Integrated chassis domain control method and system based on clustering algorithm

Abstract

This application relates to the field of data processing, and more particularly to an integrated chassis domain control method and system based on clustering algorithms. The method includes: constructing pothole road condition features during vehicle operation; these features include pothole elevation differences and the electric vehicle's electric braking response sequence when passing through potholes; using an improved clustering algorithm to cluster the pothole road condition features, generating pothole road condition clusters representing different electric braking response sequences; determining the clustering distance based on the pothole elevation differences and the electric vehicle's electric braking response sequence when passing through potholes; and optimizing the electric braking cooperative control strategy in the electric vehicle chassis domain electric braking system based on a braking cooperative adjustment factor. This application has the effect of adaptively and dynamically optimizing the electric braking cooperative control strategy.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to an integrated chassis domain control method and system based on clustering algorithms. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the power systems of electric and hybrid vehicles are evolving towards greater integration and domain control. Integrated chassis domain controllers are gradually integrating subsystems such as electric drive, electric braking, regenerative energy recovery, and battery management to achieve global optimization and coordinated control of the vehicle's energy. During vehicle operation, under complex conditions such as potholes and damaged road surfaces, if the braking torque distribution and energy recovery strategies fail to accurately match real-time road conditions, energy recovery efficiency will be significantly reduced, leading to problems such as braking jerking and superimposed mechanical shocks.

[0003] In the existing technology, the braking control system of some electric vehicles uses road surface pre-aiming sensors (such as lidar or vision cameras) to obtain the geometric features of potholes, and combines them with basic parameters such as vehicle speed and acceleration to classify road conditions using traditional clustering algorithms, and then matches them with a preset electric braking cooperative control strategy.

[0004] However, when the above methods automatically classify complex road conditions based on clustering algorithms, they do not integrate the road geometric features with the multi-dimensional dynamic response features of the electric vehicle's electric braking system into a model. This results in the clustering results being disconnected from the actual electric braking / energy recovery performance, and the generated electric braking coordinated control strategy lacks accurate data support. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides an integrated chassis domain control method and system based on clustering algorithms.

[0006] Firstly, this application provides an integrated chassis domain control method based on a clustering algorithm, employing the following technical solution: An integrated chassis domain control method based on a clustering algorithm includes the following steps: constructing pothole road condition features during vehicle operation; the pothole road condition features include the pothole height difference and the electric braking response sequence of the electric vehicle passing through the pothole; using an improved K-means clustering algorithm to cluster the pothole road condition features, generating pothole road condition clusters representing different electric braking response sequences; wherein, the distance function used for clustering is determined based on the pothole height difference and the electric braking response sequence of the electric vehicle passing through the pothole; calculating the braking smoothness index based on the acquired vehicle acceleration data, and calculating the braking smoothness index corresponding to each cluster at the time of vehicle manufacture. The system determines the baseline recovery efficiency for various pothole conditions; it calculates the distance from real-time pothole features to each pothole cluster based on the distance function, assigns the real-time pothole features to the pothole cluster with the smallest distance, and takes this pothole cluster as the target cluster; it calculates the braking coordination adjustment factor based on the baseline recovery efficiency and braking smoothness index of the target cluster; it optimizes and adjusts the electric braking coordination control strategy for the corresponding pothole condition category in the electric braking system of the electric vehicle chassis domain based on the braking coordination adjustment factor; after one optimization adjustment, it continues to compare the deviation rate between the collected actual braking smoothness index and the baseline recovery efficiency; if the deviation rate is greater than a preset threshold, it dynamically updates the baseline recovery efficiency of the corresponding pothole cluster based on the deviation rate.

[0007] Optionally, the characteristics of pothole road conditions during vehicle operation are constructed, including: acquiring test data of the vehicle during various pothole road condition tests during the factory testing phase and using it as a sample; processing the sample data to construct multi-dimensional features about pothole height difference and electric braking response sequence; and constructing a feature set.

[0008] Optionally, an improved K-means clustering algorithm is used to cluster the pothole road condition features, including initializing cluster centers. The method for initializing cluster centers is as follows: sort the samples in ascending order according to the pothole height difference, and calculate the difference sequence of pothole height difference between adjacent samples; set a mutation threshold based on the mean and standard deviation of the difference sequence; take samples with pothole height difference greater than the mutation threshold as gradient mutation points, and determine the initial number of clusters based on the gradient mutation points; divide the sorted samples into several height difference intervals according to the gradient mutation points; within each height difference interval, calculate the root mean square value of each electric braking response sequence in each sample as the average response amplitude of each drive motor under the pothole road condition corresponding to that height difference interval, and calculate the mean of the average response amplitude of each drive motor to obtain the comprehensive electric braking response index of each sample; select the sample with the largest comprehensive electric braking response index in each height difference interval as the initial cluster center of that cluster. The initial cluster center is a multi-dimensional feature vector, including the pothole height difference and the electric braking response sequence of each drive motor.

[0009] Optionally, the distance can be calculated as follows: calculate the difference in pothole height between the real-time sample and the current cluster center of each pothole condition cluster and normalize it as a height difference term; calculate the dynamic time-normalized distance of the electric braking response sequence between the real-time sample and the current cluster center of each pothole condition cluster and normalize it as a braking response difference term; and obtain the distance between the real-time sample and the current cluster center of each pothole condition cluster based on the height difference term and the braking response difference term.

[0010] Optionally, the difference in pothole height difference between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as a height difference term. The normalization method is as follows: the absolute value of the ratio of the difference in pothole height difference between the real-time sample and the current cluster center of each pothole condition cluster to the maximum pothole height difference in the sample is used as the normalization result.

[0011] Optionally, the dynamic time warping distance between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as a braking response difference term. The normalization method is as follows: calculate the dynamic time warping distance between the real-time sample and the current cluster center of each pothole condition cluster; calculate the maximum dynamic time warping distance between any two samples in the feature set constructed during the factory testing phase; and use the ratio of the dynamic time warping distance to the maximum dynamic time warping distance as the normalization result.

[0012] Optionally, an improved K-means clustering algorithm is used to cluster the pothole road condition features, including: for each pothole road condition cluster, based on all samples assigned to that pothole road condition cluster, repeatedly updating the pothole height difference of the cluster center and the electric braking response sequence of the cluster center until the change in the pothole height difference of the cluster center and the change in the electric braking response sequence of the cluster center reach the preset convergence condition.

[0013] Secondly, this application provides an integrated chassis domain control system based on clustering algorithms, employing the following technical solution: An integrated chassis domain control system based on clustering algorithm includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the integrated chassis domain control method based on clustering algorithm described above.

[0014] The beneficial effects are: the above-mentioned integrated chassis domain control method based on clustering algorithm is generated into a computer program and stored in the memory so that it can be loaded and executed by the processor, thereby making the system easy to use by creating a system based on the memory and processor.

[0015] This application has the following technical effects: 1. First, a composite road condition feature model is constructed, simultaneously incorporating pothole elevation differences and electric braking response sequences, to overcome the limitation of single geometric parameters in characterizing electric drive excitation. Second, an improved K-means algorithm is employed, and a composite distance function integrating elevation difference differences and DTW distance of braking response time series is designed for precise clustering, generating pothole road condition clusters representing different response modes, achieving deep coupling between geometric features and dynamic response. Finally, a braking coordination adjustment factor is calculated based on the benchmark recovery efficiency of the real-time matching target cluster and the current braking smoothness index, thereby adaptively optimizing the electric braking coordination control strategy and dynamically updating the benchmark value when the deviation exceeds the threshold. This establishes a logical link between "road condition perception - feature clustering - strategy execution - feedback iteration," enabling braking torque distribution and energy recovery strategies to accurately match real-time complex pothole conditions, effectively eliminating the problem of classification and response disconnect in traditional methods, and significantly improving the adaptive coordination capabilities of braking smoothness, energy recovery efficiency, and chassis domain control.

[0016] 2. This application introduces a dynamic time warping algorithm to quantify the braking response difference term of the four-wheel response sequence, and constructs a normalized distance by combining it with the pothole height difference term. This enables the algorithm to accurately distinguish the differences in electric braking response caused by "the same pothole height difference but different vehicle loads or frequencies of minor bumps". Simultaneously, the initial cluster number is adaptively determined based on the statistical mutation points of the height difference gradient, and the sample with the largest comprehensive electric braking response index is selected as the initial cluster center, ensuring high representativeness of the clustering starting point and fast convergence. Combined with a lightweight strategy, this improved clustering scheme, while meeting the fixed-point computing power limitations of automotive-grade MCUs and the low-power sleep architecture, reduces invalid timing alignment operations, achieving low-latency and high-precision partitioning of high-dimensional road condition features.

[0017] 3. During long-term vehicle operation, the system continuously monitors the relative deviation between the actual braking smoothness index and the factory benchmark. When the deviation rate exceeds a preset threshold, it automatically triggers a smooth update of the benchmark recovery efficiency based on a weighted fusion formula, dynamically bringing the energy recovery evaluation standard closer to the current actual health state of the system. This mechanism not only avoids the disconnect between the control target and the actual execution capability caused by hardware aging or electrical characteristic drift, but also ensures that the electric braking torque compensation always conforms to the physical boundaries and safety margins of the chassis actuators. As a result, the integrated chassis domain controller can maintain optimal braking smoothness and energy recovery efficiency throughout the entire vehicle life cycle, significantly improving control accuracy under complex operating conditions, long-term system robustness, and overall vehicle energy economy. Attached Figure Description

[0018] Figure 1 This is a flowchart of the invention method of the integrated chassis domain control method based on clustering algorithm according to the embodiments of this application. Detailed Implementation

[0019] This application discloses an integrated chassis domain control method based on a clustering algorithm, referring to... Figure 1 This includes steps S1-S4: S1: Construct the characteristics of potholes during vehicle operation; the characteristics of potholes include the pothole elevation difference and the electric braking response sequence when the electric vehicle passes through the pothole.

[0020] During the factory testing phase, test data of the vehicle during various road conditions with potholes were acquired and used as samples; the samples were processed to construct multi-dimensional features about pothole height difference and electric braking response sequence; and a feature set was constructed.

[0021] In one embodiment, the method for calculating the pothole height difference is as follows: A laser rangefinder installed on the vehicle chassis scans and collects data on the pothole area and the surrounding road surface. For example, a laser rangefinder module (four sets in total, corresponding to the areas below the front left, front right, rear left, and rear right drive motors respectively, ensuring coverage of the wheel trajectory and the surrounding road surface) is installed on the vehicle chassis to accurately collect the physical height data of the pothole area, with a sampling frequency of 10Hz.

[0022] First, based on the spatial distribution of laser ranging points, a region growing algorithm is used to divide the set of ranging points at the edge of the pit. and the set of internal ranging points on the edge Using the distance measurement values ​​of smooth road surfaces as a benchmark, distance measurement points that deviate from the benchmark by more than a set threshold (which can be set according to actual implementation, such as 10cm) and are continuously distributed are identified as pothole areas, thus obtaining an edge distance measurement point set composed of points located at the pothole boundaries. Points located inside the boundary form the internal ranging point set. Calculate the mean laser ranging values ​​for the two types of point sets respectively. (mean of edge distance measurement) and (Internal distance mean), pothole elevation difference It is the absolute value of the difference between the two means, i.e. , This represents the maximum value function.

[0023] In one embodiment, the method for constructing the electric braking response sequence of each wheel is as follows: An electrical parameter acquisition module installed on the four-wheel electric drive / braking actuator of the vehicle collects the time-series data of the electric braking response during the vehicle's passage through a pothole. Electric braking response refers to braking torque, used to quantify the dynamic response characteristics of the electric drive system under pothole impact. The sampling frequency is set to 100Hz, the acquisition duration is 0.5s, and each wheel's electric drive / braking unit can obtain 50 time-series sampling points. The electric braking response characteristics of each wheel correspond to a sequence. , Indicates the first The wheels, i=1,2,3,4 correspond to the electric drive / braking units of the four wheels: front left, front right, rear left, and rear right, respectively.

[0024] In one embodiment, vehicle body acceleration data also needs to be collected for subsequent braking smoothness index calculation. Specifically, using an onboard triaxial accelerometer, based on the sampling time of the collected drive motor data, the vertical acceleration data of the vehicle body during the passage through the pothole is simultaneously collected, and the acceleration time-series data is standardized and normalized. The acceleration time series data, with time points t=1,2,…,50, are used to quantify the degree of vehicle body vibration.

[0025] Specifically, pothole road conditions Including pothole elevation differences Electric braking response sequence .Right now And record the characteristics of the potholes and road conditions. Braking comfort index . A feature set consisting of pothole road condition features .

[0026] S2: An improved K-means clustering algorithm is used to cluster the pothole road condition features to generate pothole road condition clusters representing different electric braking response sequences. The distance function used for clustering is determined based on the pothole height difference and the electric braking response sequence of the electric vehicle when passing through the pothole. The braking comfort index is calculated based on the acquired vehicle acceleration data. The baseline recovery efficiency for each type of pothole road condition is determined based on the average braking comfort index of each cluster at the time of vehicle leaving the factory.

[0027] In one embodiment, the method for calculating the braking comfort index based on the acquired vehicle acceleration data is as follows: The peak value in the normalized acceleration time-series data is obtained; the root mean square (RMS) value of the acceleration time-series data sequence is calculated; the sum of the peak value and the RMS value is calculated, and the absolute value of the sum is divided by 2 to obtain the vibration level; finally, the inverse of the vibration level is calculated, and the exponential function value of the inverse is obtained, thus yielding the braking comfort index. .

[0028] Suppose that several pothole road condition samples were collected during the factory testing phase, and three typical pothole road condition clusters were obtained through improved K-means clustering. Cluster 1 corresponds to shallow and high-frequency small bumps (e.g., pothole height difference of about 3cm, electric braking response sequence fluctuates frequently but with small amplitude), and the current cluster center of each pothole road condition cluster is marked as "shallow bumps"; Cluster 2 corresponds to medium-depth standard potholes (exemplary height difference of about 6cm, four-wheel electric braking response amplitude is symmetrical and fluctuates moderately), and is marked as "medium potholes"; Cluster 3 corresponds to deeper potholes with strong impact (exemplary height difference of about 10cm, rear wheel motor braking torque response peak is significantly greater than front wheel), and is marked as "deep potholes".

[0029] For example, the following definitions are provided: frequent fluctuations are defined as the number of local extreme points in the electric braking response sequence within a unit time (0.5s) not less than 8; small amplitude is defined as the normalized RMS value of the response sequence being less than 0.5, medium amplitude is defined as being between 0.5 and 2.0, and large amplitude is defined as being greater than 2.0; the DTW distance of the electric braking response sequence of any two diagonally opposite wheels is defined as not greater than 0.3, or the difference in RMS response between the left and right wheels is less than 0.5 and the ratio of the RMS response between the front and rear wheels is ∈ [0.8, 1.2]; the rear wheel response is defined as significantly greater than the front wheel response when the ratio of the mean RMS response of the rear wheel to the mean RMS response of the front wheel is not less than 1.4, or the difference between the peak value of the maximum braking torque response of the rear wheel and the peak value of the maximum braking torque response of the front wheel is greater than 8 Nm.

[0030] Each label is defined by the sample features (pothole height difference, four-wheel electric braking response sequence) of the corresponding cluster center and the average braking smoothness index of that cluster, serving as the basis for subsequent real-time classification and electric braking torque distribution strategy invocation. The initial cluster center is obtained by sorting the samples in ascending order according to pothole height difference, resulting in an ordered height difference sequence. Calculate the difference sequence of pit height differences between adjacent samples, and calculate the ordered height difference sequence. Height difference between adjacent samples This characterizes the degree of difference in elevation between adjacent pits, i.e. .

[0031] A mutation threshold is set based on the mean and standard deviation of the difference sequence; samples with pit height differences greater than the mutation threshold are designated as gradient mutation points, and the initial number of clusters is determined based on these gradient mutation points. For example, calculate all mean and standard deviation Setting mutation thresholds based on statistical criteria Screening for significantly different neighboring samples: mutation threshold Where λ is a preset coefficient, preferably 2 or 3, which can be fine-tuned during the calibration phase according to the vehicle model positioning during actual deployment. Finally, the statistics satisfy... Number of gradient mutation points Initial number of clusters .

[0032] At this point, the initial cluster number has been completed. The selection of values.

[0033] from Based on the above mutation points, select from the samples One sample was used as the initial cluster center. To facilitate subsequent clustering and improve accuracy, the initial cluster centers should cover all road condition feature distributions, and each cluster center should represent the typical road conditions and corresponding suspension response of the corresponding cluster.

[0034] Specifically, depending on the selection ordered height difference sequence divided by time value With gradient mutation points, Evenly divided into For each interval, perform the following operations to filter the initial cluster center (i.e., select one sample from each interval as the initial cluster center).

[0035] Based on the gradient mutation points, the sorted samples are divided into several height difference intervals. Within each height difference interval, the root mean square value of each electric braking response sequence in each sample is calculated as the average response amplitude of each drive motor under the pothole road conditions corresponding to that height difference interval. The mean value of the average response amplitude of each drive motor is calculated to obtain the comprehensive electric braking response index of each sample.

[0036] For example, calculate the root mean square value of the electric braking response sequence of the i-th wheel for each sample within the height difference interval. ( This represents the average electric braking response amplitude of the corresponding wheel in the sample under pothole conditions. Thus, the first... The first sample within each height difference interval Average response amplitude of each wheel Calculate the comprehensive electric braking response index for each sample. Characterizing that the sample in the first The overall dynamic response of the electric braking system within a certain elevation difference range is representative, and its calculation formula is as follows: .

[0037] Each cluster center obtained Structural and pothole road characteristics Consistency, that is: .in, For the first The elevation difference of the pits at the center of the initial cluster; For the first The initial cluster center The electric braking response sequence of each wheel. A category index for pothole condition clusters. The index remains consistent throughout the initialization, iterative update, and real-time control phases.

[0038] The sample with the largest comprehensive electric braking response index within each elevation difference interval is selected as the initial cluster center for that cluster. The initial cluster center includes the pothole elevation difference and the electric braking response sequence of the four-wheel electric drive / braking unit. Specifically, within the elevation difference interval... The largest sample is used as the initial cluster center of the cluster. , The largest sample represents the most significant dynamic excitation of the vehicle's electric drive-braking system by the pothole road condition, and its electric braking coordinated response mode has the best operating condition differentiation. Using such samples as the initial cluster centers helps the algorithm quickly gather samples with similar electric braking response characteristics, ensuring the efficiency of the clustering iteration process and the accuracy of the final road condition cluster division.

[0039] Electric braking response sequences are time-series data. Directly using the raw time-series data makes it difficult to quantify the overall dynamic response amplitude of a single wheel's electric drive / braking unit under specific pothole conditions. Therefore, the root mean square (RMS) value is introduced to comprehensively characterize the fluctuation intensity of the electric braking response time-series data. This involves traversing the set of pothole condition features. All Individual sample features Calculate each sample and Cluster center distance The samples are assigned to the cluster with the smallest distance.

[0040] In one embodiment, the distance is calculated as follows: the difference between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as a height difference term. The normalization method is as follows: the absolute value of the ratio of the difference between the real-time sample and the current cluster center of each pothole condition cluster to the maximum pothole height difference in the sample is used as the normalization result.

[0041] The dynamic time warping distance between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as the braking response difference term. The normalization method is as follows: calculate the dynamic time warping distance between the real-time sample and the current cluster center of each pothole condition cluster; calculate the maximum dynamic time warping distance between any two samples in the feature set constructed during the factory testing phase; and use the ratio of the dynamic time warping distance to the maximum dynamic time warping distance as the normalization result.

[0042] For example, the expression for the distance function can be: In the formula, Indicates the characteristics of potholes in road conditions and The distance between them; This represents the weighting coefficient of the pothole elevation difference in the distance function; and These represent the characteristics of potholes on the road. and The difference in elevation between the pits; This represents the maximum pothole elevation difference recorded in historical data, used to map the distance value of the pothole elevation difference between 0 and 1. For the above-mentioned height difference items.

[0043] DTW (Dynamic Time Warping) represents the dynamic time warping function; Indicates the characteristics of potholes in road conditions The Characteristics of the electric braking response sequence of each wheel; Indicates the characteristics of potholes in road conditions The The electric braking response sequence characteristics of each wheel. 4 refers to the fact that each pothole road condition has electric braking response sequence characteristics of 4 wheels. This represents the maximum dynamic time warping distance between any two samples in the feature set constructed during the factory testing phase. It is used to map the numerical range of the dynamic time warping distance term to between 0 and 1, ensuring the uniqueness and determinism of the normalized benchmark during online matching. This refers to the aforementioned braking response difference term.

[0044] This represents the quantification of differences in the core features of potholes, enabling the differentiation of pothole depth differences. Simultaneously... This represents the quantification of differences in the detailed features of electric braking response. By calculating the DTW distance of the time sequence features of the electric braking response of the same wheel in two samples, and using the mean, the coordinated electric braking response of the four wheels is quantified into a single value. This ensures that scenarios with "potholes of the same depth but different electric braking responses" can be effectively distinguished, such as shallow potholes but different vehicle loads, or shallow potholes but different actual minor bump characteristics, which lead to differences in electric braking response.

[0045] It can be set to a fixed threshold or a variable threshold. In the fixed threshold implementation, A preset fixed value is taken, for example, 0.6; in the variable threshold implementation, for example, The setup method can be: Among them, it can make The value ranges from 0.5 to 0.7. The purpose of a value greater than 0.5 is to ensure that the clustering does not deviate from the geometric essence of the road surface. At this time, the difference in elevation plays a dominant role in the distance calculation. The algorithm can force samples with the same geometric features to be anchored in the same cluster. The value less than 0.7 is to preserve the sensitivity of the electric braking timing features to system attenuation / load changes. This range is an exemplary preferred value and is not the only necessary condition for realizing the invention. Indicates the height difference of characteristic pits The coefficient of variation reflects the degree of dispersion of the pothole height difference in the sample; This represents the hyperbolic tangent function, used to convert the aforementioned coefficient of variation. Mapping to a value between 0 and 1 avoids anomalies caused by the coefficient of variation exceeding 1 or even larger. The standard deviation of the sample pit height difference divided by the mean of the sample pit height difference is used as the height difference. The coefficient of variation; when the mean of the pit height difference in the sample is 0, the coefficient of variation is set to 0 by default.

[0046] when When the potholes are larger, the road conditions are more complex, indicating that the difference in pothole elevation is more valuable and requires increased precision. Strengthen its dominant role in clustering; conversely, When smaller, The value is slightly increased from 0.5, which maintains the weight advantage of pothole elevation difference as the main feature of clustering, while retaining the auxiliary role of electric braking response feature in distinguishing subdivided road conditions.

[0047] The cluster center can be updated iteratively. The method is as follows: For each pothole condition cluster, based on all samples assigned to that pothole condition cluster, repeatedly update the pothole elevation difference and the electric braking response sequence of the cluster center until the change in the pothole elevation difference and the change in the electric braking response sequence of the cluster center reach the preset convergence condition. The method for calculating the change in pit height difference is as follows: The difference in pit height difference of the same cluster center in two adjacent iterations is calculated using the absolute difference method: that is, take the pit height difference value of the cluster center obtained in the current iteration, subtract the pit height difference value of the same cluster center obtained in the previous iteration, and then take the absolute value of the difference to obtain the change in pit height difference.

[0048] The method for calculating the change in the electric braking response sequence is as follows: The mean square error method is used to calculate the difference in the electric braking response sequences of the four wheels at the same cluster center in two adjacent iterations: For each wheel, the difference between the current iteration cluster center response sequence and the previous iteration cluster center response sequence is calculated at the corresponding sampling point; the average of the squared differences of all sampling points for each wheel is taken to obtain the change in the response sequence of that wheel; the arithmetic mean of the change in the response sequences of the four wheels is taken to obtain the change in the electric braking response sequence of the cluster center.

[0049] In one embodiment, the convergence condition is that the changes in pothole elevation and the changes in the electric braking response sequence at the cluster center are both less than a preset convergence threshold. This convergence threshold is an empirical value and can be adjusted according to the specific implementation, for example, 1×10⁻⁻⁶. 4 .

[0050] In one embodiment, the method for updating the cluster center pit height difference is: taking the mean of all samples in the cluster as the new cluster center pit height difference; the method for updating the cluster center electric braking response sequence is: taking the time mean of the electric braking response sequence corresponding to each sample in the cluster as the new electric braking response sequence.

[0051] The improved K-means algorithm addresses the problem that traditional clustering cannot simultaneously consider both "coarse features of pothole conditions (pothole elevation difference H)" and "detailed features of electric braking response (time-series data of four-wheel electric drive / braking units)" by fusing the distance function of features and the pre-selection logic of the number of clusters K. This improves the matching degree between cluster division and actual road conditions and the dynamic response of the electric braking system. The core clustering parameter (number of clusters K) is adjustable, and the selection of the initial cluster centers during the clustering process is based on the K value as a prerequisite constraint, providing a clear adjustment target for the feedback design closed-loop optimization based on the braking comfort index evaluation in step S3.

[0052] In a preferred embodiment, before initiating a full DTW path search, the real-time electric braking response sequence and cluster center sequence are first obtained. Lower bound distance. The lower bound distance function is dynamically time-warped for quickly filtering non-matching candidate clusters; existing techniques will not be elaborated upon here. If this lower bound distance is greater than the minimum distance threshold in the current matching process, the system directly determines that the candidate cluster is not an optimal match and immediately skips subsequent steps. The complexity is fully regularized. This pre-screening mechanism reduces invalid timing alignment operations. This indicates that the computational cost of the algorithm increases with the square of N.

[0053] By employing Sakoe-Chiba band constraints, the time warping path is strictly limited to the search bandwidth. Inside (of which) (where the number of sampling points in the sequence is used), reducing the single computation complexity from Down to The existing technology will not be elaborated further. At the same time, all timing data undergoes 16-bit fixed-point quantization before entering the normalization module. The calculation process only uses addition, subtraction and shift operations to fully adapt to the fixed-point DSP core of automotive-grade MCUs and avoid clock cycle loss and power consumption fluctuations caused by floating-point units.

[0054] Only when the lidar or visual pre-aiming system confirms that a pothole feature enters the corresponding wheel's trajectory within the next 0.5 seconds, does the chassis domain controller wake up the DTW calculation thread via the real-time operating system task scheduler. Under flat road or normal driving conditions, this feature matching module is in a low-power sleep state, only polling the pre-aiming flag via hardware interrupts. This architecture ensures that the average CPU load of the chassis domain controller MCU remains low, and the worst-case execution time meets the ASIL-B (Automotive Safety Integrity Level B) functional safety requirements for deterministic response. An engineering-level lightweight optimization strategy is introduced to calculate the DTW distance for braking response timing features. Specifically, in the risk assessment levels of the ISO 26262 international standard "Road Vehicles Functional Safety," ASIL-B is a medium safety risk level, requiring the system to have corresponding diagnostic coverage (≥90%), fault tolerance time interval (FTTI), and hardware random failure rate (≤10⁻). 7 Safety mechanisms such as / h are applicable to electronic and electrical systems that have a moderate impact on driving safety, such as chassis domain control.

[0055] S3: Based on the distance calculation, the real-time pothole road condition features are distributed to the pothole road condition clusters with the smallest distance, and this pothole road condition cluster is taken as the target cluster; based on the baseline recovery efficiency and braking smoothness index of the target cluster, the braking coordination adjustment factor is calculated; based on the braking coordination adjustment factor, the electric braking coordination control strategy corresponding to the pothole road condition category in the electric braking system of the electric vehicle chassis domain is optimized and adjusted.

[0056] Based on the baseline recovery efficiency and braking smoothness index of the target cluster, a braking coordination adjustment factor is calculated. The calculation method is as follows: obtain the baseline recovery efficiency of the target cluster calibrated at the time of vehicle delivery and compare it with the current braking smoothness index; calculate the relative deviation between the current braking smoothness index and the baseline recovery efficiency; in response to the absolute value of the relative deviation being greater than a preset tolerance threshold, generate a braking coordination adjustment factor based on the relative deviation, and perform upper and lower limit processing on the braking coordination adjustment factor to constrain it within a preset safety adjustment range.

[0057] In one embodiment, the determination of the baseline recovery efficiency is performed after the improved K-means clustering algorithm has iteratively converged. Specifically, when the change in cluster centers converges to a preset threshold, all samples in the feature set are assigned to the finally determined pothole condition clusters; subsequently, all samples within each cluster are statistically analyzed and denoted as follows: ,get The corresponding braking comfort index is used to calculate its arithmetic mean, which is then defined as the baseline recovery efficiency for the corresponding cluster. This baseline recovery efficiency characterizes the baseline performance level of electric braking / energy recovery for this type of pothole road condition in the initial state of the vehicle after leaving the factory.

[0058] For example, the update cycle of the electric braking coordinated control strategy for the chassis domain electric braking system of electric vehicles can be set according to the corresponding vehicle inspection and annual inspection time or the regular maintenance time of unmanned urban delivery vehicles. For ordinary passenger cars, the update cycle can be set to every 12 months or every 20,000 kilometers (whichever comes first); for high-frequency unmanned urban delivery vehicles, the electric braking coordinated control strategy can be evaluated and updated every 3 months or every 10,000 kilometers. When each update cycle arrives, the system automatically triggers the chassis domain control module to collect the actual operating data of recent road conditions through potholes, and performs classification matching, braking smoothness index comparison and braking coordination adjustment factor calculation, thereby completing the adaptive optimization of the electric braking coordinated control strategy.

[0059] During the update phase of the aforementioned electric braking coordinated control strategy, characteristic data of the current road conditions involving potholes are collected, including data from road surface pre-aiming sensors (such as LiDAR / cameras), dynamic response sequence characteristics of the electric drive / braking system, and vehicle longitudinal acceleration / deceleration characteristics. The characteristics of the current road conditions involving potholes are obtained as follows: and its corresponding braking smoothness index .

[0060] Calculate the current characteristics of potholes in the road. Distance from each cluster center distance The current pothole conditions are assigned to the nearest cluster (pothole condition classification), and the baseline recovery efficiency of the nearest cluster is calculated to determine the pothole condition. Braking coordination adjustment factor for pothole-like road conditions .

[0061] In one embodiment, the expression for the braking coordination adjustment factor can be: ;in Represents the amplitude limiting function. This indicates the lower limit calibrated based on the physical boundaries of the chassis actuators and the braking safety margin. This represents the upper limit calibrated based on the physical boundaries of the chassis actuators and the braking safety margin. For example, it is taken as... =−0.3, =0.2, To prevent a zero threshold, for example, a threshold of 10⁻³ is used. , and These are all hyperparameters and can be adjusted according to the actual application scenario. For the... Pothole-like road conditions This indicates the corresponding baseline recovery efficiency; Indicates braking smoothness index. Braking coordination adjustment factor. It quantifies in real time the degree to which the current performance deviates from the factory benchmark. The larger the value, the more severe the system performance degradation and the greater the torque compensation. The braking coordination adjustment factor is calculated multiple times for the same pothole road condition category. Take the arithmetic mean to get the final result. Braking coordination adjustment factor for pothole-like road conditions .

[0062] The electric braking coordinated control strategy for the corresponding pothole road condition category in the electric braking system of the electric vehicle chassis domain is optimized and adjusted. The optimization and adjustment method is as follows: The electric braking torque coefficient is calculated and then sent as a real-time control command from the chassis domain controller to the braking actuator to obtain the electric braking torque coefficient used under all road conditions. The electric braking torque coefficient is used to characterize the electric braking torque distribution ratio, and its acquisition method is common knowledge. The optimized electric braking torque coefficient is calculated. , In the formula, The optimized electric braking torque coefficient; For the first Braking coordination adjustment factor under pothole-like road conditions; The maximum adjustable electric braking torque coefficient can be determined through bench testing and real-vehicle road calibration during the factory calibration phase and stored in the non-volatile memory of the chassis domain controller. For example, the value is... It can be adjusted according to the actual application scenario. This represents the electric braking torque coefficient used under the current road conditions.

[0063] Based on the classification of pothole conditions, the system adaptively adjusts the electric braking torque coefficient for each corresponding road condition category. While ensuring the adjusted electric braking torque coefficient does not exceed the limit, the system determines the actual amount of electric braking torque coefficient that should be increased, thus obtaining the actual torque compensation amount required.

[0064] S4: After one optimization adjustment, continue to compare the deviation rate between the actual braking smoothness index and the benchmark recovery efficiency. If the deviation rate is greater than the preset threshold, the benchmark recovery efficiency of the corresponding pothole road condition cluster will be dynamically updated based on the deviation rate.

[0065] After the electric braking compensation is adjusted, new pothole and road condition characteristics are collected simultaneously during vehicle operation. Calculate the average braking ride comfort index for the same type of pothole road condition after adjustment, and use it as the braking ride comfort index for the same type of pothole road condition after adjustment. , Indicates the first Various potholes and road conditions. This will be used as input data for calculating the deviation rate in step S4, thereby determining whether the baseline recovery efficiency is triggered. Weighted update.

[0066] Specifically, calculate the deviation rate. The formula can be: ; Indicates the first The baseline recovery efficiency corresponding to various pothole road conditions.

[0067] In one embodiment, a threshold is set for when the desired effect is not achieved. Based on specific engineering requirements, such as That is, a deviation rate ≤ 10% is considered acceptable; if If the adjustment achieves the desired effect, the system remains unchanged; if If the adjustment fails to meet the target, it indicates that the current electric drive-braking system is experiencing significant performance degradation or electrical characteristic drift. In this case, the baseline recovery efficiency for the corresponding pothole road condition classification should be appropriately adjusted. This makes the evaluation standard for energy recovery in such pothole road conditions more dynamically adaptable, providing a more reasonable reference benchmark for the iterative adjustment of subsequent electric braking coordinated control strategies, i.e., updating the benchmark recovery efficiency. Referring to the ISO 2631-1 vibration comfort standard, a 10% ride comfort deviation corresponds to the "perceptible but acceptable" boundary of human subjective perception, and is therefore set as the critical value for triggering benchmark updates.

[0068] Specifically, the corresponding number The baseline recovery efficiency update formula for each pothole condition category (cluster) is as follows: In the formula, Indicates the updated number The baseline recovery efficiency for each pothole road condition classification; This represents the deviation rate, which is used in the formula to characterize the degree of tolerance. This represents the braking smoothness index under the same pothole conditions after adjustment. The weights of the original benchmark and actual performance are dynamically allocated based on the deviation rate, achieving smooth iteration of the benchmark strength. A larger deviation rate indicates a more significant performance degradation in the electric drive-braking system, a lower compatibility with the original factory benchmark, and a greater need to consider the current actual recovery efficiency. Perform a baseline update.

[0069] To ensure the numerical stability of the weighted fusion process, the actual calculation is performed using... When the amplitude is limited, When the system determines that the electric drive-braking system has experienced a significant performance drift, it directly uses the current measured average value. As a new benchmark, namely .

[0070] This application achieves dynamic adaptation of the baseline recovery efficiency, effectively overcoming the shortcomings of traditional technologies that adhere to fixed factory calibration values ​​and cannot cope with the performance degradation of the electric drive-braking system throughout its entire life cycle. The baseline is dynamically updated based on the deviation rate of the actual adjustment effect, ensuring that the recovery efficiency evaluation standard always aligns with the real-time health status of the vehicle chassis's electrical system. This avoids a disconnect between the baseline and actual capabilities caused by motor demagnetization, inverter aging, or increased battery internal resistance, ensuring that subsequent optimization goals for electric braking torque distribution and energy recovery strategies have continuous real-world adaptability and control precision.

[0071] This application also discloses an integrated chassis domain control system based on a clustering algorithm, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the integrated chassis domain control method based on the clustering algorithm according to this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0072] In this application, the aforementioned memory can be any tangible medium that contains or stores a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as resistive random access memory (DRAM), dynamic random access memory (DRAM), static random access memory (SRAM), etc., or any other medium that can be used to store desired information and can be accessed by an application program, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.

[0073] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. An integrated chassis domain control method based on clustering algorithm, characterized in that, Including the following steps: Construct the characteristics of pothole road conditions during vehicle operation; the characteristics of pothole road conditions include the pothole elevation difference and the electric braking response sequence when the electric vehicle passes through the pothole; Constructing the characteristics of potholes and road conditions during vehicle travel, including: During the factory testing phase, test data of the vehicle under various pothole road conditions is acquired and used as samples; the samples are processed to construct multi-dimensional features about pothole height difference and electric braking response sequence; and a feature set is constructed. An improved K-means clustering algorithm is used to cluster pothole road condition features, generating pothole road condition clusters representing different electric braking response sequences. The distance function used for clustering is determined based on the pothole elevation difference and the electric braking response sequence of the electric vehicle passing through the pothole. The improved K-means clustering algorithm for pothole road condition features includes initializing cluster centers. The method for initializing cluster centers is as follows: The samples are sorted in ascending order according to the difference in pit height, and the difference sequence of pit height between adjacent samples is calculated. A mutation threshold is set based on the mean and standard deviation of the difference sequence. Samples with pit height difference greater than the mutation threshold are taken as gradient mutation points, and the initial number of clusters is determined based on the gradient mutation points. Based on the gradient mutation points, the sorted samples are divided into several high-difference intervals. Within each high-difference interval, the root mean square value of each electric braking response sequence in each sample is calculated as the average response amplitude of each drive motor under the pothole road conditions corresponding to that high-difference interval. The mean value of the average response amplitude of each drive motor is calculated to obtain the comprehensive electric braking response index of each sample. The sample with the largest comprehensive electric braking response index in each elevation difference interval is selected as the initial cluster center of the cluster. The initial cluster center is a multi-dimensional feature vector, including the pothole elevation difference and the electric braking response sequence of each drive motor. The braking smoothness index is calculated based on the acquired vehicle acceleration data. The baseline recovery efficiency for various pothole road conditions is determined based on the average braking smoothness index of each cluster at the time of vehicle leaving the factory. Based on the distance function, the distance from the real-time pothole condition feature to each pothole condition cluster is calculated. The real-time pothole condition feature is assigned to the pothole condition cluster with the smallest distance and this pothole condition cluster is taken as the target cluster. Based on the baseline recovery efficiency and braking smoothness index of the target cluster, the braking coordination adjustment factor is calculated. Based on the braking coordination adjustment factor, the electric braking coordination control strategy corresponding to the pothole condition category in the electric braking system of the electric vehicle chassis domain is optimized and adjusted. After one optimization adjustment, the deviation rate between the actual braking smoothness index and the benchmark recovery efficiency is compared. If the deviation rate is greater than the preset threshold, the benchmark recovery efficiency of the corresponding pothole road condition cluster is dynamically updated based on the deviation rate.

2. The integrated chassis domain control method based on clustering algorithm according to claim 1, characterized in that, The distance is calculated as follows: Calculate the difference in pothole height between the real-time sample and the current cluster center of each pothole condition cluster, and normalize it as a height difference term; Calculate the dynamic time-normalized distance between the real-time sample and the current cluster center of each pothole condition cluster, and normalize it as the braking response difference term. The distance between the real-time sample and the current cluster center of each pothole condition cluster is obtained based on the height difference term and the braking response difference term.

3. The integrated chassis domain control method based on clustering algorithm according to claim 2, characterized in that, The difference in pothole height between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as a height difference term. The normalization method is as follows: the absolute value of the ratio of the difference in pothole height between the real-time sample and the current cluster center of each pothole condition cluster to the maximum pothole height difference in the sample is used as the normalization result.

4. The integrated chassis domain control method based on clustering algorithm according to claim 2, characterized in that, The dynamic time-normalized distance between the real-time sample and the current cluster center of each pothole condition cluster is calculated and normalized as the braking response difference term. The normalization method is as follows: Calculate the dynamic time warping distance between the real-time sample and the current cluster center of each pothole road condition cluster; calculate the maximum dynamic time warping distance between any two samples in the feature set constructed during the factory testing phase; and use the ratio of the dynamic time warping distance to the maximum dynamic time warping distance as the normalized result.

5. The integrated chassis domain control method based on clustering algorithm according to claim 1, characterized in that, An improved K-means clustering algorithm is used to cluster pothole road condition features. This includes: for each pothole road condition cluster, based on all samples assigned to that pothole road condition cluster, repeatedly updating the pothole height difference and the electric braking response sequence of the cluster center until the change in pothole height difference and the change in the electric braking response sequence of the cluster center reach the preset convergence condition.

6. An integrated chassis domain control system based on clustering algorithm, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the integrated chassis domain control method based on a clustering algorithm according to any one of claims 1-5.

Images