequency error signal.

[0202] Point cloud matching calculations are performed on the lidar data to obtain the second error signal;

[0203] The second error signal is decomposed into a second preset frequency domain range to obtain the second prediction error signal;

[0204] The second prediction error signal is processed using a second Kalman filter to obtain a high-frequency error signal.

[0205] In one possible implementation, the alignment module 32 is used to perform frequency band separation processing on the low-frequency error signal and the high-frequency error signal respectively, so as to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

[0206] In one possible implementation, the alignment module 32 is specifically used to perform frequency band separation on the low-frequency error signal using a low-pass filter and on the high-frequency error signal using a band-pass filter, so as to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

[0207] In one possible implementation, the lateral error calibration module 33 is used to weight the first signal and the second signal according to the terrain stiffness coefficient to obtain the target lateral error; the terrain stiffness coefficient characterizes the stiffness of the terrain traversed by the target vehicle.

[0208] In one possible implementation, a coefficient determination module is also included to obtain the terrain stiffness coefficient:

[0209] Obtain terrain data of the target vehicle's location, and the first stiffness coefficient corresponding to the terrain represented by the terrain data;

[0210] The terrain stiffness coefficient is determined based on the first stiffness coefficient and the preset mapping relationship; wherein, the preset mapping relationship is constructed based on the second historical data, which includes historical terrain data and historical coefficient error data, and the preset mapping relationship characterizes the relationship between terrain type and coefficient error.

[0211] In one possible implementation, the lateral error calibration module 33 is specifically used to determine the first weighting coefficient based on the terrain stiffness coefficient, stiffness threshold, and adjustment factor.

[0212] The weighted average of the first and second signals is calculated based on the first weighting coefficient and used as the target lateral error.

[0213] In one possible implementation, it also includes: a correction module, used for:

[0214] The observation vector is determined based on IMU data using the observation model; the observation model is used to observe the velocity and angular velocity of the target vehicle.

[0215] Based on the target's lateral error, the observation vector is updated using a Kalman filter-based algorithm to obtain the state vector;

[0216] The lateral error compensation value is calculated based on the state vector; the lateral error compensation value is used to correct the lateral error of the target vehicle.

[0217] It should be noted that the lateral error calibration device provided in this application embodiment has the same beneficial effects as the lateral error calibration method provided in the above embodiments, and therefore will not be described again.

[0218] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, this application embodiment also provides a vehicle, which includes a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to execute the lateral error calibration method described in any of the above embodiments to correct the lateral error of the vehicle.

[0219] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, a photovoltaic cleaning robot includes: a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to execute the lateral error calibration method described in any of the above embodiments to correct the lateral error of the photovoltaic cleaning robot.

[0220] In one possible implementation, the photovoltaic cleaning robot may also include: tracks; and a motor connected to the tracks.

[0221] Based on the lateral error calibration method provided in the above embodiments, in one possible implementation, see [link to relevant documentation]. Figure 4 The figure is a schematic diagram of a control device provided in an embodiment of this application.

[0222] The control device may include a memory 411 and a processor 412. For example... Figure 4 As shown, the memory can be random access memory (RAM), flash memory, read-only memory (ROM), EPROM, non-volatile read-only memory (Electronic Programmable ROM), registers, hard disks, removable disks, etc.

[0223] The memory 411 can store computer instructions. When the computer instructions stored in the memory 411 are executed by the processor 412, the processor 412 can be used to perform the lateral error calibration method. The memory 411 can also store data, such as information like the preset lateral error involved in the above embodiments.

[0224] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape) or a semiconductor medium (e.g., solid-state disk (SSD)).

[0225] This application also provides a readable storage medium for storing the methods provided in the above embodiments. Examples include random access memory (RAM), flash memory, read-only memory (ROM), EPROM, non-volatile read-only memory (EPROM), registers, hard disks, removable disks, or any other form of storage medium in the art.

[0226] In the embodiments of this application, the terms "first" and "second" (if they exist) are used only as name identifiers and do not represent the order of first and second.

[0227] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Regarding the methods disclosed in the embodiments, since they correspond to the product embodiments disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the description of the product embodiments.

[0228] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A lateral error calibration method, characterized in that, The method includes: If the current lateral error of the target vehicle is greater than the preset lateral error, determine the low-frequency error signal and the high-frequency error signal; The low-frequency error signal and the high-frequency error signal are aligned to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time. The target lateral error is determined based on the first signal and the second signal; the target lateral error is used to calibrate the target vehicle.

2. The lateral error calibration method according to claim 1, characterized in that, The determination of the low-frequency error signal and the high-frequency error signal includes: Obtain the vehicle motion data of the target vehicle; The low-frequency error signal and the high-frequency error signal are determined from the low-frequency channel and the high-frequency channel, respectively, based on the vehicle motion data.

3. The lateral error calibration method according to claim 1, characterized in that, The vehicle motion data includes inertial measurement unit (IMU) data, pressure sensor data, and motor current data of the target vehicle; it also includes the low-frequency error signal obtained through the following methods: The ground pressure distribution and slip angle of the target vehicle are determined by using a pre-trained slip rate-frequency response model based on the IMU data, the pressure sensor data, and the current data. The pre-trained slip rate-frequency response model is obtained by training the slip rate-frequency response model with first historical data, which includes historical IMU data, historical pressure sensor data, and historical current data. The low-frequency error signal is determined based on the ground pressure distribution value, the slip angle, and the current lateral error.

4. The lateral error calibration method according to claim 3, characterized in that, The step of determining the low-frequency error signal based on the ground pressure distribution value, the slip angle, and the current lateral error includes: The current lateral error is corrected based on the ground pressure distribution value and the slip angle to obtain a first error signal; The first error signal is decomposed into a first preset frequency domain range to obtain the first prediction error signal; The first prediction error signal is processed using a first Kalman filter to obtain the low-frequency error signal.

5. The lateral error calibration method according to claim 2, characterized in that, The vehicle motion data includes lidar data; it also includes the high-frequency error signal obtained through the following methods: Point cloud matching calculations are performed on the lidar data to obtain a second error signal; The second error signal is decomposed into a second preset frequency domain range to obtain the second prediction error signal; The second prediction error signal is processed using a second Kalman filter to obtain the high-frequency error signal.

6. The lateral error calibration method according to claim 1, characterized in that, The step of aligning the low-frequency error signal and the high-frequency error signal to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time includes: The low-frequency error signal and the high-frequency error signal are subjected to frequency band separation processing respectively to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal at the same time.

7. The lateral error calibration method according to claim 6, characterized in that, The step of performing frequency band separation processing on the low-frequency error signal and the high-frequency error signal respectively to obtain a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal includes: By using a low-pass filter to perform frequency band separation on the low-frequency error signal and a band-pass filter to perform frequency band separation on the high-frequency error signal, a first signal corresponding to the low-frequency error signal and a second signal corresponding to the high-frequency error signal are obtained at the same time.

8. The lateral error calibration method according to any one of claims 1-7, characterized in that, Determining the target lateral error based on the first signal and the second signal includes: The first signal and the second signal are weighted according to the terrain stiffness coefficient to obtain the target lateral error; the terrain stiffness coefficient characterizes the stiffness of the terrain traversed by the target vehicle.

9. The lateral error calibration method according to claim 8, characterized in that, This also includes obtaining the terrain stiffness coefficient through the following methods: Acquire terrain data of the location of the target vehicle, and a first stiffness coefficient corresponding to the terrain represented by the terrain data; The terrain stiffness coefficient is determined based on the first stiffness coefficient and the preset mapping relationship; wherein, the preset mapping relationship is constructed based on the second historical data, the second historical data includes historical terrain data and historical coefficient error data, and the preset mapping relationship characterizes the relationship between terrain type and coefficient error.

10. The lateral error calibration method according to claim 8, characterized in that, The step of weighting the first and second signals according to the terrain stiffness coefficient to obtain the target lateral error includes: The first weighting coefficient is determined based on the terrain stiffness coefficient, stiffness threshold, and adjustment factor. The weighted average of the first signal and the second signal is calculated based on the first weighting coefficient, and is used as the target lateral error.

11. The lateral error calibration method according to any one of claims 1-7, characterized in that, Also includes: An observation vector is determined based on the IMU data using an observation model; the observation model is used to observe the velocity and angular velocity of the target vehicle. Based on the target lateral error, the observation vector is updated using a Kalman filter-based algorithm to obtain the state vector; The lateral error compensation value is calculated based on the state vector; the lateral error compensation value is used to correct the lateral error of the target vehicle.

12. A vehicle, characterized in that, The vehicle includes a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to perform the lateral error calibration method as described in any one of claims 1-11 to correct the lateral error of the vehicle.

13. A photovoltaic cleaning robot, characterized in that, include: The system includes a controller, a pressure sensor, a lidar, an inertial measurement unit, and a motor; the controller is used to perform the lateral error calibration method as described in any one of claims 1-11 to correct the lateral error of the photovoltaic cleaning robot.

14. The photovoltaic cleaning robot according to claim 13, characterized in that, Also includes: Tracks; the motor is connected to the tracks.

Images