A method and system for online identification and compensation of response hysteresis in wire-controlled brake actuators
By decomposing the hysteresis of the brake-by-wire actuator into multiple physical components and compensating for it under online identification and safety constraints, the problem of time-varying hysteresis characteristics of the brake-by-wire actuator is solved, and high-precision and safe braking response under different operating conditions is achieved.
Patent Information
- Application Number
- CN202611444331.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-17
- Publication Date
- 2026-11-17
AI Technical Summary
The response hysteresis characteristics of brake-by-wire actuators change dynamically with braking direction, temperature, power supply voltage, braking pressure and wear condition. Existing compensation strategies based on fixed time delay parameters have reduced compensation accuracy after changes in operating conditions or long-term use, which cannot meet the requirements of intelligent driving for consistent braking response and safety.
The total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis. Direction-dependent hysteresis parameter sets are established for each, and online updates and limited compensation of the hysteresis parameters are achieved through preset excitation conditions, online identification of residuals, and constraints on the physical boundaries of the parameters.
It significantly improves the accuracy and stability of hysteresis compensation, ensuring high-precision braking response under different operating conditions and meeting the safety and consistency requirements of intelligent driving functions.
Smart Images

Figure CN123104585A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of brake-by-wire technology, and in particular to a method and system for online identification and compensation of response hysteresis in brake-by-wire actuators. Background Technology
[0002] Brake-by-wire systems replace traditional hydraulic or mechanical transmissions with electronic control signals, generating frictional braking force through actuators. This is an important development direction for braking systems in intelligent driving and new energy vehicles. The brake-by-wire actuator receives commands for target pressure, target clamping force, or target braking torque, and establishes braking force through a motor, transmission mechanism, and hydraulic or mechanical path. Its response quality directly determines the vehicle's braking distance and stability, and is also the basis for the reliable triggering of functions such as automatic emergency braking and adaptive cruise control.
[0003] The response process of a brake-by-wire actuator is subject to various hysteresis factors: transmission delay of commands in the communication and scheduling stages, motor current settling time, idle travel and contact gaps in the mechanical transmission chain, propagation and pressure build-up time in the hydraulic circuit, and loading / unloading hysteresis introduced by friction pairs and elastic elements. These factors cause the actuator output to exhibit characteristics of time delay, dead zone, hysteresis, and dynamic gain drift in response to commands. To improve response speed, existing technologies typically employ a fixed time delay model in conjunction with a Smith predictor, or design a predictive compensation control strategy based on a given system time delay. These schemes equate hysteresis to a fixed time delay parameter, achieving certain effects under specific operating conditions and within a specific lifespan.
[0004] However, in practical applications, the response hysteresis of brake-by-wire actuators is not a fixed value, but dynamically changes with braking direction, temperature, power supply voltage, braking pressure or clamping force, and wear condition. Existing compensation strategies based on fixed time delay parameters can work under specific calibration conditions, but the compensation accuracy decreases significantly after changes in operating conditions or long-term use, and cannot meet the requirements of intelligent driving for consistent braking response and safety.
[0005] Therefore, an online identification and compensation scheme is needed that can adapt to the time-varying characteristics of actuator hysteresis, and continuously maintain high-precision braking response while ensuring functional safety. Summary of the Invention
[0006] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing an online identification and compensation method and system for response hysteresis in line-controlled brake actuators.
[0007] The objective of this invention can be achieved through the following technical solutions:
[0008] Firstly, to address the aforementioned technical problems, this invention provides a method for online identification and compensation of response hysteresis in a line-controlled brake actuator, comprising:
[0009] Acquire the response timing of braking commands, motor current, actuator displacement, and braking output parameters;
[0010] Based on the aforementioned response timing, the total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis.
[0011] Based on the electronic control delay, the mechanical idle travel hysteresis, and the braking force establishment hysteresis, a set of direction-related hysteresis parameters is established according to the braking force establishment phase, the braking force steady-state maintenance phase, and the braking force release phase.
[0012] When the braking event meets the preset excitation conditions, the hysteresis parameter set is updated online to obtain candidate hysteresis parameters and online identification residuals.
[0013] The confidence level of the candidate hysteresis parameters is determined based on the online identification of residuals and parameter physical boundaries.
[0014] When the confidence level is not lower than the preset confidence level threshold, the candidate hysteresis parameter is written into the working parameter set at a limited update rate;
[0015] Based on the set of working parameters, generate the delay prediction compensation amount and the hysteresis inverse model compensation amount;
[0016] Based on the braking command, a basic braking control quantity is generated;
[0017] The time delay prediction compensation and the hysteresis inverse model compensation are fused with the basic braking control quantity within a preset safety constraint envelope to obtain the target control quantity, which is then output to the brake-by-wire actuator.
[0018] Secondly, in order to solve the above-mentioned technical problems, the present invention provides an online identification and compensation system for response hysteresis of a brake-by-wire actuator, comprising:
[0019] The timing acquisition module is used to acquire the response timing of braking commands, motor current, actuator displacement, and braking output parameters.
[0020] The hysteresis decomposition module is used to decompose the total response hysteresis of the brake-by-wire actuator into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis based on the response timing.
[0021] The parameter establishment module is used to establish a set of direction-related hysteresis parameters according to the electronic control delay, the mechanical idle travel hysteresis and the braking force establishment hysteresis, and according to the braking force establishment phase, the braking force steady-state maintenance phase and the braking force release phase.
[0022] The parameter update module is used to update the hysteresis parameter set online when the braking event meets the preset excitation conditions, so as to obtain candidate hysteresis parameters and online identification residuals;
[0023] A credibility determination module is used to determine the credibility of the candidate hysteresis parameters based on the online identified residuals and parameter physical boundaries.
[0024] The parameter writing module is used to write the candidate hysteresis parameter into the working parameter set at a limited update rate when the confidence level is not lower than a preset confidence level threshold.
[0025] The compensation quantity generation module is used to generate the delay prediction compensation quantity and the hysteresis inverse model compensation quantity based on the working parameter set.
[0026] A basic control quantity generation module is used to generate basic braking control quantities based on the braking command;
[0027] The control quantity fusion module is used to fuse the time delay prediction compensation quantity and the hysteresis inverse model compensation quantity with the basic braking control quantity within a preset safety constraint envelope to obtain the target control quantity, and output it to the line control brake actuator.
[0028] Compared with the prior art, this application has the following advantages:
[0029] The total response hysteresis of the brake-by-wire actuator is decomposed into electronic control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis. Direction-dependent hysteresis parameter sets are established for each component, so that hysteresis compensation is no longer limited to a fixed delay parameter, but can be identified and compensated online for different physical sources and different braking directions. This significantly improves the ability of the parameter model to describe the actual hysteresis characteristics of the actuator.
[0030] By setting up incentive conditions, identifying residuals online, and imposing layer-by-layer constraints on parameter physical boundaries and credibility thresholds, it is ensured that only candidate parameters that meet the requirements of sufficient incentive, residual convergence, parameters within a reasonable physical range, and credibility can enter the working parameter set at a limited update rate. This fundamentally avoids parameter drift under low incentive conditions and parameter contamination under abnormal conditions.
[0031] By integrating the time delay prediction compensation and the hysteresis inverse model compensation with the basic braking control quantity within a preset safety constraint envelope, the system can shorten the braking response time and reduce transient tracking error by utilizing online identification information, while strictly limiting the adaptive compensation quantity within the safety boundary. Online learning is only used to improve performance and will not become a necessary condition for obtaining basic braking force, thus meeting the functional safety requirements of the brake-by-wire system.
[0032] The method and system of this invention are applicable to hydraulic brake-by-wire actuators, electromechanical brake-by-wire actuators, and hybrid brake-by-wire architectures. They can maintain adaptive compensation accuracy under long-term operating conditions such as brake pad wear, temperature changes, and power supply fluctuations. They have important practical application value for intelligent driving functions with high response speed requirements, such as automatic emergency braking and adaptive cruise control. Attached Figure Description
[0033] Figure 1 This is a flowchart illustrating the online identification and compensation method for response hysteresis of the linear brake actuator of the present invention.
[0034] Figure 2 This is a block diagram of the online identification and compensation system for response hysteresis of the linear brake actuator of the present invention.
[0035] Figure 3 A multi-node response timing diagram of braking command, motor current, actuator displacement and braking output characteristics provided for this invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0037] In actual operation, the response hysteresis of a brake-by-wire actuator is formed by the superposition of multiple physical sources, and these sources dynamically change with factors such as braking direction, temperature, power supply voltage, pressure or clamping force, and wear condition. If all hysteresis is simply equated to a fixed time delay, the compensation accuracy will inevitably decrease after the operating conditions change. To address the above problems, this invention provides an online identification and compensation method for the response hysteresis of a brake-by-wire actuator. This method starts from the multi-node response timing, decomposes the total response hysteresis into electrical control time delay, mechanical idle travel hysteresis, and braking force build-up hysteresis, and realizes online updating and constrained compensation of hysteresis parameters under the protection of credibility gating and safety constraint envelope.
[0038] Reference Figure 1 This paper provides a method for online identification and compensation of response hysteresis in wire-controlled brake actuators, comprising the following steps:
[0039] Step S1: Obtain the response timing of braking command, motor current, actuator displacement, and braking output characteristics;
[0040] Step S2: Based on the response timing, the total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis.
[0041] Step S3: Based on the electronic control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis, establish direction-related hysteresis parameter sets according to the braking force establishment phase, braking force steady-state maintenance phase, and braking force release phase;
[0042] Step S4: When the braking event meets the preset excitation conditions, update the hysteresis parameter set online to obtain candidate hysteresis parameters and online identification residuals;
[0043] Step S5: Determine the credibility of candidate hysteresis parameters based on online identification of residuals and parameter physical boundaries;
[0044] Step S6: When the confidence level is not lower than the preset confidence level threshold, write the candidate hysteresis parameters into the working parameter set at a limited update rate.
[0045] Step S7: Based on the working parameter set, generate the delay prediction compensation amount and the hysteresis inverse model compensation amount;
[0046] Step S8: Generate basic braking control quantity based on braking command;
[0047] Step S9: The time delay prediction compensation and the hysteresis inverse model compensation are fused with the basic braking control quantity within the preset safety constraint envelope to obtain the target control quantity, which is then output to the brake-by-wire actuator.
[0048] The method provided in this application solves the problem that traditional fixed-delay models cannot describe the internal structure of hysteresis by establishing a multi-node response timing sequence within the braking event and decomposing the total response hysteresis into three physical components: electronic control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis. Furthermore, by establishing direction-dependent hysteresis parameter sets according to the braking force establishment phase, braking force steady-state holding phase, and braking force release phase, the asymmetric hysteresis behavior of the same actuator in the loading and unloading directions is independently characterized. The reliability problem in the online updating process of time-varying hysteresis parameters is solved through mechanisms such as preset excitation condition constraints, online identification of residuals and dual evaluation of parameter physical boundaries for reliability, and writing to the working parameter set at a limited update rate. Finally, the compensation quantity and the basic braking control quantity are fused within a preset safety constraint envelope, which improves transient response performance and ensures basic closed-loop stability and functional safety.
[0049] The steps described above are interconnected. Multi-node time-series acquisition provides the data foundation for hysteresis decomposition, hysteresis decomposition provides the structural basis for parameter set establishment, parameter set update provides the parameter source for compensation quantity generation, credibility gating ensures the quality of parameters written into the working parameter set, and constrained fusion ensures that the compensation quantity will not exceed the safety boundary. Thus, overall, a safe, reliable and long-term sustainable hysteresis adaptive compensation is achieved.
[0050] In this invention, the braking output characteristic quantity is used to characterize the braking state output by the brake-by-wire actuator, and its specific physical form depends on the actuator type. For electro-hydraulic brake-by-wire actuators, the braking output characteristic quantity can be braking pressure; for electromechanical brake-by-wire actuators, the braking output characteristic quantity can be clamping force; in specific implementations, the braking output characteristic quantity can also be equivalent braking torque, which is not limited in this application embodiment.
[0051] A braking event refers to the complete process from issuing a braking command to the recovery of the braking output characteristic quantity to a preset range, including the braking force establishment phase, the braking force steady-state maintenance phase, and the braking force release phase. Preset excitation conditions refer to the set of conditions used to determine whether the current braking event has sufficient dynamic excitation to support parameter identification. The direction-dependent hysteresis parameter set refers to one or more sets of hysteresis parameters set for the braking force establishment direction and the braking force release direction, respectively. The working parameter set is the set of parameters actually used for calculating the compensation quantity at the current moment, distinct from the candidate hysteresis parameters generated during online identification. The preset safety constraint envelope is a pre-defined actuator safety operating boundary used to limit the amplitude and rate of change of the compensation quantity and the target control quantity. Credibility is a quantitative indicator used to characterize the reliability of candidate hysteresis parameters; the higher the value, the more reliable the candidate hysteresis parameter and the more suitable it is to be included in the working parameter set.
[0052] In step S1, the response timing of braking command, motor current, actuator displacement and braking output characteristics is obtained.
[0053] The braking command is the desired braking request input from the upper layer. The motor current reflects the establishment of the motor's electromagnetic torque, the actuator displacement reflects the motion state of the mechanical transmission chain, and the braking output characteristic reflects the actual braking effect.
[0054] These four types of signals correspond to the control command layer, electrical response layer, mechanical motion layer, and braking output layer, respectively, forming a complete hysteresis transmission chain. The reason why it is necessary to obtain the response timing of the above four nodes at the same time is that only the total response hysteresis can be observed from the two endpoints of braking command and braking output, but the physical source of the hysteresis cannot be identified. Only by introducing the state information of intermediate nodes can the total delay be broken down into specific links. If only the endpoint signals are relied upon, the delay caused by mechanical idle travel will be misjudged as pure time delay, leading to errors in the calculation of subsequent compensation amounts, which may result in problems such as excessive early compensation or incorrect compensation direction.
[0055] The response timing obtained in step S1 is output to step S2. Step S2 extracts the time features of each node from the timing for hysteresis decomposition. At the same time, the response timing is also used for online update in step S4 and basic control quantity generation in step S8, providing raw data support for the entire compensation process.
[0056] In specific implementation, the braking command can be a target pressure value, a target clamping force value, or a target braking torque value; the motor current can be at least one phase current value or a bus current value among the three-phase currents; the actuator displacement can be the motor rotor angle, the lead screw displacement, or the caliper piston displacement; the braking output characteristic quantity can be the braking pressure or the clamping force; the sampling period of each signal can be determined according to the actuator type and the controller computing power, and the specific values are not limited in this application embodiment.
[0057] In practice, to ensure the time consistency of multi-node response timing, it is necessary to synchronize and align the braking command, motor current, actuator displacement, and braking output characteristics. If the sensor signals are collected through different channels and there is no unified timestamp, the time deviation between channels will be superimposed on the actual actuator hysteresis, causing serious distortion of the subsequent time delay estimation results. Therefore, in this embodiment, the same hardware time base is used to synchronously sample the signals of each channel, or the messages using different clock domains are timestamped and interpolated for alignment.
[0058] In one specific implementation, the timing acquisition module 110 acquires the above four types of signals at a frequency of 1kHz and records the hardware timestamp of each sampled data. This sampling frequency is higher than the time scale of the electronic control delay to be identified, and can capture the response start point at the millisecond level, avoiding confusion between hysteresis and first-order dynamics due to excessively long sampling periods. Compared with the scheme that only acquires the two endpoints of the command and output, this embodiment introduces an intermediate node and performs time synchronization, which enables the independent identification of the three components of electronic control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis, creating conditions for targeted sub-item compensation.
[0059] In step S2, based on the response timing, the total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis.
[0060] Total response hysteresis is the overall time delay between the issuance of a braking command and the achievement of the expected braking output response; electronic control delay refers to the time between the issuance of a braking command and the start of change in motor current, including the combined delay of command transmission, task scheduling, and electromagnetic inertia effects; mechanical idle travel hysteresis refers to the idle travel phase after the motor current responds, before the actuator displacement pushes the braking element to generate effective braking force, corresponding to the process of eliminating the gap between the brake pads or piston and the brake disc; braking force build-up hysteresis refers to the delay between the start and stable establishment of the braking output characteristic quantity after the actuator displacement has passed the idle travel, mainly including the time of mechanical elastic deformation, friction force establishment, and hydraulic pressure build-up processes.
[0061] The total response hysteresis is decomposed into three physical components. The technical principle is that hysteresis from different sources has different compensation methods. Electronic control delay is suitable for elimination through predictive compensation; mechanical idle travel hysteresis is suitable for overcoming through dead zone compensation or pre-clamping; and braking force build-up hysteresis is suitable for improvement through dynamic gain compensation and hysteresis inverse model feedforward. If only one total delay is estimated, it is impossible to apply the optimal compensation strategy for each source separately. In extreme cases, hysteresis from different sources cancels out or superimposes on each other, making the single total delay estimate unstable.
[0062] By decomposing the structure, the variation range of each component under different working conditions can be narrowed, thereby improving the stability and physical interpretability of parameter identification. Step S2 receives the response timing obtained in step S1 and outputs three hysteresis components to step S3, providing a structural basis for the establishment of the hysteresis parameter set.
[0063] In specific implementations, the response timing includes at least four time features: the time of instruction issuance, the time of current response, the time of displacement change, and the time of output response. Hysteresis decomposition can also be implemented using cross-correlation method, derivative energy method, or adaptive threshold determination method, which are not limited in this application.
[0064] In practice, the impact of actual signal noise on hysteresis decomposition needs to be considered; braking commands, motor currents, actuator displacements and braking output characteristics all contain varying degrees of measurement noise during the sampling process; if the transition edge of the original signal is directly used as the time feature extraction point, noise may trigger threshold misjudgment, resulting in inaccurate identification of the start and end times of each component.
[0065] Therefore, in this embodiment, before extracting time features, low-pass filtering or moving average processing is performed on the signals of each channel. However, the selection of the filter cutoff frequency must ensure that the phase characteristics of the original signal are not changed.
[0066] In one possible implementation, a zero-phase filtering method is used to preprocess the slowly varying signal, filtering out high-frequency noise while keeping the time position of the signal edge unchanged. After such preprocessing, the response signals of each node output by the timing acquisition module 110 can truly reflect the physical response process of the actuator, and the four moments extracted by the hysteresis decomposition module 120 on this basis have high repeatability. Compared with the scheme of directly performing differential detection on the original signal, this embodiment has stronger anti-noise capability and can still stably identify the response start point in edge scenarios where the braking output characteristic quantity rises slowly, thereby avoiding the misdetection of noise peaks as the braking force establishment start point.
[0067] In step S3, based on the electronic control delay, mechanical idle travel hysteresis, and braking force establishment hysteresis, a set of direction-related hysteresis parameters is established according to the braking force establishment phase, braking force steady-state maintenance phase, and braking force release phase.
[0068] The braking force establishment phase refers to the phase in which the braking command causes the braking output quantity to rise from its initial value to its target value; the braking force steady-state maintenance phase refers to the phase in which the braking output quantity remains near its target value; and the braking force release phase refers to the phase in which the braking output quantity falls from its target value to its initial value.
[0069] The direction-dependent hysteresis parameter set refers to the parameter set set for the braking force establishment direction and braking force release direction respectively, because the hysteresis characteristics of the same actuator are often different on the loading and unloading paths; the friction and elastic deformation in the mechanical transmission system are direction-dependent.
[0070] During the braking force build-up process, the transmission chain needs to overcome static friction and compress the elastic element; during the braking force release process, the direction of elastic element rebound and the direction of friction are opposite to those during the build-up process, resulting in different output values for the same input command on the loading and unloading paths; if the parameters of the loading and unloading paths are mixed, the same set of parameters cannot accurately describe the hysteresis behavior in both directions at the same time, and the compensation effect will decrease.
[0071] The technical principle of this step lies in modeling and updating the upper and lower branches of the hysteresis loop separately, so that the parameter set can automatically match the hysteresis characteristics of the current path when the braking direction changes. Step S3 receives the three hysteresis components output by step S2, maps the corresponding hysteresis components to the corresponding parameter items according to the phase of the current braking event, and outputs the direction-related hysteresis parameter set. This parameter set is used as the object of online updating in step S4 and serves as a bridge between parameter identification and compensation quantity generation. In specific implementation, the direction-related hysteresis parameter set can include types such as time delay parameters, dead zone parameters, and dynamic parameters. Time delay parameters are used to describe electronic control time delay, dead zone parameters are used to describe mechanical idle travel hysteresis, and dynamic parameters are used to describe braking force establishment hysteresis. Each type of parameter distinguishes between parameters corresponding to the braking force establishment phase and parameters corresponding to the braking force release phase.
[0072] In practice, the distinction between the braking force establishment phase, the braking force steady-state maintenance phase, and the braking force release phase can be achieved by monitoring the rate of change and target value of the braking output characteristic quantity. When the rate of change of the braking output characteristic quantity is greater than the preset rate of change threshold, it is determined that the current phase is either the braking force establishment phase or the braking force release phase. When the braking output characteristic quantity remains near the target value and the rate of change is less than the preset rate of change threshold within a preset time, it is determined that the current phase is the braking force steady-state maintenance phase.
[0073] The phase division results directly control the application range of the parameter set: when the braking force establishes the phase, the parameter update module uses the parameters corresponding to the braking force establishment direction to participate in the identification; when the braking force releases the phase, the parameters corresponding to the braking force release direction are used to participate in the identification; when the braking force maintains the steady state phase, the direction-related parameters are not updated to avoid misjudging steady-state drift as hysteresis change.
[0074] In one specific implementation, the parameter set is stored in non-volatile memory in the form of a lookup table. Each entry in the table corresponds to a braking phase and a parameter value of a hysteresis type. The lookup table is partially refreshed based on the online update results after each braking event, rather than being completely rewritten, thus ensuring the continuity of the parameter set at different life stages. Compared with the single parameter set scheme, the direction-dependent parameter set can more accurately reflect the actual hysteresis behavior of the actuator under frequent alternating pressure build-up and release conditions.
[0075] In step S4, when the braking event meets the preset excitation conditions, the hysteresis parameter set is updated online to obtain candidate hysteresis parameters and online identification residuals. The preset excitation conditions are the criteria for determining whether the current braking event is suitable for parameter identification.
[0076] Online update refers to the process of recursively correcting the hysteresis parameter set based on the actual response data collected during or after a braking event to obtain candidate hysteresis parameters. Online identification residual refers to the deviation between the actual braking output characteristic and the braking output characteristic predicted based on the current parameter model. The effectiveness of parameter identification depends on the amount of information carried in the data. If the braking command changes slowly and the braking output changes very little, the identifiability of each hysteresis component in the response data is very low, and using such data for parameter update will produce large variance or even drift.
[0077] By setting preset excitation conditions, parameter updates are only initiated when both the command and output change significantly enough, which is equivalent to requiring the input signal to cover the effective frequency band required for the actuator's dynamic characteristics in the frequency domain. Simultaneously, the online identification residual reflects the model's fit to the current operating condition: the smaller the residual, the more accurately the candidate hysteresis parameters describe the current hysteresis characteristics; the larger the residual, the more likely the parameter update has not yet converged, or that new, unmodeled factors have appeared in the current operating condition. Step S4 receives the direction-related hysteresis parameter set established in step S3, performs online updates when the preset excitation conditions are met, and outputs candidate hysteresis parameters and online identification residuals. The candidate hysteresis parameters are used in step S5 for credibility assessment, and the online identification residuals are also used in step S5 as input for credibility calculation.
[0078] In practice, online updates can be implemented using a recursive least squares algorithm with a forgetting factor. This algorithm recursively corrects the parameters in the hysteresis parameter set, where the typical value of the forgetting factor can be selected between 0.95 and 0.995. The closer the forgetting factor is to 1, the longer the algorithm remembers historical data, and the smoother the parameter updates. The smaller the forgetting factor, the greater the weight the algorithm gives to recent data, and the faster it tracks time-varying parameters, but the more sensitive it is to noise. Those skilled in the art can select an appropriate forgetting factor according to the application scenario of the actuator.
[0079] In one specific implementation, when the vehicle is under normal braking conditions, the forgetting factor is set to 0.98; when a rapid brake pad wear rate or a large temperature change is detected, the forgetting factor is adjusted to 0.95 to accelerate parameter tracking; candidate hysteresis parameters obtained through online updates will not be directly involved in compensation calculations before passing a credibility assessment, but will first undergo a credibility check; if the excitation level of the current braking event is insufficient, the update will be abandoned directly, and the original parameters will remain unchanged; this mechanism ensures the safety and stability of parameter updates and avoids parameter jumps caused by bad data.
[0080] In step S5, the confidence level of the candidate hysteresis parameters is determined based on the online identification of residuals and parameter physical boundaries. The confidence level is used to quantify whether the candidate hysteresis parameters can be safely used for compensation control.
[0081] The physical boundaries of the parameters are the pre-defined maximum and minimum possible values of the hysteresis parameters in physical terms. If a candidate hysteresis parameter exceeds the physical boundaries, it indicates an anomaly in the identification process; the online identification residuals reflect the deviation between the model output and the actual output.
[0082] The technical principle of credibility assessment is that only candidate parameters that simultaneously meet the two conditions of good model fit and physically reasonable parameters are reliable. If the residual is large but the parameter is still within the physical boundary, it indicates that the model may not have converged yet, and the parameter update direction may still be correct but the magnitude is too large. If the residual is small but the parameter has exceeded the physical boundary, it indicates that the parameter identification is driven by some unmodeled singularities, which may be false identification results caused by sensor failure or actuator abnormality. Therefore, combining these two dimensions for credibility calculation can effectively filter out most unreliable candidate parameters. Step S5 receives the candidate hysteresis parameters and online identification residuals output from step S4, combines them with the preset parameter physical boundary, calculates and outputs the credibility to step S6. Step S6 uses this credibility to make a gating decision.
[0083] In a specific implementation, the credibility can be obtained by weighted summation of the residual index and the boundary index; the residual index is a normalized index obtained by comparing the online identification residual with a preset residual threshold, and the boundary index is a normalized index obtained by comparing the candidate hysteresis parameter with the physical boundary of the parameter; the weighting coefficient can be calibrated according to the actuator type and operating conditions, and this application embodiment does not limit this.
[0084] In practical implementation, the setting of physical boundaries for parameters needs to be combined with the specific structure and physical laws of the actuator; for example, the electronic control delay will not be less than one control cycle, nor greater than a preset safety limit, such as 100 milliseconds; the displacement corresponding to the mechanical idle travel hysteresis will not exceed the upper limit of the actuator's total stroke; the dead zone parameter will not be greater than the equivalent input corresponding to the maximum allowable clearance; these boundary values are obtained through offline calibration and stored in the controller's read-only memory area; when calculating the confidence level, the confidence determination module 150, in addition to checking the value of the current candidate parameter, can also check the candidate parameter relative to the value of the current candidate parameter. The difference between the identification results of adjacent braking events; if the candidate parameters identified in this event differ too much from the parameter values of the previous valid event and there is no clear physical reason, such as a sudden temperature change, then the confidence level should be reduced even if the residuals and boundary indicators are normal; in one specific implementation, the confidence level is a real number between 0 and 1, and the preset confidence threshold can be set to 0.75. When the confidence level is lower than this value, the working parameter set is not updated; in this way, even if a single identification result is abnormal, it will not immediately affect the parameters used by the compensation control, and the system exhibits strong fault tolerance.
[0085] In step S6, when the confidence level is not lower than the preset confidence level threshold, the candidate hysteresis parameters are written into the working parameter set at a limited update rate.
[0086] The working parameter set is the actual set of parameters used for compensation calculation, distinct from the candidate hysteresis parameters temporarily stored during online identification. The constrained update rate refers to the maximum allowable change in each step when writing candidate hysteresis parameters into the working parameter set, used to limit the amount of compensation change caused by a single update. Even if the candidate hysteresis parameters pass the reliability assessment, directly replacing the working parameters with the latest candidate parameters may still cause a sudden jump in the control quantity. The introduction of the constrained update rate is equivalent to applying a low-pass filter to the parameter update process, causing the working parameter set to slide slowly along the candidate parameter direction rather than abruptly changing. This ensures that the compensation quantity remains stable even after the parameter identification results are stored. During residual fluctuations, the change in compensation is continuous and smooth, without causing any impact on the braking process. At the same time, the limited update rate makes the working parameter set itself a slow variable, with a time scale significantly larger than the dynamic time scale of the current closed loop and pressure closed loop, forming a good spectral separation with the robust feedback control. Step S6 receives the confidence level output from step S5 and the candidate hysteresis parameters output from step S4. When the confidence level is not lower than the preset confidence level threshold, the candidate hysteresis parameters are written into the working parameter set at a limited update rate to provide the latest reliable parameters for step S7. If the confidence level is lower than the threshold, the original working parameter set remains unchanged.
[0087] In a specific implementation, the restricted update rate can be set to the maximum allowable change between each braking event, or it can be set to the rate of change of the parameter per unit time; for example, for the delay parameter, the magnitude of a single update can be set to no more than 10% of the current value, and this application embodiment does not limit this.
[0088] In practice, the parameter writing module 160 needs to manage the version and status of the working parameter set. During the vehicle startup phase, the working parameter set loads the most recently saved calibration parameters from the non-volatile memory. During the operation phase, the working parameter set is updated in a restricted manner based on the confidence assessment results. To prevent the parameter set from drifting continuously due to excessively frequent parameter updates, parameter writing is only allowed when high-confidence candidate parameters are obtained in multiple consecutive braking events.
[0089] In one specific implementation, a parameter stabilization counter is added: when the difference between the candidate hysteresis parameter and the current working parameter set is less than a preset threshold, the counter increments; when the difference is greater than the preset threshold, the counter is reset to zero; only when the counter reaches a preset number, such as 10 consecutive braking events, is a parameter write performed at a limited update rate. This slow confirmation mechanism, combined with the limited update rate, can effectively suppress the impact of occasional abnormal data on the working parameter set. At the same time, the write operation records a log, including the update time, candidate parameter values, credibility assessment results, and corresponding braking event characteristics, providing a basis for post-event diagnosis and tracing. Through the above mechanism, the working parameter set maintains a stable evolution throughout the vehicle's entire life cycle, and will not experience unpredictable jumps due to a single abnormal event, thus improving response speed while ensuring control stability and safety.
[0090] In step S7, based on the operating parameter set, the delay prediction compensation amount and the hysteresis inverse model compensation amount are generated. The delay prediction compensation amount is a compensation term used to offset the effects of electronic control delay and pure delay.
[0091] Hysteresis inverse model compensation is a compensation term used to offset the effects of dead zone, hysteresis, and backlash. It maps the desired output to a feedforward input that takes hysteresis into account by constructing the inverse model of the hysteresis model. Delay prediction compensation and hysteresis inverse model compensation are complementary in principle. Delay prediction compensation is equivalent to adding a lead element to the command channel, making the current command transmission time reference the future target value, thereby eliminating phase lag caused by communication and scheduling delays. Hysteresis inverse model compensation is equivalent to connecting an anti-nonlinear element in series in the control channel, matching the actual output after dead zone and hysteresis with the desired output. Alignment; if there is only time delay compensation and no hysteresis compensation, the input change within the dead zone will not produce any output response, and the compensation cannot overcome the gap dead zone; if there is only hysteresis compensation and no time delay compensation, the execution of the inverse model feedforward signal itself will also have a time delay, and the compensation effect will be partially canceled by the delay time; step S7 receives the working parameter set written in step S6, and at the same time receives the braking command, and outputs the time delay prediction compensation amount and the hysteresis inverse model compensation amount to step S9; the above compensation amount and the basic braking control amount generated in step S8 are merged in step S9 to form the final target control amount.
[0092] In specific implementation, the time delay prediction compensation amount can be predicted by the changing trend of the target braking output characterization quantity and the currently estimated time delay parameter, and the hysteresis inverse model compensation amount can be calculated by the Prandtl-Ishlinskii inverse model; the Prandtl-Ishlinskii inverse model is constructed by the weighted superposition of several Play operators, and its weight coefficients are mapped from the time delay parameter and dead zone parameter in the working parameter set; the embodiments of this application do not limit this.
[0093] In practical implementation, the calculation of the hysteresis inverse model compensation amount needs to avoid generating unreasonable feedforward amounts outside the model boundary. When constructing the Prandtl-Ishlinskii inverse model, it is necessary to first perform dead-zone compensation on the target braking output representation amount based on the dead-zone parameters in the working parameter set, and then obtain the feedforward compensation amount through superposition calculation using the inverse Play operator. When there are multiple sets of phase-related parameters in the working parameter set, the system selects the corresponding parameter set for inverse model calculation based on the phase to which the current braking event belongs. In the braking force establishment phase, the parameter set corresponding to the braking force establishment phase is selected; in the braking force release phase, the parameter set corresponding to the braking force release phase is selected; if the braking event is in a steady-state holding phase, the compensation amount of the previous phase output is maintained, and no abrupt adjustment is made to the compensation amount.
[0094] In one specific implementation, the delay prediction compensation is further limited: when the rate of change of braking command exceeds the preset upper limit, the increment of prediction compensation is limited to avoid the prediction algorithm outputting an excessively large advance correction under severe operating conditions; through this combination of phase- and direction-separated inverse model feedforward and prediction compensation, the compensation can overcome the nonlinear effects of idle travel and hysteresis, and also offset the pure delay effects of communication and scheduling links. The two types of compensation complement each other and act on the actuator input, significantly shortening the effective braking force establishment time.
[0095] In step S8, a basic braking control quantity is generated based on the braking command.
[0096] The basic braking control quantity is a control quantity generated based on the braking command and the pre-calibrated mapping relationship, independent of the online identification results. It represents the most basic control output of the actuator under fixed calibration parameters and is the bottom line guarantee for functional safety. Any adaptive compensation algorithm is at risk of failure, such as parameter identification divergence, sensor failure, and compensation quantity exceeding the limit. The existence of the basic braking control quantity ensures that even if the adaptive compensation quantity is completely removed, the actuator can still respond to the braking command based on the calibrated mapping relationship and produce a control effect that meets the basic braking requirements. This is essentially a concrete implementation of the safety principle that "online learning can only improve performance and cannot be a necessary condition for obtaining basic braking force". Separating the basic control quantity from the adaptive compensation quantity makes the adaptive part a supervised performance enhancement module, and its failure does not affect the basic braking function. Step S8 receives the braking command and outputs the basic braking control quantity to step S9, which is then merged with the two compensation quantities generated in step S7 in step S9. The basic control quantity is not affected by the working parameter set and reliability and always operates independently.
[0097] In specific implementation, the basic braking control quantity can be obtained by looking up a table or calculating a formula; the pre-calibration mapping relationship can be established separately for different actuator types: for electro-hydraulic wire-controlled brake actuators, the basic mapping relationship is the relationship between the target braking pressure and the target torque of the motor; for electromechanical wire-controlled brake actuators, the basic mapping relationship is the relationship between the target clamping force and the target current or target displacement of the motor; the embodiments of this application do not limit this.
[0098] In specific implementation, the generation path of the basic braking control quantity is independent of the generation path of the compensation quantity, and the two run in different modules or threads in the software architecture. In high-security embedded software development, the basic control quantity generation module 180 can be deployed in the security kernel, while the compensation quantity generation module 170 can be deployed in the performance kernel. The two kernels communicate through a protected interface. In the functional safety monitoring logic, if the performance kernel detects its own abnormal operation, parameter out-of-bounds, or calculation timeout, the security kernel directly takes over the output control. During the above takeover process, the calculation of the target control quantity no longer includes the compensation quantity, but is only output by the basic braking control quantity after safety verification, avoiding the impact of adaptive compensation failure on the basic braking function. In this embodiment, the generation cycle of the basic braking control quantity is consistent with the generation cycle of the compensation quantity to ensure the smoothness of instruction switching during safety takeover. Compared with the scheme of completely powering off the actuator in case of an abnormality, this embodiment can clear the adaptive increment while retaining the basic braking capability, ensuring both driving safety and the continuity of braking function. This design logic complements the redundant actuator architecture and can achieve safe braking under the condition that any actuator channel is working normally.
[0099] In step S9, the time delay prediction compensation and the hysteresis inverse model compensation are fused with the basic braking control quantity within the preset safety constraint envelope to obtain the target control quantity, which is then output to the brake-by-wire actuator.
[0100] The preset safety constraint envelope is a set of constraints imposed on the amplitude and rate of change of the target control quantity, including the upper limit of motor current, the upper limit of actuator displacement, the upper limit of the rate of change of braking output characteristic quantity, and the upper limit of compensation quantity. The technical principle of compensation quantity fusion lies in the combination of superposition and amplitude limiting. First, the time delay prediction compensation quantity and the hysteresis inverse model compensation quantity are superimposed on the basic braking control quantity to achieve the additive effect of compensation. Then, the superposition result is checked by the safety constraint envelope. If it exceeds the preset boundary, amplitude limiting is performed. This preserves the improvement of response speed by the compensation quantity while preventing excessive compensation quantity from causing actuator overload or braking shock. The boundary value of the safety constraint envelope is usually determined comprehensively based on the actuator hardware capability, vehicle dynamics limitations, and functional safety requirements. Step S9 receives the time delay prediction compensation amount and hysteresis inverse model compensation amount generated in step S7, and the basic braking control amount generated in step S8, and fuses them within the preset safety constraint envelope, outputting the target control amount to the actuator; this output is the final execution result of the entire method; in specific implementation, the preset safety constraint envelope may include the maximum absolute value, maximum rate of change, and maximum acceleration of the target control amount, etc.; for example, the upper limit of motor current can be set to 150% of the rated current, the upper limit of actuator displacement can be set to 95% of the maximum allowable stroke, the upper limit of the rate of change of braking output characteristic quantity can be set to 120% of the rated rate of change, and the upper limit of compensation amount can be set to 30% of the basic braking control amount, which is not limited in this embodiment.
[0101] In practice, the control quantity fusion module 190 algebraically superimposes the basic braking control quantity with the two compensation quantities, and then performs a compensation quantity upper limit check and a total control quantity boundary check. The purpose of the compensation quantity upper limit check is to prevent the single compensation increment from being too large. The purpose of the total control quantity boundary check is to prevent the target control quantity from exceeding the safety limits of the actuator or the whole vehicle. During the check process, if the superimposed control quantity exceeds the preset safety constraint envelope, it is corrected in the order of limiting the compensation quantity first and then limiting the total control quantity.
[0102] In one specific implementation, when the target control quantity exceeds the upper limit of the motor current, the target control quantity is clamped to the upper limit value, and an over-limit flag is output, which is sent to the diagnostic system for recording. If the over-limit state lasts for more than a preset time, the adaptive compensation is determined to be abnormal, the parameter update is frozen and the compensation quantity is canceled, and the output of the pure basic control quantity is switched. Through this multi-level limiting and abnormal diagnosis linkage mechanism, the compensation fusion algorithm can ensure that even in the case of serious errors in the identified parameters or sensor drift, the instructions received by the actuator are always within the safety boundary, fundamentally avoiding overshoot or instability caused by excessive compensation.
[0103] In one optional implementation, the response timing includes the time when the braking command is issued, the time when the motor current responds, the time when the actuator displacement changes, and the time when the braking output characteristic quantity responds.
[0104] Specifically, the electronic control delay is obtained by subtracting the response time of the motor current from the issuance time of the braking command; the mechanical idle travel delay is obtained by subtracting the change time of the actuator displacement from the response time of the motor current; and the braking force build-up delay is obtained by subtracting the response time of the braking output characteristic quantity from the change time of the actuator displacement.
[0105] The total response hysteresis is equal to the sum of the electronic control delay, the mechanical idle travel hysteresis, and the braking force build-up hysteresis. Through the above multi-node timing decomposition, the single total delay is subdivided into three components with clear physical meanings, enabling subsequent compensation to apply control separately for different sources of hysteresis.
[0106] For example, when the actuator's idle stroke increases due to brake pad wear, the increased mechanical idle stroke hysteresis can be directly identified in the identification results, without having to attribute all changes to a total delay as in the endpoint model. Compared to the traditional fixed delay model, the decomposition method provided by this preferred embodiment is physically more interpretable, making it easier for calibration engineers to determine whether parameter changes are reasonable based on the actual hardware conditions.
[0107] In one optional implementation, the hysteresis parameter set includes a time delay parameter corresponding to the braking force establishment phase, a time delay parameter corresponding to the braking force release phase, a dead zone parameter corresponding to the braking force establishment phase, a dead zone parameter corresponding to the braking force release phase, a dynamic parameter corresponding to the braking force establishment phase, and a dynamic parameter corresponding to the braking force release phase. The time delay parameter characterizes the electronic control delay, the dead zone parameter characterizes the mechanical idle travel hysteresis, and the dynamic parameter characterizes the braking force establishment hysteresis. That is, the time delay parameter corresponding to the braking force establishment phase and the time delay parameter corresponding to the braking force release phase respectively describe the values of the electronic control delay of the actuator in the two directions of establishing and releasing braking force; the dead zone parameter corresponding to the braking force establishment phase and the dead zone parameter corresponding to the braking force release phase respectively describe the values of the mechanical idle travel hysteresis in the two directions; and the dynamic parameter corresponding to the braking force establishment phase and the dynamic parameter corresponding to the braking force release phase respectively describe the values of the braking force establishment hysteresis in the two directions.
[0108] By classifying parameters by both phase and hysteresis type, the system can independently update various hysteresis parameters in the braking force establishment direction and braking force release direction. This parameter organization method is consistent with the actual physical behavior of the actuator: the frictional force directions on the loading path and unloading path are different, and the idle travel clearance is also different at different wear stages.
[0109] This preferred embodiment improves the adaptability of the hysteresis model to different operating conditions by storing and updating parameters independently in both directions, while avoiding the negative impact of parameter updates in one direction on the other.
[0110] In one optional implementation, the preset excitation condition includes: the rate of change of the braking command is greater than a preset rate of change threshold, and the change in the braking output representation is greater than a preset output change threshold. That is, only when significant dynamic changes occur on both the command side and the output side is the current braking event considered to have sufficient excitation information to initiate parameter updates; this condition essentially excludes inefficient updates in slow braking and steady-state maintenance scenarios.
[0111] In natural driving, the speed and depth at which the driver presses the brake pedal usually meet the above conditions; however, in the fine-tuning of adaptive cruise control, the change in command may be small, and the system automatically skips parameter updates. In this way, this preferred embodiment ensures that the data used for parameter updates have sufficient excitation information without affecting the user's normal vehicle use, thereby improving the numerical stability of online identification.
[0112] In one optional implementation, updating the hysteresis parameter set online includes: using a recursive least squares algorithm with a forgetting factor to recursively derive the hysteresis parameter set to obtain the candidate hysteresis parameters and the online identification residual. The recursive process of this algorithm is as follows: within the current sampling period, firstly, the gain matrix is calculated using the covariance matrix of the parameter estimation error from the previous sampling period, the regression vector of the current sampling period, and the forgetting factor; then, the parameter vector estimate is corrected using the difference between the measured value of the braking output representation quantity in the current sampling period and the predicted value calculated based on the parameter vector estimate from the previous sampling period, i.e., the innovation, to obtain the parameter vector estimate for the current sampling period; subsequently, the covariance matrix of the parameter estimation error is updated; finally, the above innovation is directly used as the online identification residual for the current sampling period, wherein the forgetting factor is used to adjust the influence of historical data on the current parameter estimate, and its value ranges from 0.95 to 0.995.
[0113] The core formulas involved in the above recursive process are as follows:
[0114] The formula for updating the parameter vector estimate is: ;
[0115] The formula for calculating the gain matrix is: ;
[0116] The formula for updating the covariance matrix is: ;
[0117] The formula for calculating residuals in online identification is: ;
[0118] The relationship between the above formulas is as follows: Gain matrix From the covariance matrix Regression vector and forgetting factor Jointly determined; parameter vector estimate exist Based on this, through the gain matrix With new information The product is corrected; the covariance matrix The update is then determined by the gain matrix. Regression vector covariance matrix at the previous time step Joint decision-making; online residual identification That is, the information itself, which is the measured value. Compared with the predicted value difference.
[0119] in, Indicates the first The estimated value of the hysteresis parameter vector for each sampling period, which constitutes a candidate hysteresis parameter vector, whose elements include the time delay parameter, dead zone parameter and dynamic parameter; Indicates the first Estimated values of the hysteresis parameter vector for each sampling period; Indicates the first The gain matrix for each sampling period is used to adjust the magnitude of the correction of the innovation to the parameter estimate. Indicates the first The regression vector for each sampling period is constructed from signals such as braking commands, motor current, actuator displacement, and braking output characteristics at the current and historical moments. Represents the regression vector Transpose of; Indicates the first Measured values of braking output characterization parameters for each sampling period; Indicates the first The covariance matrix of the parameter estimation error for each sampling period is used to describe the uncertainty of the parameter estimation. Indicates the first The covariance matrix of parameter estimation errors for each sampling period; represents the forgetting factor, with a value ranging from 0.95 to 0.995; I represents the identity matrix with the same dimension as the parameter vector; Indicates the first The online identification residual for each sampling period is the deviation between the measured value of the braking output characteristic and the predicted value calculated based on the current parameter estimate.
[0120] The algorithm performs real-time recursive calculations of the regression values within each sampling period. Its computational complexity is proportional to the square of the number of parameters, which can meet the real-time requirements on embedded controllers. The introduction of the forgetting factor enables the algorithm to gradually reduce the influence of historical data on the current estimate, thereby continuously tracking parameter changes under time-varying conditions such as brake pad wear and temperature drift. Compared with the ordinary least squares algorithm without the forgetting factor, the recursive least squares algorithm in this preferred embodiment will not exhibit the phenomenon of sluggish parameter updates caused by the accumulation of historical data during long-term operation.
[0121] In one optional implementation, determining the credibility of the candidate hysteresis parameter based on the online identified residual and the physical boundary of the parameter includes: comparing the online identified residual with a preset residual threshold to determine a residual index; comparing the candidate hysteresis parameter with the physical boundary of the parameter to determine a boundary index; and weighted summing the residual index and the boundary index to obtain the credibility.
[0122] The residual index is used to quantify how close the online identification residual is to a preset residual threshold. When the online identification residual does not exceed the preset residual threshold, the residual index takes a higher value. When the online identification residual exceeds the preset residual threshold, the residual index decreases as the residual increases.
[0123] Boundary indices are used to quantify the degree of deviation of candidate hysteresis parameters from the physical boundaries of the parameters. When all candidate hysteresis parameters are within the physical boundaries of the parameters, the boundary indices take higher values; when candidate hysteresis parameters exceed the physical boundaries of the parameters, the boundary indices decrease as the degree of deviation increases.
[0124] The reliability is a weighted sum of the residual index and the boundary index. Specifically, it is obtained by multiplying each index by its respective weighting coefficient and then summing the results. Both weighting coefficients are positive numbers, and their sum is 1. The specific values can be calibrated based on the actuator type and operating conditions. In practice, higher residual and boundary indexes result in higher reliability; conversely, lower residual or boundary indexes result in lower reliability.
[0125] This preferred embodiment uses a dual-index joint determination to ensure that the update direction is consistent with the actual hysteresis change, and that the update result does not exceed the physically feasible range of the actuator.
[0126] In one optional implementation, generating the hysteresis inverse model compensation amount includes: determining a target value of the braking output characterization quantity based on the braking command; constructing a hysteresis inverse model based on the working parameter set, wherein the hysteresis inverse model is a Prandtl-Ishlinskii inverse model; and inputting the target value into the hysteresis inverse model to obtain the hysteresis inverse model compensation amount.
[0127] Specifically, the Prandtl-Ishlinskii model is a hysteresis model composed of a weighted superposition of multiple Play operators, where each Play operator has its own threshold and weight; the corresponding Prandtl-Ishlinskii inverse model takes the desired target output as input and calculates the feedforward input that can offset the original hysteresis effect through the weighted superposition of inverse Play operators, which is the hysteresis inverse model compensation amount.
[0128] When constructing the inverse model, the weights and thresholds of each operator in the inverse model are obtained by mapping the time delay parameters, dead zone parameters and dynamic parameters in the working parameter set; when the working parameter set is updated, the parameters of the inverse model are updated synchronously to achieve adaptive adjustment of the inverse model.
[0129] In this preferred embodiment, the target braking output characteristic quantity is first determined according to the braking command, then the inverse model is constructed using the working parameter set, and finally the target value is input into the inverse model to obtain the compensation quantity. The Prandtl-Ishlinskii model has an analytical inverse model, which can describe dead zone and hysteresis nonlinearity with a smaller number of operators, making it easy to implement in embedded real-time systems. Compared with the lookup table type inverse model, the analytical inverse model does not have interpolation errors and does not require a large amount of storage space, making it more suitable for resource-constrained automotive-grade controllers.
[0130] Reference Figure 2 This invention provides an online identification and compensation system 100 for response hysteresis of a line-controlled brake actuator. The system includes a timing acquisition module 110, a hysteresis decomposition module 120, a parameter establishment module 130, a parameter update module 140, a reliability determination module 150, a parameter writing module 160, a compensation quantity generation module 170, a basic control quantity generation module 180, and a control quantity fusion module 190.
[0131] The timing acquisition module 110 is used to acquire the response timing of braking commands, motor current, actuator displacement, and braking output parameters. The braking commands originate from the vehicle controller or brake control unit; the motor current originates from the actuator motor current sensor; the actuator displacement originates from a displacement sensor or motor rotor position sensor; and the braking output parameters originate from a pressure sensor or clamping force sensor. The timing acquisition module 110 performs time-synchronized acquisition of the above multiple signals, uniformly timestamps them with high precision, and outputs the time-synchronized multi-node response data.
[0132] The hysteresis decomposition module 120 is used to decompose the total response hysteresis of the brake-by-wire actuator into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis based on the response timing output by the timing acquisition module 110. Specifically, the electrical control delay represents the time from the issuance of the braking command to the start of motor current response; the mechanical idle travel hysteresis represents the time from the start of motor current response to the start of actuator displacement change; and the braking force build-up hysteresis represents the time from the start of actuator displacement change to the start of braking output response. The hysteresis decomposition module 120 outputs the decomposition results to the parameter establishment module 130.
[0133] The parameter establishment module 130 is used to establish a set of direction-related hysteresis parameters based on the electronic control delay, mechanical travel hysteresis, and braking force establishment hysteresis, according to the braking force establishment phase, braking force steady-state holding phase, and braking force release phase. This parameter set contains multiple parameter items corresponding to different braking phases and different hysteresis types, and organizes the parameter items according to the phase of the current braking event, and outputs them to the parameter update module 140.
[0134] The parameter update module 140 is used to update the hysteresis parameter set online when the braking event meets the preset excitation conditions, thereby obtaining candidate hysteresis parameters and online identification residuals. The parameter update module 140 can use a recursive least squares algorithm with a forgetting factor to recursively update the hysteresis parameter set. During the update process, the candidate hysteresis parameters and the residual information generated by each recursion are retained and output to the confidence determination module 150.
[0135] The credibility determination module 150 is used to determine the credibility of candidate hysteresis parameters based on the online identified residuals and parameter physical boundaries. The credibility determination module 150 compares the online identified residuals with a preset residual threshold to obtain a residual index; compares the candidate hysteresis parameters with the parameter physical boundaries to obtain a boundary index; and then weights and sums the residual index and the boundary index to obtain a comprehensive credibility, and outputs the credibility to the parameter writing module 160.
[0136] The parameter writing module 160 is used to write candidate hysteresis parameters into the working parameter set at a limited update rate when the confidence level is not lower than a preset confidence level threshold. The parameter writing module 160 performs the parameter writing operation when the confidence level meets the condition, and keeps the original working parameter set unchanged when the confidence level does not meet the condition. The written working parameter set is output to the compensation quantity generation module 170.
[0137] The compensation quantity generation module 170 generates a time delay prediction compensation quantity and a hysteresis inverse model compensation quantity based on the operating parameter set. The time delay prediction compensation quantity is calculated based on the changing trend of the target braking output characteristic quantity and the time delay parameters in the current operating parameter set. The hysteresis inverse model compensation quantity is obtained by inversely transforming the target braking output characteristic quantity using the Prandtl-Ishlinskii inverse model. The compensation quantity generation module 170 outputs both compensation quantities to the control quantity fusion module 190.
[0138] The basic control quantity generation module 180 is used to generate basic braking control quantities based on braking commands. These basic braking control quantities are obtained based on a safety-calibrated basic mapping relationship, do not rely on online identification results, and serve as a baseline guarantee for functional safety. The basic control quantity generation module 180 outputs the basic braking control quantities to the control quantity fusion module 190.
[0139] The control quantity fusion module 190 is used to fuse the time delay prediction compensation quantity and the hysteresis inverse model compensation quantity with the basic braking control quantity within a preset safety constraint envelope to obtain the target control quantity, and output the target control quantity to the brake-by-wire actuator. During the fusion process, the control quantity fusion module 190 performs safety logic such as compensation quantity limiting, total control quantity boundary verification, and over-limit flag output.
[0140] The above system achieves online decomposition of hysteresis components, reliability determination, and constrained compensation fusion through the collaborative work of its modules, and has the same or corresponding technical effects as the above method.
[0141] This application also provides an electronic device, including a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method; the electronic device may be a vehicle controller, a brake control unit, or an independent adaptive compensation controller in a brake-by-wire system; the electronic device may also include a communication interface for data interaction with the sensors and actuators of the brake-by-wire actuator; the memory may be volatile memory or non-volatile memory, and the processor may be one or more, which is not limited in this application embodiment.
[0142] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method; the computer-readable storage medium includes, but is not limited to, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks; when the computer program stored on the storage medium is executed by a processor, it can realize the above-described hysteresis online identification and compensation function.
[0143] When the vehicle is in a high-frequency adjustment state of the anti-lock braking system or the electronic stability system, the wheel-end braking pressure is constantly adjusted by the upper-level stability control algorithm. The input command changes in both directions multiple times in a short period of time. Under this condition, the hysteresis loop of the actuator presents a complex secondary loop path, and each pressure adjustment cycle is short, making it difficult to extract stable and reliable hysteresis parameters from a single pressure adjustment process.
[0144] In the above scenario, the method of this embodiment judges the data quality by setting the rate of change threshold and the output change threshold in the preset excitation conditions: although the rate of change of the braking command may meet the threshold requirements, if the change of the braking output characteristic quantity is too small in one adjustment cycle, the excitation conditions are not met, and the parameter update module automatically skips this parameter update. In addition, the active signal of the anti-lock braking system or the vehicle electronic stability system can be directly used as the prohibition condition for parameter update. When the signal is valid, even if the conventional excitation conditions are met, a new parameter update will not be started. In this way, during the high-frequency pressure adjustment process on low-adhesion road surfaces, the working parameter set maintains the last valid update value before the anti-lock braking system or the vehicle electronic stability system is activated, avoiding the misidentification of the adjustment action of the upper-level stability control as a change in the actuator hysteresis characteristic.
[0145] When a vehicle is in a low-temperature environment, such as when the ambient temperature is below -20 degrees Celsius, the viscosity of the hydraulic brake fluid and the friction characteristics of the electromechanical actuator grease change significantly. The electronic control delay and the hysteresis of the braking force build-up will increase significantly. In high-temperature scenarios, such as after continuous downhill driving or repeated high-intensity braking, the caliper temperature rises to above 300 degrees Celsius. The change in the friction coefficient of the brake pads and the tendency of the hydraulic system to have air resistance intensify, and the hysteresis characteristics change in the opposite direction.
[0146] In the aforementioned extreme temperature scenarios, the method of this embodiment automatically adjusts the forgetting factor to a smaller value when the temperature signal reaches a preset high temperature threshold or low temperature threshold, enabling the online identification algorithm to quickly track parameter drift caused by temperature changes. Simultaneously, the physical boundaries of the parameters can be segmented and calibrated according to temperature, for example, setting different boundary ranges in the low temperature zone, normal temperature zone, and high temperature zone to avoid candidate hysteresis parameters being incorrectly accepted within the normal boundaries under extreme temperatures. This embodiment, through the aforementioned temperature adaptive adjustment mechanism, ensures that the hysteresis parameter set remains within an effective range under different temperature conditions.
[0147] As brake pad wear increases, actuator free travel increases significantly, and mechanical free travel hysteresis continues to grow. In this scenario, the multi-node decomposition method of this embodiment can directly observe the free travel trend from the time difference between the actuator displacement change moment and the brake output response moment. When the identification result of mechanical free travel hysteresis continues to increase monotonically, and is accompanied by a synchronous increase in dead zone parameters, the system can determine it as a brake pad wear characteristic, rather than a change in electronic control delay. Using this information, the vehicle can remind the driver to check or replace the brake pads at an appropriate maintenance cycle. At the same time, the direction-related parameter set updates the dead zone parameters of the braking force establishment direction and release direction separately in the later stage of wear. Even if the wear on both sides is inconsistent, the parameter set can accurately reflect the free travel change on each side. The above-mentioned wear adaptive mechanism extends the effectiveness of the actuator parameter model throughout its entire life cycle and avoids compensation inaccuracies caused by wear accumulation.
[0148] When a vehicle brakes on a separated road surface with different coefficients of adhesion on the left and right sides, the four-wheel brake-by-wire actuators receive different target braking output parameters, resulting in significant differences in the operating points of each actuator. In this scenario, the preset excitation conditions and parameter update rates of each actuator can be configured independently to avoid mutual interference between the left and right actuators due to their different operating points. When the braking output parameter of one actuator fails to reach the target value due to adhesion limitations, the parameter update of that actuator is suspended, while the parameter update of the other actuator can proceed normally. In this embodiment, under the coordination of the vehicle controller, each actuator manages its own set of operating parameters, thereby ensuring the independent adaptive capability of each wheel during braking on a separated road surface while preventing the low-adhesion side from contaminating the hysteresis parameters due to excessive tracking errors. By locking the parameter updates of the target actuator only in case of a fault or anomaly, rather than locking all actuators uniformly, this embodiment maximizes the online adaptive performance of the normal actuators.
[0149] To ensure the functional safety of the brake-by-wire system, this embodiment also sets several preconditions and exit mechanisms. Regarding the preconditions for parameter update initiation, the parameter update module is only allowed to initiate online updates when the vehicle speed is below the preset speed limit and the braking system is not in a fault-degraded state. If the driver does not press the brake pedal but the system receives an abnormal braking command, or the braking force request is significantly inconsistent with the current vehicle state, the sensor data is considered unreliable, and parameter updates are directly prohibited. Before outputting the compensation amount, it is also necessary to check whether the actuator temperature is within the normal operating range and whether the power supply voltage is within the preset voltage window. If the actuator temperature exceeds the safe operating range, or the power supply voltage is below the minimum operating voltage, all compensation amounts are immediately revoked, and only the basic braking control amount is output. Regarding exit conditions, adaptive compensation is automatically exited and the working parameter set is frozen when the following situations occur: the online identification residual continuously exceeds the preset residual threshold for a preset duration; the candidate hysteresis parameters fail the reliability assessment within a preset number of times; the number of compensation amount exceeding the limit exceeds the preset number of times; or the sensor self-test report is abnormal or data synchronization times out. Under the above circumstances, the basic control quantity path continues to operate, and the target control quantity switches to the output of the basic braking control quantity. Instead of making an instantaneous step switch, the compensation quantity is canceled in a short time by using a ramp transition or smooth amplitude limiting method to prevent the sudden change in braking force from causing longitudinal impact on the vehicle.
[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for online identification and compensation of response hysteresis in a brake-by-wire actuator, characterized in that, include: Acquire the response timing of braking commands, motor current, actuator displacement, and braking output parameters; Based on the aforementioned response timing, the total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis. Based on the electronic control delay, the mechanical idle travel hysteresis, and the braking force establishment hysteresis, a set of direction-related hysteresis parameters is established according to the braking force establishment phase, the braking force steady-state maintenance phase, and the braking force release phase. When the braking event meets the preset excitation conditions, the hysteresis parameter set is updated online to obtain candidate hysteresis parameters and online identification residuals. The confidence level of the candidate hysteresis parameters is determined based on the online identification of residuals and parameter physical boundaries. When the confidence level is not lower than the preset confidence level threshold, the candidate hysteresis parameter is written into the working parameter set at a limited update rate; Based on the set of working parameters, generate the delay prediction compensation amount and the hysteresis inverse model compensation amount; Based on the braking command, a basic braking control quantity is generated; The time delay prediction compensation and the hysteresis inverse model compensation are fused with the basic braking control quantity within a preset safety constraint envelope to obtain the target control quantity, which is then output to the brake-by-wire actuator.
2. The method according to claim 1, characterized in that, The response timing includes the moment when the braking command is issued, the moment when the motor current responds, the moment when the actuator displacement changes, and the moment when the braking output characteristic quantity responds. The total response hysteresis of the brake-by-wire actuator is decomposed into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis, including: The electronic control delay is determined based on the timing of the braking command and the response timing of the motor current. The mechanical idle travel hysteresis is determined based on the response time of the motor current and the change time of the actuator displacement; The braking force build-up hysteresis is determined based on the timing of the change in actuator displacement and the response timing of the braking output characteristic quantity.
3. The method according to claim 1, characterized in that, The hysteresis parameter set includes the time delay parameter corresponding to the braking force establishment phase, the time delay parameter corresponding to the braking force release phase, the dead zone parameter corresponding to the braking force establishment phase, the dead zone parameter corresponding to the braking force release phase, the dynamic parameter corresponding to the braking force establishment phase, and the dynamic parameter corresponding to the braking force release phase. The time delay parameter is used to characterize the electronic control time delay, the dead zone parameter is used to characterize the mechanical idle travel hysteresis, and the dynamic parameter is used to characterize the braking force build-up hysteresis.
4. The method according to claim 1, characterized in that, The preset excitation conditions include: the rate of change of the braking command is greater than a preset rate of change threshold, and the amount of change of the braking output characteristic quantity is greater than a preset output change threshold.
5. The method according to claim 1, characterized in that, Online updating of the hysteresis parameter set includes: A recursive least squares algorithm with a forgetting factor is used to recursively derive the candidate hysteresis parameters and the online identification residuals from the set of hysteresis parameters.
6. The method according to claim 1, characterized in that, Based on the online identification of residuals and parameter physical boundaries, the confidence level of the candidate hysteresis parameters is determined, including: The online identified residual is compared with a preset residual threshold to determine the residual index; The candidate hysteresis parameters are compared with the physical boundaries of the parameters to determine the boundary indices; The confidence level is obtained by weighted summation of the residual index and the boundary index.
7. The method according to claim 1, characterized in that, Generate hysteresis inverse model compensation quantities, including: Determine the target value of the braking output characteristic quantity based on the braking command; Based on the aforementioned set of working parameters, a hysteresis inverse model is constructed, which is the Prandtl-Ishlinskii inverse model. The target value is input into the hysteresis inverse model to obtain the compensation amount of the hysteresis inverse model.
8. A system for online identification and compensation of response hysteresis in a brake-by-wire actuator, characterized in that, include: The timing acquisition module is used to acquire the response timing of braking commands, motor current, actuator displacement, and braking output parameters. The hysteresis decomposition module is used to decompose the total response hysteresis of the brake-by-wire actuator into electrical control delay, mechanical idle travel hysteresis, and braking force build-up hysteresis based on the response timing. The parameter establishment module is used to establish a set of direction-related hysteresis parameters according to the electronic control delay, the mechanical idle travel hysteresis and the braking force establishment hysteresis, and according to the braking force establishment phase, the braking force steady-state maintenance phase and the braking force release phase. The parameter update module is used to update the hysteresis parameter set online when the braking event meets the preset excitation conditions, so as to obtain candidate hysteresis parameters and online identification residuals; A credibility determination module is used to determine the credibility of the candidate hysteresis parameters based on the online identified residuals and parameter physical boundaries. The parameter writing module is used to write the candidate hysteresis parameter into the working parameter set at a limited update rate when the confidence level is not lower than a preset confidence level threshold. The compensation quantity generation module is used to generate the delay prediction compensation quantity and the hysteresis inverse model compensation quantity based on the working parameter set. A basic control quantity generation module is used to generate basic braking control quantities based on the braking command; The control quantity fusion module is used to fuse the time delay prediction compensation quantity and the hysteresis inverse model compensation quantity with the basic braking control quantity within a preset safety constraint envelope to obtain the target control quantity, and output it to the line control brake actuator.
