Steering compensation method and device for autonomous vehicle, vehicle, medium and product
Patent Information
- Application Number
- CN202611327802.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-31
- Publication Date
- 2026-10-23
AI Technical Summary
[0002]相关技术中,前后轮独立转向车辆的前轴和后轴均可独立转向并产生横摆力矩,装配误差、转角传感器安装误差、转向拉杆间隙、轮胎磨损、执行器回中误差以及长期机械漂移,均可能导致“转向指令为零”与“车轮几何转角为零”不一致,即出现转向零位偏差
将前后轴等效转向零位偏差变换到轴距加权共模和差模空间后,两种模态在直线稳态及命令映射条件下主要对应不同的闭环残差,有助于降低前后轴零位偏差估计之间的耦合。
Smart Images

Figure CN122957326A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle autonomous driving control, specifically to a steering compensation method, device, vehicle, medium, and product for an autonomous vehicle. Background Technology
[0002] In related technologies, the front and rear axles of vehicles with independent steering can both steer independently and generate yaw moments. Assembly errors, angle sensor installation errors, steering tie rod clearance, tire wear, actuator centering errors, and long-term mechanical drift can all lead to a discrepancy between "zero steering command" and "zero wheel geometric angle," resulting in steering zero-position deviation. If the front axle zero-position deviation and rear axle zero-position deviation are directly expressed as two physical quantities, they often couple with each other in the straight-line driving closed loop: front and rear axle offset in the same direction mainly manifests as steady-state heading residuals related to the overall vehicle attitude, while front and rear axle offset in opposite directions mainly manifests as equivalent curvature residuals. These residuals are absorbed into the steering command difference between the front and rear axles by the heading closed loop of the lateral controller, leaving almost no obvious trace in the vehicle's lateral and heading errors, leading to a decrease in the accuracy of vehicle control.
[0003] Therefore, how to perform modal transformation and decoupling identification on the equivalent steering zero-position deviation of the front and rear axles for vehicles with independent front and rear wheel steering, and generate front and rear axle compensation amounts accordingly to improve the control accuracy of autonomous vehicles, has become an urgent problem to be solved. Summary of the Invention
[0004] In this embodiment, the equivalent steering zero-position deviation of the coupled front and rear axles is transformed into wheelbase-weighted common mode and differential mode, which are then identified online using different closed-loop residuals. Finally, they are restored to front axle compensation and rear axle compensation and injected into steering commands, thereby improving the control accuracy of vehicles with independent front and rear wheel steering.
[0005] In a first aspect, this application provides a steering compensation method for an autonomous vehicle. The method includes: acquiring, in real time, the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated while it is driving stably along a straight path; calculating the difference between the front axle steering command and the rear axle steering command to obtain a steering command difference; using the effective heading angle error, updating the common-mode zero-position deviation estimate of the vehicle to be compensated periodically, wherein the common-mode zero-position deviation estimate is used to characterize the same-direction offset of the front and rear axles, and the effective heading angle error is obtained by limiting the heading angle error through a dead zone threshold and a saturation threshold; using the steering command difference that satisfies steady-state conditions, updating the differential-mode zero-position deviation estimate of the vehicle to be compensated periodically, wherein the differential-mode zero-position deviation estimate is used to characterize the opposite-direction offset of the front and rear axles; and restoring the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, so as to perform steering compensation on the front and rear axles of the vehicle to be compensated respectively.
[0006] In conjunction with the first aspect, in one embodiment of this application, the step of using the effective heading angle error to periodically update the common-mode zero-position deviation estimate of the vehicle to be compensated includes: The common-mode zero-point deviation estimate is calculated using the following formula:
[0007]
[0008] in, Indicates update cycle The corresponding common mode update amount, This represents the effective heading angle error after being limited by dead zone threshold and saturation threshold. This represents the common-mode learning gain, used to adjust the convergence speed. This represents the common-mode update weight generated based on the lateral position error, and this represents the common-mode zero-position deviation from the previous update cycle. This indicates the common-mode zero-point deviation in the current update cycle.
[0009] In conjunction with the first aspect, in one embodiment of this application, the step of using the steering command difference that satisfies the steady-state condition to periodically update the differential mode zero-position deviation estimate of the vehicle to be compensated includes: The differential mode zero-point deviation estimate is calculated using the following formula:
[0010] in, This represents the differential learning gain, used to adjust the convergence speed. This indicates the differential mode zero-point deviation from the previous update cycle. This indicates the differential modulus zero-point deviation in the current update cycle. Indicates the update cycle. This represents the steering command difference that satisfies the steady-state condition.
[0011] In conjunction with the first aspect, in one embodiment of this application, the step of restoring the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount includes: The front axle compensation and rear axle compensation are calculated using the following formula:
[0012] in, Indicates the front axle compensation amount. Indicates the rear axle compensation amount. Indicates wheelbase. This indicates the distance from the front axle to the center of gravity of the vehicle to be compensated. This indicates the distance from the rear axle to the center of gravity of the vehicle to be compensated.
[0013] In conjunction with the first aspect, in one embodiment of this application, weight value It is selected using the following formula:
[0014] in, The lateral position error of the vehicle to be compensated is obtained by projecting the vehicle's position relative to the matching point of the reference path onto the local normal direction, representing the deviation between the vehicle and the lane centerline. This represents the preset full confidence threshold for lateral error. This indicates the preset lateral error stopping learning threshold.
[0015] In conjunction with the first aspect, in one embodiment of this application, after restoring the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, the method further includes: converting the front axle compensation amount... With the front axle steering command The commands are summed to obtain the independent steering control command for the front axle; the rear axle compensation amount is then calculated. Rear axle steering command The commands are summed to obtain the independent steering control command for the rear axle; the independent steering control command for the front axle and the independent steering control command for the rear axle are respectively limited to a preset steering range, and the limited independent steering control command for the front axle and the limited independent steering control command for the rear axle are sent to the vehicle to be compensated for execution.
[0016] Secondly, this application provides a steering compensation device for an autonomous vehicle, the device comprising: The parameter acquisition module is configured to acquire the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated in real time while the vehicle to be compensated is driving stably along a straight path. The calculation module is configured to calculate the difference between the front axle steering command and the rear axle steering command to obtain the steering command difference; The deviation compensation module is configured to use the effective heading angle error to update the common mode zero-position deviation estimate of the vehicle to be compensated on a cycle basis. The common mode zero-position deviation estimate is used to characterize the same-direction offset of the front axle and the rear axle. The effective heading angle error is obtained by limiting the heading angle error through a dead zone threshold and a saturation threshold. The deviation compensation module is configured to update the differential zero-position deviation estimate of the vehicle to be compensated on a cycle using the steering command difference that satisfies the steady-state condition, wherein the differential zero-position deviation estimate is used to characterize the reverse bias of the front axle and the rear axle. The deviation compensation module is configured to restore the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, so as to perform steering compensation on the front axle and the rear axle of the vehicle to be compensated, respectively.
[0017] In conjunction with the second aspect, in one embodiment of this application, the deviation compensation module is configured as follows: The common-mode zero-point deviation estimate is calculated using the following formula:
[0018]
[0019] in, Indicates update cycle The corresponding common mode update amount, This represents the effective heading angle error after being limited by dead zone threshold and saturation threshold. This represents the common-mode learning gain, used to adjust the convergence speed. This represents the common-mode update weight generated based on the lateral position error, and this represents the common-mode zero-position deviation from the previous update cycle. This indicates the common-mode zero-point deviation in the current update cycle.
[0020] In conjunction with the second aspect, in one embodiment of this application, the deviation compensation module is configured as follows: The differential mode zero-point deviation estimate is calculated using the following formula:
[0021] in, This represents the differential learning gain, used to adjust the convergence speed. This indicates the differential mode zero-point deviation from the previous update cycle. This indicates the differential modulus zero-point deviation in the current update cycle. Indicates the update cycle. This represents the steering command difference that satisfies the steady-state condition.
[0022] In conjunction with the second aspect, in one embodiment of this application, the deviation compensation module is configured as follows: The front axle compensation and rear axle compensation are calculated using the following formula:
[0023] in, Indicates the front axle compensation amount. Indicates the rear axle compensation amount. Indicates wheelbase. This indicates the distance from the front axle to the center of gravity of the vehicle to be compensated. This indicates the distance from the rear axle to the center of gravity of the vehicle to be compensated.
[0024] In conjunction with the second aspect, in one embodiment of this application, weight value It is selected using the following formula:
[0025] in, The lateral position error of the vehicle to be compensated is obtained by projecting the vehicle's position relative to the matching point of the reference path onto the local normal direction, representing the deviation between the vehicle and the lane centerline. This represents the preset full confidence threshold for lateral error. This indicates the preset lateral error stopping learning threshold.
[0026] In conjunction with the second aspect, in one embodiment of this application, the deviation compensation module is configured to: increase the front axle compensation amount With the front axle steering command The commands are summed to obtain the independent steering control command for the front axle; the rear axle compensation amount is then calculated. Rear axle steering command The commands are summed to obtain the independent steering control command for the rear axle; the independent steering control command for the front axle and the independent steering control command for the rear axle are respectively limited to a preset steering range, and the limited independent steering control command for the front axle and the limited independent steering control command for the rear axle are sent to the vehicle to be compensated for execution.
[0027] Thirdly, this application provides an autonomous driving vehicle, including: a vehicle controller, a memory, and a bus; the vehicle controller is communicatively connected to the memory via the bus, the memory stores a computer program, and when the computer program is executed by the processor of the vehicle controller, it implements the method described in any of the above embodiments.
[0028] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed, can implement the method as described in any embodiment of the first aspect.
[0029] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed, can implement the method as described in any embodiment of the first aspect.
[0030] Therefore, the beneficial effects of the embodiments of this application are as follows: After transforming the equivalent steering zero-position deviation of the front and rear axles to the wheelbase-weighted common-mode and differential-mode spaces, the two modes mainly correspond to different closed-loop residuals under linear steady-state and command mapping conditions, which helps to reduce the coupling between the estimation of the zero-position deviation of the front and rear axles.
[0031] In the independent steering closed-loop control of the front and rear wheels, the differential mode zero bias is balanced by the controller through the original steering command difference between the front and rear axles. Its effect is not necessarily directly manifested as lateral error or heading error. This application uses the original steering command difference between the front and rear axles after closed-loop balancing as the differential mode observation, which can identify the differential mode zero bias that is difficult to separate by relying solely on path error. When the straight-line command difference is zero and there are no other continuous biases, after the differential mode compensation converges, the steady-state estimate of the original steering command difference between the front and rear axles can approach zero; when there is a non-zero nominal baseline, it converges to the corresponding nominal baseline.
[0032] This application utilizes the existing path error, reference curvature, and controller commands during normal vehicle operation to continuously track slow-changing equivalent steering deviations caused by tire wear, load changes, or actuator drift, without the need for dedicated zero-position calibration fixtures.
[0033] By employing multiple constraints, including straight line conditions, velocity conditions, error anomaly protection, stable duration, minimum average formation time, cumulative increment limit for this operation, total compensation limit, and slope injection, transient, noise, and abnormal operating condition mislearning are suppressed, and steering step jumps caused by compensation abrupt changes are avoided.
[0034] Save the front and rear axle compensation values that meet the validity requirements. Use these values to initialize the mode during subsequent vehicle operation to shorten the convergence time. Attached Figure Description
[0035] Figure 1This is one of the flowcharts illustrating a steering compensation method according to an embodiment of this application; Figure 2 This is a second flowchart illustrating a steering compensation method according to an embodiment of this application; Figure 3 This is a third flowchart illustrating a steering compensation method according to an embodiment of this application; Figure 4 This is a schematic diagram of a steering compensation device according to an embodiment of this application; Figure 5 This is a schematic diagram of an autonomous vehicle shown in an embodiment of this application. Detailed Implementation
[0036] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the protection scope of this application.
[0037] The embodiments of this application can be applied to scenarios where, without adding dedicated zero-position calibration equipment, the steering zero-position deviation of a vehicle with independent front and rear wheel steering can be identified and compensated online using the lateral error, heading error, reference line curvature, and the original steering commands of the front and rear axles output by the lateral controller generated when the autonomous vehicle travels along a reference path.
[0038] To address the aforementioned technical issues, this application does not directly identify the coupled front axle compensation and rear axle compensation simultaneously. Instead, it transforms the equivalent steering zero-position deviation of the front and rear axles into a wheelbase-weighted common-mode zero-position deviation representing the unidirectional offset of the front and rear axles and a differential-mode zero-position deviation representing the inverse offset of the front and rear axles. The common-mode is updated by the effective heading angle error under straight-line steady state, while the differential-mode is updated by the steady-state estimate of the original steering command difference between the front and rear axles formed when the absolute straight line is stable. The two modes are then restored to front and rear axle compensation values, which are then injected into the chassis steering command after amplitude limiting and slope smoothing, and saved as reusable zero-position compensation parameters for subsequent operation.
[0039] Therefore, all the technical problems solved by this application include: how to decompose the mutually coupled and difficult-to-distinguish zero-position deviations of the front and rear axles into wheelbase-weighted common modes and differential modes that can be learned online and separately for vehicles with independent front and rear wheel steering; how to learn different modes using heading error and the difference in front and rear axle steering commands within the controller under straight-line, steady-state, and low-noise operating conditions; and how to stably and reliably convert the learning results into front and rear axle steering command compensation to form a closed loop that can be implemented in engineering.
[0040] It should be noted that the vehicle to be compensated as referred to in this application can be an autonomous vehicle, which includes unmanned transport vehicles whose front and rear axles can receive steering commands separately, four-wheel steering mobile platforms, independent steering transport vehicles, and vehicles with the same steering topology. This application does not limit the type of autonomous vehicle.
[0041] The threshold, gain, and limiting parameters in this application can be calibrated according to vehicle and site conditions. Unless otherwise specified, All variables adopt a unified steering command coordinate system for the target vehicle, which can represent percentage steering commands, equivalent wheel angles, or dimensionless command values after actuator calibration and mapping. The positive direction of each variable is consistent with the vehicle's existing steering control interface. The zero-position deviation referred to in this application is the equivalent steering zero-position deviation characterized by the closed-loop residual, and is not limited to a single mechanical source.
[0042] It is understood that the executing entity of this application is the control module in an autonomous vehicle with independent control of the front and rear wheels, or it can be the vehicle controller. Those skilled in the art can adjust the executing entity according to the actual situation of the vehicle, and this application does not limit it in this regard.
[0043] The following example, using a vehicle controller as the executing entity, illustrates a steering compensation method for an autonomous vehicle provided by some embodiments of this application.
[0044] To explain the relationship between front axle compensation, rear axle compensation, common mode zero deviation, and differential mode zero deviation, the specific process of determining this relationship based on the physical characteristics of the automatically controlled vehicle will be described in detail below.
[0045] The core idea of this application is to avoid directly identifying the mutually coupled front axle compensation amounts simultaneously. and rear axle compensation Instead, the equivalent steering zero-point deviation of the front and rear axles is represented as two reversible modes, β and c. β represents the wheelbase-weighted common-mode zero-point deviation estimate, mainly corresponding to the equivalent side-bias residual formed by the front and rear axles being offset in the same direction; c represents the differential-mode zero-point deviation estimate, mainly corresponding to the equivalent curvature residual formed by the front and rear axles being offset in opposite directions. The observability of the two modes is based on linear, steady-state, control assignment, and command mapping conditions.
[0046] The above modal division is derived from the kinematics of a vehicle with independent front and rear wheel steering, assuming the distance from the front axle to the center of gravity is... The distance from the rear axle to the center of gravity is equivalent wheelbase The geometric rotation angles of the front and rear axles are respectively The equivalent sideslip angle of the vehicle's velocity direction relative to the vehicle's longitudinal axis is: Let the longitudinal and lateral components of the center-of-gravity velocity in the vehicle coordinate system be respectively... , The yaw rate is .
[0047] In the single-track model without sideslip, it is assumed that each wheel rolls purely along its own wheel plane without lateral slippage. Based on this, the constraint relationship between the geometric rotation angles of the front and rear axles and the vehicle's center of gravity velocity and yaw rate is established.
[0048]
[0049] Subtracting equation (2) from equation (1) yields ; by two formulas and We can get the weighted sum. Therefore, we have:
[0050] in, The actual curvature of the vehicle's motion (characterizing the degree of curvature of the trajectory drawn by the vehicle's center of mass on the ground).
[0051] Under near-straight-line conditions with small turning angles, negligible tire sideslip, and vehicle speed within the calibration learning range, equations (3) and (4) can be approximated as:
[0052] Therefore, when the zero-position deviation manifests as a fixed offset in the same direction between the front and rear axles, this deviation is mainly mediated by the wheelbase weighted combination term. It affects the vehicle's equivalent sideslip angle, thereby changing the vehicle's lateral kinematics characteristics. In the straight-line driving closed loop, this manifests as the steady-state heading residual (i.e., heading angle error) required to maintain path tracking.
[0053] When the zero-position deviation manifests as a fixed offset in opposite directions for the front and rear axles, this deviation is mainly manifested through the difference in steering angle between the front and rear axles. The equivalent curvature of the vehicle's motion is affected, which in straight-line driving means that the controller needs to actively balance the steering commands of the front and rear axles to maintain directional stability.
[0054] Therefore, this application defines the following in modal space:
[0055] Equation (7) corresponds to the weighted unidirectional compensation component in the equivalent lateral slip shown in equation (5), and equation (8) corresponds to the front-to-back axis relative compensation component in the equivalent curvature shown in equation (6). Inverting the above binary linear relationship yields the front-to-back axis compensation quantities, which are then reduced from two modes:
[0056] in, This indicates the upper limit of the absolute value of the zero-position compensation amount for the front and rear axles. and These represent the front axle compensation amount and the rear axle compensation amount, respectively; the final range of the chassis steering command execution is determined by... and Further restrictions apply.
[0057] Therefore, it can be seen that the common-mode zero-position deviation β acts on the front and rear axles with the same magnitude; the differential-mode zero-position deviation c is calculated according to the normalized half-axis distance. / L and / L is distributed to the front and rear axles in opposite directions. This transformation is reversible under small steering angle and low sideslip kinematic model. The common mode and differential mode mainly correspond to the equivalent sideslip channel and equivalent curvature channel, respectively, which helps to reduce the coupling of zero-position deviation estimation between the front and rear axles. However, the two channels may still affect each other under dynamic conditions or when the compensation amount triggers the amplitude limit.
[0058] Accordingly, the initial value of the front axle compensation is obtained during the initialization phase. Initial value of rear axle compensation Then, the internal modal initial values are transformed using the following formula:
[0059] in, This represents the initial value of the common-mode zero-point deviation. This represents the initial value of the differential mode zero-position deviation.
[0060] In other words, during stable straight-line driving, the front and rear axle compensation values are learned in real time. First, it's determined whether the vehicle is in a suitable driving state for identification, including that the zero-position compensation function is enabled, the vehicle is driving stably along an approximately straight reference path, the vehicle speed is within the allowable learning range, and the lateral and heading errors do not exceed preset limits. Once these conditions are met, the front and rear axle offsets in the same direction and opposite directions are updated respectively: the same-direction offset is mainly corrected based on the heading angle error during straight-line driving, while the opposite-direction offset is mainly corrected based on the stable average value of the original steering command difference between the front and rear axles over a period of time during straight-line driving. During the update process, limits are set on the error amplitude, duration, single change, and cumulative change to avoid misinterpreting cornering, control transients, or positioning noise as mechanical zero-position deviation. The updated common-mode and differential-mode zero-position deviations are then converted into front and rear axle compensation values, which are then superimposed onto the chassis steering commands after amplitude limiting and smoothing.
[0061] like Figure 1 As shown, the steering compensation method for autonomous vehicles provided in this application includes the following steps: S110: Obtain the vehicle's driving path and front and rear axle steering commands; S120: Determine if the zero-position deviation learning function is enabled, and check if the vehicle speed is within the calibration learning range, if the reference path is reliable, and if the vehicle is in a stable tracking state; if the conditions are not met, execute S130 to reset the effective observation state and retain the existing compensation. If the conditions are met, S140: Perform dead zone and saturation processing on the heading angle error, and update the common-mode zero-position deviation estimate periodically; S150: Update the differential-mode zero-position deviation estimate periodically based on the steering command difference estimate that meets the steady-state conditions; S160: Restore the two modes to front and rear axle compensation amounts; S170: After limiting and slope smoothing the compensation amounts, inject the front and rear axle steering commands, and save the compensation parameters that meet the validity conditions.
[0062] Based on the above derivation process, the reasoning process of converting common-mode zero-position deviation and differential-mode zero-position deviation into front axle compensation and rear axle compensation is clarified.
[0063] As a specific embodiment of this application, the system implementing this application functionally includes: an initialization unit, a vehicle status acquisition unit, a working condition determination unit, a lateral control unit, a steering zero deviation identification unit, a compensation injection unit, and a compensation result storage unit.
[0064] The initialization unit is used to obtain the initial values of the function enable state, learning gain, and zero-position deviation of the front and rear axles of the vehicle, and transforms them into the initial values of the wheelbase weighted common mode and differential mode according to equations (11) and (12). The working condition determination unit is used to provide the lateral error. Heading angle error Reference line curvature Reference path status, vehicle operating status, and vehicle speed The lateral control unit first calculates the original steering commands for the front and rear axles without zero-bias compensation. , The zero-bias identification unit uses the above information to update the common mode. and differential mode .
[0065] The compensation injection unit restores the wheelbase weighted common mode and differential mode to the front and rear axle compensation amounts according to equations (9) and (10), and after amplitude limiting and slope smoothing, superimposes them onto the original steering commands of the front and rear axles and sends them to the actuators. The compensation result storage unit is used to store the front and rear axle compensation amounts after validity confirmation, and uses them as the initial value of zero position deviation during subsequent vehicle operation.
[0066] As a specific explanation of the modal decomposition of zero offset between the front and rear axles in this application, the aforementioned equations (3) to (6) give the kinematic basis of a vehicle with independent steering of the front and rear wheels. Combining the definition of path tracking error, it can be further written as follows under small-angle straight-line conditions:
[0067] As can be seen from equation (13), the common-mode zero-position deviation changes the equivalent sideslip angle. ; When the lateral motion reaches a steady state along a reliable straight line, the rate of change of the lateral error Approaching zero, the common-mode residual can be found in the heading angle error. An approximate steady-state observation is formed. From equation (14), it can be seen that the differential mode bias changes the equivalent curvature. Under absolute straight-line conditions, this offset is mainly achieved through closed-loop trimming based on the difference in initial steering commands between the front and rear axles. Therefore, this application updates the common-mode and differential-mode zero-position deviations using the effective heading angle error and the estimated value of the difference in initial steering commands between the front and rear axles that meets the steady-state conditions, respectively.
[0068] Therefore, the modal decomposition in this application makes the common mode and differential mode correspond mainly to different closed-loop observations under the linear steady-state and command mapping conditions, which helps to reduce the coupling between the front and rear axle zero-position deviation estimates. The differential mode learning utilizes the original steering command difference inside the controller to identify the equivalent steering deviation that has been balanced in the closed loop but has not been significantly retained in the path error; the two modes are restored to the front and rear axle compensation amounts in the output stage, which facilitates entry into the steering execution link and parameter configuration.
[0069] The steering compensation method for autonomous vehicles in the embodiments of this application will be described in detail below, such as... Figure 2 As shown, the method specifically includes: S210: While the vehicle to be compensated is traveling stably along a straight path, acquire the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated in real time. S230: Calculate the difference between the front axle steering command and the rear axle steering command to obtain the steering command difference.
[0070] As a specific embodiment of this application, before executing S210-S230, the identification unit first checks the basic zero-position deviation learning conditions in each control cycle:
[0071] in, Indicates whether the zero-point deviation learning function is enabled; the indicator function is in Take 1 at time. and These represent the minimum and maximum learning speeds, respectively. The reliable straight road segment marker is determined by transient conditions such as reference curvature, look-ahead path continuity, curvature change, stable tracking status, and path switching.
[0072] If the basic learning conditions are not met, the effective observation time for both common-mode zero-position deviation and differential-mode zero-position deviation is reset, and the learning status is not updated in this cycle. This process prevents parking, low-speed movement, curves, path changes, and other obviously non-linear operating conditions from being incorrectly learned as mechanical zero-position deviation.
[0073] After determining that the current road segment and driving state meet the learning conditions for zero deviation, the actual heading angle, reference heading angle, and front axle steering command of the vehicle to be compensated are acquired in real time. and rear axle steering command And calculate the heading angle error. Steering command difference .
[0074] After achieving the shortest effective observation time, a dead zone and saturation are applied to the heading angle error:
[0075] in, Effective heading angle error after being limited by dead zone threshold and saturation threshold Indicates the direction of the heading angle error. This indicates that the heading angle error has stopped updating and the saturation threshold has been reached. The dead zone threshold represents the heading angle error. Used to filter out small heading error noise Used to suppress the effects of abnormally large heading errors.
[0076] As one implementation method for obtaining the heading angle error in this application, the actual heading angle and the reference heading angle are obtained, and the difference between the actual heading angle and the reference heading angle is calculated to obtain the heading angle error.
[0077] Another implementation method, as an observation mechanism for common-mode residuals, involves the lateral error rate of change when the lateral motion reaches a steady state along a reliable straight line with a small rotation angle. Approaching zero, while the lateral position error It can itself remain a non-zero constant value. If there is uncompensated common-mode residual... This residual will form an approximate steady-state residual in the heading angle error through the equivalent sideslip angle. Therefore, the heading angle error can be used as an observation of the common mode residual.
[0078] In the presence of common mode residuals In this case, the residual will act as a constant disturbance, affecting the vehicle steering control through the equivalent sideslip angle. Due to the existence of this disturbance, the control algorithm must generate an approximately constant heading angle error to maintain the vehicle traveling along the target straight line. This is done to counteract the disturbance and maintain straight-line driving. According to the principles of kinematic and dynamic control, the heading angle error in this steady state... With common mode residual There exists an approximate linear relationship as shown in equation (17):
[0079] in, This is a proportionality coefficient characterizing the approximate mapping relationship between the common-mode residual and the heading angle error under steady state. This application uses this relationship to demonstrate that the heading angle error can be used as an observation of the common-mode residual without needing to determine... The specific value of the value and its impact on the closed-loop update scale are comprehensively reflected by the calibration of the common-mode learning gain.
[0080] Therefore, the heading angle error detected under straight steady state Can be used as common mode residual Based on the effective observations, the common-mode learning process in step S140 actually utilizes these observations and, in conjunction with a preset learning gain, iteratively identifies and updates the residuals. After closed-loop learning and updating, the control system in step S160 restores the learned common-mode zero-position deviation to the front and rear axle compensation amounts, ultimately realizing the common-mode residual... The convergence approaches zero.
[0081] S240 uses the effective heading angle error to update the common mode zero-position deviation estimate of the vehicle to be compensated on a cycle-by-cycle basis.
[0082] Understandably, the common-mode zero-position deviation estimate is used to characterize the same-direction offset of the front and rear axles, and the effective heading angle error is obtained by limiting the heading angle error through dead zone threshold and saturation threshold.
[0083] In one embodiment of this application, the common-mode zero-point deviation is calculated using the following formula:
[0084] in, Indicates update cycle The corresponding common mode update amount, This represents the effective heading angle error after being limited by dead zone threshold and saturation threshold. This represents the common-mode learning gain, used to adjust the convergence speed. Represents the common-mode update weights generated based on the lateral position error. This represents the estimated common-mode null offset value from the previous update period. This represents the estimated common-mode null offset value for the current update period. These represent the lower and upper limits of the common-mode zero-point deviation for this operation, respectively. This indicates the cumulative increment limit relative to the starting baseline.
[0085] Furthermore, weight value It is selected using the following formula:
[0086] in, The lateral position error of the vehicle to be compensated is obtained by projecting the vehicle's position relative to the matching point of the reference path onto the local normal direction, representing the deviation between the vehicle and the lane centerline. This represents the preset full confidence threshold for lateral error. This indicates the preset lateral error stop learning threshold.
[0087] In other words, the larger the lateral error, the more likely it is to be in a collision, disturbance recovery, or control transient state, and the lower the weight of the learned signal will be. Excessive lateral position error indicates abnormal tracking; learning is only performed when the lateral position error is relatively small.
[0088] S250, using the steering command difference that satisfies the steady-state condition, update the differential mode zero-position deviation estimate of the vehicle to be compensated on a cycle-by-cycle basis.
[0089] It should be noted that the differential zero-position deviation estimate is used to characterize the reverse bias of the front and rear axles. The steady-state conditions include the processing steps of formulas (23), (24) and (25) described later.
[0090] Specifically, differential mode zero-position deviation The core observation is not the path error itself, but the difference in the original steering commands between the front and rear axles:
[0091] in and This is the original controller output before zero-bias compensation is injected. When the vehicle is on an absolutely straight line and the path error is stable, if the controller requires a non-zero output for a long period of time... This is necessary to maintain the heading, indicating that there is an uncompensated differential mode residual between the front and rear axles. Its kinematic relationship can be directly obtained from equation (14): and In a steady-state straight line, the effective front and rear axle rotation angle difference should approach zero. If the current uncompensated differential residual is denoted as... Then we have:
[0092] When there is a proportional relationship between the steering command and the actual wheel angle, the differential learning gain... Based on the steering command mapping relationship and closed-loop response calibration of the target vehicle, the steering command difference estimate that meets the steady-state condition is used to update the differential mode zero position deviation on a cycle-by-cycle basis.
[0093] Differential mode zero-position deviation uses a more stringent linearity condition than common mode zero-position deviation:
[0094] in, This represents the curvature threshold of a straight line.
[0095] Simultaneously, error stability is required:
[0096] If the above conditions are not met, reset the effective observation status of the differential mode zero-position deviation.
[0097] After meeting the differential modulus zero-position deviation observation conditions, first check the command difference change rate:
[0098] in, This indicates the difference in the currently collected steering command. This indicates the difference in steering commands within the previous learning cycle. This represents the threshold for the rate of change of steering command difference.
[0099] If the rate of change of the command difference exceeds the threshold, it is determined that the command difference between the front and rear axes is still in a transient balancing state, the effective observation state of the differential mode is reset, and accumulation starts again from a new stable sample.
[0100] The instruction difference remains stable and continuous. After (i.e., instruction difference stabilization waiting time), for The steady-state average value can be obtained using a first-order low-pass filter.
[0101] in, This indicates the filtering time parameter used when forming the steady-state average value of the command difference.
[0102] As another embodiment of first-order low-pass filtering, the first-order low-pass filter can be replaced by any low-pass / steady-state estimator such as sliding window mean, median filtering, or recursive least squares. The filtering time parameter, window length, and minimum formation time are all adjustable implementation parameters.
[0103] In one embodiment of this application, the differential mode zero-position deviation is calculated using the following formula:
[0104] in, This represents the differential learning gain, used to adjust the convergence speed. This represents the estimated value of the differential modulus zero-point deviation from the previous update period. This represents the estimated value of the differential modulus zero-point deviation for the current update period. Indicates the update cycle. This represents the difference in steering commands that satisfy steady-state conditions.
[0105] Specifically, the steady-state average formation time is less than When the minimum formation time of the steady-state average value of the instruction difference is reached, the differential mode is not updated. After sufficient averaging, it is updated according to the following formula, and the cumulative increment limit of this run is applied:
[0106] in, This is the initial value of the differential mode at the start of this run. For differential mode learning gain, This indicates the cumulative increment limit of the differential mode relative to the starting baseline during this operation.
[0107] This update direction forms a closed-loop negative feedback: when the differential zero-position deviation causes the controller to continuously output the front and rear axle steering command difference in linear steady state, equation (27) gradually updates c, which is then restored by equations (9) and (10) and injected with the front and rear axle steering commands, causing the steering command difference to gradually decrease. When the nominal linear command difference is zero and there are no other continuous biases, the estimated value converges to zero; when there is a non-zero nominal baseline, it converges to the corresponding nominal baseline.
[0108] In another embodiment of the differential mode observation, besides using In addition to the steady-state average value, the differential residual can also be represented by the yaw rate residual under linear steady state, the integral term of the front and rear axle command difference, or the steady-state control quantity used by the controller for trimming the heading. Their common feature is that they are internal control quantities that are trimmed in real time by the closed loop, do not leave obvious residues in the path error, but are non-zero for a long time.
[0109] S260 restores the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, so that the steering compensation of the front axle and the rear axle of the vehicle to be compensated can be performed separately.
[0110] In one embodiment of this application, the front axle compensation amount and the rear axle compensation amount are calculated using the following formula:
[0111] in, Indicates the front axle compensation amount. Indicates the rear axle compensation amount. Indicates wheelbase. This indicates the distance from the front axle to the center of gravity of the vehicle to be compensated. This indicates the distance from the rear axle to the center of gravity of the vehicle to be compensated.
[0112] After obtaining the front axle and rear axle compensation values from S260, these values need to be converted into corresponding independent steering control commands. Specifically: front axle compensation Front axle steering command The commands are summed to obtain independent steering control commands for the front axle, and the compensation amount for the rear axle is also calculated. Rear axle steering command The independent steering control commands for the rear axle are summed to obtain the independent steering control commands for the front and rear axles. Then, the independent steering control commands for the front and rear axles are limited to preset steering ranges, and the limited independent steering control commands for the front and rear axles are sent to the vehicle to be compensated for execution.
[0113] In other words, autonomous vehicle control is executed during the chassis command generation stage to achieve independent steering of the front and rear axles. The online learning results are smoothed and then directly enter the vehicle steering execution link.
[0114]
[0115] in, This indicates an independent steering control command for the front axle. This indicates an independent steering control command for the rear axle. This indicates the lower limit of the chassis steering command. Indicates the upper limit of chassis steering commands. This represents the estimated front axle compensation amount (i.e., the front axle compensation amount obtained through the aforementioned process). This represents the estimated value of the rear axle compensation (i.e., the rear axle compensation obtained through the aforementioned process). This indicates a front axle steering command. This indicates a rear axle steering command.
[0116] As a specific implementation of differential learning, such as Figure 3 As shown: S310 acquires the steering command difference between the front and rear axles when the vehicle is in an absolutely straight line and the error is stable; S320 determines whether the rate of change of the command difference is less than the threshold. If not, S330 is executed to reset the differential mode observation state and returns to S310 after stabilization. If yes, S340 is executed to wait for the stabilization duration to reach the threshold; then S350 is executed to form a steady-state estimate of the steering command difference, and S360 determines whether its formation time has reached the threshold. If not, it continues to form a steady-state estimate. If it has, S370 is executed to update the differential mode zero-position deviation and apply a cumulative increment limit.
[0117] When the common-mode and differential-mode learning results meet the requirements of numerical validity, amplitude limiting constraints, and stability, the corresponding front and rear axle compensation values can be saved as the vehicle's zero-position compensation parameters. During subsequent vehicle operation, the saved compensation values are used as initial values for the front and rear axle zero-position deviations, which are then transformed into initial values for wheelbase-weighted common-mode and differential-mode. This process eliminates the need for the vehicle to relearn from zero after each startup and allows online learning to continue tracking slow drifts during long-term operation.
[0118] It is understandable that modal decomposition can be extended to three-axis and above multi-axis independent steering platforms. For three-axis and above independent steering platforms, full-rank independent modal transformations can be constructed based on the geometric relationships and control allocation relationships of each axis to achieve the identification and compensation of each independent mode.
[0119] In one embodiment of this application, the following abnormal situations and security measures are also included: This application includes the following exception and safety handling: Zero value rollback is used when initial values are missing or unreadable; the learning speed is not within […]. , When the path does not belong to a reliable straight line or the function is turned off, reset the valid observation status and stop updating; stop updating when the lateral error or heading angle error exceeds the stop learning threshold; differential mode learning only occurs when... The process is initiated when the rate of change of the instruction difference meets the requirements and the steady-state estimation formation time reaches the threshold; the cumulative changes of the common mode and the differential mode relative to the starting baseline are respectively affected by... and Limitations; the amount of compensation between the front and rear axles after restoration is limited to... Ultimately, chassis commands are limited to The actual injection compensation amount is based on the slope rate. Changes; only retain compensation results that meet the requirements of numerical validity, amplitude limiting, and stability.
[0120] Therefore, the innovation of this application's embodiment lies in: transforming the mutually coupled front and rear axle zero-position deviations from the physical axle space to the front and rear axle wheelbase weighted common mode and front and rear axle differential mode spaces for identification, and then restoring them to the front and rear axle compensation outputs according to the front and rear half-wheelbase weights, so that different modes correspond to different observable residuals. Common mode learning and differential mode learning use different observation sources: the common mode uses the heading error under straight steady state combined with the lateral error confidence level, and the differential mode uses the steady-state average value of the original front and rear axle steering command difference of the controller under absolute straight steady state. Differential mode learning explicitly requires absolute straightness, stable command difference rate of change, stable duration, minimum formation time of the average value, and limit of the cumulative increment of the current operation, so as to both identify the differential mode residuals of the closed-loop trim and avoid mislearning the legal cornering differential steering as zero-position deviation. The online learning results are limited, slope smoothed, and saved before entering the front and rear axle steering command link, forming a continuously reusable compensation closed loop.
[0121] The above describes a specific embodiment of a steering compensation method for an autonomous vehicle. The following describes a steering compensation device for an autonomous vehicle.
[0122] like Figure 4 As shown, some embodiments of this application provide a steering compensation device for an autonomous vehicle, the device including: a parameter acquisition module 410, a calculation module 420 and a deviation compensation module 430.
[0123] The parameter acquisition module is configured to acquire the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated in real time while the vehicle is traveling stably along a straight path. The calculation module is configured to calculate the difference between the front axle steering command and the rear axle steering command to obtain the steering command difference. The deviation compensation module is configured to use the effective heading angle error to update the common-mode zero-position deviation estimate of the vehicle to be compensated periodically, wherein the common-mode zero-position deviation estimate is used to characterize the unidirectional offset of the front and rear axles, and the effective heading angle error is... The deviation compensation module is configured to update the differential zero-position deviation estimate of the vehicle to be compensated periodically using the steering command difference that satisfies the steady-state condition, wherein the differential zero-position deviation estimate is used to characterize the reverse offset of the front axle and the rear axle; the deviation compensation module is configured to restore the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, so as to perform steering compensation on the front axle and the rear axle of the vehicle to be compensated respectively.
[0124] In the embodiments of this application, Figure 4 The module shown can achieve Figure 1 , Figure 2 and Figure 3 Each process in the method embodiment. Figure 4 The operations and / or functions of each module in the document are respectively designed to achieve... Figure 1 , Figure 2 and Figure 3 The corresponding processes are described in the method embodiments above. For details, please refer to the descriptions in the above method embodiments; to avoid repetition, detailed descriptions are appropriately omitted here.
[0125] like Figure 5 As shown, this application provides an electronic device 500, including: a vehicle controller 510, a memory 520 and a bus 530. The vehicle controller is connected to the memory via the bus. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the method as described in any of the above embodiments. For details, please refer to the description in the above method embodiments. To avoid repetition, detailed descriptions are appropriately omitted here.
[0126] The bus is used to enable direct communication between these components. In this embodiment, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), an On-Premises Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or any conventional processor.
[0127] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory stores computer-readable instructions, which, when executed by the processor, can perform the methods described in the above embodiments.
[0128] Understandable. Figure 5 The structure shown is for illustrative purposes only and may include structures larger than [other structures]. Figure 5The more or fewer components shown, or having the same Figure 5 The different configurations shown. Figure 5 The components shown can be implemented using hardware, software, or a combination thereof.
[0129] This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a vehicle controller, it implements any of the methods described in all the above embodiments. For details, please refer to the descriptions in the above method embodiments. To avoid repetition, detailed descriptions are appropriately omitted here.
[0130] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
Claims
1. A steering compensation method for an autonomous vehicle, characterized in that, The method includes: While the vehicle to be compensated is traveling stably along a straight path, the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated are acquired in real time. Calculate the difference between the front axle steering command and the rear axle steering command to obtain the steering command difference; Using the effective heading angle error, the common mode zero-position deviation estimate of the vehicle to be compensated is updated periodically, wherein the common mode zero-position deviation estimate is used to characterize the same-direction offset of the front axle and the rear axle, and the effective heading angle error is obtained by limiting the heading angle error through dead zone threshold and saturation threshold; Using the steering command difference that satisfies the steady-state condition, the differential mode zero-position deviation estimate of the vehicle to be compensated is updated periodically, wherein the differential mode zero-position deviation estimate is used to characterize the reverse bias of the front axle and the rear axle. The common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate are restored to the front axle compensation amount and the rear axle compensation amount, so that the steering compensation is performed on the front axle and the rear axle of the vehicle to be compensated, respectively.
2. The method according to claim 1, characterized in that, The step of using the effective heading angle error to update the common-mode zero-position deviation estimate of the vehicle to be compensated on a cycle-by-cycle basis includes: The common-mode zero-point deviation estimate is calculated using the following formula: in, Indicates update cycle The corresponding common mode update amount, This represents the effective heading angle error after being limited by dead zone threshold and saturation threshold. This represents the common-mode learning gain, used to adjust the convergence speed. This represents the common-mode update weights generated based on the lateral position error. This represents the estimated common-mode null offset value from the previous update period. This represents the estimated common-mode zero-point deviation for the current update period.
3. The method according to claim 2, characterized in that, The step of using the steering command difference that satisfies the steady-state condition to update the differential mode zero-position deviation estimate of the vehicle to be compensated on a cycle-by-cycle basis includes: The differential mode zero-point deviation estimate is calculated using the following formula: in, This represents the differential learning gain, used to adjust the convergence speed. This indicates the differential mode zero-point deviation from the previous update cycle. This represents the estimated value of the differential modulus zero-point deviation for the current update period. Indicates the update cycle. This represents the difference in steering commands that satisfy steady-state conditions.
4. The method according to claim 3, characterized in that, The step of restoring the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount includes: The front axle compensation and rear axle compensation are calculated using the following formula: in, Indicates the front axle compensation amount. Indicates the rear axle compensation amount. Indicates wheelbase. This indicates the distance from the front axle to the center of gravity of the vehicle to be compensated. This indicates the distance from the rear axle to the center of gravity of the vehicle to be compensated.
5. The method according to claim 4, characterized in that, weight value It is selected using the following formula: in, The lateral position error of the vehicle to be compensated is obtained by projecting the vehicle's position relative to the matching point of the reference path onto the local normal direction, representing the deviation between the vehicle and the lane centerline. This represents the preset full confidence threshold for lateral error. This indicates the preset lateral error stop learning threshold.
6. The method according to claim 5, characterized in that, After restoring the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation and rear axle compensation, the method further includes: front axle compensation With the front axle steering command The commands are summed to obtain independent steering control commands for the front axle. Rear axle compensation Rear axle steering command The commands are summed to obtain independent steering control commands for the rear axle. The independent steering control commands for the front axle and the rear axle are each limited to a preset steering range, and the limited independent steering control commands for the front axle and the rear axle are sent to the vehicle to be compensated for execution.
7. A steering compensation device for an autonomous vehicle, characterized in that, The device includes: The parameter acquisition module is configured to acquire the heading angle error, front axle steering command, and rear axle steering command of the vehicle to be compensated in real time while the vehicle to be compensated is driving stably along a straight path. The calculation module is configured to calculate the difference between the front axle steering command and the rear axle steering command to obtain the steering command difference; The deviation compensation module is configured to use the effective heading angle error to update the common mode zero-position deviation estimate of the vehicle to be compensated on a cycle basis. The common mode zero-position deviation estimate is used to characterize the same-direction offset of the front axle and the rear axle. The effective heading angle error is obtained by limiting the heading angle error through a dead zone threshold and a saturation threshold. The deviation compensation module is configured to update the differential zero-position deviation estimate of the vehicle to be compensated on a cycle using the steering command difference that satisfies the steady-state condition, wherein the differential zero-position deviation estimate is used to characterize the reverse bias of the front axle and the rear axle. The deviation compensation module is configured to restore the common-mode zero-position deviation estimate and the differential-mode zero-position deviation estimate to the front axle compensation amount and the rear axle compensation amount, so as to perform steering compensation on the front axle and the rear axle of the vehicle to be compensated, respectively.
8. An autonomous vehicle, characterized in that, include: Vehicle controller, memory, and bus; The vehicle controller is communicatively connected to the memory via the bus. The memory stores a computer program, which, when executed by the processor of the vehicle controller, implements the method described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, can implement the method as claimed in any one of claims 1-6.
10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed, it can implement the method as claimed in any one of claims 1-6.
