Angular module vehicle lateral trajectory self-calibration method

Abstract

The present disclosure relates to the technical field of vehicle chassis technology, and particularly relates to a kind of angle module vehicle horizontal trajectory self-calibration method, the method comprises: after vehicle enters horizontal state, real-time obtains horizontal deviation state and operating state information, and horizontal deviation state and operating state information are input into trained artificial intelligence model, and the original torque correction of each wheel hub motor is output;Original torque correction is coordinated to handle so that the sum of torque correction of four wheel hub motors is zero, to obtain actual torque correction;According to actual torque correction, the driving torque of each wheel hub motor is respectively corrected, target torque is generated and sent to corresponding angle module controller, to execute differentiated torque output;During self-calibration process, each angle module keeps target rotation angle unchanged.The method can effectively identify and correct the trajectory deviation of angle module vehicle when horizontal, improve the accuracy and stability of horizontal straight driving.

Description

Technical Field

[0001] This disclosure relates to the field of vehicle drive-by-wire chassis technology, and in particular to a self-calibration method for the lateral trajectory of a corner module vehicle. Background Technology

[0002] With the development of drive-by-wire chassis, four-wheel independent steering and distributed drive technology, the vehicle's steering and drive actuators can be arranged in the four corner modules, enabling the vehicle to achieve various motion forms such as lateral driving, diagonal driving and turning on the spot.

[0003] In lateral driving mode, the four corner modules of the vehicle rotate to the lateral position at a set angle, and the vehicle moves laterally by the hub motors. In existing control methods, once the feedback value of the rotation angle of each wheel reaches the target value, the vehicle is usually considered to have entered a stable lateral driving state, and the driving torque is distributed to the four hub motors in the same way, in a fixed proportion, or in a manner based on a preset dynamic model.

[0004] However, due to factors such as steering angle zero-position calibration error, steering actuator deviation, reducer clearance, and mechanism deformation, the actual direction of wheel movement may deviate from the ideal lateral direction. Simultaneously, differences in vertical load, tire wear, and lateral slip characteristics among the wheels can cause the same torque to produce different driving forces. These deviations cannot be detected through steering angle feedback signals and are difficult to identify and correct using only closed-loop steering angle control. Furthermore, existing torque distribution methods largely rely on preset models or fixed rules, failing to adjust the driving torque of each wheel online based on the vehicle's actual lateral deviation, thus limiting their adaptability to individual vehicle differences and changes in operating conditions. Summary of the Invention

[0005] In view of this, the present disclosure aims to provide a self-calibration method for the lateral trajectory of a corner module vehicle, in order to solve the problem that when the actual movement direction of the wheel deviates from the ideal lateral direction due to factors such as zero-position error of the corner, mechanical clearance, and tire condition differences during the lateral movement of the corner module vehicle, the existing corner closed-loop control cannot identify and correct the deviation, thus causing the lateral trajectory of the vehicle to be inaccurate.

[0006] To achieve the above objectives, the technical solution disclosed herein is implemented as follows:

[0007] A method for self-calibrating the lateral trajectory of a corner module vehicle includes:

[0008] After the vehicle enters the lateral movement state, the lateral deviation status and the vehicle's operating status information are acquired in real time.

[0009] The lateral deviation state and the vehicle's operating status information are input into a trained artificial intelligence model, which then outputs the original torque correction amount for each of the four hub motors.

[0010] The original torque correction amount is coordinated to make the sum of the torque correction amounts of the four hub motors zero, thus obtaining the actual torque correction amount of each hub motor.

[0011] Based on the actual torque correction amount of each hub motor, the drive torque of the corresponding hub motor is corrected respectively, and the target torque of each hub motor is generated.

[0012] The target torque of each hub motor is sent to the corresponding corner module controller to drive the corresponding hub motor to perform differentiated torque output; wherein, during the lateral trajectory self-calibration process, the target rotation angle of the four corner modules remains unchanged.

[0013] Furthermore, the lateral deviation state includes the path deviation of the vehicle's current position relative to the ideal lateral path, the heading angle deviation of the vehicle's current heading angle relative to the initial heading angle, and the yaw rate of the vehicle.

[0014] Furthermore, the real-time acquisition of the vehicle's lateral deviation status includes:

[0015] Obtain the vehicle's initial position and initial heading angle;

[0016] The ideal lateral direction is determined based on the initial heading angle, and the ideal lateral path is established along the ideal lateral direction starting from the initial position.

[0017] Based on the vehicle's current position and initial position, the path deviation of the vehicle relative to the ideal lateral path is determined, and the heading angle deviation is determined based on the vehicle's current heading angle and initial heading angle.

[0018] Furthermore, the path deviation is the projected distance of the difference between the vehicle's current position and its initial position along the longitudinal direction of the vehicle body; the heading angle deviation is the difference between the vehicle's current heading angle and its initial heading angle.

[0019] Furthermore, the trained artificial intelligence model is a fully connected neural network.

[0020] Furthermore, the process of coordinating the original torque correction to make the sum of the torque corrections of the four hub motors zero includes:

[0021] Calculate the average of the four original torque corrections;

[0022] The original torque correction amount for each hub motor is subtracted from the average value to obtain the actual torque correction amount for the corresponding hub motor.

[0023] Furthermore, generating the target torque for each hub motor includes:

[0024] The actual torque of each hub motor is superimposed with the corresponding actual torque correction amount to obtain the preliminary target torque;

[0025] The target torque is obtained by limiting the torque range and the torque change rate of the preliminary target torque.

[0026] Furthermore, it also includes: after executing differentiated torque output, reacquiring the lateral deviation state of the vehicle, and evaluating the correction effect based on the reacquiring lateral deviation state;

[0027] When the correction effect indicates a reduction in lateral deviation, retain the current torque correction values ​​for each group.

[0028] When the correction effect indicates an increase in lateral deviation, cancel the current torque correction for each group and restore the previous effective torque.

[0029] Furthermore, the evaluation of the correction effect based on the re-acquired lateral deviation state includes:

[0030] The lateral deviation evaluation value is calculated based on the reacquired path deviation, heading angle deviation, and yaw rate.

[0031] When the evaluation value decreases, the correction effect is determined to be a reduction in horizontal deviation;

[0032] When the evaluation value increases continuously, the correction effect is determined to be an increase in horizontal deviation.

[0033] Furthermore, it also includes: when the vehicle satisfies the following conditions in multiple consecutive control cycles: path deviation is less than a first threshold, heading angle deviation is less than a second threshold, and yaw rate is less than a third threshold, the vehicle is determined to have entered a stable lateral state.

[0034] The torque correction of each hub motor during the stable lateral movement period is averaged to obtain the four-wheel torque calibration parameters under the corresponding working conditions.

[0035] Store the four-wheel torque calibration parameters in relation to the current operating conditions.

[0036] Furthermore, when the vehicle re-enters a working condition that is the same as or similar to the stored working condition, the stored four-wheel torque calibration parameters are called as the initial torque correction amount.

[0037] Furthermore, the trained artificial intelligence model is trained using an offline supervised learning method. During training, vehicle operating state information collected in simulation is used as training samples, and the optimal torque correction amount obtained by the optimization algorithm is used as the training label. The training samples are obtained by setting different operating conditions in vehicle dynamics simulation, and the training label is the optimal torque correction amount that reduces the lateral deviation state obtained by the optimization algorithm.

[0038] Furthermore, the loss function for offline supervised learning is a weighted sum of mean squared error and torque change smoothing term.

[0039] Furthermore, it also includes: when the trained artificial intelligence model outputs abnormally, the lateral deviation of the vehicle continues to increase, or any corner module malfunctions, the current torque correction amount is stopped, and the previous set of effective torque or the original four-wheel drive torque is restored.

[0040] Furthermore, if the lateral deviation of the vehicle continues to increase after the previous set of effective torques is restored, speed-limited lateral movement or lateral movement cessation shall be implemented.

[0041] Compared with the prior art, this disclosure can achieve the following beneficial effects:

[0042] After the vehicle enters a lateral movement state, the system acquires real-time information on the vehicle's lateral deviation and operating status. This information is then input into a trained artificial intelligence model to output the original torque correction values ​​for each hub motor. These original torque correction values ​​are then coordinated to ensure the sum of the torque correction values ​​for all four hub motors is zero, yielding the actual torque correction value for each hub motor. Based on these actual torque correction values, the drive torque of the corresponding hub motor is adjusted to execute differentiated torque output. Furthermore, during the self-calibration process, each corner module maintains a constant target steering angle. This solves the problem that existing corner closed-loop control systems cannot identify and correct deviations caused by factors such as zero-position steering angle error, mechanical backlash, and tire condition differences during lateral vehicle movement, leading to inaccurate lateral trajectory. Attached Figure Description

[0043] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure. In the drawings:

[0044] Figure 1 A flowchart of the self-calibration method for the lateral trajectory of a vehicle using an angle module provided in this embodiment of the disclosure;

[0045] Figure 2 This is a schematic diagram of the lateral movement position of a vehicle provided in an embodiment of this disclosure. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute a limitation thereof. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of this disclosure. However, those skilled in the art will readily recognize that some features may be omitted in different circumstances, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to this disclosure are not shown or described in the specification. This is to avoid obscuring the core parts of this disclosure with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this disclosure can be combined to form various implementations. Furthermore, the steps or actions in the method description can be rearranged or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various orders in the specification and drawings are merely for the clear description of a particular embodiment and do not imply a mandatory order, unless otherwise stated that a particular order must be followed.

[0048] In the description of this disclosure, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this disclosure and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this disclosure, unless otherwise stated, "a plurality of" means two or more.

[0049] In the description of this disclosure, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linkage" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art will understand the specific meaning of the above terms in this disclosure based on the specific circumstances.

[0050] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0051] like Figure 1 As shown, this embodiment provides a self-calibration method for the lateral trajectory of a corner module vehicle, including the following steps:

[0052] Step S1: Establish vehicle lateral movement status and acquire control information.

[0053] In some embodiments, a corner module vehicle refers to a vehicle in which the steering actuator and hub motor are respectively arranged in the four corner modules of the vehicle. Each corner module can independently control the wheel angle and driving torque, enabling the vehicle to achieve various motion forms such as lateral driving, diagonal driving and turning on the spot.

[0054] Specifically, when the vehicle needs to enter lateral driving mode, the vehicle controller receives the lateral mode request, lateral direction, and target speed, and sends the corresponding target lateral turning angles to the four corner modules (left front, right front, left rear, and right rear). In this embodiment, the target turning angles for the four corner modules can be set to -90°, 90°, 90°, and -90°, respectively. After entering lateral driving mode, the target turning angles of the four corner modules remain unchanged, and the hub motors output torque according to the vehicle's original drive commands, causing the vehicle to begin lateral movement. It should be noted that this disclosure does not recalculate the vehicle's resultant speed and the original drive torque of each hub motor, but directly calls the resultant speed signal and torque signal from the vehicle's existing control system.

[0055] The moment when the vehicle began to cross the road is recorded as The vehicle controller records the vehicle's initial position. and initial heading angle Simultaneously, the vehicle controller acquires the vehicle's lateral deviation status and operational status information in real time. The lateral deviation status includes the vehicle's current position. Vehicle heading angle yaw rate Operational status information includes the combined speed of the vehicle. The system can obtain the actual torque of the four hub motors and the rotational speed of the four wheels. Additionally, it can acquire the vertical load on each wheel and the operating status of the hub motors for fault diagnosis and torque coordination. (Torque of each wheel...) Rotation speed and vertical load They are represented as follows:

[0056] ;

[0057] ;

[0058] ;

[0059] in, , , , These represent the left front wheel, right front wheel, left rear wheel, and right rear wheel, respectively. , , , These represent the torque of the corresponding hub motor; , , , These represent the rotational speeds of the corresponding wheels; , , , These represent the vertical loads of the corresponding wheels.

[0060] Step S2: Determine the ideal lateral path and calculate the vehicle lateral deviation.

[0061] In some embodiments, the lateral deviation status of the vehicle is acquired in real time, including acquiring the vehicle's initial position and initial heading angle. An ideal lateral direction is determined based on the initial heading angle, and an ideal lateral path is established along the ideal lateral direction starting from the initial position. The path deviation of the vehicle relative to the ideal lateral path is determined based on the vehicle's current position and initial position, and the heading angle deviation is determined based on the vehicle's current heading angle and initial heading angle.

[0062] Specifically, the longitudinal unit vector of the vehicle body and the ideal lateral unit vector are determined based on the initial heading angle. In this embodiment, the longitudinal unit vector of the vehicle body can be set as:

[0063] ;

[0064] The ideal horizontal unit vector is:

[0065] ;

[0066] in This represents the longitudinal unit vector of the vehicle body. Indicates the direction of lateral movement of the vehicle. Used to distinguish between two opposite lateral movement directions. Starting from the initial position. Starting from, with Establish an ideal lateral path for the direction.

[0067] like Figure 2 As shown, the ideal lateral path is a path starting from the initial position. Start, along the ideal lateral direction An extended straight line. In the k-th control cycle, the vehicle has actually traveled to its current position. The deviation of this position relative to the ideal lateral path in the longitudinal direction of the vehicle body is the path deviation. Current heading angle of the vehicle Relative to the initial heading angle The offset is the heading angle deviation. .

[0068] In the k-th control cycle, obtain the current position of the vehicle. (k), and calculate the undesired path deviation generated by the vehicle along the longitudinal direction of the vehicle body:

[0069] ;

[0070] in, Indicates the first Path deviation per control cycle This represents the longitudinal unit vector of the vehicle body. This represents the transpose of the longitudinal unit vector of the vehicle body. Indicates the first The current position of the vehicle in each control cycle This indicates the vehicle's initial position.

[0071] Simultaneously obtain the vehicle's current heading angle and yaw rate The heading angle deviation is:

[0072] ;

[0073] in, Indicates the first The heading angle deviation per control cycle Indicates the first The vehicle's current heading angle for each control cycle. This indicates the vehicle's initial heading angle.

[0074] The vehicle's lateral deviation is determined by path deviation, heading angle deviation, and yaw rate.

[0075] ;

[0076] in, Indicates the first Vehicle lateral deviation status for each control cycle Indicates the first Path deviation per control cycle Indicates the first The heading angle deviation per control cycle Indicates the first The yaw rate for each control cycle.

[0077] When driving in an ideal direction, the vehicle only travels along... The direction of the car is translated, but the orientation of the front of the car remains unchanged. , and All values ​​should be close to zero. This lateral deviation state is used by the subsequent artificial intelligence model to calculate the torque correction amount for the four hub motors.

[0078] In some embodiments, path deviation is the projected distance of the difference between the vehicle's current position and its initial position along the longitudinal direction of the vehicle body; heading angle deviation is the difference between the vehicle's current heading angle and its initial heading angle. That is, path deviation. Heading angle deviation .

[0079] Step S3: Generate four-wheel torque correction using a lightweight fully connected neural network.

[0080] In some embodiments, lateral deviation status and vehicle operating status information are input into a trained artificial intelligence model, which then outputs the original torque correction values ​​for each of the four hub motors. The operating status information includes the vehicle's resultant speed, the actual torque of the four hub motors, and the rotational speed of the four wheels.

[0081] Specifically, based on the vehicle lateral path deviation obtained in step S2 , heading angle deviation and yaw rate By combining the vehicle's current combined speed, the actual torque of the four hub motors, and the rotational speed of the four wheels, a neural network input vector is constructed:

[0082]

[0083] ;

[0084] in, Indicates the first The neural network input vector for each control cycle, Indicates the first The combined speed of the vehicle in each control cycle, Indicates the first Path deviation per control cycle For the first The heading angle deviation per control cycle Indicates the first The yaw rate per control cycle , , , They represent the first The actual torque of the left front, right front, left rear, and right rear hub motors in each control cycle. , , , They represent the first The rotational speeds of the left front, right front, left rear, and right rear wheels are controlled in one cycle.

[0085] Each component of the above input vector is normalized before entering the network. The network mapping relationship is expressed as:

[0086] ;

[0087] in, (·) denotes a multilayer perceptron. This represents the network parameters obtained after training. Indicates the first The original torque correction amount for the four wheels in each control cycle. They represent the first The original torque correction amount of the left front, right front, left rear, and right rear hub motors for each control cycle.

[0088] In some embodiments, the trained artificial intelligence model is a fully connected neural network. Specifically, the trained artificial intelligence model can be a lightweight fully connected feedforward neural network, i.e., a multilayer perceptron. Exemplarily, the network structure can be: 12 input nodes, 32 neurons, 16 neurons, and 4 output nodes, including one input layer, two hidden layers, and one output layer. The input layer contains 12 input nodes, the two hidden layers contain 32 neurons and 16 neurons respectively, and the output layer contains 4 output nodes. Exemplarily, the two hidden layers can use a hyperbolic tangent activation function, and the output layer can use a linear activation function to output a continuous torque correction amount that can be positive or negative.

[0089] In some embodiments, the original torque correction amounts are coordinated to make the sum of the torque correction amounts of the four hub motors zero, thus obtaining the actual torque correction amount of each hub motor. Specifically, the coordination process is a total amount coordination process, which includes: calculating the average value of the four original torque correction amounts; and subtracting the average value from the original torque correction amount of each hub motor to obtain the actual torque correction amount of the corresponding hub motor. It should be noted that the total amount is the sum of the original torque correction amounts of the four hub motors; the total amount coordination process means adjusting the torque distribution relationship between the four hub motors without changing the total drive torque of the four hub motors, so that the sum of the actual torque correction amounts of the four hub motors is zero.

[0090] Specifically, in order to ensure that the model primarily adjusts the torque distribution relationship between the four wheels without significantly altering the vehicle's original total lateral driving force, the network output is processed using a zero-sum method:

[0091] ;

[0092] in, For the first The first control cycle The actual torque correction amount for each hub motor For the first The first control cycle The original torque correction amount for each hub motor For hub motor index, These represent left front, right front, left back, and right back, respectively.

[0093] Therefore, the sum of the torque corrections for the four hub motors is zero. When When >0, increase the driving torque of the corresponding hub motor; when When the torque is less than 0, the driving torque of the corresponding hub motor is reduced. The model reduces path deviation, heading angle deviation, and yaw motion during vehicle lateral movement by changing the relative magnitude of the torques of the four wheels. This step does not change the target steering angle of the four corner modules, nor does it recalculate the vehicle's lateral direction.

[0094] In some embodiments, the trained artificial intelligence model is trained using offline supervised learning. During training, vehicle operating state information collected through simulation is used as training samples, and the optimal torque correction amount obtained through an optimization algorithm is used as the training label. The training samples are obtained by collecting data under different operating conditions in vehicle dynamics simulation, and the training label is the optimal torque correction amount that reduces lateral deviation, obtained through the optimization algorithm.

[0095] Specifically, firstly, different operating conditions are set in the vehicle dynamics simulation, including different lateral speeds, vehicle loads, steering zero-position errors, tire conditions, and road adhesion conditions, and the network input states are collected. For each set of input states, a constraint optimization algorithm is used to find the optimal conditions for reducing... , and Optimal torque correction for four wheels And use it as a label for supervised training.

[0096] In some embodiments, the loss function for offline supervised learning is a weighted sum of mean squared error and a torque variation smoothing term. Specifically, the network training loss function uses a weighted sum of mean squared error and a torque variation smoothing term:

[0097] ;

[0098] Where L is the loss function value, and N is the total number of training samples. For the neural network Torque correction amount for each sample output For the first The optimal torque correction label for each sample, where λ is the weighting coefficient of the smoothing term.

[0099] For example, training can employ the Adam optimization algorithm, with an initial learning rate of 0.001 and a batch size of 128, and early stopping using a validation set to suppress overfitting. After training, the network parameters are embedded into the vehicle controller. During vehicle operation, the network only performs online forward computation and does not undergo real-time training.

[0100] Step S4: Constrain the torque correction amount and execute differentiated drive.

[0101] In some embodiments, generating the target torque for each hub motor includes: superimposing the actual torque of each hub motor with the corresponding actual torque correction amount to obtain a preliminary target torque; and applying torque range limitations and torque change rate limitations to the preliminary target torque to obtain the target torque.

[0102] Specifically, the actual torque of each hub motor is superimposed with the corresponding actual torque correction amount to obtain the preliminary target torque:

[0103] ;

[0104] in, For the first The first control cycle The initial target torque of the hub motor, For the first The actual torque of each hub motor.

[0105] To ensure the safe operation of the hub motor and the vehicle, the initial target torque is limited by its range and rate of change.

[0106] , ;

[0107] in, and The first The minimum and maximum allowable torque of each hub motor. This represents the maximum allowable torque variation in a single control cycle. When a hub motor experiences overheating, malfunction, or reduced available torque, the controller reduces the torque correction for that wheel and re-coordinates among other hub motors with torque margins to keep the total drive torque of all four wheels as stable as possible.

[0108] When a hub motor experiences overheating, malfunction, or reduced available torque, the controller reduces the torque correction for that wheel and re-coordinates among other hub motors with torque margins to keep the total drive torque of the four wheels as stable as possible.

[0109] After constraint processing, the final target torque for the four hub motors is determined:

[0110] ;

[0111] The vehicle controller sends the four final target torques to the left front, right front, left rear, and right rear corner module controllers, respectively. Each corner module controller then controls the corresponding wheel hub motors to perform differentiated drive. During this process, the four corner modules maintain the target turning angle corresponding to the lateral movement mode, and the drive direction of the four wheel hub motors remains unchanged in the original lateral direction. Only the magnitude of the drive torque of each wheel is adjusted to reduce path deviation, heading angle deviation, and yaw motion during vehicle lateral movement.

[0112] When the neural network output is abnormal, the vehicle's lateral deviation continues to increase, or any corner module malfunctions, the current torque correction amount is stopped, and the previous set of effective torque or the original four-wheel drive torque is restored; if necessary, speed-limited lateral movement or lateral movement is stopped.

[0113] Step S5: Evaluate the horizontal correction effect and update the calibration parameters.

[0114] In some embodiments, the corner module vehicle lateral trajectory self-calibration method may further include: after performing differentiated torque output, re-acquiring the vehicle's lateral deviation state, and evaluating the correction effect based on the re-acquiring lateral deviation state. When the correction effect indicates that the lateral deviation decreases, the current torque correction amounts are retained. When the correction effect indicates that the lateral deviation increases, the current torque correction amounts are revoked, and the previous effective torque is restored.

[0115] Specifically, after the four wheel hub motors execute differentiated target torques, the vehicle controller continues to acquire the vehicle's current position, heading angle, and yaw rate, and then... Figure 2 The lateral path deviation, heading angle deviation, and yaw rate are recalculated to determine whether the torque correction improves the vehicle's deviation from the ideal lateral path.

[0116] Horizontal deviation evaluation value The calculation is based on the weighted sum of path deviation, heading angle deviation, and yaw rate, i.e.:

[0117] ;

[0118] in, For the first The evaluation value of the lateral deviation for each control cycle. , and These can be weighted coefficients for path deviation, heading angle deviation, and yaw rate, respectively. This evaluation value comprehensively reflects the weighted combined effects of the above path deviation, heading angle deviation, and yaw rate; the smaller the value, the closer the vehicle is to the ideal lateral movement state.

[0119] Considering that a single increase in the evaluation value may be affected by noise or instantaneous disturbances and cannot reliably reflect the changing trend of the correction effect, this disclosure uses a continuous increase in the evaluation value as the basis for judging the deterioration of the correction effect. When the evaluation value decreases, it indicates that the current torque correction direction is effective, and the correction effect is determined to be a reduction in lateral deviation; when the evaluation value increases continuously, it indicates that the current torque correction direction is ineffective, and the correction effect is determined to be an increase in lateral deviation. In this case, the current correction amount should be canceled and the previous set of effective torques should be restored.

[0120] In some embodiments, the self-calibration method for the lateral trajectory of the corner module vehicle further includes: determining that the vehicle has entered a stable lateral state when the vehicle satisfies the following conditions for multiple consecutive control cycles: path deviation less than a first threshold, heading angle deviation less than a second threshold, and yaw rate less than a third threshold. The torque correction values ​​of each hub motor during the stable lateral period are averaged to obtain the four-wheel torque calibration parameters under the corresponding operating condition. The four-wheel torque calibration parameters are then stored in relation to the current operating condition.

[0121] Specifically, a vehicle is deemed to have entered a stable lateral movement state when it meets the following conditions for multiple consecutive control cycles:

[0122] ;

[0123] in, , and These are the allowable thresholds for path deviation, heading angle deviation, and yaw rate, respectively.

[0124] After the vehicle reaches a stable lateral movement state, the torque correction values ​​of each wheel hub motor during the stable lateral movement period are averaged to form the four-wheel torque calibration parameters for the corresponding operating conditions:

[0125] ;

[0126] in, For the first The stable torque calibration value of each hub motor The number of stable control cycles participating in the average. For the first The first control cycle The actual torque correction amount for each hub motor.

[0127] The vehicle controller stores the four-wheel calibration parameters along with the corresponding operating conditions such as lateral direction, vehicle speed, vehicle load, and road conditions.

[0128] In some embodiments, when the vehicle re-enters an operating condition that is the same as or similar to a stored operating condition, the stored four-wheel torque calibration parameters are invoked as the initial torque correction amount.

[0129] Specifically, when the vehicle re-enters the same or similar operating conditions, the stored calibration parameters are used as the initial torque correction amount, and then the artificial intelligence model makes small adjustments based on the real-time lateral deviation. When the torque correction amount reaches the allowable boundary but the lateral deviation has not decreased, the calibration parameters are stopped from being updated, and speed-limited lateral movement, lateral movement is stopped, or the steering system is prompted to recalibrate.

[0130] In some embodiments, the corner module vehicle lateral trajectory self-calibration method may further include an anomaly handling step: when the trained artificial intelligence model outputs an anomaly, the vehicle's lateral deviation continues to increase, or any corner module malfunctions, the current torque correction amount is stopped, and the previous set of effective torques or the original four-wheel drive torque is restored. If the vehicle's lateral deviation continues to increase after restoring the previous set of effective torques, speed-limited lateral movement or lateral movement is stopped.

[0131] During operation, the trained artificial intelligence model may deviate from the reasonable range in output torque correction due to sudden changes in sensor signals, overflow of network forward computation, or abnormal model parameters. A continuous increase in the vehicle's lateral deviation indicates that the current torque correction direction has failed to effectively correct the vehicle's trajectory and has instead exacerbated the deviation. A malfunction in any corner module means that the actuator at that corner cannot respond normally to control commands. If the current torque correction is continued to be used when any of these abnormal situations occur, it may lead to further deviation of the vehicle's lateral trajectory, instability of the vehicle's posture, or even safety problems. Therefore, this disclosure first stops using the current torque correction and restores the previous effective torque or the original four-wheel drive torque, allowing the vehicle to quickly exit the abnormal correction state and return to a known, safe driving mode, thereby ensuring the vehicle's basic lateral capability and system safety.

[0132] After restoring the previous set of effective torque, the system continues to monitor the vehicle's lateral deviation. If the lateral deviation continues to increase, it indicates that the current anomaly is not caused by a single torque correction anomaly, but may stem from a more serious systemic problem, such as severe steering zero-point drift, severe tire wear, or a sudden change in road surface adhesion conditions. In this case, continuing to maintain the current lateral speed may cause the vehicle to deviate further from the intended path, increasing the risk of collision. Therefore, this disclosure further implements speed-limited lateral movement or lateral movement cessation. Speed-limited lateral movement refers to continuing lateral movement after reducing the vehicle's lateral speed to reduce the trajectory deviation per unit time, giving the driver or higher-level decision-making system more reaction time; lateral movement cessation refers to completely stopping lateral movement, placing the vehicle at a standstill to avoid more serious trajectory deviation or safety accidents caused by continued lateral movement.

[0133] Through the aforementioned graded anomaly handling mechanism, this disclosure can take corresponding safety measures under anomaly conditions of varying severity, achieving a step-by-step safety guarantee from "restoring effective torque" to "speed-limited lateral movement" and then to "stopping lateral movement." When the deviation stops increasing after the previous set of effective torques is restored, the system can continue to operate in safe mode; when the deviation continues to increase after restoration, the system automatically degrades to speed-limited lateral movement or stops lateral movement. This graded handling method avoids overreaction in minor anomalies (such as directly stopping driving) while ensuring effective risk control in severe anomalies, effectively improving the robustness and safety of the system.

[0134] The above technical solution, by acquiring the lateral deviation status—composed of path deviation, heading angle deviation, and yaw rate—in real time after the vehicle enters the lateral movement state, directly reflects the vehicle's actual motion state. This solves the problem that existing cornering closed-loop control relies solely on cornering feedback and cannot identify actual motion deviations caused by steering zero-position deviation, mechanical backlash, mechanism deformation, and tire side slip. The lateral deviation status and vehicle operating status information are input into a trained artificial intelligence model to output the original torque correction for each wheel hub motor. This model can adapt to assembly errors, load variations, tire condition differences, and changes in road surface adhesion conditions in different vehicles. Compared to existing methods that rely on preset vehicle models or fixed allocation rules, it has stronger nonlinear fitting and generalization capabilities. The original torque correction amount is processed by overall coordination to make the sum of the torque correction amounts of the four hub motors zero, thus obtaining the actual torque correction amount of each hub motor. This ensures that the total drive torque remains basically constant during the adjustment of the torque distribution relationship of the four wheels, avoiding significant fluctuations in the vehicle's lateral speed and improving the smoothness of the lateral movement. Furthermore, the drive torque of the corresponding hub motors is adjusted according to the actual torque correction amount to execute differentiated torque output. During the self-calibration process, each corner module maintains a constant target steering angle, avoiding frequent steering mechanism movements, reducing mechanical shock and actuator wear, and simultaneously reducing the coupling between steering control and lateral trajectory control.

[0135] Furthermore, this disclosure continuously evaluates the correction effect after executing differentiated torque output. When the correction is effective, the current correction amount is retained and stored; when the correction is ineffective, it is revoked and the previous set of effective torque is restored. After the vehicle enters a stable lateral movement state, the four-wheel torque calibration parameters under the corresponding operating condition are stored in relation to the current operating condition. When the vehicle re-enters the same or similar operating condition, the stored calibration parameters are called as the initial torque correction amount, reducing the time for repeated calibration and improving control efficiency. At the same time, by establishing a hierarchical anomaly handling mechanism, when the artificial intelligence model output is abnormal, the vehicle lateral deviation continues to increase, or the corner module fails, the current torque correction amount is first stopped and the previous set of effective torque or the original drive torque is restored. If the deviation continues to increase after restoration, speed-limited lateral movement or lateral movement is stopped. This achieves a step-by-step safety guarantee from "restoring effective torque" to "speed-limited lateral movement" and then to "stopping lateral movement," which avoids overreaction in the case of minor anomalies and ensures effective risk control in the case of serious anomalies, significantly improving the stability, reliability, and safety of the system. Therefore, this disclosure solves the problem that existing cornering closed-loop control cannot identify and correct the deviation when the actual movement direction of the wheels deviates from the ideal lateral direction due to factors such as corner zero-position error, mechanical clearance, and tire condition differences during the lateral movement of corner module vehicles, thus causing the vehicle's lateral trajectory to be inaccurate. This improves the accuracy, stability, and adaptability to changes in vehicle condition of corner module vehicles in lateral straight-line driving.

[0136] Although embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this disclosure.

[0137] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Any other corresponding changes and modifications made based on the technical concept of this disclosure should be included within the scope of protection of the claims of this disclosure.

Claims

1. A method for self-calibrating the lateral trajectory of a corner module vehicle, characterized in that, include: After the vehicle enters the lateral movement state, the lateral deviation status and the vehicle's operating status information are acquired in real time. The lateral deviation state and the vehicle's operating state information are input into a trained artificial intelligence model, which then outputs the original torque correction amount for each of the four hub motors. The original torque correction amount is coordinated to make the sum of the torque correction amounts of the four hub motors zero, thus obtaining the actual torque correction amount of each hub motor. Based on the actual torque correction amount of each hub motor, the drive torque of the corresponding hub motor is corrected respectively, and the target torque of each hub motor is generated. The target torque of each hub motor is sent to the corresponding corner module controller to drive the corresponding hub motor to perform differentiated torque output; wherein, during the lateral trajectory self-calibration process, the target rotation angle of the four corner modules remains unchanged.

2. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, The lateral deviation state includes the path deviation of the vehicle's current position relative to the ideal lateral path, the heading angle deviation of the vehicle's current heading angle relative to the initial heading angle, and the yaw rate of the vehicle.

3. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 2, characterized in that, The real-time acquisition of the vehicle's lateral deviation status includes: Obtain the vehicle's initial position and initial heading angle; The ideal lateral direction is determined based on the initial heading angle, and the ideal lateral path is established along the ideal lateral direction starting from the initial position. Based on the vehicle's current position and initial position, the path deviation of the vehicle relative to the ideal lateral path is determined, and the heading angle deviation is determined based on the vehicle's current heading angle and initial heading angle.

4. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 3, characterized in that, The path deviation is the projection distance of the difference between the vehicle's current position and its initial position onto the longitudinal direction of the vehicle body; the heading angle deviation is the difference between the vehicle's current heading angle and its initial heading angle.

5. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, The trained artificial intelligence model is a fully connected neural network.

6. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, The process of coordinating the original torque correction to make the sum of the torque corrections of the four hub motors zero includes: Calculate the average of the four original torque corrections; The original torque correction amount for each hub motor is subtracted from the average value to obtain the actual torque correction amount for the corresponding hub motor.

7. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, The generation of the target torque for each hub motor includes: The actual torque of each hub motor is superimposed with the corresponding actual torque correction amount to obtain the preliminary target torque; The target torque is obtained by limiting the torque range and the torque change rate of the preliminary target torque.

8. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 2, characterized in that, Also includes: After executing the differentiated torque output, the lateral deviation state of the vehicle is reacquired, and the correction effect is evaluated based on the reacquired lateral deviation state. When the correction effect indicates a reduction in lateral deviation, retain the current torque correction values ​​for each group. When the correction effect indicates an increase in lateral deviation, cancel the current torque correction for each group and restore the previous effective torque.

9. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 8, characterized in that, The evaluation of the correction effect based on the re-acquired lateral deviation state includes: The lateral deviation evaluation value is calculated based on the reacquired path deviation, heading angle deviation, and yaw rate. When the evaluation value decreases, the correction effect is determined to be a reduction in horizontal deviation; When the evaluation value increases continuously, the correction effect is determined to be an increase in horizontal deviation.

10. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 2, characterized in that, Also includes: When the vehicle satisfies the following conditions in multiple consecutive control cycles: path deviation is less than the first threshold, heading angle deviation is less than the second threshold, and yaw rate is less than the third threshold, the vehicle is determined to have entered a stable lateral state. The torque correction of each hub motor during the stable lateral movement period is averaged to obtain the four-wheel torque calibration parameters under the corresponding working conditions. Store the four-wheel torque calibration parameters in relation to the current operating conditions.

11. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 10, characterized in that, When the vehicle re-enters a working condition that is the same as or similar to the stored working condition, the stored four-wheel torque calibration parameters are used as the initial torque correction amount.

12. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, The trained artificial intelligence model is trained using an offline supervised learning method. During training, vehicle operating state information collected in simulation is used as training samples, and the optimal torque correction amount obtained by the optimization algorithm is used as the training label. The training samples are obtained by setting different operating conditions in vehicle dynamics simulation, and the training label is the optimal torque correction amount that reduces the lateral deviation state obtained by the optimization algorithm.

13. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 12, characterized in that, The loss function for offline supervised learning is a weighted sum of mean squared error and torque change smoothing term.

14. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 1, characterized in that, Also includes: When the trained artificial intelligence model outputs an abnormality, the lateral deviation of the vehicle continues to increase, or any corner module malfunctions, the current torque correction amount is stopped, and the previous set of effective torque or the original four-wheel drive torque is restored.

15. The self-calibration method for the lateral trajectory of a vehicle using a corner module according to claim 14, characterized in that, If the lateral deviation of the vehicle continues to increase after the previous set of effective torques is restored, speed-limited lateral movement or lateral movement cessation shall be implemented.

Images