A suspension self-adaptive adjusting method for new energy vehicles

Abstract

The present application belongs to the technical field of suspension adjustment, and specifically relates to a suspension self-adaptive adjustment method for a new energy vehicle, which comprises the following steps: collecting electric drive chassis and road surface sensing data of the new energy vehicle, identifying a driving condition and processing in a thread pool; analyzing and extracting suspension electric control load information, and generating parameter adjustment queues after modeling in combination with topological bionic subtractive parameters; optimizing by using an improved sparrow search algorithm, constructing a damping stiffness mapping relationship, and outputting a suspension concurrent control sequence; performing multi-parameter collaborative scheduling and vehicle body posture pre-compensation, dividing a working condition interval, calibrating an actuator mode, and obtaining an initial adjustment instruction; dissipating road surface high-frequency vibration through a passive damping channel, balancing control efficiency and energy consumption, and forming an intermediate instruction; collecting battery and road surface data, calculating a matching deviation dynamic compensation, iteratively correcting parameters, and outputting a suspension control instruction. In the present application, multi-source intelligent algorithms are fused for closed-loop control, and full-condition self-adaptation is achieved while ensuring smoothness, energy saving and maneuverability.

Description

Technical Field

[0001] This invention belongs to the field of suspension adjustment technology, specifically a method for adaptive suspension adjustment in new energy vehicles. Background Technology

[0002] Traditional adaptive suspension adjustment methods are not fully adapted to the unique characteristics of electric drives and batteries in new energy vehicles, resulting in low control precision and a lack of full-condition adaptive capability. They rely heavily on single sensor signals, failing to integrate core features such as battery center of gravity shift and electric drive transmission losses. This leads to coarse condition identification and a lack of precise, thread-based load distribution. Furthermore, the absence of topological biomimetic structure optimization and intelligent algorithm-based global optimization prevents coordinated adaptation between suspension structure and electronic control parameters. The lack of a load self-learning iteration mechanism makes it difficult to dynamically compensate for deviations in complex scenarios such as battery charge fluctuations and sudden road condition changes. Control commands are lagging, failing to balance ride comfort, handling stability, and energy economy. The adaptability and intelligence levels fall far short of the requirements for new energy vehicle use. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention proposes an adaptive suspension adjustment method for new energy vehicles. This invention primarily addresses the problems of poor adaptability to operating conditions, lag in parameter tuning, and inability to adapt to the characteristics of new energy vehicles in traditional suspension systems.

[0004] The present invention provides a method for adaptive suspension adjustment in new energy vehicles, comprising: S1: collecting the status signals of the electric drive chassis of the new energy vehicle and the original road perception data, identifying the driving conditions through the dimension of the vehicle's electronic control signals, distributing them to the corresponding independent control thread pool according to the exclusive mapping rules, and outputting the distributed original dataset.

[0005] S2: Based on the original dataset of the split, the suspension electronic load information is parsed and converted into a suspension adjustment data model. The suspension links and wheel end components are subjected to topological bionic material reduction, and the suspension parameters are output in a batch adjustment queue.

[0006] S3: Perform feature coupling modeling on the batch adjustment queue of suspension parameters, use an improved sparrow search algorithm to perform global optimization of the suspension control sequence, perform unsupervised clustering iteration based on driving characteristics, establish a damping stiffness mapping relationship, and generate a concurrent suspension control sequence.

[0007] S4: Perform multi-parameter priority collaborative scheduling on the concurrent control sequence of the suspension, deduce the attitude pre-compensation logic based on the timing characteristics of the motor torque, divide the working condition interval and calibrate the actuator mode to obtain the initial adjustment command of the suspension.

[0008] S5: Based on the initial adjustment command, the passive damping diversion channel dissipates high-frequency vibrations of the road surface, balances the suspension adjustment efficiency and energy consumption level according to the working condition interval rules, and outputs the intermediate control command of the suspension.

[0009] S6: Real-time acquisition of battery data and road surface undulation data, calculation of matching deviation to dynamically compensate for intermediate control commands, continuous correction of parameters, and output of new energy vehicle suspension adjustment commands.

[0010] According to the adaptive suspension adjustment method for new energy vehicles provided by the present invention, the specific steps for outputting the split original dataset in step S1 are as follows: S11: Perform time-series alignment and anomaly denoising filtering on the status signals of the electric drive chassis of new energy vehicles and the raw data of road perception, convert the raw sensor data into standardized fused data, and output a standardized raw fused dataset.

[0011] S12: Construct a feature space for the vehicle's electronic control signals based on the standardized original fusion dataset, extract battery load and motor torque, identify driving conditions through feature threshold matching, and output a feature identification dataset.

[0012] S13: Based on the feature identification dataset, retrieve the new energy vehicle-specific working condition-suspension parameter mapping rules, complete data partitioning and classification and link encapsulation according to working condition categories, and output the working condition classification and encapsulation dataset.

[0013] S14: Based on the classification and encapsulation of the dataset according to the working conditions, match and assign the corresponding data of each working condition to the preset independent control thread pool, complete the data distribution and regularization, and output the split original dataset.

[0014] According to the present invention, a method for adaptive suspension adjustment in new energy vehicles, in step S2, the specific steps for outputting the batch adjustment queue of suspension parameters are as follows: S21: Extract the timing, amplitude, and coupling characteristic values ​​of the suspension electronic control load from the original dataset, remove outlier data, perform timing interpolation to complete the dataset, and output the original numerical set of the suspension electronic control load.

[0015] S22: Perform in-depth numerical extrapolation based on the original numerical set of suspension electronic control load, cluster and screen the parameters of the whole domain according to the working condition characteristic threshold, and perform data block encapsulation and time-series link cache arrangement under different stress scenarios to output a type of suspension calibration cache dataset.

[0016] S23: The classification suspension tuning cache dataset is incorporated into the topological bionic computing link. The stress field is discretely decomposed for the suspension link and wheel end components. The dynamic deduction of the material ratio of the structure is dynamically derived through iterative optimization. The result is a set of bionic material reduction optimized suspension structural parameters.

[0017] S24: Based on the biomimetic subtractive optimization suspension structure parameter set, perform global parameter boundary threshold verification and timing logic rearrangement, adapt to the electronic control system interface specification to complete the unified and regularized numerical format, and output the suspension parameter batch calibration queue.

[0018] According to the present invention, a method for adaptive suspension adjustment in new energy vehicles is provided, S31: collecting the original time-series data of the batch adjustment queue of suspension parameters, extracting the battery center of gravity offset, electric drive transmission path loss and end-vehicle-cloud chassis topology connection features, and outputting a high-dimensional feature coupling dataset.

[0019] S32: Based on the high-dimensional feature coupling dataset, a multivariate constraint coupling mechanism model is built. The suspension control sequence is used as the decision variable. The sparrow search algorithm rules are improved and a composite objective function is set to output the global optimization initial model.

[0020] S33: Based on the global optimization initial model, perform population initialization, position mutation and extreme value update, screen the convergent optimal individual, and output the standard sample set of the optimal suspension control sequence.

[0021] S34: Associate the optimal suspension control sequence sample set with the vehicle operation feature vector, use density clustering for unsupervised iterative partitioning, aggregate samples of the same working condition and complete convergence verification, and output the working condition feature cluster group.

[0022] S35: Based on the characteristics of the working conditions, cluster and group the suspension damping and stiffness parameters to statistically analyze the time-series changes. Then, fit the linkage law through nonlinear regression, construct the parameter linkage function, and output the damping-stiffness dynamic correlation mapping function.

[0023] S36: Substitute the real-time battery center of gravity, electric drive path, and chassis topology-related features into the damping-stiffness correlation mapping function, combine the real-time coupling state, split the control items according to the parallel scheduling logic, and output the complete sequence of suspension concurrent control.

[0024] According to the adaptive suspension adjustment method for new energy vehicles provided by the present invention, the specific steps for outputting the global optimization initial model in step S32 are as follows: Based on the high-dimensional feature coupling dataset, dimensionality normalization and feature denoising are performed to remove redundant noise features and output a standardized modeling input feature matrix.

[0025] Based on the standardized modeling input feature matrix, the constraint relationship among the three is established. The suspension control sequence is used as the optimization decision variable. Multivariate coupling constraint equations and boundary conditions are established to construct a complete model and output a multivariate constraint coupling mechanism model.

[0026] Based on the multivariate constraint coupling mechanism model, the logic of the traditional sparrow search algorithm is deconstructed, and the reconstruction and iterative update rules of the suspension parameter adjustment scenario are adapted to output the sparrow search algorithm framework.

[0027] Based on the sparrow search algorithm framework and driving comfort and handling performance indicators, a weighted composite objective function is constructed, and parameter calibration is completed by matching the model constraint boundary to output the global optimization initial model.

[0028] According to the method for adaptive suspension adjustment in new energy vehicles provided by the present invention, the specific steps for generating the initial suspension adjustment command in step S4 are as follows: S41: Analyze the temporal characteristics of stiffness, damping, and vehicle height parameters in the concurrent control sequence of the suspension, sort and coordinate the multi-parameter priority according to the vehicle stability performance weight, and output the multi-parameter priority coordinated scheduling weight matrix.

[0029] S42: Based on the multi-parameter priority collaborative scheduling weight matrix and the dynamic timing characteristics of motor torque, the vehicle body attitude deviation compensation mechanism is deduced, a closed-loop pre-compensation operation logic is constructed, and the vehicle body attitude pre-compensation deduction model is output.

[0030] S43: Use the vehicle body attitude pre-compensation simulation model to divide the critical boundaries of the driving condition range, calibrate the actuator segmented working mode switching parameters, and output the actuator segmented calibration parameter set.

[0031] S44: Perform global optimization tuning of the passive damping diversion channel configuration parameters based on the actuator segment calibration parameter set, integrate the compensation logic with the actuator calibration results, and output the initial suspension adjustment command.

[0032] According to the adaptive suspension adjustment method for new energy vehicles provided by the present invention, the specific steps in step S42 of outputting the vehicle body attitude pre-compensation deduction model are as follows: Collect dynamic time-series data of motor torque and perform noise reduction and smoothing processing. Extract instantaneous torque value, rate of change and time-series gradient features. After normalization, match the dimensions with the weight matrix to output a torque time-series-weight coupled dataset.

[0033] Based on the torque time-weighted coupling dataset, a correlation function between torque and suspension parameter weights is constructed. The mapping relationship between vehicle body attitude deviation and the two is derived through linear regression. The deviation generation coefficient is quantified, and the attitude deviation correlation derivation formula is output.

[0034] By inputting the closed-loop feedback correction coefficient into the attitude deviation correlation derivation formula, setting the compensation response threshold and timing window, a closed-loop operation logic of deviation prediction, parameter correction, and feedback verification is built, and a vehicle attitude pre-compensation derivation model is output.

[0035] According to the present invention, a method for adaptive suspension adjustment in new energy vehicles, the specific steps for outputting the intermediate suspension control command in step S5 are as follows: S51: Analyze the damping and stiffness adjustment parameters and timing logic in the initial suspension adjustment command, and perform dimensional matching between the command parameters and the channel structure parameters by collecting the configuration parameters of the passive damping shunt channel, and output the command channel parameter matching dataset.

[0036] S52: Based on the instruction channel parameter matching dataset, the damping control component in the initial adjustment instruction is imported into the passive damping shunt channel, and filtering and dissipation calculations are performed on the high-frequency vibration signal of the road surface to output the correction adjustment instruction.

[0037] S53: The input of the correction and adjustment command is based on the preset working condition interval division rules. The quantitative indicators of suspension adjustment efficiency and energy consumption under different working conditions are quantified, a two-way constraint evaluation model is established, the optimal compromise ratio is dynamically calculated, and the efficiency and energy consumption balance control parameter set is output.

[0038] S54: Optimize timing logic based on efficiency and energy consumption balance control parameter set and correction adjustment instructions, and output suspension intermediate control instructions adapted to the current working conditions.

[0039] According to the present invention, a method for adaptive suspension adjustment in new energy vehicles is provided. In step S6, the specific steps for outputting the final adaptive adjustment command for the new energy vehicle suspension are as follows: S61: Real-time acquisition of battery state of charge, motor output power, electronic control system response delay and road surface undulation height time series data, extraction of each data feature component and completion of dimension alignment, output of standardized multi-source working condition real-time dataset.

[0040] S62: Input the standardized multi-source working condition real-time dataset into the preset deviation calculation model, calculate the matching deviation between the suspension intermediate control command and the actual working condition, the time delay deviation in the parameter linkage process, quantify the deviation amplitude and timing characteristics, and output the dual deviation quantization dataset.

[0041] S63: Perform dynamic compensation calculations on the intermediate control commands of the suspension based on the dual-deviation quantization dataset, correct the parameter amplitude and timing trigger nodes, build a load self-learning iterative model, import the deviation data to initialize the model parameters into the load self-learning iterative model, and output the initial self-learning model.

[0042] S64: Input real-time operating condition deviation data into the initial self-learning model for iterative training, correct the model parameters, and output the suspension adjustment command for new energy vehicles.

[0043] According to the method for adaptive suspension adjustment in new energy vehicles provided by the present invention, the specific steps for outputting the initial self-learning model in step S63 are as follows: A dynamic compensation correlation equation is established between the dual deviation and the suspension control parameters. The quantified data of the dual deviation is input into the compensation correlation equation to calculate the compensation amount. The parameter amplitude of the suspension intermediate control command is accurately corrected, and the compensated suspension intermediate control command is output.

[0044] Based on the intermediate control command of the compensated suspension, a load self-learning iterative logic architecture is designed, the iterative convergence condition and timing update rule are set, a closed-loop iterative mechanism for deviation feedback and parameter correction is constructed, and the load self-learning iterative model framework is output.

[0045] The initial weights and bias correction coefficients of the load self-learning iterative model framework are determined by nonlinear fitting, and the initial self-learning model is output.

[0046] This invention provides a method for adaptive suspension adjustment in new energy vehicles, which integrates multi-source intelligent algorithms for closed-loop control, achieving adaptive adjustment under all operating conditions while balancing smoothness, energy saving, and handling.

[0047] The beneficial effects of this invention are as follows: 1. This invention constructs a six-dimensional electronic control feature space and combines it with a threshold matrix to accurately identify 12 typical driving conditions. Then, the BKDR hash mapping algorithm distributes the driving condition data to an independent control thread pool for isolated computation. Relying on a multi-source information full-domain acquisition and refined driving condition identification mechanism, it breaks the limitation of traditional suspension relying solely on a single sensor signal for control. The independent computation of the thread pool avoids the problem of mutual interference between multiple driving condition data, enabling real-time matching of complex driving scenarios such as constant speed cruising, bumpy roads, turning, and climbing. This ensures that the suspension stiffness and damping adjustment parameters always conform to the real-time force characteristics of the driving conditions, significantly improving vehicle ride comfort and steering stability. It adapts to the unique characteristics of electric drive and batteries in new energy vehicles, solving the pain point that traditional suspensions cannot adapt to the load distribution and power output characteristics of new energy vehicles.

[0048] 2. This invention performs iterative optimization of suspension links and wheel-end supports through material reduction, simultaneously combining battery center of gravity offset, electric drive transmission loss, and end-vehicle-cloud topology architecture to construct a high-dimensional coupled model. An improved sparrow search algorithm is used for global optimization, coupled with density clustering to establish a dynamic correlation mapping relationship between damping and stiffness. By combining structural mechanics optimization with intelligent algorithm global optimization, lightweight design is achieved while ensuring suspension structural strength and driving safety, reducing vehicle energy consumption and unsprung mass. Furthermore, it breaks through the traditional fixed-parameter tuning mode, relying on multi-feature coupling constraints to complete the collaborative parameter optimization of multiple suspension units, accurately quantifying the nonlinear linkage law of damping and stiffness, and achieving deep adaptation of structural configuration, mechanical performance, and electronic control adjustment parameters. This balances the lightweight requirements of the entire vehicle with the dynamic adjustment accuracy of the suspension, meeting the energy-saving and intelligent development needs of new energy vehicles.

[0049] 3. This invention calibrates the segmented working mode of the actuator and tunes the configuration parameters of the passive damping channel through multi-parameter priority collaborative scheduling and vehicle posture pre-compensation logic deduction. It then optimizes the control commands by combining road surface high-frequency vibration dissipation and efficiency-energy consumption balance rules. Finally, it collects real-time data such as battery charge and motor power, and continuously dynamically compensates and adjusts the commands through deviation calculation and a self-learning iterative model. Relying on a feedforward prediction, real-time feedback, and self-learning closed-loop architecture, it can dynamically balance the suspension adjustment response efficiency and energy consumption loss, effectively dissipating road surface high-frequency vibration interference. Simultaneously, it autonomously adapts to complex scenarios such as battery charge fluctuations, electronic control delay drift, and sudden changes in road conditions, continuously correcting control parameter deviations. It can adapt to vehicle aging, road condition changes, and driving habit differences over a long period without manual calibration, significantly reducing the cost of later suspension calibration and maintenance, and improving the reliability and intelligence level of adaptive suspension control throughout the entire life cycle of new energy vehicles. Attached Figure Description

[0050] The invention will now be further described with reference to the accompanying drawings.

[0051] Figure 1 This is a flowchart illustrating the steps of a suspension adaptive adjustment method for new energy vehicles provided in an embodiment of the present invention; Figure 2 This is a flowchart of a suspension adaptive adjustment method for new energy vehicles provided in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the steps for generating a concurrent control sequence for suspension provided in an embodiment of the present invention. Detailed Implementation

[0052] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below according to specific embodiments.

[0053] like Figures 1 to 3 As shown in the figure, an embodiment of the present invention provides a method for adaptive suspension adjustment in new energy vehicles, the method comprising: S1: Collects the status signals of the electric drive chassis of new energy vehicles and the raw road perception data. It identifies driving conditions through the vehicle's electronic control signals and distributes the data to corresponding independent control thread pools according to specific mapping rules, outputting a distributed raw dataset. The distributed raw dataset contains the status signals of the electric drive chassis of new energy vehicles and road perception data. This dataset is first parsed to extract suspension electronic control load-related information. Then, these parsed electrical signal parameters are converted into suspension adjustment data models that can be used for suspension tuning. Finally, the data is classified and cached in the tuning queue and combined with topological bionic subtractive processing to output a batch tuning queue for suspension parameters.

[0054] S11: Perform time-series alignment and anomaly denoising filtering on the status signals of the electric drive chassis of new energy vehicles and the raw data of road perception, convert the raw sensor data into standardized fused data, and output a standardized raw fused dataset.

[0055] The vehicle electric drive chassis status signals for new energy vehicles refer to the battery pack cell voltage / current, drive motor three-phase current / speed, brake master cylinder pressure, and ESP vehicle stability system acceleration / angular velocity signals collected via the vehicle's CAN bus. Road perception raw data refers to forward millimeter-wave radar point cloud data, monocular camera road texture images, and lidar elevation data. The vehicle electric drive chassis signals are collected via the OBD-II interface to the CAN analyzer at a sampling rate of 500Hz, while road perception data is synchronously collected via the vehicle domain controller's Ethernet interface at a sampling rate of 100Hz. Time alignment uses a timestamp-based linear interpolation method. Anomaly denoising uses a sliding window mid-range filter. Dimensional unification uses Min-Max normalization. All collected data are linearly interpolated and aligned using a 10ms unified timestamp. Impulse noise and outliers are removed using a sliding window mid-range filter, and all signals are normalized to the [0,1] interval, outputting a standardized raw fusion dataset.

[0056] S12: Construct a feature space for the vehicle's electronic control signals based on the standardized original fusion dataset, extract battery load and motor torque, identify driving conditions through feature threshold matching, and output a feature identification dataset.

[0057] A feature space for the vehicle's electronic control signals is constructed based on the standardized original fusion dataset. Multi-dimensional correlation parameters are extracted to identify driving conditions, and a feature identification dataset with driving condition attribute labels is output. The feature space for the vehicle's electronic control signals is a 6-dimensional feature vector constructed with the unique electric drive, battery, and regenerative braking signals of new energy vehicles as its core. It includes the battery pack load distribution coefficient, motor torque fluctuation amplitude, regenerative braking power ratio, vehicle pitch rate, lateral acceleration, and road roughness index. Corresponding signals are first extracted from the standardized original fusion data, and each feature value is statistically calculated using a sliding window with a length of 100ms and a step size of 20ms. The formula for the battery pack load distribution coefficient is: ,in, This is the battery pack load distribution coefficient. , Let be the voltage and current of the i-th battery cell, and n be the total number of battery cells. The formula for the motor torque fluctuation amplitude is: ,in, This represents the amplitude of motor torque fluctuation. Let the motor torque be at the j-th sampling point. Let N be the average torque within the window, and N be the number of sampling points within the window. A multi-dimensional threshold judgment matrix is ​​used to identify the working condition. The formula is: Where S is the working condition determination result, For the m-th eigenvalue, , The operating condition threshold is defined as the corresponding feature. When all features fall within the corresponding interval, the operating condition is determined. A threshold matrix of 12 typical operating conditions is preset, including constant speed cruising, rapid acceleration, rapid deceleration, brake energy recovery, climbing, descending, bumpy road surface, and turning. Finally, a feature identification dataset containing timestamps, 6-dimensional feature vectors, and operating condition attribute labels is generated.

[0058] S13: Based on the feature identification dataset with operating condition attribute labels, retrieve the new energy vehicle-specific operating condition-suspension parameter mapping rules to complete data partitioning, classification, and link encapsulation, and output the operating condition classification encapsulated dataset. The new energy vehicle-specific operating condition-suspension parameter mapping rules are established through real-vehicle calibration and bench testing, determining the optimal parameter ranges and data processing priority rules for suspension stiffness, damping, and vehicle height under different driving conditions. These rules are stored in the non-volatile memory of the vehicle chassis ECU and can be remotely updated via OTA. First, obtain the operating condition labels and feature data from the feature identification dataset. Then, retrieve the pre-stored operating condition-suspension parameter mapping table from the chassis ECU's Flash memory. The mapping rules are represented by a two-dimensional correlation matrix, with the formula: ,in, For the i-th driving condition, Let be the j-th eigenvalue, and M be the mapping matrix. To determine the suspension parameter ranges under corresponding working conditions and features, each data frame is divided into corresponding working condition data groups based on its working condition label. After attaching the corresponding suspension parameter range information, the data is encapsulated in a unified binary stream format with a total length of 60 bytes. A checksum is added to ensure data transmission reliability, generating a structured encapsulated dataset classified by working condition.

[0059] S14: Based on the classification and encapsulation of the dataset according to the working conditions, match and assign the corresponding data of each working condition to the preset independent control thread pool, complete the data distribution and regularization, and output the split original dataset.

[0060] The independent control thread pool data distribution unit classifies and encapsulates the dataset according to operating conditions, matching and assigning each data item corresponding to a specific operating condition to a pre-defined independent control thread pool. This completes data distribution and normalization, outputting the distributed original dataset. The independent control thread pool consists of 12 independent threads pre-created in the chassis domain controller. Each thread corresponds to a specific driving condition, possessing independent memory space and computing resources. Threads synchronize using semaphores and do not interfere with each other. The unit first obtains the encapsulated data frames from the operating condition classification and encapsulation dataset, then retrieves a pre-defined mapping table between thread pool IDs and operating condition labels from the domain controller's thread management module. Data distribution uses the BKDR hash mapping algorithm, with the following formula: Where S is the operating condition label. The BKDR hash function is used, N=12 is the total number of thread pools, TID is the target thread pool ID, and data transmission adopts a producer-consumer model. It is implemented through a circular buffer of 1024 data frames. After parsing the working condition label of each data frame, the corresponding thread pool ID is calculated by hash mapping algorithm. The data frame is written into the circular buffer of the corresponding thread pool. The consumer threads in the thread pool read data from the buffer in real time and perform flow control to avoid data overflow, thus completing the accurate distribution and regularization of all data and generating an independent split original dataset for each thread pool.

[0061] S2: Parse the suspension electronic load information based on the original dataset, convert it into a suspension adjustment data model, classify and cache it in the calibration queue, perform topological bionic material subtraction on the suspension links and wheel end components, and output the suspension parameter batch calibration queue.

[0062] S21: Extract the timing, amplitude, and coupling characteristic values ​​of the suspension electronic control load from the offloaded raw dataset, remove outlier data, and perform timing interpolation to complete the dataset, outputting the original numerical set of the suspension electronic control load. The offloaded raw dataset refers to the multi-channel timing raw sampling data collected in real time by the suspension electronic control controller, wheel-end pressure sensors, and attitude acquisition unit during vehicle operation, including timestamps, vertical loads, lateral loads, pitch coupling correlation variables, and other low-level sampling sequences. The suspension electronic control load timing sequence is a discrete sequence of load changes over time, collected at equal time intervals. The amplitude is the quantized value of the load extreme value and mean value within a single period, and the coupling characteristic values ​​characterize the interference correlation coefficients between loads in the multi-link suspension. Data acquisition is achieved by real-time polling of the suspension sensing hardware via the vehicle's CAN bus, with millisecond-level data storage at a fixed sampling frequency Fs. The function implementation first constructs an outlier discrimination formula for the original load sequence, expressed as: ,in, This represents the load value at a single sampling point. The mean of the sequence. The standard deviation of the sample is used. Points satisfying the formula are identified as outliers and removed. A linear time-series interpolation formula is used: , , These are valid time points before and after the missing period. , To correspond to the load values, missing sampling bits in the timing sequence are filled in. After anomaly removal, interpolation completion, timing alignment, and numerical accuracy normalization, the multi-channel load correlation is standardized to form a standardized data carrier with continuous timing, no abnormal jumps, and unified dimensions, generating the original numerical set of suspension electronic control loads.

[0063] S22: Deep numerical extrapolation is conducted based on the original numerical set of suspension electronic loads. Typical driving conditions such as high-speed cruising, low-speed heavy load, bumpy road conditions, and cornering are used as classification benchmarks. The inherent force characteristic boundaries of each condition are used as screening criteria. All controllable parameters across the entire domain, including suspension damping adjustment coefficients, wheel end travel parameters, and lateral support force parameters, are deeply decomposed and deconstructed. According to the parameter's own force variation law and feature similarity criteria, adaptive clustering and screening of all parameters across the entire domain are completed, automatically grouping parameters with similar mechanical response characteristics and consistent adaptability to the same feature group. For multi-dimensional heterogeneous parameters with different physical properties and different dimensional standards, dimensional merging is implemented to smooth out dimensional differences and numerical range deviations between various parameters, unifying them into the same computational analysis range. Based on different road surface driving conditions and actual suspension stress scenarios, the various types of data after clustering are encapsulated into structured independent blocks. Each data block is bound with a unique working condition identifier, time sequence timestamp, and attribute label. The complete time sequence link is arranged strictly according to the timing rhythm of the suspension electronic control system's calibration response. Dedicated memory cache partitions are divided, and data call priorities, read and write indexes, and trigger call conditions are set. The data block storage, time sequence link arrangement, and hierarchical cache arrangement are completed in an orderly manner, generating a hierarchical, well-classified, and readily accessible typed suspension calibration cache dataset.

[0064] S23: The categorized suspension tuning cache dataset is integrated into the topological bionic computation chain. Load feature data that has been classified and time-ordered within the cache is directly retrieved as the core computational input. Based on the topological bionic design concept, fully automated finite element mesh discretization is performed on key load-bearing components such as suspension links and wheel-end connection supports. The continuous solid structure is decomposed into several regularized micro-mesh units, establishing a comprehensive structural mechanics discrete analysis system. Combining the mechanical properties of the structural materials with the vehicle's driving load conditions, the stress distribution, deformation trend, and load-bearing contribution of each mesh unit are analyzed. Based on the constraints of ensuring the overall mechanical strength and deformation stability of the structure, and with structural lightweighting and material reduction as the core optimization objective, multiple rounds of iterative optimization calculations are conducted. During the iteration process, the load-bearing capacity and stress level of each unit are continuously compared, gradually identifying and eliminating redundant structural areas with low load-bearing contribution and small stress load. The optimal structural material reduction ratio scheme that adapts to the stress requirements of all working conditions is dynamically deduced. After each round of calculation, the overall mechanical performance is verified to meet the vehicle's safety standards. This process is repeated until the structural parameters tend to stabilize and converge. Simultaneously, based on the structural configuration characteristics after material reduction, the related mechanical parameters such as suspension link stiffness and wheel end support stiffness are calibrated to ensure that structural deformation and electronic control adjustment logic are matched and adapted, and output a set of biomimetic material reduction optimized suspension structural parameters.

[0065] S24: Based on the biomimetic subtractive optimization of the suspension structure parameter set, perform full-domain parameter boundary threshold verification and timing logic rearrangement. In advance, based on the suspension system design specifications and the capabilities of the electronic control hardware, define reasonable upper and lower limit boundary ranges for all structural parameters and electronic control adjustment parameters. Perform full-coverage boundary compliance verification on all optimized calibration parameters one by one, accurately screening for abnormal parameters exceeding the threshold range. Slightly exceeding limits are smoothly corrected, and severely abnormal parameters are marked and isolated, retaining only valid parameters that conform to hardware and design standards. Combining the actual driving rhythm of the vehicle, the response delay characteristics of the suspension electronic control unit, and the linkage control logic between various components, break the original disordered parameter arrangement, reorganize and reconstruct the arrangement order of all parameters, and establish the sequential linkage constraints and timing matching relationships between parameters. Strictly adhere to the external interface communication specifications of the vehicle electronic control system, unify the data encoding format, data bit width standard, and signal interaction rules of all parameters, and complete the unified and standardized adaptation of heterogeneous parameters to electronically recognizable standard values. Based on the batch division of calibration tasks, the priority of system responses, and the hardware port mapping rules, the standardized parameters are batch-grouped and integrated. The queue execution timing interval, batch scheduling trigger conditions, and abnormal parameter jump handling rules are configured to build a complete queue scheduling operation logic. After the entire process of parameter verification, timing reconstruction, interface adaptation, batch grouping, and queue rule configuration, a suspension parameter batch calibration queue that can be directly scheduled and executed by the electronic control system is finally generated.

[0066] S3: Perform feature coupling modeling of battery center of gravity, electric drive transmission path and end-vehicle-cloud chassis topology for the batch adjustment queue of suspension parameters. Use an improved sparrow search algorithm to perform global optimization of suspension control sequence. Integrate driving operation features to complete unsupervised clustering iteration, establish damping-stiffness correlation mapping relationship, and generate suspension concurrent control sequence.

[0067] S31: Collect raw time-series data of the suspension parameter batch calibration queue. The suspension parameter batch calibration queue refers to an ordered set of parameters such as suspension damping, stiffness, and adjustment delay, arranged in a time sequence under multiple operating conditions. Data is collected in real-time via the vehicle chassis CAN bus and the suspension controller timing sampling interface, capturing raw time-series samples from multi-condition road tests and calibrations. Simultaneously, three types of core features are extracted and quantified. The battery center of gravity spatial offset is defined as the coordinate deviation of the actual battery pack installation position relative to the theoretical center of gravity of the vehicle in the longitudinal, lateral, and vertical three-dimensional spaces. , This represents the vertical coordinate deviation. Lateral coordinate deviation, To determine the vertical coordinate deviation, data was obtained by measuring the battery mounting points using a vehicle-wide 3D coordinate measuring machine and performing differential calculations based on the vehicle's reference center of gravity. Electric drive transmission path link loss parameters were also analyzed. ,in, For transmission delay, For power loss, The line impedance loss is obtained by sampling the electric drive controller signal and measuring the link impedance with a link impedance tester, and the characteristics of the connection relationship between the terminal-vehicle-cloud chassis topology nodes are also described. ,in, For the number of topology nodes, For cross-platform data interaction frequency, Link correlation is obtained based on vehicle network topology scanning and cloud node access log statistics. To eliminate the differences in the dimensions of multi-source features, extreme value normalization is used to normalize the single-dimensional original feature x. i Standardization is performed, and a high-dimensional feature coupling matrix is ​​constructed by matrix concatenation. It performs outlier removal, missing data interpolation and completion, and multi-source feature cross-coupling correlation calculation to fully solidify the correlation mapping relationship and dimensional arrangement rules of three types of features: battery center of gravity, electric drive transmission path, and end-vehicle-cloud chassis topology. It outputs high-dimensional feature coupling data with standardized structure and unified dimensions that can be directly used for subsequent modeling and calculation.

[0068] S32: Based on the high-dimensional feature coupling dataset, feature dimension normalization and feature denoising are performed. Redundant noise features with no physical meaning or weak correlation interference are screened and removed. The feature dimension distribution and numerical range are regularized, and a standardized modeling input feature matrix adapted to the modeling requirements is output. The standardized modeling input feature matrix establishes the intrinsic constraint relationship between battery center of gravity offset, electric drive transmission path loss, and end-vehicle-cloud chassis topology. The suspension control sequence composed of suspension damping, stiffness, and timing adjustment is defined as the optimization decision variable. x i (i=1,2,…,n) represent the damping adjustment, stiffness adjustment, and timing control parameters of each suspension unit, respectively. The upper and lower bounds X of the decision variables are pre-set based on the chassis hardware adjustment limits. max X min Based on the feature association mechanism, multivariate coupling constraint equations and parameter value boundary conditions are established, constructing a mathematical association architecture of mutual constraints and dynamic linkages among variables, and then building a logically self-consistent and constraint-complete multivariate constraint coupling mechanism model. On this basis, the inherent operational logic of the traditional sparrow search algorithm is deconstructed, and combined with the working condition adaptation requirements and parameter coupling characteristics of batch suspension parameter tuning, the core rules of population iterative update, location search, and perception foraging are specifically reconstructed. The algorithm structure is adapted to the parameter optimization mechanism of this scenario, resulting in a specialized sparrow search algorithm framework. Based on this sparrow search algorithm framework, two performance evaluation indicators, vehicle driving comfort (S) and handling stability (M), are introduced, and a composite objective function is constructed through a weighted approach. Satisfying weight constraints S is calculated from the root mean square of the vehicle's vertical acceleration, and M is jointly quantified from the vehicle's roll angle and steering lag. The constraint boundaries and variable value ranges of the multivariate constraint coupling mechanism model are matched to complete the precise calibration and configuration of the algorithm's hyperparameters and iteration parameters. Feature coupling constraints and decision variable boundary constraints are embedded to complete the full adaptation of the algorithm's population parameters, iteration rules, objective function, and constraints. The output is an improved sparrow search algorithm global optimization initial model with a complete architecture, fixed parameters, and adapted to the constraints of this scenario.

[0069] S33: Based on the improved sparrow search algorithm's global optimization initial model, a global optimization operation is performed. Using the algorithm's preset population size and individual encoding rules as a baseline, the population initialization formula is called to randomly generate multiple sets of initial sparrow individuals covering the entire feasible domain of suspension parameters. Each individual uniquely corresponds to a set of suspension damping and stiffness timing control parameters. The entire process—population position mutation, adaptive search step size adjustment, discoverer-led search, follower-synchronous iteration, and vigilant boundary warning—is executed sequentially according to the algorithm's reconstruction rules. In each iteration, the composite objective function F... obj The algorithm calculates the comprehensive performance evaluation value of each individual in real time, and updates and retains the global optimal and local extreme solutions synchronously. A dual termination condition is set: a maximum number of iterations and a convergence threshold for the objective function. When the change in the optimal value between adjacent iterations is less than the convergence threshold or the maximum number of iterations is reached, the algorithm is considered to have converged and the loop terminates. Physical constraint verification is performed on all converged optimal individuals. Abnormal outliers and invalid control sequences are eliminated by comparing them against suspension hardware adjustment limits, battery center of gravity load constraints, and electric drive transmission matching constraints. Individual samples with compliant parameters, suitable operating conditions, and optimal comprehensive performance are retained. Parameter format standardization, operating condition dimension classification, and sample dimension alignment are uniformly completed, outputting a standard sample set of optimal suspension control sequences that covers multiple driving conditions, is evenly distributed, and can be directly used as the basis for cluster analysis.

[0070] S34: Using the standard sample set of the optimal suspension control sequence as the basic sample library, it associates real-time vehicle speed, road excitation intensity, vehicle pitch and roll attitude, overall vehicle load status, steering angle, and acceleration / deceleration rate with driving operation feature vectors. All these operational features are acquired in real-time through onboard attitude sensors, speed sensors, and road perception modules. A unique corresponding operating condition feature label is matched to each suspension control sequence, achieving precise binding between control parameters and driving condition features. Iterative partitioning is performed using a density-based unsupervised clustering algorithm, defining the Euclidean distance formula between sample i and sample j in their feature spaces. Where p is the dimension of the working condition feature, and d is the preset neighborhood distance threshold. neigh With the minimum cluster density threshold, when d ij ≤d neighTwo samples are determined to be similar samples in the neighborhood and assigned to the same cluster. During the iteration process, the cluster center coordinates and sample affiliation are continuously updated, and a cluster convergence verification mechanism is introduced simultaneously. The clustering convergence status is judged by monitoring the changes in intra-cluster feature variance and inter-cluster spatial distance. Iteration stops when the cluster structure is stable and the inter-cluster boundaries are clear. Isolated noise samples and invalid scattered clusters with too few samples are removed. The effective clusters are regularized, labeled and classified according to driving condition attributes. The driving condition feature attributes and sample coverage of each cluster are clearly defined. The driving condition feature clustering group is output with clear cluster boundaries, clear driving condition attributes and regular sample affiliation. S35: Based on the driving condition feature clustering grouping, each independent driving condition cluster is used as the smallest analysis unit. All optimal suspension control sequence samples within the cluster are extracted. The temporal fluctuation amplitude, synchronous increase and decrease trend and dynamic lag correlation characteristics of the suspension damping parameter C and stiffness parameter K are statistically analyzed. Both parameters are taken from the standardized adjustment components within the control sequence samples. A high-order nonlinear polynomial with multiple feature correction terms is used for regression fitting to construct the basic formula for the damping-stiffness correlation mapping. Where a0, a1, a2, , , The regression fitting coefficients to be solved are embedded with the battery centroid offset. The electric drive link loss L and the vehicle-cloud topology characteristics T are used as cross-condition correction terms to achieve multi-factor coupling correction. For each cluster of samples, the fitting coefficients are solved successively, and residual analysis, fitting accuracy verification, and significance verification are performed. The fitting deviation of the parameter range is corrected, and the interference of abnormal samples on the fitting results is eliminated. The fitting results of all clusters under all conditions are integrated for unified normalization correction. The applicable parameter range, operating condition adaptation range, and error tolerance threshold of the mapping function are clarified. The nonlinear linkage law of suspension damping and stiffness changing with the vehicle coupling characteristics is fully quantified, and a damping-stiffness dynamic correlation mapping function with cross-condition adaptability, clear physical constraints, and controllable calculation accuracy is output.

[0071] S36: Real-time online acquisition of three types of core features that dynamically change under vehicle driving conditions, including battery center of gravity offset feature. The electric drive transmission path loss characteristic quantity L is updated in real time using the aforementioned measured and bus-acquisition methods. The terminal-vehicle-cloud chassis topology node association status characteristic quantity T is dynamically captured and updated through real-time interaction logs between the vehicle and the cloud, ensuring that all characteristic quantities are synchronized with the actual operating conditions of the vehicle. The real-time collected data... The L and T features are substituted into the damping-stiffness dynamic correlation mapping function to complete the real-time coupling calculation of multiple features. According to the built-in working condition correction, center of gravity compensation, and electric drive loss correction logic of the function, the optimal matching values ​​of suspension damping and stiffness and the dynamic adjustment constraint boundary under the current working condition are solved in real time. Based on the multi-task parallel control mechanism of the end-vehicle-cloud chassis topology architecture, the control requirements of the front and rear axles and left and right independent suspension units of the whole vehicle are decomposed into multiple independent control items according to the parallel task scheduling logic. The priority sorting and timing arrangement of each control item are combined with the topology node instruction transmission delay and chassis actuator response lag time. According to the damping-stiffness mapping constraint and the coupling constraint between battery center of gravity and electric drive path, the rationality of the control parameters is checked one by one, and the control commands with parameter conflicts and timing misalignments are corrected. The multi-unit parameter collaborative matching and timing logic alignment are completed. All control items and timing logic are integrated according to the standardized instruction format that the chassis controller can recognize. Finally, a complete sequence of concurrent suspension control is output, which is time-ordered, parameter-matched, multi-unit collaboratively controllable, and can be directly sent to the chassis actuator.

[0072] S4: Perform priority coordinated scheduling of multiple parameters such as stiffness, damping, and vehicle height for the concurrent control sequence of the suspension, integrate the dynamic timing characteristics of motor torque to complete the pre-compensation logic deduction of vehicle attitude, complete the boundary division of the working condition interval and the calibration of the segmented working mode of the actuator, complete the full-domain tuning of the configuration parameters of the passive damping diversion channel, and generate the initial adjustment command of the suspension.

[0073] S41: The suspension concurrent control sequence is a continuous control sequence formed by the time-sequential arrangement of stiffness, damping, and vehicle height synchronous adjustment parameters of multiple suspension units. Based on the vehicle attitude sensor, height sensor, and shock absorber status acquisition module, continuous time-series sampling data of stiffness, damping, and vehicle height are collected synchronously in real time. The stiffness parameter characterizes the deformation resistance stiffness coefficient of the suspension elastic element, the damping parameter is the vibration attenuation energy dissipation coefficient of the shock absorber, and the vehicle height parameter is the time-series sampling value of the vertical distance between the vehicle frame and the ground. Based on the stiffness time sequence... , Here, n represents the real-time sampled value of the suspension stiffness at the i-th sampling time, and n is the total length of the time-series sampling, representing the damping time-series sequence. , The real-time sampled value of suspension damping at the i-th sampling time, and the time series sequence of vehicle height. , The vertical distance between the vehicle body and the ground at the i-th sampling time is the sampled value. The weight of the vehicle stability performance is determined by the multi-dimensional evaluation coefficients assigned based on ride comfort, roll stability, and pitch stability. The performance contribution weights of the three types of parameters are respectively and satisfy the following conditions. , Weights are assigned to ride comfort. The weights corresponding to the roll stability are... The weights corresponding to pitch stability are determined. The time-series characteristic variance formula is used to analyze the fluctuation dispersion of the three types of parameters. Then, the comprehensive contribution S is calculated by weighting the fluctuation amplitude with the stability performance weights. i Press S i The numerical values ​​are used to rank the priority levels of multiple parameters. The collaborative scheduling allocation ratio among parameters is solved through collaborative normalization, and the matrix is ​​arranged in a regularized manner according to the temporal and parameter dimensions. Finally, the temporal fluctuation trends, amplitude changes, and phase correlation characteristics of the three types of parameters within the suspension concurrent control sequence are analyzed frame by frame. The priority levels are divided based on the comprehensive contribution. The scheduling and assignment of each parameter adjustment component are completed according to the collaborative allocation ratio. A standardized structure is generated and output, with complete dimensions, matched weights, and can be directly used for subsequent attitude compensation simulations. This is a two-dimensional temporal structured matrix that carries the adjustment priority and collaborative allocation ratio of the three types of parameters.

[0074] S42: Based on the multi-parameter priority collaborative scheduling weight matrix, the dynamic timing characteristics of motor torque are collected in real time from the CAN bus of the electric drive controller—that is, the instantaneous amplitude, rate of change, and transmission lag delay timing parameters of the electric drive motor output torque as the driving conditions change. The real-time offset of the actual pitch angle and roll angle of the vehicle body relative to the theoretical reference attitude is sampled through gyroscopes and attitude tilt sensors. Torque change rate Vehicle body pitch deviation yaw deviation Overall posture deviation The compensation correction coefficient β and the weight matrix correlation factor W are used. A pre-compensation model is constructed by deriving formulas to clarify the coupling mechanism and establish the torque-attitude deviation coupling relationship. ,in, This represents the instantaneous change in motor torque. To adjust the three suspension parameters incrementally, a closed-loop feedback iterative formula is used. The pre-compensation amount is calculated, and the transmission mechanism of motor torque lag to attitude deviation is deduced through time-series differential analysis. A closed-loop pre-compensation calculation logic combining feedforward prediction and real-time feedback is established, integrating advanced prediction and real-time deviation correction into a closed-loop calculation architecture. Using the weight matrix as the coupling constraint benchmark, the inherent transmission law of vehicle attitude deviation caused by transient changes in motor torque through the electric drive link and suspension parameter linkage is analyzed. The time-series mechanism of attitude deviation generation and attenuation is sorted out. The coupling relationship, feedback iteration rules and closed-loop calculation logic are integrated to build a quantifiable mathematical deduction model of attitude deviation and compensation amount, and output a complete and real-time calculation vehicle attitude pre-compensation deduction model.

[0075] S43: Based on the vehicle posture pre-compensation model, and combined with the characteristic data of vehicle speed, road surface excitation, steering angle, and acceleration under various road conditions and driving behaviors collected during real-vehicle road tests, the inherent performance parameters of the actuators are retrieved from the chassis actuator factory calibration library. The critical boundary of the driving condition interval refers to the characteristic threshold dividing line that distinguishes between smooth roads, bumpy roads, steering, acceleration, and deceleration, among other typical conditions. The actuator segmented working mode switching parameters are the configuration parameters of the suspension damper and height actuator corresponding to different working ranges, including working gear, response rate, adjustment threshold, and switching hysteresis. The actuator segmented calibration parameter set is a complete set of actuator calibration configuration parameters divided according to the working range and matched with different working modes. The working condition feature vector is... Where v is vehicle speed, a is longitudinal acceleration, δ is steering angle, q is road excitation, critical boundary threshold Ψ is the operating condition, and actuator gear parameters are... Response latency Switching hysteresis threshold The process involves formula derivation to complete the division of operating conditions and parameter calibration: Operating condition feature clustering boundary formulas are used to divide the critical boundaries of operating condition intervals. A pre-compensation model is used to calculate the extreme values ​​of attitude deviations and compensation requirements under different operating conditions, establishing segmented switching judgment criteria and solving for the actuator adjustment thresholds and switching hysteresis parameters corresponding to each operating condition. All driving operating condition feature samples are substituted into the pre-compensation model to calculate the variation patterns of vehicle attitude deviation amplitude and compensation requirement intensity under different driving states. Based on the distribution patterns of operating condition features, various driving operating condition intervals are divided and the critical boundary thresholds are accurately calibrated. The actuator working mode parameters are matched with the compensation requirements of each interval. All segmented configuration parameters are categorized and organized according to the operating condition intervals, outputting a clearly defined, mode-matched actuator segmented calibration parameter set that can be directly used for parameter tuning.

[0076] S44: Extract the configuration parameters of the passive damping diversion channel from the damper structural design parameter library, namely, the structural and control configuration parameters such as the orifice diameter, number of passages, and damping fluid viscosity matching ratio of the passive damper diversion loop. Clarify that the feasible domain boundary of the parameters is determined by hardware structural limits and fluid dynamic constraints. The global optimization tuning refers to the intelligent tuning process that traverses and searches for the optimal configuration matching parameters within the feasible domain. The initial suspension adjustment command is a timing data package of initial stiffness, damping, and height control commands issued to the suspension actuator. Define the diversion channel configuration parameters. Find the objective function and constraints. , , The upper and lower boundaries of the configuration parameters are defined. A particle swarm optimization algorithm is employed to initialize the particle population, iteratively update particle positions, calculate the objective function value for each generation of particles, and select the globally optimal configuration parameters. This is combined with closed-loop pre-compensation logic and actuator segmented calibration parameters, and integrated through fusion computation. Global optimization tuning is performed on the configuration parameters of the passive damping shunt channel, traversing the feasible domain of parameters to select the optimal configuration combination. The matching between the parameters and the actuator calibration parameters is verified. The closed-loop pre-compensation logic is integrated to complete the collaborative correction of the adjustment parameters. Stiffness, damping, and vehicle height adjustment parameters are integrated according to a timing format recognizable by the actuator. The final output is an initial suspension adjustment command that can be directly sent to the suspension actuator and adapted to the current operating conditions.

[0077] S5: Based on the initial adjustment command, the passive damping diversion channel dissipates high-frequency vibrations of the road surface, balances the suspension adjustment efficiency and energy consumption level according to the working condition interval rules, and outputs the intermediate control command of the suspension.

[0078] S51: Analyzes the core components of the initial suspension adjustment command, precisely disassembling the embedded damping and stiffness control values, as well as the parameter execution sequence logic. Clarifies the execution sequence nodes, durations, and hierarchical triggering relationships of each adjustment parameter, and defines the sequential linkage constraints of different parameters within the adjustment cycle. Simultaneously, it completes the full-dimensional configuration parameter acquisition of the passive damping shunting channel, covering key structural and control parameters such as channel flow orifice diameter, shunting damping coefficient, fluid path impedance, structural natural resonant frequency, and shunting flow threshold. Based on data standardization and normalization algorithms, it eliminates dimensional differences, misalignments, and sampling frequency mismatches between the command adjustment parameters and the channel configuration parameters. Based on the suspension control timing mapping relationship and parameter physical coupling association rules, the damping and stiffness adjustment parameters are paired one-to-one with the configuration parameters of each diversion channel to construct a structured mapping relationship between parameters. Invalid and redundant parameter dimensions and abnormal sampling data points are eliminated. Through matrix recombination and feature alignment, accurate bidirectional dimension matching is achieved. The data format, timing scale, and parameter representation caliber are unified. The field definitions, sampling intervals, and parameter association index relationships within the dataset are standardized. This forms a well-structured, dimensionally aligned command channel parameter matching dataset that can be directly used for subsequent vibration dissipation calculations. This provides a standard data input basis for the passive damping diversion channel to access adjustment commands and carry out high-frequency vibration filtering calculations, ensuring the effectiveness and accuracy of subsequent parameter coupling calculations.

[0079] S52: Based on the command channel parameter matching dataset, extract the independent damping control component from the initial suspension adjustment command, remove interference from stiffness and vehicle height-independent control parameters, and lock the damping timing control sequence used for vibration suppression separately. This damping control component is then fully imported into the passive damping shunt channel mathematical operation model according to the channel parameter matching mapping relationship. Combining the inherent structural characteristics of the channel and the fluid dynamics response law, the original high-frequency vibration signals of the road surface collected in real time during vehicle operation are incorporated, covering high-frequency disturbance components such as road surface unevenness excitation, high-frequency tire bumps, and chassis micro-vibrations. A high-order filtering algorithm combined with the inherent attenuation characteristics of the passive damping shunt is used to perform layered filtering and energy dissipation calculations on the high-frequency vibration signals of the road surface. High-frequency vibration amplitudes exceeding the comfort threshold are attenuated step by step, correcting the vibration transmission phase lag problem, suppressing the energy superposition effect in the resonant frequency band, and preserving the low-frequency normal road condition excitation signal without excessive attenuation. During the calculation process, the channel configuration parameter constraint boundary is strictly followed to limit the upper and lower limits of the damping control component output, so as to avoid abnormal suspension operation caused by parameter over-limit. The vibration amplitude attenuation rate, phase offset and energy consumption fluctuation amplitude before and after filtering are checked simultaneously. Abnormal data points are eliminated, and the damping control timing value and timing trigger node are corrected to form a correction adjustment command that has been optimized for high-frequency vibration dissipation and meets the vibration suppression effect.

[0080] S53: The adjustment command is input into a pre-calibrated driving condition interval division rule system. The rules cover four core operating condition dimensions: vehicle speed range, road surface grade, vehicle load, and driving posture, clearly defining the critical boundary thresholds and state judgment criteria for each operating condition. Based on the operating condition interval assignment results, suitable suspension adjustment efficiency evaluation indicators and energy consumption quantitative indicators are selected for the current driving condition. Adjustment efficiency includes quantitative dimensions such as parameter response delay, posture correction rate, and operating condition adaptation sensitivity; energy consumption includes calculation dimensions such as damping adjustment power consumption, channel shunting energy loss, and actuator drive energy consumption. A two-way constraint evaluation model for efficiency and energy consumption is constructed based on these two types of quantitative indicators. A minimum efficiency threshold and a maximum energy consumption constraint red line are set, and a weighted balance coefficient is introduced to represent the priority weights of both. A multi-objective constraint objective function and inequality constraint conditions are established. An iterative optimization algorithm is used to traverse and solve within the feasible parameter domain, dynamically balancing the contradictory relationship between suspension adjustment response speed and overall energy consumption. The optimal control ratio coefficient and timing allocation ratio of damping and stiffness parameters under different operating conditions are solved, and the parameter adjustment range is constrained to not exceed the hardware execution limit. The parameter ratio schemes of efficiency priority, energy consumption priority or balanced adaptation under various operating conditions are solidified. The quantitative index values, constraint coefficients and optimal ratio parameters are organized to form a structured and directly callable set of efficiency and energy consumption balance control parameters.

[0081] S54: The suspension control timing logic is globally optimized and reconstructed using the efficiency and energy consumption balance control parameter set as a constraint benchmark and correction adjustment commands. Based on core parameters such as the ratio coefficient, response delay constraint, and energy consumption limit in the balance control parameter set, the execution timing arrangement of damping and stiffness parameters within the correction adjustment commands is re-organized. The trigger time, duration, and interval of each control link are adjusted, and the switching logic between multi-parameter synchronous control and step-by-step control is optimized. Strict adherence to the operating condition range adaptation rules is maintained, matching the vibration characteristics, load characteristics, and driving requirements of the current driving state. Redundant and repetitive control actions in the commands are eliminated, similar adjustment commands with similar timing nodes are merged, and unreasonable timing arrangements that are ahead or behind are corrected. This ensures that the control timing meets both the requirements for continuous dissipation of high-frequency vibration and the requirements for efficiency and energy consumption balance constraints. The parameters, timing intervals, execution priorities, and operating condition compatibility of the optimized instructions are verified synchronously to ensure that all indicators fall within the constraint threshold range, avoiding issues such as timing conflicts, parameter overshoot, and energy consumption exceeding limits. The instructions are repackaged and standardized according to the standardized instruction format, solidifying the final control parameter values ​​and timing execution logic. This generates intermediate suspension control instructions that are fully adapted to the current driving conditions, while also considering dynamic suppression, efficiency adjustment, and energy consumption control. These instructions can be directly sent to the chassis execution unit to complete the actual suspension control operation.

[0082] S6: Real-time acquisition of battery charge, motor power, electronic control delay and road surface undulation data; calculation of matching deviation and linkage delay deviation to dynamically compensate for intermediate control commands; construction of load self-learning iterative model to continuously correct parameters; output of new energy vehicle suspension adaptive final adjustment command.

[0083] S61: Real-time synchronous acquisition of four core time-series sensor data points during the operation of new energy vehicles: battery state of charge, motor output power, electronic control system response delay, and road surface undulation height. Continuous sampling is performed throughout the entire process at a unified sampling frequency of the vehicle chassis sensing system, avoiding timing misalignment issues caused by inconsistent sampling periods of different sensor hardware. Battery state of charge reflects the remaining capacity and output load capacity of the power battery, directly related to the vehicle's load supply boundary. Motor output power characterizes the real-time power output level of the electric drive system, determining the magnitude of dynamic load changes in the chassis. Electronic control system response delay represents the inherent lag characteristic from the issuance to execution of chassis control commands, and is the core cause of timing control deviations. Road surface undulation height reflects the intensity of external road condition excitation, directly affecting the magnitude of suspension vibration input. Abandoning the simple raw data retention model, based on time-domain feature decomposition and trend term separation algorithms, random noise, instantaneous interference pulses, and steady-state redundant components are removed from the original time-series sequence, accurately extracting effective feature components such as trend features, fluctuation amplitude features, and phase lag features. To address the issues of different data units, inconsistent dimensional structures, and mismatched time scales among the four types of data, a normalization mapping and time-series interpolation alignment algorithm is adopted to unify the data value range and time coordinate benchmark, establish a one-to-one correspondence structure of multi-source data feature dimensions, eliminate outlier sampling points, and standardize the data storage format and feature arrangement logic. This results in a standardized multi-source real-time dataset with a regular structure, synchronized time sequence, unified dimensions, and direct access to the deviation calculation stage. This provides regular and reliable original feature input support for the accurate calculation of subsequent working condition matching deviations and time delay deviations.

[0084] S62: The standardized multi-source real-time dataset is fully integrated into the pre-built and fixed deviation calculation model's internal computational space. This model incorporates a base feature library of operating conditions, a theoretical suspension control baseline curve, and an electronic control delay standard threshold system, possessing the capability to couple and compare multi-source operating condition features with control commands. Using operating condition features such as battery charge, motor power, and road surface undulations as benchmarks, the model performs a full-domain comparison calculation between the preset parameters of the suspension intermediate control commands and the actual vehicle operating conditions. It calculates the numerical difference between the theoretical control parameters and the actual operating condition requirements, defining this as the operating condition matching deviation, characterizing the degree of mismatch between the suspension control parameters and the actual driving load and road condition excitation. Combining the inherent response delay characteristics of the electronic control system, it tracks the entire link timing interval of control command signal transmission, logical operation, and execution response, comparing it with the ideal lag-free timing standard to quantify the timing offset generated during parameter linkage control, forming a linkage delay deviation index. The matching deviation and time delay deviation are quantified in a refined manner from four dimensions: amplitude, fluctuation trend, time offset phase, and duration of action. The single numerical representation method is abandoned, and the dynamic time-series characteristics of the deviation change with the driving conditions are fully preserved. The quantification results and time-series correlation information of the two types of deviations are organized according to the structured data format, invalid weak deviation components are eliminated, the data field and time sequence arrangement relationship is standardized, and the dual deviation quantization dataset is output.

[0085] S63: Based on the dual-deviation quantization dataset as the core calculation basis, the inherent constraint boundary of the suspension intermediate control command and the parameter adjustment limit threshold are introduced to construct a deviation-parameter linkage dynamic compensation calculation mechanism, and establish a nonlinear compensation mapping relationship between matching deviation, time delay deviation and suspension damping and stiffness control parameters. The parameter amplitude of the intermediate control command is directionally corrected according to the compensation calculation rules, and the output magnitude of the control parameters is fine-tuned according to the magnitude of the working condition matching deviation to match the actual load and road condition requirements. Combined with the linkage time delay deviation correction command timing trigger node, the timing misalignment caused by the electronic control transmission lag is compensated, ensuring precise synchronization between the control command issuance time and the chassis working condition change time. Based on the completion of command dynamic compensation, and considering the application characteristics of time-varying vehicle dynamic load and road condition switching, a self-learning iterative model architecture for the load is independently built. The model input dimension, output representation, iteration update rules, and convergence judgment conditions are defined, and a parameter self-correction closed-loop logic based on historical deviation data is constructed. Using the dual-bias quantization dataset as the core sample source for model initialization, key parameters such as the model's internal weight coefficients, load correction factors, and bias adaptive adjustment coefficients are initialized. Parameter adaptation and calibration are completed in accordance with physical constraints and suspension control mechanisms, avoiding problems such as slow model convergence and poor adaptability caused by random initial parameter assignment. The basic structure of the model and the initial parameter system are solidified, forming an initial self-learning model with bias perception, load adaptation, and iterative optimization capabilities.

[0086] S64: Real-time updated operating condition deviation time-series data generated during vehicle operation is continuously input into the initial self-learning model. Based on the model's preset iterative update mechanism and convergence constraints, a normalized closed-loop iterative training process is initiated. The model uses real-time operating condition deviation data as driving input, comparing it with historical deviation samples, compensation and control records, and vehicle load variation patterns. Iteratively adjusts core parameters such as internal weight coefficients, load correction factors, and time-series compensation coefficients layer by layer, continuously narrowing the error difference between the model's theoretical output and the actual suspension control requirements of the vehicle. Strict upper and lower limits for parameter adjustment are set during the iteration process to avoid over-correction and failure. A convergence criterion is introduced: when the deviation error fluctuation range enters a set threshold range, the optimal parameter combination of the model is stably locked, terminating the invalid iteration process. After multiple rounds of real-time operating condition data iterative training and parameter adaptive correction, the model can autonomously adapt to complex driving scenarios such as battery charge fluctuations, motor power switching, road condition variations, and electronic control delay drift, possessing adaptive control capabilities of real-time perception, autonomous learning, and dynamic correction. Based on the optimized and converged self-learning model, combined with dynamic compensation logic and operating condition constraint rules, the system integrates the full-dimensional control parameters of suspension damping, stiffness, and vehicle height with the timing execution logic. According to multiple indicators such as vehicle ride comfort, handling stability, and energy consumption economy, the system encapsulates and generates new energy vehicle suspension adjustment commands that are adaptable to all operating condition changes and have adaptive correction capabilities. These commands can be directly sent to the chassis actuators to complete precise control.

[0087] In summary, this embodiment provides a method for adaptive suspension adjustment in new energy vehicles. Through multi-parameter collaborative scheduling, attitude pre-compensation, and load self-learning dynamic correction mechanisms, it achieves full-process adaptive closed-loop intelligent control of the suspension. This invention uses multi-parameter priority collaborative scheduling and vehicle attitude pre-compensation logic deduction to calibrate the actuator's segmented working mode and tune the passive damping channel configuration parameters. It then combines road surface high-frequency vibration dissipation and efficiency-energy consumption balance rules to optimize control commands. Finally, it collects real-time data such as battery charge and motor power, and continuously dynamically compensates for adjustment commands through deviation calculation and a self-learning iterative model. Relying on a feedforward prediction, real-time feedback, and self-learning closed-loop architecture, it can dynamically balance suspension adjustment response efficiency and energy consumption loss, effectively dissipate road surface high-frequency vibration interference, autonomously adapt to complex scenarios such as battery charge fluctuations, electronic control delay drift, and sudden road condition changes, continuously correcting control parameter deviations. It can adapt to vehicle aging, road condition changes, and driving habit differences over a long period without manual calibration, significantly reducing the cost of later suspension calibration and maintenance, and improving the reliability and intelligence level of adaptive suspension control throughout the entire lifecycle of new energy vehicles.

[0088] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for adaptive suspension adjustment in new energy vehicles, characterized in that, include: S1: Collects the status signals of the electric drive chassis of new energy vehicles and the raw data of road perception. Identifies the driving conditions through the dimension of vehicle electronic control signals, distributes them to the corresponding independent control thread pool according to the exclusive mapping rules, and outputs the split raw dataset. S2: Based on the original dataset of the diversion, the suspension electronic control load information is parsed, converted into a suspension adjustment data model, and topological bionic material reduction is performed on the suspension link and wheel end components to output a batch adjustment queue of suspension parameters; S3: Perform feature coupling modeling on the batch adjustment queue of suspension parameters, use an improved sparrow search algorithm to globally optimize the suspension control sequence, perform unsupervised clustering iteration based on driving characteristics, establish a damping stiffness mapping relationship, and generate a concurrent suspension control sequence. The specific steps are as follows: S31: Collect raw time-series data of the batch calibration queue of suspension parameters, extract battery center of gravity offset, electric drive transmission path loss and end-vehicle-cloud chassis topology connection features, and output high-dimensional feature coupling dataset. S32: Based on the high-dimensional feature coupling dataset, a multivariate constraint coupling mechanism model is built. The suspension control sequence is used as the decision variable. The sparrow search algorithm rules are improved and a composite objective function is set to output the global optimization initial model. S33: Based on the global optimization initial model, perform population initialization, position mutation and extreme value update, screen the convergent optimal individual, and output the standard sample set of the optimal suspension control sequence. S34: Associate the standard sample set of the optimal suspension control sequence with the vehicle operation feature vector, perform unsupervised iterative partitioning using density clustering, aggregate samples of the same working condition and complete convergence verification, and output the working condition feature clustering group. S35: Based on the operating condition characteristics, cluster and group the time-series variation characteristics of suspension damping and stiffness parameters, fit the linkage law through nonlinear regression, construct the parameter linkage function, and output the damping-stiffness dynamic correlation mapping function. S36: Substitute the real-time battery center of gravity, electric drive path, and chassis topology related features into the damping-stiffness related mapping function, combine the real-time coupling state, split the control items according to the parallel scheduling logic, and output the complete sequence of suspension concurrent control. S4: Perform multi-parameter priority collaborative scheduling on the suspension concurrent control sequence, deduce the attitude pre-compensation logic based on the motor torque timing characteristics, divide the working condition interval and calibrate the actuator mode to obtain the initial suspension adjustment command; S5: According to the initial adjustment command, the passive damping diversion channel dissipates high-frequency vibrations of the road surface, balances the suspension adjustment efficiency and energy consumption level according to the working condition interval rules, and outputs the intermediate control command of the suspension. S6: Real-time acquisition of battery data and road surface undulation data, calculation of matching deviation to dynamically compensate for intermediate control commands, continuous correction of parameters, and output of new energy vehicle suspension adjustment commands.

2. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S1, the specific steps for outputting the split original dataset are as follows: S11: Perform time-series alignment and anomaly denoising filtering on the status signals of the electric drive chassis of new energy vehicles and the raw data of road perception, convert the raw sensor data into standardized fused data, and output a standardized raw fused dataset. S12: Construct a feature space for the vehicle's electronic control signal dimension based on the standardized original fusion dataset, extract battery load and motor torque, identify driving conditions through feature threshold matching, and output a feature identification dataset. S13: Based on the feature identification dataset, retrieve the new energy vehicle-specific working condition-suspension parameter mapping rules, complete the data partitioning and classification and link encapsulation according to the working condition category, and output the working condition classification and encapsulation dataset. S14: Based on the described working condition classification and encapsulation dataset, match and assign each data item corresponding to each working condition to a preset independent control thread pool, complete data distribution and regularization, and output the split original dataset.

3. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S2, the specific steps for outputting the batch calibration queue of suspension parameters are as follows: S21: Extract the timing, amplitude, and coupling characteristic values ​​of the suspension electronic control load from the original dataset, remove outlier data and perform timing interpolation to complete the dataset, and output the original numerical set of the suspension electronic control load. S22: Perform deep numerical deduction based on the original numerical set of suspension electronic control load, perform clustering and dimensionality merging of the full domain parameters according to the working condition characteristic threshold, encapsulate data blocks and arrange time-series link caches under different stress scenarios, and output a type-specific suspension calibration cache dataset. S23: The classified suspension tuning cache dataset is fed into the topological bionic computing link. The stress field is discretely decomposed for the suspension link and wheel end components. The dynamic deduction of the material reduction ratio of the structure is dynamically performed through iterative optimization. The bionic material reduction optimized suspension structure parameter set is output. S24: Based on the biomimetic subtractive optimized suspension structure parameter set, perform global parameter boundary threshold verification and timing logic rearrangement, adapt to the electronic control system interface specification to complete the unified and regularized numerical format, and output the suspension parameter batch calibration queue.

4. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S32, the specific steps for outputting the initial global optimization model are as follows: Based on the high-dimensional feature coupling dataset, dimensionality normalization and feature denoising are performed to remove redundant noise features and output a standardized modeling input feature matrix. Based on the standardized modeling input feature matrix, the constraint relationship among the three is established, the suspension control sequence is used as the optimization decision variable, multivariate coupling constraint equations and boundary conditions are established, a complete model is constructed, and a multivariate constraint coupling mechanism model is output. Based on the multivariate constraint coupling mechanism model, the logic of the traditional sparrow search algorithm is decomposed, and the reconstruction and iterative update rules of the suspension parameter adjustment scenario are adapted to output the sparrow search algorithm framework. Based on the sparrow search algorithm framework and driving smoothness and handling performance indicators, a weighted composite objective function is constructed, and parameter calibration is completed by matching the model constraint boundary to output the global optimization initial model.

5. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S4, the specific steps for generating the initial suspension adjustment command are as follows: S41: Analyze the temporal characteristics of stiffness, damping, and vehicle height parameters in the concurrent control sequence of the suspension, sort and coordinate the multi-parameter priority according to the vehicle stability performance weight, and output the multi-parameter priority coordination scheduling weight matrix. S42: Based on the multi-parameter priority collaborative scheduling weight matrix and the dynamic timing characteristics of motor torque, the vehicle posture deviation compensation mechanism is deduced, a closed-loop pre-compensation operation logic is constructed, and the vehicle posture pre-compensation deduction model is output. S43: Use the vehicle body attitude pre-compensation deduction model to divide the critical boundaries of the driving condition interval, calibrate the actuator segmented working switching mode parameters, and output the actuator segmented calibration parameter set; S44: Perform global optimization tuning of the passive damping diversion channel configuration parameters based on the actuator segment calibration parameter set, integrate the compensation logic with the actuator calibration results, and output the initial suspension adjustment command.

6. The method for adaptive suspension adjustment in new energy vehicles according to claim 5, characterized in that: In step S42, the specific steps for outputting the vehicle body attitude pre-compensation deduction model are as follows: Collect dynamic time-series data of motor torque and perform noise reduction and smoothing processing. Extract instantaneous torque value, rate of change and time-series gradient features. After normalization, match them with the dimension of the weight matrix to output torque time-series-weight coupled dataset. Based on the torque time-weighted coupling dataset, a correlation function between torque and suspension parameter weights is constructed. The mapping relationship between vehicle body attitude deviation and the two is derived through linear regression. The deviation generation coefficient is quantified, and the attitude deviation correlation derivation formula is output. By inputting the closed-loop feedback correction coefficient into the attitude deviation correlation derivation formula, setting the compensation response threshold and timing window, a closed-loop operation logic of deviation prediction, parameter correction, and feedback verification is built, and a vehicle attitude pre-compensation derivation model is output.

7. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S5, the specific steps for outputting the suspension intermediate adjustment command are as follows: S51: Analyze the damping and stiffness adjustment parameters and timing logic in the initial suspension adjustment command, and perform dimensional matching between the command parameters and the channel structure parameters by collecting the configuration parameters of the passive damping shunt channel, and output the command channel parameter matching dataset. S52: Based on the instruction channel parameter matching dataset, the damping control component in the initial adjustment instruction is imported into the passive damping shunt channel, and a filtering and dissipation operation is performed on the high-frequency vibration signal of the road surface to output a correction adjustment instruction. S53: Call the correction and adjustment command to input the preset working condition interval division rules, quantify the quantitative indicators of suspension adjustment efficiency and energy consumption under different working conditions, establish a two-way constraint evaluation model, dynamically calculate the optimal compromise ratio, and output the efficiency and energy consumption balance control parameter set. S54: Optimize the timing logic based on the efficiency and energy consumption balance control parameter set and the correction adjustment command, and output the suspension intermediate control command adapted to the current working condition.

8. The method for adaptive suspension adjustment in new energy vehicles according to claim 1, characterized in that: In step S6, the specific steps for outputting the final adaptive adjustment command for the new energy vehicle suspension are as follows: S61: Real-time acquisition of battery state of charge, motor output power, electronic control system response delay and road surface undulation height time series data, extraction of each data feature component and completion of dimension alignment, output of standardized multi-source working condition real-time dataset; S62: Input the standardized multi-source working condition real-time dataset into the preset deviation calculation model, calculate the matching deviation between the suspension intermediate control command and the actual working condition, the time delay deviation in the parameter linkage process, quantify the deviation amplitude and timing characteristics, and output the dual deviation quantization dataset. S63: Perform dynamic compensation calculation on the suspension intermediate control command according to the dual deviation quantization dataset, correct the parameter amplitude and timing trigger node, build a load self-learning iterative model, import the deviation data initialization model parameters into the load self-learning iterative model, and output the initial self-learning model. S64: Input the real-time operating condition deviation data into the initial self-learning model for iterative training, correct the model parameters, and output the new energy vehicle suspension adjustment command.

9. A method for adaptive suspension adjustment in new energy vehicles according to claim 8, characterized in that: In step S63, the specific steps for outputting the initial self-learning model are as follows: A dynamic compensation correlation equation is established between the dual deviation and the suspension control parameters. The quantitative data of the dual deviation is input into the compensation correlation equation to calculate the compensation amount. The parameter amplitude of the suspension intermediate control command is accurately corrected, and the compensated suspension intermediate control command is output. Based on the compensation suspension intermediate control command, a load self-learning iterative logic architecture is designed, the iterative convergence condition and timing update rule are set, a closed-loop iterative mechanism for deviation feedback and parameter correction is constructed, and the load self-learning iterative model framework is output. The initial weights and bias correction coefficients of the load self-learning iterative model framework are determined by nonlinear fitting, and the initial self-learning model is output.

Images