Recursive least square vehicle load estimation method and system based on adaptive forgetting factor

Abstract

The invention relates to the technical field of vehicle load estimation, in particular to a recursive least square vehicle load estimation method and system based on an adaptive forgetting factor, and the method comprises the steps: initializing an RLS algorithm, and setting a steady-state forgetting factor; vehicle operation parameters are collected and processed in real time to calculate effective traction force and comprehensive acceleration; in each sampling period, calculating and filtering the information, if the absolute value of the filtered information continuously exceeds a threshold value, judging that the load is suddenly changed, and switching the forgetting factor to a dynamic value for accelerating convergence, otherwise, keeping a steady-state value; smoothing the switching of the forgetting factors; and executing RLS recursive calculation by using the smoothed forgetting factor, and outputting an accurate and stable vehicle load estimation value in real time. The method effectively solves the contradiction between stability and rapidity of a traditional fixed forgetting factor RLS algorithm.

Description

Technical Field

[0001] This application relates to the field of vehicle load estimation technology, specifically to a recursive least squares vehicle load estimation method and system based on an adaptive forgetting factor. Background Technology

[0002] In the field of dynamic estimation of load on commercial vehicles, recursive least squares method is widely used due to its online recursiveness.

[0003] However, existing technologies generally use a fixed forgetting factor λ, which leads to a fundamental contradiction in the performance of the algorithm under different vehicle operating conditions: if a small λ value is set, although the algorithm can quickly respond to real changes in load, it is too sensitive to sensor noise and road interference, causing the re-estimated value to fluctuate drastically under steady-state conditions, resulting in serious instability; conversely, if a large λ value is set, although the algorithm can effectively suppress noise and output smooth and reliable estimates under steady-state conditions, its "memory" of historical data is too strong. Under dynamic conditions such as actual loading and unloading of goods that cause step changes in load, the algorithm is slow to track and takes too long to converge to the true value, which cannot meet the real-time requirements.

[0004] The inherent contradiction between stability and speed makes it difficult for the RLS algorithm based on a fixed forgetting factor to achieve both high accuracy and strong robustness in complex and ever-changing real-world driving environments, thus limiting the overall performance and engineering practical value of the vehicle load estimation system. Summary of the Invention

[0005] To address the technical shortcomings of existing fixed-forgetting-factor RLS algorithms in vehicle load estimation, which cannot simultaneously ensure stability and fast response, this invention provides a recursive least squares vehicle load estimation method and system based on an adaptive forgetting factor.

[0006] In a first aspect, the present invention provides a recursive least squares method for estimating vehicle load based on an adaptive forgetting factor, comprising the following steps: S1. Set the initial estimation parameters for the RLS algorithm and set the forgetting factor for the RLS algorithm. Initialize to the first value, which is used to estimate stability under steady-state conditions; S2. Real-time acquisition of vehicle operating parameters, including engine torque, vehicle speed, vehicle acceleration, and gradient information; based on the vehicle operating parameters, calculation of the effective traction force and comprehensive acceleration at the current sampling moment; S3. For each sampling time, iteratively execute the following steps: S31. Input the effective traction force and combined acceleration as the observation data at the current moment into the RLS algorithm; S32. Calculate the innovation of the RLS algorithm at the current time, wherein the innovation is the residual between the actual value based on the observed data and the predicted value of the RLS algorithm at the current time based on historical data; perform rolling average filtering on the innovation to obtain the filtered innovation; S33. If the absolute value of the filtered information exceeds a preset first threshold within N consecutive sampling periods, it is determined that the vehicle load has changed abruptly, and the current forgetting factor is updated to a second value, which is used to accelerate convergence under dynamic conditions; otherwise, the current forgetting factor is maintained at the first value. S34. The update process of the forgetting factor is smoothed by discrete filtering to generate a smoothed forgetting factor for calculation at the next time step. S35. Using a smoothing forgetting factor, perform the recursive calculation of the RLS algorithm and update its internal state to obtain the estimated load parameters at the current time. S4. Output the latest estimated load parameters as the estimated value of the vehicle's current load.

[0007] During sudden load changes, switching the forgetting factor to an optimized second value accelerates the algorithm's tracking speed of the new load, solving the problem of slow response in traditional methods. In steady-state operation, maintaining the forgetting factor at its first value, the algorithm degenerates into classical least squares, fully utilizing historical data to suppress noise and ensuring high accuracy and smoothness of the estimation results, overcoming the fluctuation defects caused by small forgetting factors. Employing a "dual-mode switching" strategy instead of a complex continuous adjustment strategy results in clear logic, low computational cost, and ease of implementation and calibration in automotive embedded controllers.

[0008] As a further limitation of the technical solution of the present invention, in S2, the current sampling time effective traction The calculation method is as follows: Based on the force transmitted by the clutch to the gearbox shaft Gear ratio Gear transmission efficiency Bridge transmission efficiency and inertia loss Calculate the theoretical wheel-end traction force And deduct the wind resistance calculated based on the current vehicle speed. To obtain effective traction ; wind resistance The calculation formula is:

[0009] Wheel end traction The calculation formula is as follows:

[0010] in, air density, This is the drag coefficient. The vehicle's frontal area. This represents the vehicle speed at the current moment.

[0011] As a further limitation of the technical solution of the present invention, in S2, the current sampling time Comprehensive acceleration The calculation method is as follows: Acceleration measured by the vehicle's longitudinal acceleration sensor Slope acceleration calculated from slope information and the rolling resistance acceleration obtained from the rolling resistance model The synthesis is performed using the following formula: .

[0012] By incorporating gradient acceleration and rolling resistance acceleration, the effects of road slope and pavement conditions on vehicle dynamics are effectively compensated, allowing the observed acceleration to more purely reflect the dynamic response caused by load changes. This enables the load estimation algorithm to maintain stable estimation performance not only on level, smooth roads but also on inclines, declines, and various road conditions.

[0013] As a further limitation of the technical solution of the present invention, in S32, the new information The calculation formula is: ,in, Based on effective traction and combined acceleration The current observed values ​​are obtained based on the load model m = F / a. This is the prediction value of the RLS algorithm for the current time.

[0014] In S32, regarding the new information Perform rolling average filtering to obtain the filtered information. Specifically, it includes the following steps: S321. Set up a first-in-first-out queue of length L in the memory to store the information values ​​of the most recent L sampling periods. ; S322, at the current sampling time Perform the following operations: a) The currently calculated new interest value Stored at the tail of the queue; b) If the number of data items in the queue exceeds the length L, remove the oldest information value from the front of the queue; c) Calculate the arithmetic mean of all the new information values ​​in the current queue, and use this arithmetic mean as the filtered new information at the current time. .

[0015] It can smooth out instantaneous information fluctuations caused by sensor noise and high-frequency road surface disturbances, significantly reducing the probability of falsely triggering load change judgments. The filtered information is more representative of the trend changes in error, making the condition identification in step S33 more accurate and reliable, and avoiding mode switching errors caused by single-point noise.

[0016] As a further limitation of the technical solution of the present invention, in S34, a first-order low-pass digital filter is used to filter the forgetting factor. The update is smoothed to generate a smoothed forgetting factor. The recursive formula is as follows:

[0017] in, The current target forgetting factor value is determined based on the decision result of S33; when it is determined to be a load mutation, The second value is used; when the condition is determined to be steady state, The first value; These are the smoothing filter coefficients; This is the smoothed value of the forgetting factor from the previous sampling time. This is the calculated smoothing forgetting factor used for the next time step RLS recursive calculation.

[0018] This avoids the drastic changes in the internal state of the RLS algorithm that might be caused by a direct jump in the forgetting factor between the first and second values, thus preventing instantaneous oscillations or instability in the estimated output. The process of switching the algorithm's operating mode from stable to rapid response or reverse is gradual and smooth, improving the driver experience and the quality of load signal reception by downstream systems.

[0019] As a further limitation of the technical solution of the present invention, the method also includes a mechanism for returning the forgetting factor from the second value to the first value: When the system enters dynamic mode using the second value due to a sudden change in load, it continuously monitors the filtered information. ;like If the absolute value of the load estimate is lower than the second threshold for M consecutive sampling periods, then the load estimate is determined to have converged to a new steady state. Reset to the first value; wherein the second threshold is less than or equal to the first threshold.

[0020] This enables the system to not only respond quickly when the load increases, but also automatically recover to a high-precision, disturbance-resistant mode after converging to a new steady state, achieving autonomous optimization throughout its entire lifecycle. It avoids the increased steady-state noise sensitivity caused by the system remaining in a fast-response mode for extended periods after load changes, ensuring the most stable estimation is always provided when rapid tracking is not required.

[0021] As a further limitation of the technical solution of the present invention, in S35, the recursive calculation of the RLS algorithm and the updating of its internal state are performed using the smoothed forgetting factor, specifically including the following steps: S351. Based on the error covariance matrix of the previous time step, the current observed data vector, and the smoothed forgetting factor, calculate the gain vector at the current time step. ; S352, using the aforementioned gain vector and current news Update the estimated load parameters The calculation formula is:

[0022] in, This is the estimated load parameter value from the previous moment; S353, using the gain vector The current observed data vector and the smoothed forgetting factor are used to update the error covariance matrix for calculation at the next time step.

[0023] The specific role of the smoothed forgetting factor in gain calculation and covariance update is clarified, making the entire adaptive RLS algorithm complete.

[0024] Secondly, the technical solution of the present invention also provides a recursive least squares vehicle load estimation system based on an adaptive forgetting factor, comprising: The signal acquisition module is used to collect vehicle operating parameters in real time, including engine torque, vehicle speed, vehicle acceleration, and gradient information. The data processing module, connected to the signal acquisition module, is used to calculate the effective traction force and comprehensive acceleration at the current sampling moment based on the acquired vehicle operating parameters. An adaptive estimation and control module, connected to the data processing module, includes the following units configured to perform iterative operations to estimate the vehicle load: An observation input unit is used to receive the effective traction force and combined acceleration as observation data at the current moment; The information processing unit is used to calculate the information of the recursive least squares algorithm at the current time and perform rolling average filtering on the information to obtain the filtered information; the information is the residual between the actual value based on the observation data and the predicted value of the RLS algorithm at the current time based on historical data. The forgetting factor decision unit is used to make dual-mode decisions based on the filtered information: if the absolute value of the filtered information exceeds a preset first threshold within N consecutive sampling periods, it is determined that the vehicle load has changed abruptly, and a second value is output as the current target forgetting factor; otherwise, it is determined that the vehicle is in a steady-state condition, and a first value is output as the current target forgetting factor; the first value is used to estimate stability under steady-state conditions, and the second value is used to accelerate convergence under dynamic conditions. The smoothing filter unit is used to smooth the update process of the current target forgetting factor by using discrete filtering to generate a smoothed forgetting factor for calculation at the next time step. The RLS core computing unit is used to perform recursive calculations of the RLS algorithm and update its internal state using the smoothed forgetting factor to obtain the estimated load parameters at the current moment. The output module, connected to the adaptive estimation and control module, is used to output the latest estimated load parameters as the estimated value of the vehicle's current load.

[0025] As a further limitation of the technical solution of the present invention, the system is integrated into the vehicle's transmission control unit or vehicle controller.

[0026] As can be seen from the above technical solutions, this application has the following advantages: This application sets up an adaptive forgetting factor adjustment mechanism, which dynamically adjusts the forgetting factor according to the working conditions to achieve the optimal balance between stability and speed; the working condition identification based on information monitoring accurately judges the timing of dynamic load changes by monitoring information changes in real time; the rolling average filtering process filters out measurement noise interference, improving the accuracy and reliability of working condition judgment; the discrete filtering smooth switching avoids system oscillations caused by sudden changes in the forgetting factor, ensuring a smooth transition; the dual working mode intelligent switching adaptively switches between stable mode and fast response mode to take into account the needs of different working conditions. Attached Figure Description

[0027] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the method provided in an embodiment of the present invention.

[0029] Figure 2 A block diagram of a system provided in an embodiment of the present invention.

[0030] Figure 3 This is a simulation diagram, in which Figure 3 (a) in the graph is a global performance comparison chart. Figure 3 (b) in the figure is an enlarged view of the dynamic response region (690s-1400s). Detailed Implementation

[0031] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0033] like Figure 1 As shown, this embodiment of the invention provides a recursive least squares vehicle load estimation method based on an adaptive forgetting factor, including the following steps: S1. Set the initial estimation parameters for the RLS algorithm and set the forgetting factor for the RLS algorithm. Initialize to the first value, which is used to estimate stability under steady-state conditions; S2. Real-time acquisition of vehicle operating parameters, including engine torque, vehicle speed, vehicle acceleration, and gradient information; calculation of the current sampling time based on the vehicle operating parameters. effective traction With comprehensive acceleration ; S3, for each sampling time The following steps are executed iteratively: S31, Effective traction force With comprehensive acceleration The observation data at the current moment is input into the RLS algorithm; S32. Calculate the current information of the RLS algorithm. The new information The residual between the actual value based on observed data and the predicted value for the current moment by the RLS algorithm based on historical data; for the information... Perform rolling average filtering to obtain the filtered information. ; S33, if the filtered information If the absolute value of the forgetting factor exceeds a preset first threshold within N consecutive sampling periods, it is determined that a sudden change has occurred in the vehicle's load, and the current forgetting factor is adjusted accordingly. Update to the second value, which is used to accelerate convergence under dynamic conditions; otherwise, maintain the current forgetting factor. The first value; S34. The forgetting factor is processed using discrete filtering. The update process is smoothed to generate a smoothed forgetting factor for calculation in the next time step. ; S35. Using a smoothing forgetting factor The RLS algorithm is recursively calculated and its internal state is updated to obtain the estimated load parameters at the current time. ; S4. Output the latest estimated load parameters as the estimated value of the vehicle's current load.

[0034] By calculating the vehicle's effective traction force F and comprehensive acceleration a, and using the recursive least squares (RLS) method with the formula m=F / a to identify load parameters, a dual-mode adaptive forgetting factor adjustment mechanism is proposed. Under steady-state conditions, the forgetting factor λ is maintained at 1 to ensure estimation stability. When a sudden change in load is detected through innovation monitoring, λ is adaptively switched to an optimized fast response value, preferably 0.95, to accelerate the tracking of the actual load. The forgetting factor λ of the RLS algorithm is adaptively adjusted: the innovation ε is monitored in real time. After rolling average filtering, if its absolute value exceeds a preset threshold within N consecutive sampling periods, λ is switched from the default value of 1 to a fast response value to improve dynamic response speed; otherwise, λ=1 is maintained to ensure steady-state stability. Discrete filtering is used to achieve a smooth transition during the switching process. This dual-mode switching strategy, through continuous adjustment of the forgetting factor, effectively solves the engineering problem that the fixed forgetting factor RLS cannot simultaneously achieve stability and fast convergence when the vehicle is powered on and the load is increased or decreased. This invention effectively solves the problem that the fixed forgetting factor RLS cannot balance stability and speed, and improves the accuracy and adaptability of load estimation.

[0035] In some embodiments, in S2, the current sampling time effective traction The calculation method is as follows: Based on the force transmitted by the clutch to the gearbox shaft Gear ratio Gear transmission efficiency Bridge transmission efficiency and inertia loss Calculate the theoretical wheel-end traction force And deduct the wind resistance calculated based on the current vehicle speed. To obtain effective traction ; wind resistance The calculation formula is:

[0036] Wheel end traction The calculation formula is as follows:

[0037] in, air density, This is the drag coefficient. The vehicle's frontal area. This represents the vehicle speed at the current moment.

[0038] In some embodiments, in S2, the current sampling time Comprehensive acceleration The calculation method is as follows: Acceleration measured by the vehicle's longitudinal acceleration sensor Slope acceleration calculated from slope information and the rolling resistance acceleration obtained from the rolling resistance model The synthesis is performed using the following formula: .

[0039] In this embodiment of the invention, the innovation of the RLS algorithm is monitored in real time and obtained by rolling average filtering. When the absolute value of λ exceeds a preset threshold N times consecutively, it is determined that a sudden change in vehicle load has occurred, and λ is switched from the default value of 1 in stable mode to the optimized value in fast response mode; otherwise, the stable mode of λ=1 is maintained. Discrete filtering is used in the switching process to ensure a smooth transition. The optimized value is a value that achieves the best balance between dynamic convergence speed and steady-state estimation smoothness through simulation and real vehicle testing. In S32, the innovation... The calculation formula is: ,in, Based on effective traction and combined acceleration The current observed values ​​are obtained based on the load model m = F / a. This is the prediction value of the RLS algorithm for the current time.

[0040] Under stable operating conditions, λ=1 ensures the algorithm has strong anti-interference capabilities and smooth, reliable output results. When the load changes abruptly, λ switches to the optimized value, allowing the algorithm to quickly forget historical data and rapidly track real load changes. It specifically solves the engineering problem of slow convergence or large fluctuations in traditional fixed-forgetting-factor RLS when the vehicle is loaded or unloaded during operation without power-on / off reset. Strong robustness: The adaptive mechanism effectively distinguishes between real load changes and measurement noise, adapting to complex operating conditions. The dual-mode switching strategy has low computational complexity and clear logic, far superior to complex methods requiring continuous online calculation of λ, and is easily deployed in existing vehicle control units.

[0041] In some embodiments, in S32, the new information... Perform rolling average filtering to obtain the filtered information. Specifically, it includes the following steps: S321. Set up a first-in-first-out queue of length L in the memory to store the information values ​​of the most recent L sampling periods. ; S322, at the current sampling time Perform the following operations: a) The currently calculated new interest value Stored at the tail of the queue; b) If the number of data items in the queue exceeds the length L, remove the oldest information value from the front of the queue; c) Calculate the arithmetic mean of all the new information values ​​in the current queue, and use this arithmetic mean as the filtered new information at the current time. .

[0042] In some embodiments, in S34, a first-order low-pass digital filter is used to filter the forgetting factor. The update is smoothed to generate a smoothed forgetting factor. The recursive formula is as follows:

[0043] in, The current target forgetting factor value is determined based on the decision result of S33; when it is determined to be a load mutation, The second value is used; when the condition is determined to be steady state, The first value; These are the smoothing filter coefficients; This is the smoothed value of the forgetting factor from the previous sampling time. This is the calculated smoothing forgetting factor used for the next time step RLS recursive calculation.

[0044] A first-order low-pass digital filter has the following transfer function:

[0045] in, The value ranges from 0.8 to 0.95.

[0046] In some embodiments, the method further includes a mechanism for returning the forgetting factor from a second value to a first value: When the system enters dynamic mode using the second value due to a sudden change in load, it continuously monitors the filtered information. ;like If the absolute value of the load estimate is lower than the second threshold for M consecutive sampling periods, then the load estimate is determined to have converged to a new steady state. Reset to the first value; wherein the second threshold is less than or equal to the first threshold.

[0047] The determination of the second value is based on the balance optimization of simulation and real vehicle testing: under the condition of sudden load change, this value should enable the system to converge the load estimation error to within 5% within 20 seconds, while maintaining smooth output in steady state; preferably, the second value is 0.95.

[0048] In some embodiments, in S35, a smoothed forgetting factor is used. Performing the recursive computation of the RLS algorithm and updating its internal state specifically includes the following steps: S351, Based on the error covariance matrix of the previous time step Current observation data vector and the smoothed forgetting factor Calculate the gain vector at the current time. The calculation formula is:

[0049] Among them, the observation data vector Based on the current comprehensive acceleration Construction, in the model In the process of identification, The corresponding parameter to be estimated is the load capacity. ; S352, using the aforementioned gain vector and current news Update the estimated load parameters The calculation formula is:

[0050] in, This is the estimated load parameter value from the previous moment; S353, using the gain vector Current observation data vector and the smoothed forgetting factor Update the error covariance matrix It is used for calculation in the next time step, and the calculation formula is:

[0051] in, It is an identity matrix.

[0052] like Figure 2 As shown, this embodiment of the invention also provides a recursive least squares vehicle load estimation system based on an adaptive forgetting factor, comprising: The signal acquisition module is used to collect vehicle operating parameters in real time, including engine torque, vehicle speed, vehicle acceleration, and gradient information. The data processing module, connected to the signal acquisition module, is used to calculate the effective traction force and comprehensive acceleration at the current sampling moment based on the acquired vehicle operating parameters. An adaptive estimation and control module, connected to the data processing module, includes the following units configured to perform iterative operations to estimate the vehicle load: An observation input unit is used to receive the effective traction force and combined acceleration as observation data at the current moment; The information processing unit is used to calculate the information of the recursive least squares algorithm at the current time and perform rolling average filtering on the information to obtain the filtered information; the information is the residual between the actual value based on the observation data and the predicted value of the RLS algorithm at the current time based on historical data. The forgetting factor decision unit is used to make dual-mode decisions based on the filtered information: if the absolute value of the filtered information exceeds a preset first threshold within N consecutive sampling periods, it is determined that the vehicle load has changed abruptly, and a second value is output as the current target forgetting factor; otherwise, it is determined that the vehicle is in a steady-state condition, and a first value is output as the current target forgetting factor; the first value is used to estimate stability under steady-state conditions, and the second value is used to accelerate convergence under dynamic conditions. The smoothing filter unit is used to smooth the update process of the current target forgetting factor by using discrete filtering to generate a smoothed forgetting factor for calculation at the next time step. The RLS core computing unit is used to perform recursive calculations of the RLS algorithm and update its internal state using the smoothed forgetting factor to obtain the estimated load parameters at the current moment. The output module, connected to the adaptive estimation and control module, is used to output the latest estimated load parameters as the estimated value of the vehicle's current load.

[0053] In this embodiment of the invention, the data processing module includes: The traction calculation submodule is used to calculate the theoretical wheel-end traction force based on engine torque, transmission system speed ratio, transmission efficiency, and wheel rolling radius, and subtract the wind resistance calculated based on the current vehicle speed to obtain the effective traction force; the wind resistance... The calculation formula is:

[0054] in, air density, This is the drag coefficient. The vehicle's frontal area. This represents the vehicle speed at the current moment.

[0055] The acceleration synthesis submodule is used to process the acceleration measured by the vehicle's longitudinal acceleration sensor. Slope acceleration calculated from slope information and the rolling resistance acceleration obtained from the rolling resistance model The synthesis is performed using the following formula: .

[0056] The forgetting factor decision unit also includes a return control submodule, which is used to realize the return of the forgetting factor from the second value to the first value; The return control submodule is configured to: when the system enters the dynamic mode using the second value due to a sudden change in load, continuously monitor the filtered information; if the absolute value of the filtered information is lower than the second threshold within M consecutive sampling periods, it is determined that the load estimation has converged to a new steady state, and the forgetting factor decision unit is controlled to output the first value as the current target forgetting factor; wherein, the second threshold is less than or equal to the first threshold.

[0057] In this embodiment of the invention, the RLS core computing unit is configured to perform the following operations: Based on the error covariance matrix of the previous time step, the current observed data vector, and the smoothed forgetting factor, calculate the gain vector at the current time step. ; Using the gain vector and current news Update the estimated load parameters ; Using the gain vector The current observed data vector and the smoothed forgetting factor are used to update the error covariance matrix for calculation at the next time step.

[0058] In this embodiment of the invention, the system is integrated into the vehicle's transmission control unit or vehicle controller. It receives signals such as engine torque, engine speed, and vehicle speed via the vehicle's CAN bus, and simultaneously connects to a three-axis accelerometer to acquire vehicle motion state information. The system core employs a vehicle-grade microprocessor with sufficient computing power to execute the RLS algorithm in real time. The adaptive forgetting factor adjustment process is implemented through embedded software, specifically including: Initialize the RLS parameters, setting the initial value of λ to 1; Real-time reading of sensor data to calculate effective traction force F and overall acceleration a; Perform RLS calculation and monitor the new information, and determine whether to trigger λ adjustment after rolling average filtering; Discrete filtering is used to achieve smooth switching of λ, ensuring system stability.

[0059] like Figure 3 As shown, simulation verification was performed using data collected from a real vehicle. The operating conditions were set as follows: the vehicle ran fully loaded for 690 seconds, then stopped, underwent a coupling / uncoupling operation, and continued to run empty. The comparison images are presented using both a global view and a magnified view of specific areas. Figure 3 In the global graph of (a), the three curves are represented as follows: The black solid line represents the calculated load value when λ=1 is fixed, which shows good stability but slow response when the load changes. Light gray solid line: The adaptive λ method proposed in this invention maintains excellent performance across the entire operating range; The dotted line represents the calculated load value when λ=0.95 is fixed. It responds quickly to load changes but has large steady-state fluctuations.

[0060] Figure 3 The enlarged view in (b) highlights the dynamic response process after the 690s detachment operation, showing: At point A, the method with λ=0.95 provides a rapid response to load changes; At point B, the adaptive λ method achieves fast tracking; At point C, the method with λ=1 slowly converges to a load that has a certain difference from the actual load.

[0061] Comparative analysis shows that the adaptive λ method proposed in this invention maintains the stability of the λ=1 method while possessing a fast response capability comparable to the λ=0.95 method, effectively solving the inherent contradiction of the fixed forgetting factor RLS algorithm.

[0062] An adaptive switching mechanism is employed, which switches from a default stable mode to a fast response mode when innovation is continuously abnormal, enabling rapid tracking of load fluctuations. This application adopts a dual-mode switching strategy based on explicit criteria. This strategy is simpler, more reliable, and easier to implement in vehicle controllers. The forgetting factor in fast response mode is selected as λ=0.95, which is based on a comprehensive result of theoretical analysis, simulation comparison, and real-vehicle verification. Engineering problem background: If the vehicle is loaded or unloaded during operation without power-on / power-off reset, the traditional fixed λ RLS faces a dilemma: a small λ (such as 0.9) is needed for fast convergence, but this will lead to large steady-state fluctuations; a large λ (such as 0.995) is needed to maintain steady-state smoothness, but this will lead to slow tracking speed.

[0063] Simulation comparison and optimization: Through simulation tests comparing λ=0.90, 0.95, 0.98, it was found that λ=0.95 achieved the best balance between convergence speed and steady-state estimation smoothness after load abrupt change.

[0064] Real vehicle testing verification: Real vehicle road tests show that under standard detachment and reattachment conditions, when the system switches to dynamic mode with λ=0.95, it can converge the load estimation error to within 5% in about 20 seconds when the learning conditions are met, and the process is smooth.

[0065] In practical applications, the following calibration ranges are preferred for key parameters: Forgetting factor default value (stable mode): λ=1; Forgetting factor adjustment value (fast response mode): λ=0.95 (its preferred range is 0.93~0.97); Number of consecutive judgment cycles: N=10; Discrete filter coefficients: α = 0.9; New information threshold: calibrated based on specific vehicle models and sensor accuracy.

[0066] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium such as a USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, or other media capable of storing program code. It includes several instructions to cause a computer terminal (which may be a personal computer, server, or a second terminal, network terminal, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0067] In the embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0068] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0069] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0070] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A recursive least squares method for estimating vehicle load based on an adaptive forgetting factor, characterized in that, Includes the following steps: S1. Set the initial estimation parameters for the RLS algorithm and set the forgetting factor for the RLS algorithm. Initialize to the first value, which is used to estimate stability under steady-state conditions; S2. Real-time acquisition of vehicle operating parameters, including engine torque, vehicle speed, vehicle acceleration, and gradient information; based on the vehicle operating parameters, calculation of the effective traction force and comprehensive acceleration at the current sampling moment; S3. For each sampling time, iteratively execute the following steps: S31. Input the effective traction force and combined acceleration as the observation data at the current moment into the RLS algorithm; S32. Calculate the information of the RLS algorithm at the current time, wherein the information is the residual between the actual value based on the observed data and the predicted value of the RLS algorithm at the current time based on historical data; The new information is subjected to rolling average filtering to obtain filtered new information; S33. If the absolute value of the filtered information exceeds a preset first threshold within N consecutive sampling periods, it is determined that the vehicle load has changed abruptly, and the current forgetting factor is updated to a second value, which is used to accelerate convergence under dynamic conditions; otherwise, the current forgetting factor is maintained at the first value. S34. The update process of the forgetting factor is smoothed by discrete filtering to generate a smoothed forgetting factor for calculation at the next time step. S35. Using a smoothing forgetting factor, perform the recursive calculation of the RLS algorithm and update its internal state to obtain the estimated load parameters at the current time. S4. Output the latest estimated load parameters as the estimated value of the vehicle's current load.

2. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, In S2, the current sampling time effective traction The calculation method is as follows: Based on the force transmitted by the clutch to the gearbox shaft Gear ratio Gear transmission efficiency Bridge transmission efficiency and inertia loss Calculate the theoretical wheel-end traction force And deduct the wind resistance calculated based on the current vehicle speed. To obtain effective traction ; wind resistance The calculation formula is: Wheel end traction The calculation formula is as follows: in, air density, This is the drag coefficient. The vehicle's frontal area. This represents the vehicle speed at the current moment.

3. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, In S2, the current sampling time Comprehensive acceleration The calculation method is as follows: Acceleration measured by the vehicle's longitudinal acceleration sensor Slope acceleration calculated from slope information and the rolling resistance acceleration obtained from the rolling resistance model The synthesis is performed using the following formula: .

4. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, In S32, new information The calculation formula is: ,in, Based on effective traction and combined acceleration The current observed values ​​are obtained based on the load model m = F / a. This is the prediction value of the RLS algorithm for the current time.

5. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 4, characterized in that, In S32, regarding the new information Perform rolling average filtering to obtain the filtered information. Specifically, it includes the following steps: S321. Set up a first-in-first-out queue of length L in the memory to store the information values ​​of the most recent L sampling periods. ; S322, at the current sampling time Perform the following operations: a) The currently calculated new interest value Stored at the tail of the queue; b) If the number of data items in the queue exceeds the length L, remove the oldest information value from the front of the queue; c) Calculate the arithmetic mean of all the new information values ​​in the current queue, and use this arithmetic mean as the filtered new information at the current time. .

6. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, In S34, a first-order low-pass digital filter is used to filter the forgetting factor. The update is smoothed to generate a smoothed forgetting factor. The recursive formula is as follows: in, The current target forgetting factor value is determined based on the decision result of S33; when it is determined to be a load mutation, The second value is used; when the condition is determined to be steady state, The first value; These are the smoothing filter coefficients; This is the smoothed value of the forgetting factor from the previous sampling time. This is the calculated smoothing forgetting factor used for the next time step RLS recursive calculation.

7. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, The method also includes a mechanism for returning the forgetting factor from the second value to the first value: When the system enters dynamic mode using the second value due to a sudden change in load, it continuously monitors the filtered information. ;like If the absolute value of the load estimate is lower than the second threshold for M consecutive sampling periods, then the load estimate is determined to have converged to a new steady state. Reset to the first value; wherein the second threshold is less than or equal to the first threshold.

8. The recursive least squares vehicle load estimation method based on adaptive forgetting factor according to claim 1, characterized in that, In S35, the RLS algorithm is recursively calculated and its internal state is updated using a smoothed forgetting factor. This includes the following steps: S351. Based on the error covariance matrix of the previous time step, the current observed data vector, and the smoothed forgetting factor, calculate the gain vector at the current time step. ; S352, using the aforementioned gain vector and current news Update the estimated load parameters The calculation formula is: in, This is the estimated load parameter value from the previous moment; S353, using the gain vector The current observed data vector and the smoothed forgetting factor are used to update the error covariance matrix for calculation at the next time step.

9. A recursive least squares vehicle load estimation system based on an adaptive forgetting factor, characterized in that, include: The signal acquisition module is used to collect vehicle operating parameters in real time, including engine torque, vehicle speed, vehicle acceleration, and gradient information. The data processing module, connected to the signal acquisition module, is used to calculate the effective traction force and comprehensive acceleration at the current sampling moment based on the acquired vehicle operating parameters. An adaptive estimation and control module, connected to the data processing module, includes the following units configured to perform iterative operations to estimate the vehicle load: An observation input unit is used to receive the effective traction force and combined acceleration as observation data at the current moment; The information processing unit is used to calculate the information of the recursive least squares algorithm at the current time, and to perform rolling average filtering on the information to obtain the filtered information. The information is the residual between the actual value based on the observed data and the predicted value of the current moment by the RLS algorithm based on historical data; The forgetting factor decision unit is used to make dual-mode decisions based on the filtered information: if the absolute value of the filtered information exceeds a preset first threshold within N consecutive sampling periods, it is determined that the vehicle load has changed abruptly, and the second value is output as the current target forgetting factor; otherwise, it is determined that the vehicle is in a steady state, and the first value is output as the current target forgetting factor. The first value is used to estimate stability under steady-state conditions, and the second value is used to accelerate convergence under dynamic conditions. The smoothing filter unit is used to smooth the update process of the current target forgetting factor by using discrete filtering to generate a smoothed forgetting factor for calculation at the next time step. The RLS core computing unit is used to perform recursive calculations of the RLS algorithm and update its internal state using the smoothed forgetting factor to obtain the estimated load parameters at the current moment. The output module, connected to the adaptive estimation and control module, is used to output the latest estimated load parameters as the estimated value of the vehicle's current load.

10. The recursive least squares vehicle load estimation system based on adaptive forgetting factor according to claim 9, characterized in that, The system is integrated into the vehicle's transmission control unit or vehicle controller.

Images