Vehicle mass real-time estimation method based on dynamic forgetting factor

Abstract

The invention provides a dynamic forgetting factor recursive least square method (DFFRLS) real-time quality estimation method for improving the rapidity and stability of quality estimation in order to solve the problems of frequent change and measurement error of vehicle quality. The method is based on the vehicle dynamics principle, system output and observable measurement of a recursive least square (RLS) method are determined, and the estimation result is improved into a dynamic change mode related to the vehicle speed and working conditions by analyzing the influence of a forgetting factor on the estimation result. And through a Matlab / Simulink simulation test environment, the estimation effect of the FFRLS and the estimation effect of the DFFRLS are compared. A simulation result shows that under a specific working condition, the estimation time of the DFFRLS is within 2s, and the fluctuation of a mass estimation value is smaller than that of the FFRLS, which indicates that the DFFRLS is superior to the FFRLS in rapidity and stability, and can be effectively used for real-time estimation of the vehicle mass.

Description

Technical Field

[0001] The invention belongs to the technical field of automobile control, and in particular relates to a real-time estimation method of vehicle mass based on a dynamic forgetting factor. Background Art

[0002] At present, the most common method for vehicle mass estimation is the recursive least squares method and its improved method. Reference Vehicle mass and road slope estimation considering lateral motion [J / OL] The vehicle dynamics model is modified by acceleration, and the least squares algorithm with forgetting factor to enhance new data to adapt to the time-varying characteristics of the vehicle system is used to estimate the vehicle mass. Reference Joint estimation of freight vehicle mass and road slope under complex working conditions [J / OL] Aiming at the stability problem of mass estimation during the acceleration process of freight vehicles, the influence of different acceleration types on mass estimation is explored, and a joint estimation method of M estimation and recursive least squares method based on forgetting factor is proposed to achieve robust estimation of freight vehicle mass. Reference Electric vehicle ramp recognition integrating vehicle mass estimation [J / OL] By establishing a longitudinal dynamics model of the vehicle and analyzing the signal characteristics of the speed sensor under actual vehicle conditions, a least squares vehicle mass estimation strategy with forgetting factor is constructed, which realizes the direct acquisition of vehicle mass under starting conditions. The reference Vehicle Mass Estimation Based on Longitudinal Frequency Response Characteristics[J] obtains longitudinal acceleration and wheel speed information, and based on the vehicle longitudinal dynamics model, derives the amplitude-frequency function relationship between acceleration and wheel speed signals, and obtains the amplitude ratio between acceleration and wheel speed at the center of mass of the vehicle at different frequencies, and then obtains the vehicle mass through least squares fitting. The reference Vehicle Mass and Road Slope Estimation Based on Interactive Multi-Model[J] sets the working conditions suitable for accurate mass estimation, and proposes a mass estimation confidence factor calculation algorithm based on fuzzy rules, and then designs a recursive least squares vehicle mass estimation algorithm based on confidence factor to achieve online mass estimation. Summary of the invention

[0003] In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a real-time estimation method of vehicle mass based on dynamic forgetting factor, which can improve the efficiency and accuracy of real-time estimation of vehicle mass.

[0004] To achieve the above object, the present invention provides a vehicle mass real-time estimation method based on a dynamic forgetting factor, comprising the following steps:

[0005] Collect vehicle measurement data.

[0006] Based on vehicle data, a vehicle longitudinal dynamics model is constructed.

[0007] Based on the vehicle longitudinal dynamics model, a vehicle mass estimation model with the least squares method with forgetting factor is established.

[0008] Based on the vehicle mass estimation model, the dynamic forgetting factor expression is set.

[0009] Use the exhaustive method to solve the unknown parameters of the dynamic forgetting factor.

[0010] The collected vehicle data is substituted into the vehicle mass estimation model with the least squares method with forgetting factor to estimate the vehicle mass.

[0011] Furthermore, the collected vehicle data include tire radius, air density, frontal area, drag coefficient; vehicle speed, rolling resistance coefficient, slope, acceleration, and combined torque parameters.

[0012] According to the above measurement data, the vehicle longitudinal dynamics model can be constructed as follows:

[0013]

[0014] In the above formula, T dr The total torque generated by the driver's control of the accelerator and brake pedals, in Nm; r tire is the tire radius, in m; ρ is the air density, which is 1.29 kg / m^3; A is the windward area, in m^2; C W is the wind resistance coefficient; v is the vehicle speed, in m / s; ε is the rolling resistance coefficient; m is the vehicle mass, in kg; g is the acceleration of gravity, which is 9.8N / kg; α is the slope, which is 0 to 1; a is the acceleration, in m / s^2;

[0015] White noise is introduced to simulate the measurement errors of vehicle speed, acceleration and driving torque.

[0016] According to the above vehicle longitudinal dynamics model, a vehicle mass estimation model with the least squares method with forgetting factor is established: The incremental matrix update formula is:

[0017]

[0018] The covariance matrix update formula is:

[0019]

[0020] M(k) represents the estimated mass of the kth step, K(k) represents the gain of the kth step, EF(k) represents the output of the kinetic equation of the kth step, Eα(k) represents the observation of the kth step, k represents the kth step, M(k-1) represents the mass of the k-1th step, P(k-1) represents the covariance of the k-1th step, and P(k) represents the covariance of the kth step;

[0021] The initial conditions of the vehicle mass estimation model are:

[0022]

[0023] k is the number of iterations. When k = 0, P = aE. Where α is a sufficiently large positive real number, ε is a zero vector or a sufficiently small positive real vector, and E is the unit matrix;

[0024] When the kth time, the calculation method of the kth estimated value θ is: θ9k) = θ(k-1(+correction term;

[0025] According to the above vehicle mass estimation model, the range of the forgetting factor is determined to be 0.8 to 1. When the estimation starts, it is set to 0.8. As the system running time increases, the forgetting factor approaches 1. Based on this setting, the calculation formula of the forgetting factor is as follows:

[0026]

[0027] Where t is the time the vehicle stops, A, B, and C are parameters to be determined;

[0028] Use the exhaustive method to find the above parameters. The method to judge whether the parameters are reasonable is to observe the difference between the true value and the estimated value. The smaller the difference, the better the three parameters are. Thus, the calculation expression of the dynamic forgetting factor is obtained:

[0029]

[0030] According to the vehicle longitudinal dynamics model and the least squares vehicle mass estimation model with dynamic forgetting factor, the algorithm module of the vehicle mass real-time estimation system based on MATLAB is determined;

[0031] A vehicle data acquisition device comprises a data acquisition module, a calculation module, a display module and a data transmission module, wherein the data of the data acquisition module comprises the vehicle's combined torque, vehicle speed, acceleration and slope; the calculation module comprises any one of the vehicle mass estimation methods of claims 1-9; and the display module comprises a real-time display of the vehicle's mass estimation result.

[0032] The beneficial effects of the present invention include rapidity: the introduction of the dynamic forgetting factor makes the contribution of historical data gradually decrease after multiple iterations, and the model mainly relies on the latest data for estimation, so that when the system state changes, the estimated value can be quickly adjusted to adapt to the current working condition changes. This high real-time feature ensures the response speed of the model, but due to the low weight of historical data, the estimation results may fluctuate to a certain extent. Stability: With multiple iterations, the weight of historical data gradually increases, and the estimation results show higher stability and are less affected by short-term data fluctuations. By dynamically adjusting the weight of historical data, the model can effectively reduce the impact of working condition changes on the estimation accuracy and improve the robustness of the estimation. However, this also means that when the system state changes significantly, the response speed of the model may be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Flowchart provided by the present invention

[0034] Figure 2 The algorithm flow chart of the present invention is

[0035] Figure 3 RLS algorithm module diagram

[0036] Figure 4 The vehicle quality estimation model algorithm structure diagram of the present invention is

[0037] Figure 5 The algorithm module diagram of the system observable quantity Ea of the present invention is

[0038] Figure 6 The algorithm module diagram of the system output EF of the present invention is

[0039] Figure 7 This is the module diagram of the forgetting factor recursive least squares algorithm of the present invention.

[0040] Figure 8 is the estimated value of vehicle mass with a forgetting factor of 0.80

[0041] Fig. 9 is the estimated value of vehicle mass with a forgetting factor of 0.95

[0042] Fig.10 is the estimated value of vehicle mass with a forgetting factor of 0.99

[0043] Fig.11 is the estimated value of vehicle mass with dynamic forgetting factor

[0044] Fig.12 The force of the vehicle DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0046] Unless otherwise specified, the technical terms or scientific terms used in this application should be understood by those skilled in the art. The terms "first", "second", etc. used in this specification, claims and drawings are used to distinguish similar objects and do not indicate a specific order or priority. It is understood that these data can be interchanged where appropriate to implement the vehicle mass real-time estimation method described in the embodiment of the present invention. In addition, the terms "including" and "having" and any variations thereof are intended to indicate non-exclusive inclusion, and the term "plurality" indicates two or more possibilities unless otherwise specified. In the embodiment of the present invention, the character " / " is used to indicate that the front and back objects have an "or" relationship, for example, sensor A / sensor B means sensor A or sensor B. The term "and / or" is used to describe the association relationship of objects, and can represent three relationships, for example, sensor A and / or sensor B means sensor A, sensor B, or sensor A and sensor B exist at the same time. The term "corresponding" refers to an association relationship or binding relationship, for example, the vehicle load data corresponds to the vehicle mass estimation model, which means that there is an association or binding relationship between the two.

[0047] The present invention provides a vehicle mass real-time estimation method based on dynamic forgetting factor, such as Figure 1-2 The process shown.

[0048] Step 1: Vehicle longitudinal dynamics analysis; the force conditions of the vehicle are as follows: Fig.12 :

[0049] The longitudinal resistance of the vehicle is composed of braking force, air resistance, rolling resistance, and the component of the vehicle weight in the ramp direction. The driving force comes from the driving source (such as the main drive motor, engine). The driver controls the driving force and braking force of the vehicle through the accelerator and brake pedals to achieve acceleration and deceleration control of the vehicle.

[0050] Step 2: Collect data, check the consistency and completeness of the data, ensure that the data can be used and meet the data standards required by the experiment; the collected data includes the vehicle's total torque, speed, acceleration and slope.

[0051] The vehicle detection device is used to collect data from the experimental vehicle, including the combined torque of the vehicle driving force, the vehicle speed and the slope the vehicle is currently on; the collected data is transmitted to a computer equipped with a matlab / Simulink simulation test environment through a CAN network; the algorithm of the longitudinal dynamics model is set in the computer, and the relevant software modules are designed according to the algorithm. (e.g. Figure 3-4 (shown)

[0052] Taking the bus as an example, the relevant parameters are obtained as follows:

[0053] Table 1 Main parameters of buses

[0054]

[0055]

[0056] Step 3: Data preprocessing, introducing white noise, and simulating real errors;

[0057] In actual bus operation, the measurements of speed, acceleration and driving torque will be disturbed by various external factors, resulting in errors, and these errors are random.

[0058] The white noise settings are as follows:

[0059] The noise power of the vehicle speed white noise is set to 0.001 and the sampling time is set to 0.01 seconds; the noise power of the acceleration white noise is set to 0.000001 and the sampling time is set to 0.01 seconds; the noise power of the driving torque white noise is set to 0.01 and the sampling time is 0.02 seconds.

[0060] The addition of white noise can effectively bring these errors into the model and provide more representative simulation data for the subsequent estimation process.

[0061] Step 4: Based on the data, establish the vehicle longitudinal dynamics model. The model is as follows:

[0062]

[0063] Combining the vehicle longitudinal dynamics model and the least squares mass estimation model, in order to facilitate parameter estimation, the above formula is rewritten as:

[0064]

[0065] In the above formula, T dr The total torque generated by the driver's control of the accelerator and brake pedals, in Nm; r tire is the tire radius, in m; ρ is the air density, which is 1.29 kg / m^3; A is the windward area, in m^2; C W is the drag coefficient; v is the vehicle speed, unit is m / s; ε is the rolling resistance coefficient; m is the vehicle mass, unit is kg; g is the acceleration due to gravity, which is 9.8N / kg; α is the slope, which is 0~1; a is the acceleration, unit is m / s^2.

[0066] Design the vehicle longitudinal dynamics model into a matlab / Simulink algorithm module (such as Figure 4As shown), the data is immediately acquired and transmitted to the system for real-time calculation of relevant data.

[0067] Step 5: Select a model for estimating vehicle mass. The present invention uses a least squares model to estimate mass.

[0068]

[0069] M(k) represents the estimated mass of the kth step, K(k) represents the gain of the kth step, EF(k) represents the output of the kinetic equation of the kth step, Eα(k) represents the observation of the kth step, k represents the kth step, M(k-1) represents the mass of the k-1th step, P(k-1) represents the covariance of the k-1th step, and P(k) represents the covariance of the kth step;

[0070] The goal of the quality estimation formula (3) is to estimate the quality of the previous moment To calculate the k-th mass The estimated value at the previous moment is corrected by adding an adjustment amount, which is determined by the difference between the gain matrix K(k0) and the measured data.

[0071] The role of the gain matrix K(k) is to adjust the amount of change in the quality estimate, which depends on the current data difference and the dynamic forgetting factor.

[0072] The covariance matrix P(k) is used to represent the uncertainty of the estimated value and needs to be updated after each calculation to reflect the impact of new observations on the estimate.

[0073] The initial conditions of the vehicle mass estimation model described in the present invention are:

[0074]

[0075] k is the number of iterations. When k = 0, P = aE. Where α is a sufficiently large positive real number, ε is a zero vector or a sufficiently small positive real vector, and E is the unit matrix;

[0076] According to the analysis of the vehicle longitudinal dynamics model, the observable quantity Eα of the system is determined as:

[0077] E α =a+ε·g·cos(·α)+g·sin(·α)

[0078] The algorithm module on MATLAB is represented as Figure 5 shown.

[0079] According to the analysis of the vehicle longitudinal dynamics model, the output EF of the system is determined as:

[0080]

[0081] The algorithm module on MATLAB is represented as Figure 6 shown.

[0082] When it is the kth time, the calculation method of the kth estimated value θ is:

[0083] θ(k) = θ(k-1) + correction term;

[0084] The purpose of introducing correction terms is to improve the estimation accuracy of the model and dynamically adapt to possible deviations during vehicle operation. Over time, the characteristics of the vehicle system may gradually deviate from the initially set parameters. The role of the dynamic forgetting factor here is to ensure that the model pays more attention to the most recent data by attenuating the influence of old data, thereby improving the responsiveness to real-time conditions. The role of the correction term is to make appropriate adjustments to the parameters according to the current error at each estimation, further reducing the estimation error.

[0085] According to the above vehicle mass estimation model, the range of the forgetting factor is determined to be 0.8 to 1. The forgetting factor cannot be 1. When the estimation starts, it is set to 0.8. As the system running time increases, the forgetting factor approaches 1. Based on this setting, the calculation formula of the forgetting factor is as follows:

[0086]

[0087] Where t is the time the vehicle stops, A, B, and C are parameters to be determined;

[0088] Use the exhaustive method to find the above parameters. The method to judge whether the parameters are reasonable is to observe the difference between the true value and the estimated value. The smaller the difference, the better the three parameters are. Thus, the calculation expression of the dynamic forgetting factor is obtained:

[0089]

[0090] The dynamic forgetting factor determines the importance that this experiment attaches to old data and new data. It is used to control the model's balance between new and old data and ensure the accuracy and adaptability of the vehicle estimation model.

[0091] λ M The value range of is 0.8~1, and is not equal to 1; at each restart, λ M is 0.8, which is conducive to the rapidity of quality estimation; after a long time of docking, λ M It tends to 1, which is conducive to the stability of quality estimation.

[0092] The vehicle mass is estimated by the recursive least squares method, and the value of the forgetting factor is dynamically adjusted according to the vehicle operating status, so that the vehicle mass can be estimated quickly and stably under different working conditions.

[0093] The adjustment factor of the dynamic forgetting factor is 0.2, and its value range is determined to be 0.1 to 0.2. Within this range, the dynamic forgetting factor can quickly adjust the estimated situation in the early stage, and will adapt to the changing data in the later stage and tend to be stable.

[0094] The algorithm of the above vehicle mass estimation model is designed as a matlab / Simulink algorithm module (such as Figure 7 As shown), the obtained result data is passed into the system for real-time calculation. The algorithm in the system is expressed as:

[0095]

[0096] Substitute the forgetting factor into the vehicle mass estimation model, calculate the comparison between the vehicle mass at the corresponding moment and the actual vehicle mass. According to the comparison results, determine the parameter value of the expression of the forgetting factor, and dynamically associate the obtained dynamic forgetting factor with the vehicle restart time point t to achieve real-time, fast and stable estimation of vehicle mass.

[0097] The comparison results are as follows Figure 8-10 As shown in the above three figures, it can be seen that the smaller the forgetting factor, the shorter the time required for estimation, but the volatility of the estimation result is greater; the larger the forgetting factor, the longer the time required for estimation, but the volatility of the estimation result is smaller. Fig.11 As shown in the figure, the dynamic forgetting factor estimation method is both fast and stable, with significant results.

[0098] The present invention provides a device for measuring vehicle data in real time, which is installed on the vehicle to achieve real-time estimation of vehicle mass. The working process is as follows:

[0099] (1) Use a diagnostic instrument to measure the current driving state of the vehicle, and obtain the vehicle speed v, acceleration a, current road slope α, and the total torque Tdr generated by the accelerator and brake pedal control.

[0100] (2) By connecting the CAN communication card to the vehicle's CAN network interface, the CAN network data on the vehicle can be read in real time.

[0101] (3) Using Matlab / Simulink software, the above system is applied to perform real-time estimation of vehicle mass based on real-time read data.

Claims

1. A real-time vehicle mass estimation method based on dynamic forgetting factor, characterized in that: The following steps are involved: Collect vehicle measurement data; Based on vehicle data, construct a vehicle longitudinal dynamics model; Based on the vehicle longitudinal dynamics model, a vehicle mass estimation model with the least square method with forgetting factor is established; Based on the vehicle mass estimation model, a dynamic forgetting factor expression is set; Use the exhaustive method to solve the unknown parameters of the dynamic forgetting factor; The collected vehicle data is substituted into the vehicle mass estimation model with the least squares method with forgetting factor to estimate the vehicle mass.

2. The method according to claim 1, characterized in that The vehicle data includes tire radius, air density, frontal area, drag coefficient, rolling resistance coefficient; vehicle speed, slope, acceleration, and total torque.

3. The method according to claim 1, characterized in that The vehicle longitudinal dynamics model is constructed as: Where, T dr The total torque generated by the driver's control of the accelerator and brake pedals, in Nm; r tire is the tire radius, in m; ρ is the air density, which is 1.29 kg / m^3; A is the windward area, in m^2; C W is the drag coefficient; v is the vehicle speed, unit is m / s; ε is the rolling resistance coefficient; m is the vehicle mass, unit is kg; g is the acceleration due to gravity, which is 9.8N / kg; α is the slope, which is 0~1; a is the acceleration, unit is m / s^2.

4. The method according to claim 1, characterized in that: The least squares vehicle mass estimation model with forgetting factor is established as: In the above formula, M(k) represents the estimated mass of the kth step, K(k) represents the gain of the kth step, EF(k) represents the output of the kinetic equation of the kth step, Eα(k) represents the observation of the kth step, k represents the kth step, and M(k-1) represents the mass of the k-1th step.

5. The method according to claim 1, characterized in that: According to the vehicle longitudinal dynamics model and vehicle mass estimation model, the system observable quantity E is determined. α Model: E α =a+ε·g·cos(α)+g·sin(α) According to the vehicle longitudinal dynamics model and the vehicle mass estimation model, the model for determining the system output EF is:

6. The method according to claim 1, characterized in that According to the least squares vehicle mass estimation model, the dynamic forgetting factor is set as follows: the range of the forgetting factor is 0.8 to 1. When the estimation starts, it is set to 0.

8. As the system running time increases, the forgetting factor approaches 1. Based on this setting, the calculation formula of the forgetting factor is as follows: Where t is the time the vehicle stops, A, B, C are unknown parameters; the exhaustive method is used to find the above parameters. The method to judge whether the parameters are reasonable is to observe the difference between the true value and the estimated value. The smaller the difference, the better the three parameters are. Thus, the calculation expression of the dynamic forgetting factor is obtained:

7. The method according to claim 1, characterized in that According to the vehicle longitudinal dynamic model and the least squares vehicle mass estimation model with dynamic forgetting factor, the algorithm module of the vehicle mass real-time estimation system based on MATLAB is determined.

8. The method according to claim 4, characterized in that The calculation method of P(k) and K(k) in the least squares vehicle mass estimation model with forgetting factor is:

9. The method according to claim 1, characterized in that The filtering method is used to reduce the white noise interference in the data to be tested.

10. A vehicle data collection device, characterized in that: It includes a data acquisition module, a calculation module, a display module, and a data transmission module, wherein the data of the data acquisition module includes the vehicle's combined torque, vehicle speed, acceleration, and slope; the calculation module includes; any one of claims 1-9 of a real-time vehicle mass estimation method based on a dynamic forgetting factor; and the display module includes a real-time display of the vehicle's mass estimation result.

Images