Automobile EPS outer ring controller parameter dynamic optimization method and system

Abstract

The invention relates to the technical field of automobile steering, in particular to an automobile EPS outer ring controller parameter dynamic optimization method and system, and the method comprises the steps: collecting multi-source data, building a collection transmission link through PTP clock synchronization and a signal preprocessing module, and obtaining a standardized data set through data preprocessing; a control variable method and a multi-dimensional performance weighted evaluation mechanism are adopted to calibrate optimal control parameters, and decision support data adaptive to dynamic parameters are formed; constructing a four-dimensional scene coordinate-strategy adaptation model to generate a parameter optimization execution instruction, and obtaining a cross-system closed loop execution result of EPS two-stage closed loops; constructing a traceable system through structured log records and data fingerprint verification, and outputting a parameter credibility rating and three-dimensional quantitative evaluation result; and integrating the data to obtain optimized data, inputting the optimized data into a parameter iterative optimization link, and outputting an EPS dynamic parameter intelligent optimization control scheme. According to the scheme, the flexibility of outer ring control adjustment can be remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile steering technology, and in particular to a method and system for dynamically optimizing parameters of an EPS outer ring controller. BACKGROUND

[0002] With the rapid development of automobile electrification and intelligentization, mid-to-high-end vehicles generally carry automatic parking, lane keeping, emergency lane changing and other assisted driving functions. The implementation of these functions relies on the accurate and timely response of the electronic power steering system (EPS) to the instruction steering angle signal issued by the vehicle control unit (VCU), thereby completing the autonomous control of the vehicle in the lateral direction.

[0003] In the prior art, the EPS generally responds to the VCU instruction steering angle signal through a three-level closed-loop control of "steering angle closed loop - speed closed loop - current closed loop", wherein the outer ring control (steering angle closed loop and speed closed loop) directly affects the accuracy and stability of the steering control, and the core control logic thereof mostly adopts a PID control algorithm. Specifically, the steering angle closed loop receives the VCU instruction steering angle signal and the actual steering angle signal collected by the EPS sensor, calculates the deviation, and then outputs an instruction speed through a PID controller; the speed closed loop receives the instruction speed signal and the actual speed signal, and outputs an instruction current after PID operation, which drives the steering motor to complete the steering action.

[0004] However, the PID controller of the existing EPS outer ring control has significant defects: the PID control parameters (proportional coefficient Kp, integral coefficient Ki, and differential coefficient Kd) used in the steering angle closed loop and the speed closed loop are fixed constants, and the adjustment capability is limited. Since the assisted driving function covers a variety of differentiated working conditions, such as automatic parking corresponding to low vehicle speed, large steering angle, and high load steering demand, lane keeping corresponding to high vehicle speed, small steering angle, and high sensitivity steering demand, and the control performance requirements are different at different stages such as steering start, maintenance, stop, and reversal, it is difficult for fixed and unchanged PID parameters to adapt to multiple working conditions at the same time. Even if the parameters are adjusted to the optimal state, the steering performance under all working conditions cannot be improved to a good level, and problems such as response lag, overshoot oscillation, or steady-state error are likely to occur, which makes it difficult to meet the high precision and high stability requirements of assisted driving for steering control, and limits the adaptation ability and control performance improvement of the EPS system in complex scenarios. SUMMARY

[0005] The present application can significantly improve the flexibility of outer ring control adjustment.

[0006] The technical solution proposed by the present application is: a method for dynamically optimizing parameters of an EPS outer ring controller, the method comprising: collecting multi-source data, building a collection and transmission link through a PTP clock synchronization and signal preprocessing module, and obtaining a standardized data set through data preprocessing; Based on standardized datasets, the optimal control parameters are calibrated using the control variable method and a multi-dimensional performance weighted evaluation mechanism. Combined with hierarchical rule analysis and sample library mapping relationship back-inference, the parameter adaptability and control risks are perceived through bench tests and boundary condition verification, forming decision support data for dynamic parameter adaptation. Based on decision support data, a four-dimensional scene coordinate-strategy adaptation model is constructed to generate parameter optimization execution instructions, and cross-system closed-loop execution results of EPS two-level closed loop are obtained. Based on the key node data of the closed-loop execution results, a traceability system is constructed through structured log recording and data fingerprint verification, and the output parameter credibility rating and three-dimensional quantitative evaluation results are presented. By integrating standardized datasets, decision support data, closed-loop execution results, data credibility, and three-dimensional quantitative evaluation results, optimized data is obtained. The optimized data input parameters are iteratively optimized to output an intelligent optimization control scheme for EPS dynamic parameters that is adapted to multiple working conditions and complex loads.

[0007] Preferably, the specific process for obtaining the standardized dataset is as follows: Clearly define the core objectives and data requirements for EPS control, and determine the scope and core indicators for collecting EPS hardware specifications data, operating condition characteristic data, load signal data, and assisted driving function command data; A multi-source data acquisition mechanism is initiated, and a time synchronization network is built based on the PTP clock synchronization protocol. The signal preprocessing module calls the sensor data acquisition interface, the CAN bus data receiving module, and the bench condition simulation data output module to extract the corresponding types of data. Differentiated preprocessing is performed on the collected multi-source data, outlier data is removed by outlier removal rules, missing data is filled by interpolation method, and timestamp format, numerical unit and signal coding standard are unified; Based on data association keys, cross-source relationships are constructed, preprocessed valid data is integrated, and a standardized dataset with data type labels and collection timestamps is generated.

[0008] Preferably, the specific process for establishing the data acquisition and transmission link is as follows: Configure a PTP clock synchronization server to assign unique clock node IDs to EPS controllers, sensors, and bench test equipment, and establish a master-slave clock synchronization architecture. Deploy a signal preprocessing module, configure dedicated processing units for different types of signals, perform filtering and noise reduction processing on analog signals, perform edge detection and counting calibration on digital signals, and perform frame parsing and verification on CAN bus signals; A data transmission channel is established, using shielded twisted-pair cables to transmit analog and digital signals, and a CAN bus to transmit control commands and status data. Signal cables and high-voltage cables are kept at a preset safe distance to avoid interference. Configure a data transmission verification mechanism, add a checksum to the transmitted data, and trigger a retransmission mechanism when the receiving end detects a data error, thus forming a stable data acquisition and transmission link.

[0009] Preferably, the specific process for obtaining the decision support data is as follows: Based on a standardized dataset, the EPS control parameters were adjusted using the control variable method, adjusting only a single parameter each time while keeping other parameters unchanged. A multi-dimensional performance weighted evaluation mechanism is initiated to collect data on steering angle tracking accuracy, steering response delay, speed fluctuation, and current stability. A comprehensive score is calculated according to preset weights, and parameters that meet the comprehensive score are selected as candidate optimal control parameters. Deploy a hierarchical rule parsing module to divide vehicle speed into several intervals and turning angle deviation into several levels, and combine the mapping relationship in the sample library to infer the parameter range corresponding to different levels; The adaptability of candidate parameters under different working conditions is verified through bench tests, the stability of parameters is verified by simulating boundary conditions, the adaptability defects and control risks of parameters are perceived, and the candidate parameters, hierarchical mapping results and risk analysis conclusions are integrated to generate decision support data.

[0010] Preferably, the specific process for generating the parameter optimization execution instructions is as follows: Define four-dimensional scene coordinate dimensions, including working condition dimension, load dimension, response dimension, and safety dimension, and normalize the indicators of each dimension. Construct a strategy adaptation model, train parameter optimization algorithms based on decision support data, and calculate the adaptation priorities of interpolation, filtering, load compensation, and working condition prediction and pre-adjustment by inputting four-dimensional scene coordinate data; Based on the adaptation priority, core optimization strategies are selected, and combined with the parameter adaptation requirements and risk avoidance points in the decision support data, preliminary parameter optimization execution instructions are generated, including interpolation parameter calculation rules, filter coefficient values, compensation coefficient formulas, and pre-adjustment time window settings. The initial instructions undergo compliance review and feasibility verification. Instructions that are illegal or infeasible are removed, and the instruction parameters are fine-tuned to form the final parameter-optimized execution instructions.

[0011] Preferably, the specific process for obtaining the cross-system closed-loop execution result is as follows: The parameter optimization execution instructions are split into closed-loop levels and synchronized to the EPS first-level angle closed-loop control unit and the second-level speed closed-loop control unit through the software interface; After receiving the command, the cornering closed-loop control unit updates the dynamic proportional coefficient and starts the deviation calculation and command speed generation module; After receiving the command, the speed closed-loop control unit updates the dynamic PI parameters and starts the speed deviation calculation and command current generation module. Deploy the execution result acquisition module to monitor the control status of the two-level closed loop in real time, and collect actual rotation angle, motor speed, motor current and abnormal feedback information; A preliminary analysis of the performance data is conducted to calculate the deviation between the core indicators and the expected targets. The control status, performance data, and abnormal information are integrated to generate cross-system closed-loop execution results.

[0012] Preferably, the specific process for obtaining the parameter credibility rating and the three-dimensional quantitative evaluation results is as follows: Based on structured log recording of key node information throughout the entire process of data collection, transmission, processing, instruction generation, and system execution, a traceable data chain is formed; Assign a unique data fingerprint and task ID to each parameter data, establish a multi-dimensional traceability index, and support querying the entire process data by data fingerprint, task ID or scene coordinates; Construct a parameter credibility assessment model and calculate credibility scores based on data source reliability, data integrity, and data consistency. Design a three-dimensional quantitative evaluation index system, including responsiveness, stability, and security. Each index is weighted and summed according to preset weights, and the three-dimensional quantitative evaluation results are generated through standardization transformation. Output parameter credibility scores and three-dimensional quantitative evaluation results.

[0013] Preferably, the specific process for obtaining the EPS dynamic parameter intelligent optimization control scheme is as follows: Integrate standardized datasets, decision support data, cross-system closed-loop execution results, parameter reliability, and three-dimensional quantitative evaluation results to form a full-link optimization data pool; Activate the indicator diagnosis module to compare the three-dimensional quantitative evaluation results with the preset target values ​​and identify the shortcomings and core issues in parameter optimization. Multiple parameter adjustment schemes were generated to address the core issues. The effectiveness of the schemes was verified through small-scale bench testing. Ineffective schemes were iteratively eliminated, and the optimal adjustment scheme was retained. The optimal adjustment scheme is transformed into standardized control logic. The input parameters are iteratively optimized, and the parameters are dynamically adjusted in combination with real-time operating conditions and load data. The result is an intelligent optimization control scheme for EPS dynamic parameters that is adapted to multiple operating conditions and complex loads.

[0014] The present invention also provides a dynamic optimization system for the outer loop controller parameters of an automotive EPS, the system being used to execute the aforementioned dynamic optimization method for the outer loop controller parameters of an automotive EPS.

[0015] The present invention also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the aforementioned method for dynamic optimization of parameters of an automotive EPS outer loop controller.

[0016] The beneficial effects of this invention are: 1. By simulating various assisted driving conditions on a test bench, a "condition feature-optimal parameter" mapping sample library is constructed by collecting multi-source synchronous signals, providing a precise data source for dynamic parameter adaptation. Based on this, a multi-dimensional lookup table module is developed. Taking vehicle speed range and steering angle deviation level as input, combined with dual-range interpolation and a first-order low-pass filtering strategy, it can achieve real-time output and smooth transition of dynamic parameters. To accurately match control requirements under different conditions, in low-speed, large-steering-angle-deviation scenarios such as automatic parking, larger dynamic parameter values ​​are output to accelerate steering angle convergence and avoid response lag; in high-speed, small-steering-angle-deviation scenarios such as lane keeping, smaller dynamic parameter values ​​are output to suppress steering overshoot and oscillation, ensuring driving stability. Simultaneously, the interpolation and filtering strategies effectively eliminate parameter abrupt changes during condition switching, avoiding steering jerking. This solves the problem that traditional static parameters cannot accommodate multiple conditions and have poor adaptability, comprehensively improving the steering responsiveness and smoothness of EPS under different conditions.

[0017] 2. By integrating multi-source load signals such as road surface adhesion coefficient, tire pressure, and steering angular velocity, a dynamic model of steering resistance torque is built, enabling accurate calculation of actual steering load under different driving scenarios. The generated load compensation coefficient is integrated with the output parameters of the lookup table module, forming a dual optimization parameter of "condition adaptation + load compensation". This specifically addresses parameter adaptation deviations caused by differences in loads on complex road surfaces. In low-adhesion scenarios such as icy and snowy roads, or in high-load conditions such as abnormal tire pressure and high steering speed, the compensation coefficient is increased to enhance the proportional and integral effects of control parameters, increasing motor output torque to offset the load impact. In conventional load scenarios such as dry asphalt roads, the compensation coefficient is maintained within a reasonable range to avoid control instability caused by excessive parameter adjustments. Ultimately, this ensures that the EPS maintains precise steering control under complex load scenarios, reducing angle tracking errors and speed fluctuations caused by load changes, and improving the stability and reliability of the steering system.

[0018] 3. A lightweight CNN+LSTM driving condition switching prediction model, based on assisted driving function commands and historical steering data, can accurately identify the driving condition switching intention and target driving condition type 20-50ms in advance, and calculate pre-adjustment parameters by combining load compensation coefficients. Through a linear transition time window and deviation correction mechanism, parameters are smoothly pre-adjusted before driving condition switching. This effectively solves the pain point of traditional dynamic parameters' "passive response to driving conditions and delayed switching". In scenarios such as switching from lane keeping to emergency lane change, and automatic parking to urban roads, the pre-adjusted parameters ensure a smooth transition of steering action at the moment of driving condition switching, avoiding problems such as slow steering response and overshoot caused by parameter lag. The deviation correction mechanism further calibrates the parameter accuracy to ensure that the actual control effect is consistent with the expected effect of the target driving condition. Ultimately, this improves steering safety and comfort during driving condition switching in assisted driving, avoiding a decline in driving experience or safety risks caused by parameter lag. Attached Figure Description

[0019] Figure 1 A flowchart of a method for dynamic optimization of parameters of an automotive EPS outer loop controller; Figure 2 This is a flowchart illustrating the control process of a dynamic optimization method for the outer loop controller parameters of an automotive EPS (Electric Power Surgery) system. Detailed Implementation

[0020] The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention.

[0021] It is understood that the term "a" should be understood as "at least one" or "one or more," that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0022] like Figure 1 and Figure 2As shown, this solution constructs a basic control architecture of "turning angle closed loop - speed closed loop". First, it collects multi-source signals and builds a sample library of "operating condition characteristics - optimal parameters". Then, based on the sample library, it develops a multi-dimensional lookup table module to achieve dynamic basic parameter adaptation. Subsequently, it introduces steering resistance torque compensation to optimize parameter accuracy and constructs an operating condition prediction mechanism to solve the switching lag problem. Finally, after bench testing and real vehicle verification and optimization, the entire logic is integrated into a standardized algorithm and solidified into the EPS controller, forming a complete optimization system of "basic adaptation - precise calibration - risk prediction - engineering implementation", comprehensively improving the steering responsiveness, stability, and safety of EPS under multiple operating conditions and complex scenarios.

[0023] Furthermore, by constructing a two-level cascaded control architecture of "turning angle closed loop - speed closed loop" (using P control for the turning angle closed loop and PI control for the speed closed loop), and by pre-setting static control parameters based on EPS hardware specifications and basic steering requirements, the problem of EPS failing to achieve closed-loop steering without a basic control framework is solved. This provides stable underlying control logic for subsequent dynamic parameter optimization and solidifies the functional foundation of EPS outer-loop control. The specific implementation process is shown below: First, the compatibility of the EPS hardware system and interface standardization are carried out. This step must strictly comply with the automotive industry's ISO26262 functional safety standard and GB / T35363-2023 EPS technical specifications to ensure that the hardware solution can be directly integrated into the mass production process. The core hardware specifications should focus on three principles: "performance compliance + cost control + supply chain stability": The microcontroller unit (MCU) should preferably be an automotive-grade 32-bit processor supporting ASIL-B functional safety, with an operating frequency ≥100MHz, at least two independent cores for control logic and safety monitoring, and on-chip Flash capacity ≥1MB and RAM capacity ≥64KB to meet the steering system's program storage and real-time computing requirements; the motor drive module adopts an integrated IGBT solution (automotive-grade single-tube or half-bridge integrated product), with an output power range of 1.5-3kW, and must match the rated power of the steering motor (according to vehicle model classification: A). The 0-class vehicle uses a 1.5kW motor, while the B-class vehicle uses a 2.5-3kW motor. It also features overcurrent, overtemperature, and short-circuit protection functions, as well as a fault diagnosis feedback pin. The angle sensor adopts an automotive-grade CAN bus output product with an accuracy of ≤0.1° and a sampling frequency of ≥100Hz. It supports fault self-diagnosis (such as signal loss and out-of-range alarm) and the ISO15031 diagnostic protocol. The speed detection unit reuses the incremental encoder (resolution ≥1024 lines) built into the motor. It captures pulse signals through the controller timer, and the sampling frequency is set to 1kHz to ensure that it can capture the dynamic changes of the motor within the speed range of 0-3000rpm.

[0024] Standardized hardware interface connections must be implemented based on the electrical architecture specifications of mass-produced vehicles: VCU command angle signals are connected to the EPS controller via the vehicle's CAN network, with the CAN node address preset to 0x220 (compliant with SAEJ1939 protocol), data transmission baud rate set to 500kbps, and signal period 10ms; the EPS angle sensor is connected to the controller's analog input port (AI1 channel) via shielded twisted-pair cable, with a signal range of 0-5V corresponding to an angle of -90° to +90°, and a TVS transient suppression diode (such as SMBJ6.5CA) added at the interface to resist electromagnetic interference; the speed signal is connected to the controller's digital input port (DI1, DI2 channels) via the A and B phase pulses of the motor encoder, and the speed is calculated using an orthogonal decoding mode, with shielded twisted-pair cables used for the interface, and a distance of ≥200mm from high-voltage cables to avoid interference; the controller's PWM output port (PWM1 channel) is connected to the control signal input terminal of the motor drive module, with the PWM frequency set to 10kHz, and a duty cycle of 0-100% corresponding to a motor current of 0-rated current. All interfaces use waterproof connectors (IP6K9K protection rating), and conductive paste is applied to the terminals to reduce contact resistance, ultimately forming a hardware closed-loop link of "command input - signal feedback - control output", which can be directly adapted to the automated assembly and testing process of the vehicle production line.

[0025] Based on a hardware closed-loop link, a two-level serial control architecture of "angle closed loop - speed closed loop" is built. The core is to achieve precise response and stable execution of angle commands through explicit control logic allocation. The angle closed loop, as the first-level control loop, primarily aims to quickly match the angle commands from the VCU. Therefore, a proportional (P) control mode is adopted. This mode can achieve rapid adjustment by instantly amplifying the deviation signal, effectively avoiding the response lag problem that may be caused by the integral link. Specifically, the system first synchronously acquires two types of core signals: one is the angle command issued by the VCU through the vehicle's CAN network. (Signal period 10ms, accuracy 0.01°), and secondly, the actual rotation angle collected in real time by the EPS rotation sensor. (After linear calibration, the 0-5V voltage signal corresponds to a rotation angle of -90° to +90°.) The deviation between the two values ​​is then calculated. The calculation formula is: At the same time, a ±10° deviation limiting mechanism is set to prevent extreme deviations caused by sudden changes in VCU commands or sensor malfunctions, thus avoiding control instability.

[0026] proportionality coefficient The initial value is written to the controller via the calibration interface and set according to the vehicle type classification (A0 class vehicle). A-class car B-segment car ), through control formula Calculate command speed At the same time, increase the speed limit ( 80% of the motor's rated speed, i.e., A0 class vehicle. B-segment car To prevent motor overload, the integral coefficient of the closed-loop rotation angle is clearly defined to meet functional safety requirements. Differential coefficients At the same time, fault detection logic is added: when For more than 100ms, or and Upon signal loss, the security degradation mode is immediately triggered, and Set to 0 and send a fault code (0x05: Cornering closed-loop abnormality) to the VCU. Cornering closed-loop output signal. The data is transmitted to the next-level speed closed loop through the internal signal interaction interface of the software. The data update cycle is consistent with the steering angle signal cycle (10ms), ensuring the coordinated synchronization of the two-level closed loop.

[0027] The speed closed loop, as the second-level execution loop, is responsible for converting the commanded speed into the actual motor movement. It employs a proportional-integral (PI) control mode to balance response speed and steady-state accuracy. The actual speed is acquired in real time via a motor encoder. Calculate the speed deviation Then, through the PI control formula Generate command current signal The output is sent to the motor drive module to control the steering motor. Considering the small load fluctuation of the steering motor, the derivative element has limited improvement on the control effect and easily amplifies high-frequency noise, so the derivative coefficient of the speed closed loop is set. The proportional element quickly offsets instantaneous deviations, while the integral element gradually eliminates steady-state deviations, ensuring that the motor speed stably tracks the command value. The two-stage closed loop achieves real-time data transmission through signal interaction, and the output of the angle closed loop... As the input to the speed closed loop, it forms a series control logic, avoiding the problems of "slow angle response" or "large speed fluctuation" that occur in single closed-loop control.

[0028] Based on hardware specifications and basic steering requirements, the static control parameters are scientifically preset. Parameter presets must adhere to the principle of "hardware capacity limit - basic function compliance - control stability priority": for the steering angle closed-loop proportional coefficient... This needs to be set in conjunction with the steering system's gear ratio (typically 14-20:1) and the MCU's processing power, with an initial value of 0.8-1.2 (dimensionless) to ensure that... Output at time Within the 30%-50% range of the motor's rated speed, this avoids overshoot due to excessive speed while ensuring basic response speed; the speed closed-loop proportional coefficient The preset value is based on the motor's current response speed, calculated according to the motor's rated current (usually 5-15A) and rated speed (usually 1500-3000rpm), with an initial value of 0.02-0.05. This causes the speed deviation. Output current Reaching 20%-50% of the rated current; integral coefficient This requires balancing the elimination of steady-state error with the risk of integral saturation; the initial value is taken as 0.005-0.01. By using integral limiting (set to 80% of the rated current), excessive current caused by long-term deviations is avoided. All static parameters are written to non-volatile memory (EEPROM) through the controller's calibration software to ensure that parameters are not lost after power failure, meeting the usage requirements of mass production scenarios.

[0029] Finally, basic functional verification tests were conducted to ensure that the control architecture and static parameters met the closed-loop steering requirements. The tests were performed on an EPS test bench, simulating a basic driving speed of 0-60 km / h. The test bench control unit issued continuous steering angle commands from 0-30°, and the actual steering angle was simultaneously collected. Motor speed and current Data. Verification metrics include: corner response latency ≤ 100ms (from command issuance to...). achieve 90% of the speed), steady-state speed error ≤5% and The difference percentage and the motor current should not fluctuate abnormally (peak value not exceeding 120% of rated current). If the angle response is too slow, the speed can be appropriately increased. If the speed fluctuates too much, it can be fine-tuned. Reduce the proportional effect; if there is a persistent speed deviation, increase it. Enhance the integral effect. Through repeated testing and parameter fine-tuning under 20 different operating conditions, a set of stable static parameters was finally determined to ensure that the system can achieve "VCU command angle" even without dynamic optimization. →Actual turning angle The precise following of the "" completes the closed-loop steering basic function closed loop, providing a stable underlying control framework for subsequent dynamic parameter optimization combined with multiple working conditions.

[0030] Furthermore, by collecting multi-source signals such as vehicle speed, commanded steering angle, and actual steering angle deviation, typical assisted driving conditions such as automatic parking, lane keeping, and steering reversal are simulated on a test bench, and the optimal control parameters are recorded. This constructs a "condition feature-optimal parameter" mapping sample library, addressing the lack of multi-condition data support in traditional parameter design. This provides accurate data input tailored to real-world scenarios for dynamic parameter adjustment, solidifying the data source foundation for dynamic parameter adaptation. The specific implementation process is shown below: First, a standardized multi-source signal acquisition system was built to ensure high data accuracy and synchronization. The acquired parameters cover three core dimensions: operating condition characteristics, including vehicle speed (acquired via a bench speed simulation module, range 0-120km / h, accuracy 0.1km / h, sampling frequency 100Hz), VCU driver assistance function commands (such as CAN commands for automatic parking and lane keeping, period 10ms, sampling frequency 100Hz), and steering condition type (labeled as low-speed large angle, high-speed small angle, steering reversal, etc.); and control status, including commanded steering angle. Actual turning angle Angular deviation Command speed Actual speed (All data is synchronously acquired from the EPS controller via CAN bus and calibration software, with a timestamp accuracy of 1ms and a uniform sampling frequency of 100Hz); Performance outputs include steering force feedback values ​​(acquired by bench force sensors, range 0-500N, accuracy 0.5N, sampling frequency 100Hz) and steering response delay (the time difference between the calculation command issued by the high-speed data acquisition card and the actual response). All acquisition devices are synchronized via PTP clock, with a time deviation ≤0.5ms. Data is stored in real-time in a MySQL database, with each data entry containing 32 fields to meet feature extraction requirements.

[0031] Next, typical assisted driving conditions were reproduced on the EPS test bench. During the automatic parking simulation, the vehicle speed was set from 0-5 km / h, and the VCU issued continuous steering angle commands from -45° to +45° (change rate 5 degrees / s). The test bench load module applied a load of 0-15 degrees. The drag torque is used to reproduce parking scenarios with different ground friction coefficients. During lane-keeping simulation, the vehicle speed is set to 60-120 km / h, and the VCU issues small turning angle commands of ±5° (change rate 0.5 degrees / s), with a drag torque of 5-10. Simulates high-speed steering correction; during steering and reversing simulation, the vehicle speed is 10-40 km / h, the VCU issues a rapid reversing command from -30° to +30° (change rate 20 degrees / s), and the resistance torque fluctuates sinusoidally from 5-20 degrees / s. This reproduces the unevenness of the road surface and the changes in load. Three sets of parameter combinations are set for each working condition. , , Based on the static parameters, adjust by ±20% (with a step size of 0.05). Run each group 10 times consecutively, 30 seconds each time. Take the last 5 stable data as the valid sample to ensure statistical significance. The termination condition of the control variable method is: when the comprehensive score of 3 consecutive test groups is ≥90 points, or when the parameters are adjusted to ±40% and still cannot meet the standard, record the current optimal value and mark the working condition boundary.

[0032] Finally, optimal control parameter calibration and sample library construction were carried out. Multi-dimensional performance evaluation metrics were defined: corner tracking accuracy. ( Steady-state value ≤ 0.5°, steering response delay (≤100ms), speed fluctuation ( and Difference ≤ 5%), current stability ( For fluctuation ranges ≤ 1A, the weights are 40%, 30%, 20%, and 10% respectively, and the weighted comprehensive score of the working condition is calculated (scoring formula: (A score of ≥90 is optimal). For each working condition, parameters are adjusted using the controlled variable method. If the score is below 80, the adjustment range is expanded (±40%). After calibration, the data is preprocessed (outliers are removed, labels are added, and feature values ​​are calculated), ultimately constructing a "working condition feature vector (speed range, turning angle range, resistance torque level, etc.) - optimal parameter vector (…)". , , The mapping relationship between ")" is established. The initial sample library contains 8 types of working conditions and 24 sub-scenarios, with a total of 1,920 valid data entries. It supports multi-condition queries and can be continuously expanded through real vehicle road tests to provide accurate and complete data source support for dynamic parameter adaptation.

[0033] Furthermore, a multi-dimensional lookup table module (using vehicle speed and steering angle deviation as input and output dynamic control parameters) is built based on a working condition sample library, and the module is optimized for smoothness. This solves the problem that traditional static parameters cannot take into account multiple working conditions (large steering angle at low speed, small steering angle at high speed), providing EPS with basic control parameters that dynamically change with working conditions, and laying a solid foundation for timely steering response and stable control under multiple working conditions. The specific implementation process is as follows: First, the core architecture of the multi-dimensional lookup module is designed, clarifying the input-output mapping logic and data indexing rules. The module uses "vehicle speed range" and "turning angle deviation level" as core input dimensions, both employing hierarchical quantization to improve lookup efficiency and adaptation accuracy: the vehicle speed dimension is divided into 5 ranges based on typical assisted driving scenarios, namely... [0-5km / h] (Automatic parking exclusive zone) [6-40km / h] (Low speed driving / steering / reversal range) [41-60km / h] (Urban road driving range) [61-100km / h] (High-speed cruising range) [101-120km / h] (high-speed limit range), each range corresponds to the optimal parameter set for the same speed range in the sample database (e.g., Corresponding parameters , Corresponding parameters (and so on); the angular deviation dimension is based on Absolute values ​​are divided into four levels, namely... [0-1°] (minor correction level) [1°-5°] (Standard steering level) [5°-10°] (Large turning angle steering level) [10°+] (limit correction level), the level classification corresponds one-to-one with the angle deviation labels in the sample library (e.g., ...). Corresponding parameters , Corresponding parameters The module outputs three sets of dynamic control parameters, namely the closed-loop proportional coefficient of the steering angle. Speed ​​closed-loop proportional coefficient Speed ​​closed-loop integral coefficient It directly connects to the parameter input terminal of the EPS two-stage series control architecture, replacing the original static parameters.

[0034] Subsequently, the engineering implementation of the table lookup module and data index optimization were completed. Based on the MCU hardware resources of the EPS controller (Flash capacity ≥ 1MB, RAM capacity ≥ 64KB), the mapping relationship of "vehicle speed range - steering angle deviation level - optimal parameter" in the sample library was solidified into the controller storage unit in a two-dimensional array structure. The array index is generated by combining the vehicle speed range code (0-4) and the steering angle deviation level code (0-3), for example, "vehicle speed 30km / h (belonging to..."). , code 1) + =3° (belongs to) The index corresponding to "encoding 1)" is 1×4+1=5. This index allows direct access to the data stored in the array. , , The parameter values ​​have a single lookup response time of ≤1ms, meeting real-time control requirements. To avoid lookup failures under extreme conditions, boundary handling logic is added: when the vehicle speed exceeds the maximum range of the sample database (>120km / h), the parameter range [101-120km / h] is called by default; when... When it exceeds 10°, press The module calls parameters based on the level and simultaneously triggers a parameter warning signal to be fed back to the VCU. In addition, the module supports online update function, which can receive new operating condition-parameter mapping data via CAN bus, refresh the storage array in real time, and adapt to the needs of subsequent sample library expansion.

[0035] Next, we optimized the smoothness of the lookup module to address the steering jerking issue caused by sudden parameter changes during transitions between different operating conditions. The optimization scheme employs a combination strategy of "dual-interval interpolation + first-order low-pass filtering": when the actual vehicle speed or steering angle deviation falls within two grade boundaries, dynamic parameters are calculated using a linear interpolation algorithm. The interpolation formula is as follows: ; in, These are the interpolated dynamic parameter values; The optimal parameter value corresponding to the current adjacent interval / level (e.g., the vehicle speed in the previous interval). The previous level of angular deviation (corresponding parameters); The optimal parameter value corresponding to the next adjacent interval / level (e.g., the previous interval of vehicle speed). The previous level of angular deviation (corresponding parameters); The actual input value collected (vehicle speed) Or angular deviation ); This represents the maximum value of the current adjacent interval / level. It represents the maximum value of the next adjacent interval / level.

[0036] The parameter interpolation logic for the steering angle deviation boundary is consistent with the vehicle speed to ensure continuous parameter transition under boundary conditions. Simultaneously, a first-order low-pass filter is added to the parameter output; the filter formula is as follows: ; in This is the filter coefficient, with a value of 0.2-0.4 (dynamically adjusted according to the operating condition switching frequency: low-speed operating condition). Improved smoothness; high-speed operation (to ensure response speed) Output parameters for the current moment. The output parameters from the previous moment are filtered to eliminate current fluctuations and steering force impacts caused by parameter abrupt changes.

[0037] Finally, module function verification and performance calibration tests were conducted. The tests simulated multi-condition switching scenarios on the EPS test bench, covering typical steady-state operating conditions and boundary transition scenarios. In the typical steady-state operating condition test, representative combinations such as low-speed large-angle (corresponding to the low-speed range and large-angle level) and high-speed small-angle (corresponding to the high-speed range and small-angle level) were selected. The module was continuously run and its output was recorded under each operating condition. , , Dynamic parameters are used to verify key indicators such as steering response delay, steering angle tracking accuracy, and speed fluctuation, ensuring that control performance requirements are met under various operating conditions. In the operating condition boundary switching test, scenarios involving cross-range vehicle speed switching, cross-level steering angle deviation switching, and combined switching are simulated to verify the smoothness of parameter transitions. Current fluctuations must be controlled within a set threshold, and steering force must be free of noticeable jerking. Through testing with no fewer than 20 different operating condition combinations and iterative fine-tuning of the filter coefficients, the optimal filter parameter matrix is ​​finally determined. This ensures that the module can quickly respond to changes in operating conditions across the entire range while maintaining the stability and comfort of steering control, providing the EPS system with dynamic control parameters that are both timely and smooth.

[0038] Furthermore, by collecting signals such as road surface adhesion coefficient, tire pressure, and steering angular velocity, a dynamic model of steering resistance torque is constructed to calculate the actual resistance load, generate and optimize the load compensation coefficient, and combine it with the output parameters of the lookup table module to obtain accurate calibration parameters. This solves the problem of parameter adaptation deviation caused by the differences in load on complex road surfaces, providing precise control parameters for EPS load adaptation and laying a solid foundation for ensuring steering control accuracy in complex driving scenarios. The specific implementation process is as follows: First, a load-related multi-source signal acquisition system was built to ensure the real-time performance and reliability of load characteristic data. The core signals acquired include three categories: First, the road surface adhesion coefficient (obtained via the vehicle's ESC system CAN signal, with a preset signal ID of 0x230, data format of 8-bit floating point, range of 0.1-1.0, update cycle of 20ms, covering typical scenarios such as icy and snowy roads and dry asphalt roads); second, tire pressure (collected via a TPMS tire pressure monitoring system, with a range of 1.8-3.0 bar, accuracy of ±0.1 bar, sampling frequency of 10Hz, and the average pressure of both sides used as the input for load calculation); and third, steering angular velocity (calculated differentially from the steering angle sensor signal, using the formula...). ,in This represents the actual angle difference between two adjacent samples. The sampling period is 0-50° / s, and the calculation result range is 0-50° / s. High-frequency noise is eliminated by a first-order low-pass filter (cutoff frequency 10Hz). All signals interact synchronously with the EPS controller via the CAN bus, using the same PTP clock synchronization protocol as described above to ensure that the signal time deviation is ≤0.5ms, forming a data synchronization link with the input signal of the lookup table module.

[0039] Subsequently, a dynamic model of steering resistance torque was constructed to accurately calculate the actual steering load. The model is based on the mechanical characteristics of the EPS system and the influence of actual driving load, employing a multi-factor linear fitting algorithm. The core formula is: ; in: Actual steering resistance torque (unit: ); The coefficient of adhesion is a dimensionless coefficient of friction. The adhesion coefficient influence coefficient (value 8-12) According to bench tests, the larger the adhesion coefficient, the greater the contribution of the drag torque. Standard tire pressure (set to 2.5 bar). This represents the actual average tire pressure (bar). Tire pressure influence coefficient (value 3-5) / bar (when tire pressure is lower than the standard value, the drag torque increases linearly). The turning angular velocity (degrees / s) The angular velocity influence coefficient (value ranges from 0.02 to 0.05). / degree, the faster the steering speed, the greater the dynamic resistance torque); Basic resistance torque (value 2-3) (This covers fixed loads such as mechanical friction and return torque of the steering system).

[0040] Model parameters ( , , , Calibration was achieved through bench load simulation tests: different adhesion coefficients were simulated on the bench (achieved by adjusting the material of the bench friction plate), tire pressure was simulated (different tire pressure values ​​were set by the air pressure regulating device), and steering angular velocity was simulated (steering angle commands with different rates of change were issued by the bench control unit). The actual resistance torque output by the bench was collected simultaneously, and the optimal parameter values ​​were obtained by fitting using the least squares method to ensure that the model calculation error was ≤5% (compared with the actual measured resistance torque on the bench).

[0041] Next, a load compensation coefficient is generated and optimized to achieve precise matching between the load and dynamic parameters. The load compensation coefficient is designed as a dimensionless coefficient positively correlated with the drag torque. The core logic is: the greater the resistance torque, the larger the compensation coefficient. By enhancing the proportional and integral actions of the control parameters, the motor output torque is increased, thus offsetting the load effect. The compensation coefficient calculation consists of two steps: the first step is based on the resistance torque range classification, and... Divided into 3 intervals ( , , ), corresponding to the initial compensation coefficient The second step involves nonlinear optimization correction, introducing a correction factor. (When the adhesion coefficient deviates from the normal value of 0.6, the compensation coefficient is fine-tuned.) The final compensation coefficient formula is: ; Simultaneously set a compensation coefficient limit (0.8≤ (≤1.8), to avoid overcompensation under extreme loads that could lead to control instability.

[0042] Finally, the compensation coefficient and lookup table parameters are fused to output accurate calibration parameters. The fusion logic adopts a "weighted superposition of parameter components" method, and fusion formulas are designed for the three sets of dynamic parameters respectively: ; ; ; Where the integral coefficient An additional correction term of 0.1 is added to enhance the integral term's ability to compensate for steady-state load deviations. The fused calibration parameters are directly integrated into the EPS two-level control architecture, replacing the original lookup table module's output parameters, forming a dual-optimized control logic of "operating condition adaptation + load compensation".

[0043] Furthermore, by collecting assisted driving function commands and historical steering data (vehicle speed change rate, steering angle change rate, etc.), a condition switching prediction model is trained. This allows for advance parameter pre-adjustment and the establishment of a deviation correction mechanism, addressing the issue of parameter response lag during condition switching. This provides the EPS with proactive adaptation parameters before condition switching, laying a solid foundation for the smoothness and safety of assisted driving steering. The specific implementation process is as follows: First, a system for acquiring and preprocessing input signals for the prediction model is built to ensure that the data meets the requirements for model training and real-time inference. The core signals acquired are divided into two categories: one is driver assistance function command signals (acquired through the vehicle's CAN network, with a preset signal ID of 0x240, data format of 8-bit integers, covering 6 typical functions including automatic parking (0x01), lane keeping (0x02), and emergency lane change (0x03), with a command issuance cycle of 10ms, capturing function switching intentions in advance); the other is historical steering state data, selecting feature parameters strongly correlated with operating condition switching, including the rate of change of vehicle speed (…). , Range -5 5 ), rate of change of angle ( Range -30 30 degrees / s), steering duration (the duration of continuous steering under current operating conditions, ranging from 0 to 100 degrees / second), and steering speed (the duration of continuous steering under current operating conditions, ranging from 0 to 100 degrees / second). 5s), load change rate ( Based on the drag torque model mentioned above, the range is -2. 2 All collected data is stored in a sliding window of 100ms, retaining the most recent 5 seconds of historical data as model input samples, while simultaneously using 3... The criteria remove outliers and use normalization (mapping the data to the [0,1] interval) to eliminate dimensional differences and improve model training efficiency.

[0044] Subsequently, a working condition switching prediction model was constructed, which uses machine learning algorithms to accurately predict the switching intention and target working condition. The model adopts a lightweight CNN+LSTM hybrid architecture, specifically: 1 convolutional layer (16 kernels, 3×3 kernel size, stride 1, ReLU activation function) → 1 max pooling layer (pooling size 2×2) → 2 LSTM units (32 neurons per layer, dropout rate 0.2) → 1 fully connected layer (20 neurons, ReLU activation function) → output layer (2 branches: binary classification output activated by Sigmoid, target working condition output activated by Softmax), adapted to the MCU hardware resources of the EPS controller (operational frequency ≥100MHz, RAM ≥64KB), ensuring real-time inference latency ≤5ms. The training data comes from bench test data, covering 8 driving conditions and 24 sub-scenarios, totaling 100,000 samples. The driving conditions are distributed as follows: automatic parking 20%, lane keeping 25%, steering and reversing 20%, urban road driving 15%, highway cruising 10%, emergency lane changing 5%, low-speed driving 3%, and extreme driving conditions 2%. Data augmentation methods include: temporal signal noise injection (Gaussian noise, mean 0, variance 0.01), temporal perturbation (time window offset ±10ms), and parameter scaling (±5%). The model input is an 8-dimensional feature vector (assisted driving function commands + 7-dimensional historical state parameters), and the output is a binary classification result of "whether to switch driving conditions" (0=no switch, 1=switch) and the predicted target driving condition type (corresponding to the combination of the 5 vehicle speed ranges + 4 steering angle deviation levels mentioned above).

[0045] The model training process is as follows: 100,000 sets of historical data containing operating condition switching (covering different function commands, speed ranges, and load scenarios) are extracted from the sample library and divided into training, validation, and test sets in a 7:2:1 ratio. During training, the goal is to minimize the cross-entropy loss function. Gradient descent is used to iteratively optimize the model parameters. An early stopping mechanism is used to avoid overfitting, and the network layer parameter size is pruned to ensure that the model weight file size is ≤200KB, which can be stored in the MCU's Flash storage unit. Model deployment optimization uses INT8 quantization and structured pruning: quantization uses a symmetric quantization method to convert weights and activation values ​​from 32-bit floating-point to 8-bit integers; pruning targets fully connected layers and LSTM layers, with a pruning rate of 30%, preserving core feature extraction channels. After training, the model's accuracy on the test set is ≥95%, and the target operating condition prediction error is ≤1 interval / level, meeting the requirements of engineering applications.

[0046] Next, a parameter pre-adjustment strategy is designed to smoothly switch dynamic parameters in advance. When the model predicts a change in operating condition (output result is 1), based on the predicted target operating condition type, the corresponding target dynamic parameters are retrieved from the "operating condition feature-optimal parameter" mapping sample library. , , ), and combined with the current load compensation coefficient Calculate the pre-adjustment parameters: ; ; ; The parameter pre-adjustment adopts the "linear transition + time window control" method: a 50ms transition time window is set (matching the execution delay of the assisted driving function command), and the parameter value is gradually transitioned from the current parameter value to the pre-adjustment parameter in uniform steps to avoid parameter abrupt changes; if the model corrects the prediction result within the transition time window (such as canceling the switch), the transition is stopped immediately and the parameter is reverted to the original parameter to ensure control safety.

[0047] To further improve the adaptation accuracy of the pre-adjustment parameters, a deviation correction mechanism is added. The deviation between the "actual control effect under the pre-adjustment parameters" and the "expected effect under the target operating condition" is calculated in real time. Deviation indicators include steering angle tracking deviation (…). ), speed response deviation ( When the absolute value of the deviation exceeds the set threshold ( or When this occurs, proportional correction is initiated: ; ; ; Simultaneously, a correction coefficient limit is set (0.9 ≤ correction coefficient ≤ 1.1) to avoid over-correction leading to control instability. The corrected parameters are used as the final output and connected to the EPS two-level control architecture to form a closed-loop logic of "prediction-pre-adjustment-correction".

[0048] Finally, bench validation tests were conducted on the model and pre-adjustment strategies. The tests followed the principles of "covering typical scenarios + focusing on core indicators + quantifying the verification effect" to ensure the results have engineering universality. First, the test plan was planned: high-frequency operating condition switching types in assisted driving were selected, including three categories: "steady-state cruise - dynamic intervention", "low-speed steering - medium-speed driving", and "normal steering - emergency steering". Each category included sub-scenarios with different vehicle speed ranges, turning angle levels, and load states. A total of no less than 25 test cases were designed to cover the typical time intervals and load change ranges of operating condition switching. The testing was conducted in a step-by-step manner: In the single-factor verification phase, the model's prediction accuracy (requiring a prediction accuracy ≥95%, with a prediction lead time stable between 20ms and 50ms), parameter pre-adjustment smoothness (requiring no sudden changes in parameters within the transition time window, with an adjacent time difference ≤0.05), and deviation correction effectiveness (requiring deviation convergence time ≤100ms) were tested. In the composite scenario verification phase, a scenario simulating "function switching + load fluctuation" was simulated, collecting indicators such as steering response delay, steering angle tracking accuracy, speed fluctuation, and current fluctuation amplitude, requiring a response delay ≤100ms, tracking accuracy ≤0.3°, speed fluctuation ≤5%, and current fluctuation ≤0.5A. In the extreme condition verification phase, the system stability was tested for high-risk scenarios to ensure no model misjudgments, parameter failures, or steering instability. Through full-process testing and data statistical analysis, the performance differences before and after the pre-adjustment strategy were quantitatively compared, verifying the timeliness of the prediction model and the effectiveness of the pre-adjustment strategy. This solved the pain point of traditional dynamic parameters being "passive in response to operating conditions and with delays in switching," ensuring smooth transition of steering actions when switching operating conditions during assisted driving, and providing transitional protection for the safety and comfort of the EPS system.

[0049] For example, the NXPS32K344 automotive-grade MCU (160MHz clock speed, 2MB on-chip Flash, 256KB RAM) is selected to meet the requirements of complex algorithm operation; the motor drive module adopts the Infineon IRAMS16UP60B integrated IGBT solution (3kW output power, matching the rated power of the B-class car steering motor of 2.8kW), with overcurrent (threshold 30A) and overtemperature (threshold 150℃) protection functions; the angle sensor uses the Bosch BOSCH5549 series (accuracy 0.05°, sampling frequency 100Hz, supports ISO15031 diagnostic protocol); the speed detection uses a motor incremental encoder (2048 lines resolution, sampling frequency 1kHz).

[0050] The hardware interface is configured according to mass production specifications: the VCU command angle is accessed via CAN network (node ​​address 0x220, baud rate 500kbps, period 10ms); the angle sensor signal is connected to the AI1 channel via shielded twisted pair cable (0-5V corresponds to -90° to +90°), and the interface is connected in series with an SMBJ6.5CATVS diode for anti-interference; the speed signal is orthogonally decoded through the DI1 / DI2 channels, and the cable is 250mm away from the high-voltage cable; the PWM output (PWM1 channel, frequency 10kHz) is connected to the motor drive module, and the duty cycle of 100% corresponds to the motor's rated current of 18A.

[0051] Corner closed loop (first-stage loop) proportional coefficient Instruction corner Accuracy 0.01°, actual rotation angle After linear calibration (0V corresponds to -90°, 5V corresponds to +90°), the deviation is... Amplitude limit ±10°. When At that time, the commanded rotational speed (Converted to a motor speed of 1800 rpm, ≤ 80% of the rated speed of 2800 rpm).

[0052] Speed ​​closed-loop (secondary loop) parameters: proportional coefficient Integral coefficient The integral limit is 14.4A (80% of the rated current of 18A). When the speed deviation... At that time, command current Ensure that the current is within the range of 22%-50% of the rated current.

[0053] Vehicle speed is acquired via a bench speed simulation module (0-120km / h, accuracy 0.1km / h, 100Hz); VCU function commands (automatic parking 0x01, lane keeping 0x02, etc., 100Hz); steering force feedback is acquired via an HBMU9C force sensor (0-500N, accuracy 0.5N, 100Hz). All devices are synchronized via a PTP clock with a time deviation ≤0.3ms, and data is stored in a MySQL database (each record contains 32 fields, including vehicle speed, ...). , wait).

[0054] Automatic parking operation: Vehicle speed 3km / h, VCU issues a turning angle command from -45° to +45° (change rate 5° / s), and the test bench applies 8... Drag torque (simulating a slippery road surface). Adjustment. In 1.0-1.4, In the range of 0.03-0.05, In the range of 0.006-0.01, when With a steady-state value of 0.3°, a response delay of 85ms, and a current fluctuation of 0.6A, the overall score is 92 points, and the optimal parameters are recorded. , , .

[0055] High-speed lane keeping operation: vehicle speed 100km / h, VCU issues ±3° steering angle command (change rate 0.5° / s), drag torque 6N·m. Optimal parameters were determined using the controlled variable method. , , At this point, the corner tracking accuracy is 0.25°, and the rotational speed fluctuation is 2.8%.

[0056] The final sample library contains 8 types of working conditions and 24 sub-scenarios, with a total of 1920 valid data points, such as "vehicle speed 3km / h + AngErr=8° + resistance torque 8". "Corresponding parameter vector [1.3, 0.045, 0.009].

[0057] The vehicle speed is divided into 5 ranges: [0-5km / h] [6-40km / h] etc.; Steering deviation level 4: [0-1°]、 [1°-5°] etc. When the vehicle speed is 5.5km / h ( and boundary), ( and (Boundary), calculated using interpolation formulas: Vehicle speed interpolation: ; Angle deviation interpolation: ; Given the current operating conditions, which involve a combination of vehicle speed variations across different speed ranges and steering angle deviations across different levels, we prioritize interpolation based on vehicle speed (as the steering system is more sensitive to vehicle speed). Output parameters after filtering at the previous time step (Obtained by interpolation and filtering from the previous steady-state condition), first-order low-pass filter coefficients (Medium-speed operating condition), filtered parameters To avoid parameter mutations.

[0058] Drag torque model parameters: (adhesion coefficient) Influence), (Tire pressure effect) (Influence of angular velocity) .when (Icy and snowy road surface) , resistance torque Corresponding compensation coefficient .

[0059] Predictive Model and Parameter Pre-adjustment: The model input consists of 8-dimensional features (functional commands + vehicle speed change rate). etc.), when anticipating lane keeping ( , Switch to emergency lane change ( , ), call the target parameter 35ms in advance , combined Pre-adjustment parameters Within a 50ms transition time window, the value transitions from the current 1.292 to 1.299 in steps of 0.006.

[0060] In bench testing, the automatic parking to urban road switching scenario showed smooth parameter transition, current fluctuation of 0.25A, and a steering force jerking score of 8.6. In the icy / snowy road emergency lane change scenario, the response latency was 92ms, and the corner tracking accuracy was 0.35°, both meeting mass production specifications. After 1000km of real-vehicle testing, high-speed cruising steering fluctuation was ≤3%, and low-speed parking steering overshoot was ≤1.5°, indicating overall performance met standards.

[0061] The processes described above with reference to the flowcharts in the embodiments disclosed in this invention can be implemented as computer software programs. The embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wire segments, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless segments, wire segments, optical fibers, RF, etc., or any suitable combination thereof.

[0062] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0063] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the stated principles, the implementation of the present invention may have any variations or modifications.

Claims

1. A dynamic optimization method for parameters of an automobile EPS outer ring controller, characterized in that, The method comprises: Collecting multi-source data, building a collection and transmission link through a PTP clock synchronization and signal preprocessing module, and obtaining a standardized data set through data preprocessing; Based on the standardized data set, the optimal control parameters are calibrated by using the control variable method and the multi-dimensional performance weighted evaluation mechanism, combined with hierarchical rule analysis and sample library mapping relationship backstepping, the parameter adaptability and control risk are perceived through bench test verification and boundary condition checking, and the decision support data of dynamic parameter adaptation is formed; Based on the decision support data, a four-dimensional scene coordinate-strategy adaptation model is constructed to generate parameter optimization execution instructions, and the cross-system closed-loop execution results of the two-level closed loop of EPS are obtained; Based on the key node data of the closed-loop execution results, a traceable system is constructed through structured log recording and data fingerprint checking, and the parameter reliability rating and three-dimensional quantitative evaluation results are output; Integrating the standardized data set, decision support data, closed-loop execution results, and data reliability and three-dimensional quantitative evaluation results, the optimized data is obtained, which is input into the parameter iterative optimization link to output the EPS dynamic parameter intelligent optimization control scheme adapted to multiple working conditions and complex loads.

2. The method according to claim 1, wherein, The specific obtaining process of the standardized data set is as follows: Determine the EPS control core target and data demand, determine the collection range and core index of EPS hardware specification data, working condition characteristic data, load signal data and auxiliary driving function instruction data; Start the multi-source data collection mechanism, build a time synchronization network based on the PTP clock synchronization protocol, call the sensor data collection interface, CAN bus data receiving module and bench test data output module through the signal preprocessing module, and extract the corresponding type data respectively; Perform differential preprocessing on the collected multi-source data, remove abnormal data through the abnormal value elimination rule, complete the missing data by using the interpolation method, and unify the time stamp format, numerical unit and signal coding standard; Based on the data correlation key, the cross-source correlation relationship is constructed, the effective data after preprocessing is integrated, and the standardized data set with data type label and collection timestamp is generated.

3. The method of claim 2, wherein the method further comprises: The specific process of building the collection and transmission link is as follows: Configure the PTP clock synchronization server, assign unique clock node IDs to the EPS controller, sensors and bench test equipment, and establish a master-slave clock synchronization architecture; Deploy the signal preprocessing module, configure dedicated processing units for different types of signals, perform filter noise reduction processing on analog signals, perform edge detection and counting calibration on digital signals, and perform frame analysis and verification on CAN bus signals; Build a data transmission channel, use shielded twisted pair to transmit analog and digital signals, use CAN bus to transmit control instructions and state data, and keep a preset safety distance between signal cables and high-voltage cables to avoid interference; Configure the data transmission verification mechanism, add a verification code to the transmission data, trigger the retransmission mechanism when the receiving end detects data errors, and form a stable collection and transmission link.

4. The method of claim 3, wherein the method further comprises: The specific obtaining process of the decision support data is as follows: Based on the standardized data set, the EPS control parameters are adjusted by using the control variable method, only a single parameter is adjusted each time, and other parameters remain unchanged; A multi-dimensional performance weighting evaluation mechanism is started, corner tracking accuracy, steering response delay, rotational speed fluctuation, and current stability index data are collected, a comprehensive score is calculated according to a preset weight, and parameters with a comprehensive score meeting a standard are selected as candidate optimal control parameters; A hierarchical rule analysis module is deployed, vehicle speed is divided into a plurality of intervals, corner deviation is divided into a plurality of levels, and parameter ranges corresponding to different hierarchical levels are inversely deduced in combination with a mapping relationship in a sample library; Through bench test, the adaptability of the candidate parameters under different working conditions is verified, the parameter stability is checked under simulated boundary conditions, parameter adaptation defects and control risks are perceived, and decision support data is generated by integrating the candidate parameters, hierarchical mapping results, and risk analysis conclusions.

5. The method of claim 4, wherein the method further comprises: The specific process of generating the parameter optimization execution instruction is as follows: Four-dimensional scene coordinate dimensions are defined, including working condition dimension, load dimension, response dimension, and safety dimension, and each dimension index is normalized; A strategy adaptation model is constructed, a parameter optimization algorithm is trained based on the decision support data, and interpolation, filtering, load compensation, and working condition prediction and pre-adjustment adaptation priorities are calculated by inputting four-dimensional scene coordinate data; According to the adaptation priority, a core optimization strategy is selected, a preliminary parameter optimization execution instruction is generated by combining parameter adaptation requirements and risk avoidance points in the decision support data, and the instruction contains interpolation parameter calculation rules, filtering coefficient values, compensation coefficient formulas, and pre-adjustment time window settings; The preliminary instruction is reviewed for compliance and verified for feasibility, instructions that violate rules or are not feasible are removed, instruction parameters are fine-tuned, and a final parameter optimization execution instruction is formed.

6. The method of claim 5, wherein the method further comprises: The specific process of obtaining the cross-system closed-loop execution result is as follows: The parameter optimization execution instruction is split according to the closed-loop level and synchronized to the EPS one-level corner closed-loop control unit and the two-level rotational speed closed-loop control unit through a software interface; After receiving the instruction, the corner closed-loop control unit updates the dynamic proportional coefficient and starts the deviation calculation and instruction rotational speed generation module; After receiving the instruction, the rotational speed closed-loop control unit updates the dynamic PI parameter and starts the rotational speed deviation calculation and instruction current generation module; An execution result collection module is deployed to monitor the control state of the two-level closed loop in real time, collect actual corner, motor rotational speed, motor current, and abnormal feedback information; The effect data is preliminarily analyzed, the deviation of the core index from the expected target is calculated, the control state, effect data, and abnormal information are integrated, and a cross-system closed-loop execution result is generated.

7. The method of claim 6, wherein the method further comprises: The specific process of obtaining the parameter credibility rating and three-dimensional quantitative evaluation result is as follows: Based on structured log record collection, transmission, data processing, instruction generation, and system execution, key node information is collected to form a traceable data chain; A unique data fingerprint and task ID are assigned to each parameter data, a multi-dimensional traceability index is established, and full-process data can be queried through the data fingerprint, task ID, or scene coordinate; A parameter credibility evaluation model is constructed, and the credibility score is calculated based on data source reliability, data integrity, and data consistency; A three-dimensional quantitative evaluation index system is designed, including responsiveness, stability, and safety. Each index is weighted and summed according to a preset weight, and a three-dimensional quantitative evaluation result is generated through standardized conversion. The output parameter credibility score and the three-dimensional quantitative evaluation result are obtained.

8. The method of claim 7, wherein the method further comprises: The specific obtaining process of the EPS dynamic parameter intelligent optimization control scheme is as follows: Integrate the standardized data set, decision support data, cross-system closed-loop execution results, parameter credibility, and three-dimensional quantitative evaluation results to form a full-link optimization data pool; Start the index diagnosis module, compare the three-dimensional quantitative evaluation result with the preset target value, locate the parameter optimization short board and core problem, and generate multiple parameter adjustment schemes for the core problem. Through bench small-range test verification, the schemes with poor effects are iteratively eliminated, and the optimal adjustment scheme is reserved. The optimal adjustment scheme is converted into a standardized control logic, input into the parameter iterative optimization link, combined with real-time working condition and load data to dynamically adjust the parameters, and an EPS dynamic parameter intelligent optimization control scheme suitable for multiple working conditions and complex loads is output.

9. A system for dynamic optimization of parameters of an automotive EPS outer loop controller, characterized in that, The system is used to execute the parameter dynamic optimization method of the automobile EPS outer ring controller according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to realize the parameter dynamic optimization method of the automobile EPS outer ring controller according to any one of claims 1-8.

Images