A method for identifying backlash parameters of a multi-axle distributed electric drive chassis corner module

Abstract

The application discloses a kind of gear gap parameter identification methods of multi-axis distributed electric drive chassis angle module, it is related to vehicle chassis dynamics modeling, steering actuator parameter identification and distributed electric drive chassis control technical field.The method includes: according to the transmission ratio of speed reducer, establish output end ideal position model;Gear gap half-width and internal transmission state are introduced, and the gear gap internal state recursion model is established;Establish the linear mapping between output end encoder observation value and internal transmission state;Parameter identification problem is constructed with the mean square error between model predicted output and encoder measured value as target;Optimal parameters are solved using three-stage identification strategy of iterative reweighted least squares rough estimate, multiple starting point nonlinear parameter identification and full parameter joint fine dressing.The physical interpretability of angle module steering actuator gear gap model, identification stability and output prediction accuracy are improved by reducing gear gap non-smooth model parameter coupling and initial value sensitivity.

Description

Technical Field

[0001] This invention relates to the fields of vehicle chassis dynamics modeling, steering actuator parameter identification, and distributed electric drive chassis control technology, and in particular to a method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module. Background Technology

[0002] Multi-axle distributed electric drive chassis achieve driving, braking, and steering functions through multiple corner modules, featuring a large number of actuators, variable configuration, and high degree of control freedom. Compared with traditional fixed-axle vehicles, the steering center and motion configuration of this type of chassis change with the control strategy, thus placing higher demands on the dynamic modeling accuracy of the corner module steering actuators.

[0003] In angular module steering actuators, the motor output typically needs to be amplified by a worm gear reducer before being transmitted to the steering output end. Worm gear drives have advantages such as large reduction ratios, compact structure, and high load-bearing capacity, but backlash is unavoidable in practical engineering. Backlash originates from factors such as machining errors, assembly tolerances, shaft support clearance, lubrication requirements, thermal expansion compensation, and anti-jamming design.

[0004] When the input motor experiences a slight displacement or a switch between forward and reverse directions, the worm and worm wheel need to undergo a transition process from contact on one tooth surface to contact on the other. During this process, the input displacement cannot be immediately transmitted to the output end, resulting in dead zones, hysteresis, and backlash in the output rotation angle relative to the input displacement. Elastic deformation of the transmission chain, clearances in the support components, and long-term wear further exacerbate this non-ideal transmission phenomenon.

[0005] Existing methods for identifying tooth gap parameters typically employ a single filter or a one-time joint optimization approach to process all parameters. Since the tooth gap model contains recursive states and non-smooth min-max operators, and there is coupling between the tooth gap half-width, output gain, zero-position bias, and initial internal state, direct joint search is easily affected by initial values ​​and local extrema, leading to discrete identification results, insufficient physical consistency, or inadequate characterization of tooth gap features in the commutation interval. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing a backlash parameter identification method for a multi-axis distributed electric drive chassis corner module. This backlash parameter identification method for a multi-axis distributed electric drive chassis corner module is used to solve the problems of difficulty in stably modeling backlash dead zone, hysteresis and backlash in corner module worm gear reduction transmission, as well as the problems of sensitivity to initial values, discrete results and insufficient physical interpretability in the joint identification of non-smooth recursive model parameters.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for identifying backlash parameters of a multi-axis distributed electric drive chassis corner module includes the following steps:

[0009] Step 1: Collect the position feedback of the steering motor, the transmission ratio of the reducer, and the position feedback of the output encoder to establish an ideal position model of the output end; introduce the backlash half-width and the internal transmission state to establish a recursive model of the backlash internal state; establish an encoder observation mapping model based on the linear mapping between the output encoder observation value and the internal transmission state; combine the ideal position model of the output end, the recursive model of the backlash internal state, and the encoder observation mapping model to form the backlash model of the angular module steering actuator.

[0010] Step 2: Set the backlash parameters of the tooth backlash model, including the tooth backlash half-width. Output gain parameters Zero-position offset parameters and initial internal state A tooth gap parameter vector is established based on the tooth gap parameters, and the tooth gap parameters are identified to obtain the final identified tooth gap parameter vector.

[0011] Furthermore, after step 2, step 3 is performed: the tooth backlash parameters identified in step 2 are substituted into the tooth backlash model for forward simulation to reconstruct the output encoder response and evaluate the model output tracking effect.

[0012] Furthermore, the evaluation model output tracking performance is evaluated using root mean square error (RMSE), mean absolute error (MAE), and goodness of fit (R²).

[0013] Furthermore, the ideal position model of the output end is established according to the following formula (1).

[0014] ;

[0015] In equation (1), This provides the steering motor position feedback at the k-th sampling time. The gear ratio of the reducer. This represents the ideal position of the output at the k-th sampling time.

[0016] Furthermore, the recursive model of the internal state of the tooth gap is established according to the following formula (2).

[0017] (2);

[0018] In equation (2), This represents the internal propagation state at the k-th sampling time. It is half the width of the tooth gap.

[0019] Furthermore, the encoder observation mapping model is established according to the following formula (3).

[0020] ;

[0021] In equation (3), This is the output encoder observation value at the k-th sampling time. For output gain parameters; This is the zero-position offset parameter.

[0022] Furthermore, step 2 specifically involves identifying the backlash parameters to obtain the final identified backlash parameter vector through the following steps:

[0023] Step 2.1: Establish the backlash parameter vector according to the following formula (4).

[0024] (4);

[0025] In equation (4), T represents the vector transpose;

[0026] Step 2.2: Convert the backlash parameter vector Input the backlash model and output the prediction output at the k-th sampling time. And construct the backlash parameter identification objective function according to the following formula (5)

[0027] (5);

[0028] In equation (5), This represents the number of sampling points;

[0029] Step 2.3: Adjust the output gain parameters according to formulas (6) and (7). and zero offset parameters Make a preliminary estimate

[0030] (6);

[0031] (7);

[0032] In equation (6), These are the coarse estimates of the output gain parameter and the coarse estimates of the zero-position bias parameter obtained in the first stage of identification. For the first In the nth iteration The weights of each sample;

[0033] In equation (7), For the first The corresponding residual in the next iteration; For the first The estimated output gain parameter value for the next iteration; For the first The zero-position bias parameter estimate for the next iteration;

[0034] Step 2.4: Adjust the tooth gap half-width according to formulas (8), (9), and (10). and initial internal state Perform multi-starting point identification:

[0035] (8);

[0036] (9);

[0037] (10);

[0038] In equation (8), s is the initial value of the second stage; This represents the number of initial value groups for the second stage. The estimated half-width of the tooth gap is obtained after local optimization of the initial value of the s-th group. This corresponds to the initial internal state estimate;

[0039] In equation (9), This is the candidate parameter vector for the s-th group in the second stage;

[0040] In equation (10), The optimal backlash parameter vector for the second stage is selected from the S group of candidate vectors to minimize the objective function value.

[0041] Step 2.5: Perform joint fine-tuning of all backlash parameters according to formulas (11) and (12).

[0042] (11);

[0043] (12);

[0044] In equation (12), For Starting with the initial values ​​of the s-th group of disturbances in the neighborhood, the candidate tooth gap parameter vector of the s-th group is obtained through the third stage of local optimization. This is the initial value sequence number of the third-stage disturbance; The number of initial values ​​for the third-stage disturbance. This is the final backlash parameter vector selected from all S3 candidate parameter vectors that minimizes the objective function value.

[0045] Furthermore, the steering motor position feedback is uploaded by the steering motor controller to the vehicle controller via the CAN network, and the output encoder position feedback is measured by the encoder at the output shaft end of the worm gear reducer.

[0046] Furthermore, the backlash model is used to describe the dead zone, hysteresis, and backlash phenomena of the angular module steering actuator under rapid step input and slow ramp input.

[0047] The present invention has the following beneficial effects:

[0048] First, the present invention constructs a min-max saturated recursive model by using the backlash half-width and internal transmission state, which can directly characterize the dead zone, hysteresis and backlash phenomena in the commutation process of the worm gear reducer, thereby making the angular module steering actuator model more consistent with the actual transmission mechanism.

[0049] Second, the present invention models the output encoder observation as a linear mapping of the internal transmission state and includes the output gain and zero-position offset in the joint identification range, which can compensate for the actual transmission ratio deviation and encoder zero-position calibration error, and improve the model's adaptability to measured data.

[0050] Third, the present invention adopts a three-stage parameter identification strategy from coarse to fine. First, it stabilizes and estimates the linear parameters, then identifies the nonlinear parameters of the tooth gap, and finally performs joint refinement of all parameters, thereby reducing the degree of parameter coupling and the sensitivity to initial values.

[0051] Fourth, the present invention can identify backlash parameters under different steering conditions such as fast step input and slow ramp input, and obtain lower root mean square error (RMSE), mean absolute error (MAE) and higher goodness of fit (R2), thereby improving the prediction accuracy of the encoder signal at the output end.

[0052] Fifth, the backlash parameters identified by this invention can be used for dynamic modeling, steering control compensation, and vehicle control model correction of the steering actuator of the multi-axis distributed electric drive chassis angle module, providing a more reliable actuator model basis for high-degree-of-freedom chassis control.

[0053] In summary, this invention combines backlash nonlinear mechanism modeling with a staged parameter identification method to achieve an effective description and stable identification of the backlash characteristics of a worm gear kingpin steering actuator. This not only improves the model's ability to fit and reconstruct the actual output response, but also provides a reliable parameter basis for subsequent steering control compensation and vehicle control optimization, demonstrating good practicality and promotional value. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the backlash parameter identification method for the multi-axis distributed electric drive chassis corner module of the present invention.

[0055] Figure 2 This is a CATIA design drawing of the mechanical structure of the corner module of a multi-axis distributed electric drive chassis corner module according to the present invention; wherein, (a) is a three-dimensional... Figure 1 (b) is a three-dimensional object. Figure 2 .

[0056] Figure 3 This is a physical image of the corner module steering actuator of the multi-axis distributed electric drive chassis corner module in the 0°, 45° and 90° steering angle states of the backlash parameter identification method of the corner module of the present invention.

[0057] Figure 4 This is a diagram showing the relationship between the steering motor feedback derivation value and the output encoder feedback of the backlash parameter identification method for a multi-axis distributed electric drive chassis angle module according to the present invention; where (a) is test data 1; and (b) is test data 2.

[0058] Figure 5 This is a comparison chart of different methods for evaluating tracking error in the backlash parameter identification method of a multi-axis distributed electric drive chassis corner module according to the present invention; where (a) is the root mean square error RMSE; (b) is the mean absolute error MAE; and (c) is the coefficient of determination R².

[0059] Figure 6 This is a comparison of the overall tracking of the output encoder feedback on different test datasets of the backlash parameter identification method of the multi-axis distributed electric drive chassis angle module of the present invention; wherein, (a) is dataset 1; (b) is dataset 2; (c) is dataset 3; and (d) is dataset 4.

[0060] Figure 7 yes Figure 6 A magnified view of the area within the square frame; where (a), (b), (c), and (d) correspond to... Figure 6 The boxed areas in figures (a) to (d) in the text.

[0061] Figure 8 This is a summary chart of the tracking performance evaluation indicators of different method models for identifying backlash parameters of a multi-axis distributed electric drive chassis corner module according to the present invention; where (a) is the root mean square error RMSE; (b) is the mean absolute error MAE; and (c) is the coefficient of determination R². Detailed Implementation

[0062] The present invention will now be described in further detail with reference to the accompanying drawings and specific preferred embodiments.

[0063] In the description of this invention, it should be understood that the terms "left side," "right side," "upper part," "lower part," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention 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. "First," "second," etc., do not indicate the importance of the components, and therefore should not be construed as a limitation of this invention. The specific dimensions used in this embodiment are only for illustrating the technical solution and do not limit the scope of protection of this invention.

[0064] Example

[0065] like Figure 1 As shown, a backlash parameter identification method for a multi-axis distributed electric drive chassis corner module includes: Step 1, establishing a backlash model of the corner module steering actuator; Step 2, identifying backlash parameters using a three-stage method consisting of IRLS coarse estimation, multi-starting-point local optimization identification, and joint refinement of all parameters; Step 3, substituting the final identified parameters into the backlash model for forward simulation to reconstruct the output encoder response and evaluate the model output tracking effect.

[0066] Step 1: Collect the position feedback of the steering motor, the transmission ratio of the reducer, and the position feedback of the output encoder to establish the ideal position model of the output end; introduce the backlash half-width and the internal transmission state to establish the recursive model of the backlash internal state; establish the encoder observation mapping model based on the linear mapping between the output encoder observation value and the internal transmission state; combine the ideal position model of the output end, the recursive model of the backlash internal state, and the encoder observation mapping model to form the backlash model of the corner module steering actuator.

[0067] The backlash model in this embodiment is used to describe the dead zone, hysteresis, and backlash phenomena of the angular module steering actuator under fast step input and slow ramp input.

[0068] Specifically, such as Figure 1 As shown, the present invention first acquires the steering motor position feedback on the multi-axis distributed electric drive chassis angle module. and output encoder position feedback The steering motor position feedback can be obtained by the steering motor controller and uploaded to the vehicle controller via the CAN network. The encoder at the output end directly measures the actual rotation angle of the worm gear reducer output shaft.

[0069] In an ideal scenario where backlash is not considered, the transmission chain is assumed to be perfectly rigid, and the output end is ideally positioned. Position feedback from the steering motor and the gear ratio of the reducer The calculated ideal output serves as the input reference for the subsequent backlash nonlinear model. That is, the ideal position model at the output end is established using the following equation (1).

[0070] ;

[0071] In equation (1), This provides the steering motor position feedback at the k-th sampling time. The gear ratio of the reducer. This represents the ideal position of the output at the k-th sampling time.

[0072] When backlash is considered, the input displacement cannot be immediately transmitted to the output. Therefore, an internal transmission state is introduced. Describe the reachable position of the output end during actual transmission. Assume the backlash half-width is... Then the internal transit state is always restricted to [ , Within the range.

[0073] That is, the recursive model of the internal state of the tooth gap is established according to the following formula (2).

[0074] (2);

[0075] In equation (2), This represents the internal propagation state at the k-th sampling time. It is half the width of the tooth gap.

[0076] When the input displacement has not yet crossed the tooth gap boundary, the internal transmission state remains unchanged, and the output exhibits a dead zone; when the input direction changes and crosses the tooth gap boundary, the internal state continues to change with the input, thus forming an observable hysteresis.

[0077] In output encoder observation modeling, if the nominal transmission ratio is accurate and the encoder installation and zero-position calibration are perfectly ideal, the encoder observation value should be consistent with the internal transmission state. However, in actual systems, the effective transmission relationship of the reducer may deviate from the nominal transmission ratio, and there may also be a fixed offset between the encoder installation zero position and the mechanical zero position. Therefore, this invention addresses this issue by adjusting the output gain... and zero offset An encoder observation mapping model was established to enhance the model's adaptability to measured data.

[0078] The encoder observation mapping model is established according to the following formula (3).

[0079] ;

[0080] In equation (3), This is the output encoder observation value at the k-th sampling time. For output gain parameters; This is the zero-position offset parameter.

[0081] Step 2: Set the backlash parameters of the tooth backlash model. The backlash parameters include the backlash half-width. Output gain parameters Zero-position offset parameters and initial internal state A tooth backlash parameter vector is established based on the tooth backlash parameters, and the tooth backlash parameters are identified to obtain the final identified tooth backlash parameter vector.

[0082] Specifically, the tooth backlash parameters are identified through the following steps to obtain the final identified tooth backlash parameter vector:

[0083] Calculate point by point according to the recursive model of the internal state of the tooth gap. The predicted output is then obtained through the encoder observation mapping model. The mean square error between the predicted output and the measured encoder signal is used as the objective function to ensure that the identification results simultaneously meet the requirements of data fitting accuracy and physical consistency.

[0084] Specifically, step 2.1: Establish the backlash parameter vector according to the following formula (4).

[0085] (4);

[0086] In equation (4), T represents the vector transpose;

[0087] Step 2.2: Convert the backlash parameter vector Input the backlash model, and output the prediction at the k-th sampling time. And construct the backlash parameter identification objective function according to the following formula (5)

[0088] (5);

[0089] In equation (5), This represents the number of sampling points;

[0090] Step 2.3: Adjust the output gain parameters according to formulas (6) and (7). and zero offset parameters Conduct preliminary estimates for the first phase.

[0091] (6);

[0092] (7);

[0093] In equation (6), These are the coarse estimates of the output gain parameter and the coarse estimates of the zero-position bias parameter obtained in the first stage of identification. For the first In the nth iteration The weights of each sample;

[0094] In equation (7), For the first The corresponding residual in the next iteration; For the first The estimated output gain parameter value for the next iteration; For the first The zero-position bias parameter estimate for the next iteration;

[0095] The first stage employs an iterative reweighted least squares method to evaluate the linear parameters. and A preliminary estimate is made. At this stage, the backlash recursion state is not introduced; instead, it is based on the ideal output. With encoder observation The overall input-output relationship is used to construct a weighted least squares problem, and the sample weights are iteratively updated according to the residual size, thereby reducing the influence of outliers and local noise on the initial values ​​of linear parameters.

[0096] Step 2.4: Adjust the tooth gap half-width according to formulas (8), (9), and (10). and initial internal state The second phase of multi-starting point identification is carried out:

[0097] (8);

[0098] (9);

[0099] (10);

[0100] In equation (8), s is the initial value of the second stage; This represents the number of initial value groups for the second stage. The estimated half-width of the tooth gap is obtained after local optimization of the initial value of the s-th group. This corresponds to the initial internal state estimate;

[0101] In equation (9), This is the candidate parameter vector for the s-th group in the second stage.

[0102] In equation (10), This is to select the optimal backlash parameter vector for the second stage from the S group of candidate vectors, which minimizes the objective function value. The second stage is based on a fixed... and Under the condition of [condition], for the half-width of the tooth gap and initial internal state Multi-starting-point identification is performed. For each set of initial values, a local search is conducted to obtain a candidate parameter vector, and the parameter set with the smallest objective function value is selected from all candidate solutions as the identification result for this stage. This stage focuses on identifying the nonlinear characteristic parameters of the tooth gap, avoiding excessive coupling between linear and nonlinear parameters in the initial search.

[0103] Step 2.5: Perform the third-stage joint finishing of all backlash parameters according to formulas (11) and (12).

[0104] (11);

[0105] (12);

[0106] In equation (12), For Starting with the initial values ​​of the s-th group of disturbances in the neighborhood, the candidate tooth gap parameter vector of the s-th group is obtained through the third stage of local optimization. This is the initial value sequence number of the third-stage disturbance; The number of initial values ​​for the third-stage disturbance. This is the final backlash parameter vector selected from all S3 candidate parameter vectors that minimizes the objective function value.

[0107] The third stage uses the identification results from the second stage as an initial reference, within the complete parameter space. Joint refinement is performed. Initial perturbation values ​​for each group are generated within the neighborhood of the second-stage results. Local optimization further coordinates the coupling relationships between parameters, ultimately obtaining the optimal parameters that minimize the objective function value. .

[0108] The flowchart of the three-stage parameter identification algorithm is shown in Table 1 below.

[0109] Table 1. Three-stage parameter identification algorithm flow

[0110]

[0111] Wherein, IRLS stands for Iterative Reweighted Least Squares, and LocalSearch stands for Local Search.

[0112] In this embodiment, after step 2, step 3 is also performed: the tooth backlash parameters identified in step 2 are substituted into the tooth backlash model for forward simulation in order to reconstruct the encoder response at the output end and evaluate the model output tracking effect.

[0113] The root mean square error (RMSE), mean absolute error (MAE), and goodness-of-fit (R²) were used to evaluate the model's output tracking performance.

[0114] Figure 2 Figure (a) shows the first stereoscopic view of the corner module mechanical structure. Figure 2 Figure (b) is a second stereoscopic view; the two sub-figures show the assembly and relative positional relationship of the steering motor, worm gear reducer, kingpin steering mechanism and output absolute encoder from different directions.

[0115] like Figure 2 As shown, the corner module's mechanical structure was designed using CATIA. The steering actuator mainly consists of a steering motor, a worm gear reducer, a kingpin steering mechanism, and an absolute encoder at the output end. The steering motor's output torque is transmitted to the kingpin steering shaft via the worm gear reducer, and the reducer's nominal transmission ratio is... =62.

[0116] The steering motor position feedback is acquired by the steering motor controller and sent to the vehicle control unit (VCU) via the CAN network. To obtain the output measurement value, a Briter multi-turn absolute encoder is installed on the output side of the worm gear reducer. The encoder is powered by 12V, supports up to 100 turns of measurement, and communicates via the CAN bus at a baud rate of 500 kbit / s. The output encoder measurement value is also sent to the VCU via the CAN bus.

[0117] The VCU uses an EV22297 controller and is implemented with EcoCoder support. The host computer uses EcoCal for signal monitoring, data recording, and calibration testing. During the experiment, the steering motor position feedback, the output encoder measurement value, and the corresponding timestamp were recorded. The steering motor position was converted into the ideal output angle based on the nominal gear ratio for backlash parameter identification and model verification.

[0118] To evaluate the three-stage identification method, a steering test was conducted under no-load conditions. During the test, the wheels were controlled to perform different steering actions within a range of ±30°, and the steering angular velocity was set to 60° / s. The collected dataset was used for backlash parameter identification, forward simulation, and model verification.

[0119] like Figure 3 As shown, Figure 3 The diagram shows the physical postures of the steering actuator at 0°, 45°, and 90° steering angles, from left to right. The 0° posture on the left corresponds to the wheel being in the initial straight-line position; the 45° posture in the middle indicates that the steering actuator has driven the wheel to produce a mid-range steering angle; and the 90° posture on the right indicates that the wheel has completed a larger steering angle movement. The diagram also shows the installation positions of the output angle encoder and the steering motor, providing the hardware basis for obtaining steering motor position feedback and output encoder feedback.

[0120] like Figure 4As shown in Figure 1, (a) shows the relationship between the derived values ​​of the steering motor feedback and the output encoder feedback in Experimental Data 1, and (b) shows the corresponding relationship in Experimental Data 2. In the two sub-figures, the red diagonal line represents the ideal backlash-free transmission relationship, and the measured points form a band-like distribution around this ideal relationship. This band-like deviation reflects that the output response is affected by the historical motion direction during commutation, verifying the necessity of using an internal state recursive model to characterize backlash, hysteresis, and backlash characteristics.

[0121] like Figure 5 As shown, (a) compares the root mean square error (RMSE) of the three methods: the present invention, UKF, and PF; (b) compares the mean absolute error (MAE); and (c) compares the coefficient of determination (R²). The present invention's method maintains relatively low RMSE and MAE, with R² close to 1. The UKF results are similar to the present invention but with slightly larger fluctuations. The PF method shows a significant increase in error and error bar range, and a decrease in R², indicating that the present invention's method has better tracking accuracy and stability.

[0122] like Figure 6 As shown, (a) Figure 1 shows the overall tracking result of dataset 1, which includes continuous forward and reverse rotation processes with significant amplitude changes; (b) Figure 2 shows the overall tracking result of dataset 2, which includes frequent switching between positive and negative directions; (c) Figure 3 shows the overall tracking result of dataset 3, which includes multiple plateau segments and reversal segments; and (d) Figure 4 shows the overall tracking result of dataset 4, which includes continuous turning actions with different amplitudes. In each subfigure, the prediction curve of the method of this invention can fit the encoder's true value curve well and maintain stable overall tracking capability under different input forms.

[0123] like Figure 7 As shown, (a) is a magnified view of the positive commutation and plateau interval of dataset 1, (b) is a magnified view of the negative commutation interval of dataset 2, (c) is a magnified view of the plateau and reverse transition interval of dataset 3, and (d) is a magnified view of the negative commutation interval of dataset 4. Compared with the overall curve, the magnified views more clearly show that during backlash crossing and commutation, the prediction curve of the method of this invention is closer to the encoder's true curve, and UKF and PF have different degrees of deviation in the local plateau or transition interval.

[0124] like Figure 8 As shown, (a) Figure summarizes and compares the output tracking errors of different methods from the perspective of RMSE, (b) Figure summarizes and compares the absolute errors of different methods from the perspective of MAE, and (c) Figure summarizes and compares the fitting degree of different methods from the perspective of R². The results show that the method of the present invention maintains a small error index and R² is close to 1, indicating that the backlash parameters obtained by the three-stage identification method can effectively improve the accuracy and consistency of model output tracking.

[0125] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for identifying backlash parameters of a multi-axis distributed electric drive chassis corner module, characterized in that: Includes the following steps: Step 1: Collect the position feedback of the steering motor, the transmission ratio of the reducer, and the position feedback of the output encoder to establish an ideal position model of the output end; introduce the backlash half-width and the internal transmission state to establish a recursive model of the backlash internal state; establish an encoder observation mapping model based on the linear mapping between the output encoder observation value and the internal transmission state; combine the ideal position model of the output end, the recursive model of the backlash internal state, and the encoder observation mapping model to form the backlash model of the angular module steering actuator. Step 2: Set the backlash parameters of the tooth backlash model, including the tooth backlash half-width. Output gain parameters Zero-position offset parameters and initial internal state A tooth gap parameter vector is established based on the tooth gap parameters, and the tooth gap parameters are identified to obtain the final identified tooth gap parameter vector.

2. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: Step 3 is executed after step 2: the backlash parameters identified in step 2 are substituted into the backlash model for forward simulation in order to reconstruct the encoder response at the output end and evaluate the model output tracking effect.

3. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 2, characterized in that: The evaluation model output tracking performance is assessed using root mean square error (RMSE), mean absolute error (MAE), and goodness of fit (R²).

4. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis corner module according to claim 1, characterized in that: The ideal position model of the output terminal is established according to the following formula (1). ； In equation (1), This provides the steering motor position feedback at the k-th sampling time. The gear ratio of the reducer. This represents the ideal position of the output at the k-th sampling time.

5. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: The recursive model of the internal state of the tooth gap is established according to the following formula (2). （2）； In equation (2), This represents the internal propagation state at the k-th sampling time. It is half the width of the tooth gap.

6. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: The encoder observation mapping model is established according to the following formula (3). ； In equation (3), This is the output encoder observation value at the k-th sampling time. For output gain parameters; This is the zero-position offset parameter.

7. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: Step 2 specifically involves the following steps to identify the tooth backlash parameters and obtain the final identified tooth backlash parameter vector: Step 2.1: Establish the backlash parameter vector according to the following formula (4). ; (4)； In equation (4), T represents the vector transpose; Step 2.2: Convert the backlash parameter vector Input the backlash model and output the prediction output at the k-th sampling time. And construct the backlash parameter identification objective function according to the following formula (5) ; （5）； In equation (5), This represents the number of sampling points; Step 2.3: Adjust the output gain parameters according to formulas (6) and (7). and zero offset parameters Make a preliminary estimate （6）； （7）； In equation (6), These are the coarse estimates of the output gain parameter and the coarse estimates of the zero-position bias parameter obtained in the first stage of identification. For the first In the nth iteration The weights of each sample; In equation (7), For the first The corresponding residual in the next iteration; For the first The estimated output gain parameter value for the next iteration; For the first The zero-position bias parameter estimate for the next iteration; Step 2.4: Adjust the tooth gap half-width according to formulas (8), (9), and (10). and initial internal state Perform multi-starting point identification: (8)； (9)； (10)； In equation (8), s is the initial value of the second stage; This represents the number of initial value groups for the second stage. The estimated half-width of the tooth gap is obtained after local optimization of the initial value of the s-th group. This corresponds to the initial internal state estimate; In equation (9), This is the candidate parameter vector for the s-th group in the second stage; In equation (10), The optimal backlash parameter vector for the second stage is selected from the S group of candidate vectors to minimize the objective function value. Step 2.5: Perform joint fine-tuning of all backlash parameters according to formulas (11) and (12). (11)； (12)； In equation (12), For Starting with the initial values ​​of the s-th group of disturbances in the neighborhood, the candidate tooth gap parameter vector of the s-th group is obtained through the third stage of local optimization. This is the initial value sequence number of the third-stage disturbance; The number of initial values ​​for the third-stage disturbance. This is the final backlash parameter vector selected from all S3 candidate parameter vectors that minimizes the objective function value.

8. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: The position feedback of the steering motor is uploaded to the vehicle controller via the CAN network by the steering motor controller, and the position feedback of the output encoder is measured by the encoder at the output shaft of the worm gear reducer.

9. The method for identifying backlash parameters of a multi-axis distributed electric drive chassis angle module according to claim 1, characterized in that: The backlash model is used to describe the dead zone, hysteresis, and backlash phenomena of the angular module steering actuator under fast step input and slow ramp input.

Images