A vehicle control method and system based on road conditions

Abstract

The application discloses a vehicle control method and system based on road conditions, the method comprising: acquiring vehicle running state data and front road state data, generating a vehicle prediction trajectory and constructing a wheel position level road segment sequence corresponding to each wheel; determining an abnormal event, calculating the inter-axis entry time difference and the left-right entry time difference, and obtaining the abnormal event credibility in combination with a credibility evaluation model; determining the stage in which the vehicle is located according to the entry time difference, the credibility and the actual entry state, generating brake path constraint parameters, attitude correction constraint parameters and / or suspension auxiliary control parameters, and outputting brake control instructions, steering control instructions and suspension auxiliary control instructions in combination with the holding window.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, specifically to a vehicle control method and system based on road conditions. Background Technology

[0002] With the development of vehicle chassis electronic control technology, vehicle control systems have gradually evolved from single-actuator control to collaborative control of multiple actuators such as braking, steering, and suspension. Existing vehicle control schemes based on road conditions typically acquire vehicle operating status and road conditions ahead, identifying operating information such as road surface adhesion and smoothness, and then adjusting relevant control parameters for braking, steering, or suspension accordingly. These schemes can meet general control requirements when road conditions are relatively uniform and the wheels are affected by road conditions in a largely synchronized manner.

[0003] However, in real-world road conditions, vehicles frequently encounter complex situations such as localized water accumulation, bridge joints, manhole cover edges, repair strips, and short, rough road sections. These conditions are characterized by discontinuous spatial distribution, short duration, and varying timing of impact on the left and right sides or front and rear axles. Especially during scenarios like cornering braking, downhill deceleration, and lane change corrections, the left front wheel, right front wheel, left rear wheel, and right rear wheel often enter abnormal road surfaces at different times, causing inconsistent changes in the distribution of longitudinal and lateral forces on the tires within a short period. If the control system still judges based on a uniform operating condition for the entire vehicle, it is prone to problems such as premature or late control intervention, or abrupt switching of control parameters.

[0004] Meanwhile, changes in road surface adhesion and road surface smoothness often occur simultaneously. Factors such as rough road surfaces and impacts at joints can cause short-term fluctuations in signals such as wheel speed, deceleration, and yaw rate, thereby interfering with the judgment of road conditions. In existing technologies, if the braking path or attitude correction strategy is switched directly based on the single-cycle detection results, it is easy to cause repeated switching between regenerative braking and hydraulic braking, or frequent changes in attitude correction in a short period of time, thereby affecting the smoothness of vehicle operation and trajectory stability. Summary of the Invention

[0005] This application provides a vehicle control system and control method based on road conditions, in order to at least solve some of the technical problems existing in the related technologies described above.

[0006] According to a first aspect of the embodiments of this application, a vehicle control method based on road conditions is provided, comprising: A vehicle prediction trajectory is generated based on vehicle operating status data and road condition data ahead, and a wheel position level road segment sequence corresponding to the left front wheel, right front wheel, left rear wheel and right rear wheel is constructed based on the prediction trajectory. The abnormal events are determined based on the spatial correspondence of abnormal segments in each wheel position level road segment sequence, and the inter-axle entry time difference and left and right entry time difference of the abnormal events are calculated. The credibility of the abnormal event is obtained by inputting the event characteristics of the abnormal event into the credibility evaluation model. Based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry status of each wheel in response to the abnormal event, the vehicle is determined to be in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage. According to the determined stage, corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters are generated. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

[0007] As an optional approach, the wheel position level road segment sequence is constructed by including: The vehicle's center of gravity prediction trajectory is generated based on vehicle speed, steering wheel angle, steering wheel angle change rate, and road curvature. Based on the vehicle's wheelbase and track width, the predicted trajectory of the vehicle's center of gravity is converted into the predicted wheel tracks of each of the four wheels. The road segments in the road condition data ahead are mapped to each of the predicted wheel tracks to obtain the wheel position level road surface segment sequence for the corresponding wheel. For each segment in each wheel position level road surface segment sequence, the segment start distance, segment end distance, adhesion level, smoothness level, expected entry time, and expected exit time are recorded.

[0008] As an optional approach, the abnormal events are obtained by aggregating road surface segment sequences at the wheel position level for different wheels. The aggregation rules include: When multiple wheel position segments correspond to the same area in the longitudinal position, and at least one of them has a decrease in adhesion level or a decrease in smoothness level, or both the adhesion level and the smoothness level decrease simultaneously, the multiple wheel position segments are merged into the same abnormal event. The primary attachment level and primary flatness level of the aforementioned abnormal event are determined as the most unfavorable attachment level and the most unfavorable flatness level among the merged segments.

[0009] As an optional approach, the credibility evaluation model is a feedforward neural network model, which sequentially includes an input layer, a feature normalization layer, a first fully connected layer, a first activation layer, a second fully connected layer, a second activation layer, a third fully connected layer, and an output layer. The event features received by the input layer include at least the main attachment level, main smoothness level, abnormal event length, the inter-axle entry time difference, the left and right entry time difference, the duration of the abnormal event, wheel speed change amplitude statistics, longitudinal deceleration change rate statistics, yaw rate change rate statistics, vehicle speed, braking request quantity, road curvature, and consistency identifier of previous and subsequent periodic events. The first fully connected layer performs a linear transformation on the standardized feature vector to output a first hidden feature vector; the first activation layer performs a non-linear mapping on the first hidden feature vector; the second fully connected layer receives the output of the first activation layer and performs another linear transformation to obtain a second hidden feature vector; the second activation layer continues to perform a non-linear mapping on the second hidden feature vector; the third fully connected layer maps the output of the second activation layer to a single real number; the output layer outputs the credibility of the abnormal event.

[0010] As an optional approach, determining whether the vehicle is in the preparation phase, partial entry phase, effective continuous phase, or exit holding phase includes: When the earliest expected entry time of the abnormal event falls within a preset preparation time window and the confidence level reaches the preparation stage entry threshold, it is determined to be in the preparation stage. When at least one relevant wheel has entered the abnormal event and at least one relevant wheel has not yet entered the abnormal event, and the confidence level reaches the partial entry stage threshold, it is determined to be in the partial entry stage. When at least two related wheel positions have entered the abnormal event, or when the same abnormal event persists for multiple consecutive control cycles and the credibility reaches the threshold for entering the effective persistence stage, it is determined to be an effective persistence stage. When the wheel that first entered the abnormal event begins to leave the abnormal event and at least one holding window is in the open state, it is determined to exit the holding phase.

[0011] As an optional approach, generating the braking path constraint parameters and the attitude correction constraint parameters during the preparation phase includes: Based on the main adhesion level, the main smoothness level, the reliability, vehicle speed, and abnormal event type, a preset mapping table is queried to obtain the upper limit of regenerative braking growth rate, the upper limit of hydraulic braking additional pressure change rate, the upper limit of additional braking force differential, and the upper limit of steering correction change rate. The upper limit of the regenerative braking growth rate and the upper limit of the hydraulic braking additional pressure change rate are written into the braking path constraint parameters, and the upper limit of the additional braking force differential and the upper limit of the steering correction change rate are written into the attitude correction constraint parameters.

[0012] As an optional approach, braking control commands and steering control commands are generated during the partial entry phase, including: Determine the set of wheel positions that have entered the abnormal event and the set of wheel positions that have not entered the abnormal event; apply a first upper limit constraint to the braking additional amount corresponding to the set of wheel positions that have entered the abnormal event, and maintain a normal control upper limit for the set of wheel positions that have not entered the abnormal event; A unidirectional gradual constraint is applied to the attitude correction amount, so that the attitude correction amount changes in the same direction in adjacent control cycles until the stage is re-determined.

[0013] As an optional approach, opening the hold window during the effective duration phase includes: Based on the confidence level, the primary adhesion level, the primary smoothness level, the vehicle speed, the braking request amount, and the left and right entry time difference, the target regenerative braking upper limit, the target hydraulic braking distribution ratio, the additional braking force differential upper limit, and the steering correction target upper limit are generated. Open the brake path holding window, attitude correction holding window, and suspension assist holding window, and record the start time and shortest holding time of the corresponding holding window respectively.

[0014] As an optional approach, during the exit holding phase, constraints are gradually released according to the exit conditions of each holding window, including: When the holding time of the braking path holding window meets the requirements and the confidence level is lower than the exit threshold, the upper limit of regenerative braking is increased by a preset slope and the additional ratio of hydraulic braking is reduced. When the holding time of the attitude correction holding window meets the requirements and the yaw rate of change returns to the normal range, the upper limit of the additional braking force differential and the upper limit of the steering correction target are relaxed. When the suspension assist hold window holds for the required duration and the vehicle vertical response and / or suspension travel change return to normal range, a suspension assist exit command is output.

[0015] According to a second aspect of the embodiments of this application, a vehicle control system based on road conditions is also provided, comprising: The trajectory and segment construction module is configured to generate a vehicle prediction trajectory based on vehicle operating status data and road condition data ahead, and to construct wheel position level road segment sequences corresponding to the left front wheel, right front wheel, left rear wheel and right rear wheel based on the prediction trajectory. The abnormal event determination module is configured to determine abnormal events based on the spatial correspondence of abnormal segments in each wheel position level road segment sequence, and to calculate the inter-axle entry time difference and left and right entry time difference of the abnormal event. The credibility assessment module is configured to input the event characteristics of the abnormal event into the credibility assessment model to obtain the credibility of the abnormal event; The stage determination module is configured to determine whether the vehicle is in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry state of each wheel in response to the abnormal event. The control command output module is configured to generate corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters according to the determined stage. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

[0016] According to a third aspect of the embodiments of this application, an electronic device is provided, including: a processor; a memory for storing a computer program executable by the processor; wherein the processor is configured to execute the computer program in the memory to implement the method described in the first aspect.

[0017] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, which, when an executable computer program in the storage medium is executed by a processor, enables the implementation of the method described in the first aspect.

[0018] Compared to existing solutions that directly allocate control strategies based on unified vehicle operating conditions, this invention uses abnormal events as the organizing object, combining the sequential relationship of each wheel entering abnormal road surfaces, the credibility of abnormal events, and a phased holding mechanism to form a continuous processing flow oriented towards vehicle control. By constructing a wheel-position level road surface segment sequence and calculating the inter-axle entry time difference and left-right entry time difference, it can more accurately reflect the actual situation of asynchronous influence of the front and rear axles or left and right sides on the road surface in scenarios such as local water accumulation, bridge joints, single-sided repair strips, and short rough road sections. By evaluating the credibility of abnormal events and combining the control logic of the preparation stage, partial entry stage, effective duration stage, and exit holding stage, it can reduce the repeated switching of braking path and attitude correction path caused by single-cycle signal fluctuations. At the same time, by using the braking path holding window, attitude correction holding window, and suspension auxiliary holding window to constrain the duration and release timing of control commands, the control release process at the tail of abnormal events is smoother. As a result, the longitudinal deceleration change, yaw response, and trajectory correction process of the vehicle under complex road conditions are more stable, and the vehicle control process is more consistent with the actual operating state.

[0019] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Furthermore, no embodiment in this disclosure is required to achieve all the effects described above. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0021] Figure 1 This is a schematic diagram of a vehicle control method based on road conditions provided in an embodiment of this disclosure.

[0022] Figure 2 This is a schematic diagram illustrating the process of constructing a wheel position level road segment sequence according to an embodiment of this disclosure.

[0023] Figure 3 This is a schematic diagram of the model structure provided in the embodiments of this disclosure.

[0024] Figure 4 This is a schematic diagram of the control phase determination process provided in an embodiment of the present disclosure.

[0025] Figure 5 This is a schematic diagram illustrating the process of generating constraint parameters and control instructions provided in an embodiment of this disclosure.

[0026] Figure 6 This is a schematic diagram of a vehicle control system based on road conditions, provided as an embodiment of the present disclosure.

[0027] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0029] This implementation method is applicable to vehicles equipped with a vehicle controller, brake controller, steering controller, and suspension controller, and is especially suitable for electric vehicles. The vehicle has the ability to collect general status data and obtain the status of the road ahead, and can provide information such as vehicle speed, four-wheel speed, steering wheel angle, steering wheel angle change rate, longitudinal acceleration, lateral acceleration, yaw rate, braking request, steering request, regenerative braking status, hydraulic braking status, vehicle vertical acceleration, suspension travel change, road curvature, slope, and the level of adhesion and smoothness of the road ahead.

[0030] This method is applicable to conditions such as light to moderate braking on curves, downhill deceleration, lane change correction, single-sided water accumulation, local repair strips, bridge joints, short and rough road sections, and situations where there are significant differences in the timing of the front and rear axles or left and right wheels entering abnormal road surfaces. The vehicle controller can act as the main actuator, responsible for abnormal event identification, stage determination, constraint parameter generation, and window management; the brake controller, steering controller, and suspension controller execute corresponding controls according to the objectives issued by the vehicle controller.

[0031] The implementation process of the method described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Conventional adjustments or substitutions of each step by those skilled in the art without departing from the concept of this application should be included in the scope of protection of this application.

[0032] Please see Figure 1 , Figure 1 A flowchart of a vehicle control method based on road conditions according to an embodiment of the present invention is shown, such as... Figure 1 As shown, the method includes steps S1-S5: In step S1, a vehicle prediction trajectory is generated based on vehicle operating status data and road condition data ahead, and a wheel position level road segment sequence corresponding to the left front wheel, right front wheel, left rear wheel and right rear wheel is constructed based on the prediction trajectory.

[0033] In actual implementation, the vehicle controller can pre-save a set of calibration parameters, which include the upper limit of the aiming distance, control cycle, short segment length threshold, inter-axle entry time difference threshold, left and right entry time difference threshold, preparatory stage entry threshold, partial entry stage entry threshold, effective continuous stage entry threshold, exit threshold, braking path holding time, attitude correction holding time, suspension assist holding time, and several mapping table index boundaries.

[0034] The above parameters can be determined during the vehicle development stage through road tests, bench tests, and playback analysis, and written into non-volatile memory. During operation, the calibration values ​​are called. For the update coefficients, they are also determined during the calibration stage, with a value range of 0 to 1. They are used to balance the reliability of the current cycle model output with that of the previous cycle. The larger the value, the more sensitive the reliability changes; the smaller the value, the smoother the reliability changes.

[0035] Specifically, the exit threshold is less than the entry threshold of the preparation stage, the entry threshold of the preparation stage is no greater than the entry threshold of the partial entry stage, and the entry threshold of the partial entry stage is no greater than the entry threshold of the effective continuous stage. The preparation time window can be determined according to the pre-aiming time corresponding to the current vehicle speed; the inter-axle entry time difference threshold and the left and right entry time difference threshold can be determined according to the statistical results of the expected entry time difference under the target vehicle's wheelbase, track width, and common speed range; the braking path holding time, attitude correction holding time, and suspension auxiliary holding time can be calibrated according to the braking path stability, attitude convergence time, and vehicle vertical response decay time, respectively; the normal range of yaw rate change rate and the normal range of vehicle vertical response can be determined by the statistical distribution of normal road surface samples, wherein the normal range of vehicle vertical response is the statistical fluctuation range of the vehicle's vertical acceleration and / or suspension travel change under normal road surface conditions, and can be determined in conjunction with vehicle model calibration parameters; the preset slope is jointly calibrated according to the actuator's allowable rate of change and comfort constraints.

[0036] Please see Figure 2 , Figure 2 A schematic diagram illustrating the process of constructing a wheel-position level road segment sequence provided in an embodiment of this disclosure is shown. Figure 2 As shown, in step S101, the predicted trajectory of the vehicle's centroid is generated.

[0037] The vehicle controller reads vehicle operating status data such as current vehicle speed, steering wheel angle, steering wheel angle change rate, road curvature, and slope, as well as road condition data ahead, to calculate the vehicle's centroid prediction trajectory within the preview range. The road condition data used to generate the centroid prediction trajectory includes at least the road curvature and slope corresponding to each road segment within the preview range; the road condition data used for mapping subsequent road segments also includes the segment start distance, segment end distance, lateral coverage width, adhesion level, and smoothness level of each road segment.

[0038] The aiming range can be configured as a distance window that varies with vehicle speed; for example, a shorter aiming distance is used at lower vehicle speeds, and a longer aiming distance is used at higher vehicle speeds. The generation of the centroid prediction trajectory can be achieved using a vehicle geometric motion model. Specifically, the vehicle controller converts the steering wheel angle into the front wheel angle based on the vehicle's steering ratio, and within each preset time step, determines the predicted curvature based on the current vehicle speed, front wheel angle, steering wheel angle change rate, and the road curvature corresponding to the current position. Simultaneously, the predicted speed is corrected based on the slope corresponding to the current position. Then, the centroid position and heading angle of the next aiming discrete point are recursively obtained based on the predicted speed and predicted curvature. The centroid positions and heading angles of multiple aiming discrete points are arranged in chronological order, thus forming the centroid prediction trajectory of the vehicle within the aiming range.

[0039] Let the current steering wheel angle be... The rate of change of steering wheel angle The vehicle steering ratio is The vehicle wheelbase is The preset time step is , No. The road curvature corresponding to each pre-targeted discrete point is: Then the first The front wheel steering angle corresponding to each pre-aimed discrete point is: ; The steering curvature obtained from the front wheel steering angle is: ; The vehicle controller adjusts according to the current vehicle speed. Determine curvature blending coefficients ,in , It can be obtained from the preset vehicle speed-fusion coefficient mapping table; the predicted curvature is determined by the following formula: ; in, To predict curvature, This reflects the impact of the current rate of change in front wheel angle and steering wheel angle on the future bending trend of the vehicle. It reflects the constraint of road curvature on the vehicle's travel path at the current location.

[0040] Let the first The centroid positions of the pre-aimed discrete points are: The heading angle is The slope is The longitudinal acceleration is The predicted speed is then corrected using the following formula: ; in, This is the acceleration due to gravity. Furthermore, the vehicle controller recursively derives the next target discrete point based on the predicted velocity and predicted curvature: ; ; ; The vehicle controller sequentially obtains the centroid positions and heading angles of multiple pre-aiming discrete points, and arranges them in chronological order to form the centroid prediction trajectory of the vehicle within the pre-aiming range.

[0041] In this embodiment, the future path is estimated mainly by using the current vehicle speed and road curvature, while the near-field trajectory curvature is corrected by using the steering wheel angle change rate. After this processing, the centroid predicted trajectory reflects the current vehicle motion state and also retains the response to the upcoming steering change.

[0042] In step S102, predicted wheel tracks for each of the four wheels are generated. After the center of gravity prediction trajectory is generated, the vehicle controller converts the center of gravity prediction trajectory into predicted wheel tracks for the left front wheel, right front wheel, left rear wheel, and right rear wheel based on the vehicle's wheelbase and track width. The wheel track refers to the expected running path of the corresponding wheel within the pre-aiming range.

[0043] Specifically, for any predicted center of gravity point in the predicted trajectory, the vehicle controller reads the position and heading angle of that predicted point. It takes the unit vector along the heading direction as the longitudinal direction and the unit vector perpendicular to the heading direction and pointing to the left side of the vehicle as the lateral normal direction. Then, based on the distances from the center of gravity to the front axle and the rear axle, it obtains the center points of the front and rear axles along the longitudinal direction, respectively. Using half the wheelbase as the lateral offset, it offsets left and right from the center point of the front axle to obtain the wheel tracks of the left and right front wheels, and left and right from the center point of the rear axle to obtain the wheel tracks of the left and right rear wheels. If the center of gravity position is approximated by the midpoint of the wheelbase, then the distances from the center of gravity to the front and rear axles are both taken as half the vehicle's wheelbase. If the vehicle's center of gravity position has a different calibration value, then the offset is performed according to the calibrated distances from the center of gravity to the front and rear axles. For curve conditions, the vehicle controller calculates the position of the four wheel tracks point by point using discrete trajectory points, avoiding simplifying the wheel tracks into a fixed translation path, thus ensuring that the mapping result is consistent with the actual wheel running position.

[0044] In step S103, road segments are mapped to wheel position-level road surface segment sequences. The vehicle controller reads each road segment from the road state data ahead and determines whether each road segment intersects with the four predicted wheel tracks. When a road segment covers a wheel track, the road segment is written into the corresponding wheel position-level road surface segment sequence; thus, the left front wheel, right front wheel, left rear wheel, and right rear wheel each form their own wheel position-level road surface segment sequences. Each record in each wheel position-level road surface segment sequence includes at least: wheel position identifier, segment start distance, segment end distance, adhesion level, smoothness level, road curvature, slope, and lateral coverage width.

[0045] The adhesion level represents the classification of a road segment's ability to transmit longitudinal and lateral forces to the tires. It can be calculated from the road adhesion coefficient range output by the forward sensing device, map data, or road surface recognition module. For example, road segments with a road adhesion coefficient in the first preset range are categorized as high adhesion level, those in the second preset range as medium adhesion level, and those in the third preset range as low adhesion level. The boundary values ​​of the first, second, and third preset ranges are vehicle model calibration values. If the forward road condition data directly contains the adhesion level, the vehicle controller directly reads that adhesion level. If the forward road condition data includes road surface type, water accumulation indicator, icing indicator, or estimated adhesion coefficient, the vehicle controller calculates the adhesion level according to a preset adhesion level mapping table.

[0046] Specifically, the data output by the forward sensing device includes at least one of road surface type, water accumulation indicator, icing indicator, snow accumulation indicator, slippery indicator, and estimated coefficient of adhesion; the map data includes at least one of road material type, historical coefficient of adhesion, and road condition indicator; the road surface recognition module outputs an estimated coefficient of adhesion based on at least one of wheel speed, slip ratio, longitudinal acceleration, lateral acceleration, yaw rate, and braking request amount.

[0047] When multiple data sources output estimated adhesion coefficients, the vehicle controller performs a weighted average according to preset weights to obtain the road segment adhesion coefficient. : ; in, , , These are the estimated adhesion coefficients output by the forward sensing device, map data, and road surface recognition module, respectively. , , To calibrate the weights, and For data sources that do not output estimated adhesion coefficients, the corresponding weights are set to 0, and the remaining weights are normalized.

[0048] The first, second, and third preset intervals are implemented using two adhesion coefficient thresholds. Let the first adhesion threshold be... The second attachment threshold is ,and Then the third preset interval is The second preset interval is The first preset interval is .when When the adhesion falls into the first preset range, it is recorded as a high adhesion level; when it falls into the second preset range, it is recorded as a medium adhesion level; and when it falls into the third preset range, it is recorded as a low adhesion level.

[0049] Smoothness rating is used to represent the classification result of the height undulation, impact, or roughness of road segments. It can be calculated from the road surface height change, slope change rate, roughness index, or obstacle edge height output by the forward sensing device, map data, or road surface recognition module. For example, road segments with height undulation or roughness index in the first smoothness range are categorized as ordinary smoothness rating, those in the second smoothness range as slightly uneven, and those in the third smoothness range as severely uneven; the boundary values ​​of each smoothness range are vehicle model calibration values. If the forward road condition data already directly contains the smoothness rating, the vehicle controller directly reads that smoothness rating; if the forward road condition data includes road surface height change, roughness index, joint markings, manhole cover markings, or repair strip markings, the vehicle controller calculates the smoothness rating according to a preset smoothness rating mapping table.

[0050] Specifically, the data output by the forward sensing device includes at least one of the following: road surface height sampling points, road surface height change, obstacle edge height, joint markings, manhole cover markings, repair strip markings, and pothole markings; the map data includes at least one of the following: historical roughness index, bridge surface joint markings, speed bump markings, manhole cover markings, and repair strip markings; the road surface recognition module outputs a roughness index or impact strength index based on at least one of the following: vehicle vertical acceleration, suspension travel change, wheel speed fluctuation, and longitudinal acceleration change.

[0051] When a road segment includes multiple road surface height sampling points, the vehicle controller uses the difference between the maximum and minimum height values ​​within that road segment as the road surface height change. The maximum height difference between adjacent height sampling points is taken as the obstacle edge height. The vehicle controller determines the road segment smoothness index based on road surface height variation, roughness index, and obstacle edge height. : ; in, For roughness index, , , To calibrate the weights; the weights corresponding to parameters that are not output are set to 0.

[0052] The first, second, and third flatness intervals are achieved through two flatness thresholds. Let the first flatness threshold be... The second flatness threshold is ,and The first flatness range is The second flatness range is The third flatness range is .when When the surface falls within the first flatness range, it is recorded as ordinary flatness level; when it falls within the second flatness range, it is recorded as slightly unevenness level; and when it falls within the third flatness range, it is recorded as severely unevenness level.

[0053] Road curvature is used to represent the degree of curvature at the location of the centerline or corresponding wheel track of a road segment. Its value is the rate of change of the heading angle of the road centerline relative to the travel distance. When the centerline curvature of a road segment is given by map data or a forward sensing device, the vehicle controller directly reads the centerline curvature as the road curvature. When a road segment is represented by multiple discrete centerline points, the vehicle controller calculates the road curvature based on the ratio of the change in heading angle of adjacent discrete centerline points to the arc length between adjacent discrete centerline points.

[0054] The slope is used to represent the degree of longitudinal height change of a road segment along the vehicle's direction of travel. It is the ratio of the height change of the road segment to the horizontal travel distance of the road segment, or the slope angle corresponding to the ratio. When the slope of a road segment is given by map data or a forward sensing device, the vehicle controller directly reads the slope. When the road segment is represented by the starting elevation and the ending elevation, the vehicle controller calculates the slope by dividing the difference between the ending elevation and the starting elevation by the segment length.

[0055] The lateral coverage width represents the lateral coverage area of ​​a road segment perpendicular to the road's extension direction, specifically the lateral distance between the left and right boundaries of the road segment. When a road segment is represented by its left and right boundaries, the vehicle controller calculates the lateral coverage width based on the distance between the left and right boundaries in the lateral direction of the road. When a road segment is represented by its center position and half-width lateral, the vehicle controller uses twice the half-width lateral as the lateral coverage width. When determining whether a road segment covers a predicted wheel track, the vehicle controller compares the lateral coverage area of ​​the road segment with the predicted wheel track point of the corresponding wheel. If the predicted wheel track point falls within the lateral coverage area of ​​the road segment, and its projection distance along the longitudinal direction of the road is between the segment's start and end distances, then the road segment is determined to cover the predicted wheel track of that wheel.

[0056] Furthermore, the segment start distance and segment end distance are the longitudinal projection distances of the road segment along the predicted driving direction of the vehicle within the preview range. For a road segment covering a certain predicted wheel track, the vehicle controller uses the longitudinal projection distance of the starting point of the intersection of the road segment and the corresponding predicted wheel track as the segment start distance of the wheel position level segment, and the longitudinal projection distance of the ending point of the intersection as the segment end distance of the wheel position level segment; if the road segment completely covers the longitudinal range of the wheel track within the segment, then the segment start distance and segment end distance of the road segment itself are directly used.

[0057] To facilitate subsequent calculations, the vehicle controller further calculates the length of each wheel position segment, which can be determined using the following formula: ; in, Indicates the first The first round The length of each segment; Indicates the distance at which the segment ends; This indicates the starting distance of the segment. All three quantities mentioned above use length units. The segment length is used for subsequent abnormal event aggregation and credibility assessment.

[0058] In one embodiment, the vehicle controller obtains the predicted speed based on the current vehicle speed and the speed change trend of the most recent control cycles, and then calculates the expected entry and exit times for each wheel position segment. If the vehicle exhibits a significant deceleration trend, the vehicle controller corrects the predicted speed using the average longitudinal acceleration of the most recent cycles; if the road gradient is significant, the impact of the gradient on speed changes is further considered; the expected entry and exit times can be expressed as follows: ; in, Indicates the entry time. Indicates the time of departure. The above formula represents the predicted speed, and it converts the original road segment data into time-series segment data associated with specific wheels. After this step, the vehicle controller obtains the order in which each wheel will traverse which road segments within the future target range.

[0059] The aforementioned estimated entry and exit times are used to subsequently determine the inter-axle entry time difference, left and right entry time difference, and duration of the abnormal event. For the same abnormal event, the duration is primarily determined by the estimated entry and exit times of each relevant wheel position segment, rather than solely by the single spatial length and instantaneous vehicle speed.

[0060] In step S2, abnormal events are determined based on the spatial correspondence of abnormal segments in each wheel position level road segment sequence, and the inter-axle entry time difference and left and right entry time difference of the abnormal events are calculated.

[0061] The vehicle controller searches for segments with decreased adhesion level, decreased smoothness level, and simultaneous decrease in both adhesion and smoothness in the four wheel position level road surface segment sequences.

[0062] Optionally, before the search, the vehicle controller compares the adhesion level and smoothness level according to a preset order. The adhesion level is sorted from strongest to weakest according to the ability of the road segment to transmit longitudinal and lateral forces to the tire; the more unfavorable the level, the lower the adhesion. For example, high adhesion, medium adhesion, and low adhesion can be set in order from favorable to unfavorable. The smoothness level is sorted from best to worst according to the smoothness of the road segment; the more unfavorable the level, the greater the road surface undulation, impact, or roughness. For example, ordinary smoothness, slight unevenness, and severe unevenness can be set in order from favorable to unfavorable. The boundaries of these levels can be determined according to the vehicle model calibration.

[0063] A segment with decreased adhesion level refers to a segment in which the adhesion level of the current wheel position segment becomes more unfavorable compared to a preset normal adhesion level or an adjacent reference segment. The adjacent reference segment can be the segment in the same wheel position segment sequence that precedes the current wheel position segment and is the closest in longitudinal distance, or it can be a reference segment within the pre-aiming range that is marked as normal pavement. A segment with decreased smoothness level refers to a segment in which the smoothness level of the current wheel position segment becomes more unfavorable compared to a preset normal smoothness level or an adjacent reference segment. A segment with simultaneous decrease in adhesion and smoothness refers to a segment in the same wheel position segment that simultaneously meets both the adhesion level decrease condition and the smoothness level decrease condition.

[0064] When related segments from different wheel positions correspond to the same area in the longitudinal direction, and at least one of them shows a decrease in adhesion level or a decrease in smoothness level; or when both adhesion level and smoothness level change simultaneously, such as a decrease in both, these segments are grouped into the same abnormal event. An abnormal event refers to a set of related segments that will have a continuous impact on the vehicle control process; the event includes fields such as event number, event start position, event end position, set of participating wheel positions, expected entry time of each wheel, expected departure time of each wheel, primary adhesion level, and primary smoothness level.

[0065] The primary adhesion grade and primary smoothness grade are determined using a unified rule: the most unfavorable adhesion grade and the most unfavorable smoothness grade are selected from the merged segments. The most unfavorable adhesion grade represents the grade with the lowest adhesion ability, and the most unfavorable smoothness grade represents the grade with the worst smoothness.

[0066] Specifically, the vehicle controller merges all wheel-position level segments of the same abnormal event into a merged segment set. Within this merged segment set, segments are compared according to a preset adhesion level order, and the adhesion level with the lowest adhesion capability is determined as the primary adhesion level for the abnormal event. Similarly, segments are compared according to a preset smoothness level order, and the smoothness level with the worst smoothness is determined as the primary smoothness level for the abnormal event. If the merged segment set includes multiple identical worst adhesion levels or multiple identical worst smoothness levels, these identical worst levels can be directly used as the corresponding primary level.

[0067] In one implementation, the vehicle controller can convert the adhesion level and smoothness level into comparable level values. A higher adhesion level value indicates lower adhesion capability, and a higher smoothness level value indicates poorer smoothness. For a segment at the j-th wheel position level of the i-th wheel position, if its adhesion level value is greater than the reference adhesion level value, the segment is determined to be a segment with a decreasing adhesion level; if its smoothness level value is greater than the reference smoothness level value, the segment is determined to be a segment with a decreasing smoothness level; if both are true, the segment is determined to be a segment with a simultaneous decrease in adhesion and smoothness. For the same abnormal event, the primary adhesion level value is the maximum value among the adhesion level values ​​of all merged segments included in the abnormal event, and the primary smoothness level value is the maximum value among the smoothness level values ​​of all merged segments included in the abnormal event.

[0068] Among them, the abnormal event types include at least adhesion abnormal events, flatness abnormal events, and adhesion-flatness combined abnormal events; when only the adhesion level decreases, it is judged as an adhesion abnormal event; when only the flatness level decreases, it is judged as a flatness abnormal event; when both the adhesion level and the flatness level decrease, it is judged as an adhesion-flatness combined abnormal event.

[0069] Once an abnormal event is identified, the vehicle controller reads the earliest entry times of the front axle, rear axle, left axle, and right axle from the abnormal event, and calculates the inter-axle entry time difference and the left-right entry time difference; the calculation method is as follows: ; in, Indicates the first Inter-axis entry time difference for each abnormal event; Indicates the first The left and right entry time difference of an abnormal event; Indicates the earliest entry time of the front axle; Indicates the earliest entry time of the rear axle; Indicates the earliest entry time on the left; This indicates the earliest entry time on the right. The vehicle controller writes these two time differences into the event log. If the inter-axle entry time difference exceeds the inter-axle entry time difference threshold, the event has the characteristic of staggered entry from front to back; if the left and right entry time difference exceeds the left and right entry time difference threshold, the event has the characteristic of one side leading and the other side following.

[0070] The vehicle controller further determines the length and duration of the abnormal event. The length of the abnormal event is the difference between the event termination position and the event start position. The duration of the abnormal event represents the estimated time from when the abnormal event begins to affect any related wheel position to when it ends to affect all related wheel positions. Specifically, the vehicle controller reads the estimated entry and exit times of each related wheel position segment merged into the same abnormal event, takes the earliest estimated entry time as the event start time, takes the latest estimated exit time as the event end time, and determines the duration of the abnormal event by the difference between the event end time and the event start time.

[0071] In one implementation, the duration of an abnormal event can be expressed as: ,in, This represents the duration of the k-th abnormal event. This represents the earliest expected entry time among the relevant wheel-level segments merged into the k-th anomalous event. This represents the latest expected departure time among the relevant wheel position segments merged into the k-th anomalous event; this duration reflects the expected length of time the anomalous event will affect the vehicle.

[0072] When the vehicle speed change is small and the average predicted speed of the event interval is used for approximate calculation, the vehicle controller can also divide the length of the abnormal event by the average predicted speed of the event interval as an approximation of the duration. The average predicted speed of the event interval is the average predicted speed of the vehicle within the longitudinal interval corresponding to the abnormal event; when the predicted speed is lower than a preset minimum speed threshold, the vehicle controller uses the preset minimum speed threshold for calculation, or directly uses the expected entry and expected exit times of each relevant wheel position segment to calculate the duration.

[0073] Through the above processing, the vehicle controller completes the conversion from spatial segments to temporal events. For a situation where the left front wheel first runs over the area where water and repair strip overlap, the right front wheel enters later, and the rear axle enters even later, the vehicle controller can identify that the event belongs to the left and right staggered entry and also has the characteristics of inter-axle staggered entry, thus avoiding subsequent control based on the assumption that the entire vehicle enters the abnormal segment at the same time.

[0074] In step S3, the credibility of the abnormal event is obtained by inputting the event characteristics of the abnormal event into the credibility evaluation model.

[0075] According to embodiments of this disclosure, credibility refers to the degree of reliability of an abnormal event as a basis for judgment in the current control phase, denoted as... The value ranges from 0 to 1. Therefore, the confidence level indicates the degree of trust that the vehicle controller has in whether the abnormal event requires a drive phase transition. The higher the confidence level, the more stable, continuous, and suitable the event is as a basis for control phase switching.

[0076] The vehicle controller constructs an event feature vector for each abnormal event; the event feature vector includes at least: main attachment level, main smoothness level, abnormal event length, abnormal event duration, inter-axle entry time difference, left and right entry time difference, number of affected wheel positions, statistics of the amplitude of wheel speed change of the four wheels in the most recent several cycles, statistics of the longitudinal deceleration rate of change in the most recent several cycles, statistics of the yaw rate of change in the most recent several cycles, current vehicle speed, current braking request, current road curvature, and consistency identifier of events in previous and subsequent cycles.

[0077] As mentioned earlier, the primary adhesion level is converted into an adhesion level value according to a preset adhesion level order; the lower the adhesion capability, the larger the corresponding adhesion level value. The primary smoothness level is converted into a smoothness level value according to a preset smoothness level order; the worse the smoothness, the larger the corresponding smoothness level value. The abnormal event length is the difference between the event termination position and the event start position. The abnormal event duration is determined based on the difference between the earliest expected entry time and the latest expected departure time of each relevant wheel position segment merged into the same abnormal event. The inter-axle entry time difference and the left and right entry time difference are obtained according to the aforementioned time difference calculation method. The number of affected wheel positions is the number of different wheel positions merged into this abnormal event.

[0078] The statistical measures of the four-wheel wheel speed change amplitude over the most recent several control cycles are calculated based on the wheel speed changes of the four wheels over the most recent N control cycles, where N is a preset positive integer. For example, the vehicle controller can first calculate the absolute value of the wheel speed difference between adjacent control cycles for each wheel, and then calculate the maximum, average, or weighted average of the absolute values ​​for the four wheels and the N control cycles to obtain the statistical measures of the four-wheel wheel speed change amplitude. The statistical measures of the longitudinal deceleration change rate over the most recent several control cycles are calculated based on the rate of change of the vehicle's longitudinal deceleration over the most recent N control cycles, and can be the maximum absolute value, average value, or weighted average value. The statistical measures of the yaw rate change rate over the most recent several control cycles are calculated based on the rate of change of the yaw rate over the most recent N control cycles, and can be the maximum absolute value, average value, or weighted average value. The specific selection of the above maximum, average, or weighted average value can be used as a vehicle model calibration parameter.

[0079] The current vehicle speed, current braking request amount, and current road curvature are respectively taken as the vehicle speed in the current control cycle, the braking request amount given by the driver or the upper controller, and the road curvature corresponding to the location of the main abnormal event.

[0080] The consistency flag for events in previous and subsequent cycles is determined based on the matching results between the current cycle's abnormal event and the previous cycle's abnormal event. If the center position deviation between the current cycle's abnormal event and the previous cycle's abnormal event is less than a preset position threshold, and the main attachment level and main flatness level remain consistent or change within a preset level difference, then the consistency flag takes the first value; otherwise, it takes the second value. The first and second values ​​can be set to 1 and 0, respectively.

[0081] In one embodiment, confidence is generated by a feedforward neural network model deployed in the vehicle controller. Inference is performed during runtime, and training is completed during the development phase. See also... Figure 3 , Figure 3 A schematic diagram of the model structure provided in the embodiments of this disclosure is shown; as follows: Figure 3 As shown, the model consists of an input layer, a feature normalization layer, a first fully connected layer, a first activation layer, a second fully connected layer, a second activation layer, a third fully connected layer, and an output layer, which are connected in series in sequence.

[0082] The input layer receives the aforementioned event feature vectors, and the feature standardization layer scales the continuous inputs and performs fixed mapping on the discrete level inputs. Specifically, the vehicle controller converts the main attachment level and main smoothness level into level values ​​according to a preset level mapping table, and converts the consistency identifiers of preceding and following periodic events into consistency values; it uses the abnormal event length, abnormal event duration, inter-axle entry time difference, left and right entry time difference, number of affected wheel positions, statistics of four-wheel wheel speed change amplitude, statistics of longitudinal deceleration rate change, statistics of yaw rate change, current vehicle speed, current braking request, and current road curvature as continuous or quasi-continuous features, and standardizes them according to the corresponding calibration mean and calibration scale.

[0083] Specifically, standardization can be expressed as: ; in, Let m represent the original value of the m-th feature to be standardized. Represents the standardized eigenvalues. This represents the mean of the feature in the training or calibration samples. This represents the standard deviation or calibration scale of the feature in the training or calibration samples; for rank values ​​and consistency values, the vehicle controller can directly use the mapped values ​​or scale them in the same way; the processed features are arranged in a preset order to obtain the model input vector X.

[0084] ; Where M represents the number of input features. to These are the characteristic values ​​corresponding to the processed main attachment level, main smoothness level, abnormal event length, abnormal event duration, axle entry time difference, left and right entry time difference, number of affected wheel positions, statistics of four-wheel wheel speed change amplitude, statistics of longitudinal deceleration change rate, statistics of yaw rate change rate, current vehicle speed, current braking request, current road curvature, and consistency identifier of preceding and following periodic events.

[0085] The first fully connected layer performs a linear transformation on the standardized feature vector, outputting the first hidden feature vector; the first activation layer performs a non-linear mapping on the first hidden feature vector. The second fully connected layer receives the output of the first activation layer and performs another linear transformation to obtain the second hidden feature vector; the second activation layer continues to perform a non-linear mapping on the second hidden feature vector. The third fully connected layer maps the output of the second activation layer to a single real number; the output layer uses a bounded activation function to compress the real number to between 0 and 1, which serves as the original confidence level for the current period.

[0086] Specifically, the calculations for the first fully connected layer, the second fully connected layer, and the third fully connected layer can be expressed as follows: ; ; ; ; in, , , These are the weight matrices or weight vectors for the first, second, and third fully connected layers, respectively. , , These are the corresponding biases. and It is a non-linear activation function. It is a bounded activation function. The original confidence level for the current period is given; the bounded activation function can be the Sigmoid function or other functions with an output range of 0 to 1, so that the original confidence level falls between 0 and 1; the weights, biases, level mapping table, mean and calibration scale mentioned above are all fixed into the vehicle controller after the model training or vehicle model calibration is completed.

[0087] In one example, the output dimension of the first fully connected layer can be configured to be between 16 and 24 dimensions, the output dimension of the second fully connected layer can be configured to be between 8 and 12 dimensions, and the third fully connected layer outputs a single dimension. The number of layers, connection methods, and input-output relationships remain fixed.

[0088] In one embodiment, the model can be trained offline, with training samples derived from test field data and publicly available road data. After preprocessing, the collected data is first used to construct wheel-position level road segment sequences and anomalous events as described above, and then event feature vectors are extracted from them. The annotation process is based on measured vehicle responses, road video recordings, and engineering review results to determine whether each anomalous event should be used as a stage transition criterion at the corresponding time. The annotation results are converted into target confidence intervals, and a feedforward neural network model is trained using supervised training. The training process can employ common gradient descent algorithms or their improved forms, such as adaptive moment estimation methods, updating model parameters through backpropagation to gradually reduce the error between the model output and the annotation results.

[0089] After training, a set of model parameters with low false triggering and missed triggering rates on the validation data is fixed to the vehicle controller. The fixed model parameters include the weights and biases of each fully connected layer, the mean and calibration scale of continuous features, the mapping table for discrete-level features, and the rules for assigning values ​​to the consistency identifiers of events across different cycles. During operation, as described above, the vehicle controller converts event features of different dimensions and types into a unified input vector according to the fixed feature processing rules, and then executes feedforward neural network inference to obtain the original credibility of the current cycle.

[0090] In one embodiment, the vehicle controller first outputs the original confidence level of the current cycle from the feedforward neural network model in each control cycle, and then weights it with the confidence level of the previous cycle according to the update coefficient to obtain the confidence level of the current cycle. The update relationship can be expressed as: ; in, Indicates the first The reliability of each control cycle The output of the feedforward neural network model represents the first... Original reliability of each control cycle Indicates the first The reliability of each control cycle This represents the update coefficient.

[0091] Update coefficients According to the control cycle of the vehicle controller and smoothing time constant It is confirmed that the calculation formula can be expressed as: ; in, The time interval between two consecutive control cycles of the vehicle controller. The time constant for reliability smoothing. Greater than 0. Smoothing time constant The smaller, The larger the value, the greater the impact of the original credibility of the current period on the credibility of the current period; smoothing time constant The larger, The smaller the value, the higher the degree of credibility retention from the previous period.

[0092] During the vehicle development phase, data from the test track, bench tests, and public road playback can be used to determine [the vehicle's performance]. Specifically, the vehicle controller or calibration equipment sequentially selects candidate time constants from the candidate time constant set. The value is used to calculate the corresponding update coefficient according to the formula above. The confidence sequence is calculated by replaying it on the labeled data; then, different candidates are statistically analyzed. The values ​​correspond to the false trigger rate, missed trigger rate, and stage judgment delay of abnormal events. The selected value simultaneously meets the preset upper limit for false trigger rate, the preset upper limit for missed trigger rate, and the preset maximum response delay requirements. The value, will Values ​​and their corresponding values Write to the non-volatile memory of the vehicle controller.

[0093] In one implementation, if the control cycle of the vehicle controller is a fixed value, then the update coefficient is... A fixed control cycle can be used during the calibration phase. and calibration results Pre-calculated and fixed; if there are slight fluctuations in the control cycle of the vehicle controller, the vehicle controller can read the current control cycle during operation. And calculate in real time according to the above formula. This update prevents sporadic fluctuations in a single cycle from causing significant changes in reliability. For short-term wheel speed disturbances caused by local coarse segments, this update method avoids a sudden increase in reliability followed by an immediate decrease, thereby reducing repeated switching of the control phase.

[0094] In step S4, based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry status of each wheel in response to the abnormal event, it is determined whether the vehicle is in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage.

[0095] Please see Figure 4 , Figure 4 A schematic diagram of the control phase determination process provided in an embodiment of this disclosure is shown. For example... Figure 4 As shown, in step S401, the preparatory stage determination is performed.

[0096] The vehicle controller checks whether the earliest expected entry time of each abnormal event falls within a preset preparation time window; simultaneously, it compares whether the current confidence level reaches the preparation stage entry threshold. If both conditions are met, the vehicle state corresponding to the abnormal event is determined to be in the preparation stage. The preparation time window is a calibration time parameter, whose function is to determine the time range for the controller to prepare in advance. The preparation stage entry threshold is determined by calibration and reflects the minimum confidence level at which the event needs to be addressed in advance.

[0097] In step S402, partial entry stage determination is performed. The vehicle controller determines the actual entry status of each relevant wheel for each abnormal event. The actual entry status is determined based on the relationship between the current travel distance and the starting distance of the wheel position segment. Specifically, when the cumulative projected distance of the vehicle's current position along the corresponding predicted wheel track reaches or exceeds the starting distance of the wheel position segment, the wheel is considered to have entered the segment corresponding to the abnormal event; otherwise, it is determined that it has not yet entered. When at least one relevant wheel has entered and at least one relevant wheel has not yet entered the abnormal event, and the current confidence level reaches the partial entry stage threshold, it is determined to be in the partial entry stage. This definition ensures that the physical process is consistent with the control sequence, that is, a wheel is actually affected by the road surface first, and then enters the corresponding control stage.

[0098] In step S403, the effective persistence phase is determined. When at least two related wheel positions have entered the abnormal event phase, and the current confidence level reaches the effective persistence phase entry threshold, the vehicle controller determines that it has entered the effective persistence phase. Furthermore, if the number of affected wheel positions does not meet the aforementioned condition, but the same abnormal event persists in multiple consecutive control cycles, and the confidence level consistently reaches the effective persistence phase entry threshold, it can also be determined that it has entered the effective persistence phase. The effective persistence phase entry threshold is higher than the partial entry phase entry threshold to ensure that this phase is based on relatively stable event identification results.

[0099] In step S404, the exit from the holding phase is determined. When the wheel that first entered the abnormal event begins to leave the corresponding segment, and at least one of the braking path holding window, attitude correction holding window, or suspension assist holding window is open, the vehicle controller determines that it has entered the exit from the holding phase. The exit from the holding phase does not require the abnormal event to disappear immediately, nor does it require the confidence level to drop to a very low level instantaneously. The controller is allowed to gradually release the constraints in a predetermined order at the end of the event. After processing in this way, the restricted control will not be suddenly released due to a short-term recovery of wheel speed or yaw rate at the end of the event. After the holding window is opened in the effective duration phase, it can continue into the exit from the holding phase, and is released separately in the exit from the holding phase according to the exit conditions of each holding window.

[0100] In practical applications, the vehicle controller can maintain multiple abnormal events simultaneously, but only the abnormal event with the greatest impact on vehicle movement and the highest credibility is used as the main event for the driving phase determination; other abnormal events are retained in the event queue for switching when the main event exits or the intensity of a new event increases. The determination rule for the main event adopts a consistent priority sorting, for example, first comparing credibility, then comparing the main attachment level and the main flatness level, and finally comparing the event length.

[0101] In step S5, corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters are generated according to the determined stage. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

[0102] It should be noted that the brake controller, steering controller, and suspension controller all use existing interfaces to receive target or limit values ​​from the vehicle controller. The vehicle controller itself does not directly replace the internal control algorithms of each actuator, but rather generates upper-level constraint parameters and stage control flags. The types of parameters generated at different stages can differ; in the preparatory stage, braking path constraint parameters and attitude correction constraint parameters are generated first, while in the effective continuation stage, suspension auxiliary control parameters are generated or updated as needed.

[0103] Please see Figure 5 , Figure 5 A schematic diagram illustrating the process of generating constraint parameters and control instructions provided in an embodiment of this disclosure is shown. Figure 5 As shown, in step S501, constraint parameters are generated in the preparatory stage.

[0104] After entering the preparation stage, the vehicle controller queries the preset mapping table based on the main adhesion level, main flatness level, reliability, vehicle speed and abnormal event type to obtain the upper limit of regenerative braking growth rate, upper limit of hydraulic braking additional pressure change rate, upper limit of additional braking force differential and upper limit of steering correction change rate. The mapping table is preset in advance, and its index is the above input parameter classification combination, and the output is the corresponding constraint parameter.

[0105] Specifically, the upper limit of the regenerative braking rate is used to limit the rate of increase of regenerative braking commands in adjacent control cycles; the upper limit of the hydraulic braking additional pressure change rate is used to limit the rate of increase of hydraulic additional pressure; the upper limit of the additional braking force differential is used to constrain the maximum allowable difference between the left and right additional braking forces; and the upper limit of the steering correction change rate is used to constrain the rate of change of the additional steering correction amount. The vehicle controller writes the first two items into the braking path constraint parameters, the latter two items into the attitude correction constraint parameters, and sends an auxiliary pre-adjustment flag to the suspension controller to put the suspension into the corresponding auxiliary state. The auxiliary pre-adjustment flag or the suspension damping / stiffness adjustment target can be written into the hold window management module as the suspension auxiliary control parameters.

[0106] In step S502, selective constraints are applied to the partial entry phase. After entering the partial entry phase, the vehicle controller first determines the set of wheels that have entered and the set of wheels that have not entered. Wheels in the set of wheels that have entered are considered to be directly affected by the abnormal road surface, while wheels in the set of wheels that have not entered still use the normal control boundary. Subsequently, the vehicle controller applies a first upper limit constraint to the braking additional amount corresponding to the set of wheels that have entered, while maintaining the normal control upper limit for the set of wheels that have not entered. The first upper limit constraint is obtained by further tightening the braking path constraint parameters generated in the preparatory phase, and its tightening degree is related to the current confidence level and the primary adhesion level.

[0107] For attitude correction, the vehicle controller applies a unidirectional gradual constraint to the correction amount. The unidirectional gradual constraint means that within adjacent control cycles, the attitude correction amount can only gradually increase, maintain, or gradually decrease along the current correction direction, and is not allowed to reverse direction before the stage re-determination. This rule is particularly important for unilateral leader entry events, because a short-term disturbance to a single wheel can easily cause local changes in the yaw rate and wheel speed signals. If the correction direction is not restricted, the controller may change the correction trend back and forth in a short period of time.

[0108] In one example, when the left front wheel enters the section of water and repair strip first, before the right front wheel enters, the vehicle controller only applies a stricter limit to the additional braking amount of the left front wheel, while the right front and rear axle wheels maintain the normal control limit. At the same time, the additional steering correction amount remains gradual in one direction. This processing method makes the control amount consistent with the actual order in which the wheels are disturbed.

[0109] In step S503, target values ​​for the effective duration phase are generated. After entering the effective duration phase, the vehicle controller uses the current confidence level, main adhesion level, main smoothness level, vehicle speed, braking request amount, and left and right entry time difference to generate the target regenerative braking upper limit, target hydraulic braking distribution ratio, additional braking force differential upper limit, and steering correction target upper limit through a mapping table.

[0110] The target regenerative braking upper limit is used to define the maximum permissible regenerative braking force in this phase; the target hydraulic braking allocation ratio represents the proportional boundary of the total braking demand allocated to the hydraulic braking path; the additional braking force differential upper limit and the steering correction target upper limit continue to serve as attitude control-related constraints. Because the events in this phase have high credibility and high persistence, the vehicle controller can make more explicit control path adjustments in this phase.

[0111] In step S504, an execution command is generated based on the constraint parameters and the holding window. After completing the stage determination, the vehicle controller inputs the braking path constraint parameters, attitude correction constraint parameters, suspension auxiliary control parameters, and holding window status into the command generation module. Among them, the braking path constraint parameters are used to limit the upper limit of regenerative braking force, the rate of change of hydraulic braking additional pressure, and the difference in additional braking force between the left and right sides; the attitude correction constraint parameters are used to limit the steering correction amount and its rate of change; the suspension auxiliary control parameters are used to characterize whether the suspension enters the auxiliary state and its exit conditions; and the hold window is used to limit the duration and release timing of the above-mentioned constraint parameters. Subsequently, the vehicle controller generates braking control commands, steering control commands, and suspension auxiliary control commands that meet the requirements of the constraint parameters and the hold window based on the current driver braking request, the current steering input, and the feedback status of each actuator, and sends them to the braking controller, steering controller, and suspension controller for execution, respectively.

[0112] In some embodiments, the vehicle controller opens three holding windows during the effective sustained phase: a braking path holding window, a posture correction holding window, and a suspension assist holding window. For each holding window, the vehicle controller records its start time and shortest holding duration. The braking path holding window corresponds to the holding of braking path constraint parameters, the posture correction holding window corresponds to the holding of posture correction constraint parameters, and the suspension assist holding window corresponds to the holding of suspension assist control parameters. The three holding windows are independent of each other, yet are jointly managed by the current primary abnormal event.

[0113] The minimum holding time can be pre-calibrated; the braking path holding time is determined based on the braking response continuity requirements; the attitude correction holding time is determined based on the trajectory correction stability requirements; and the suspension assist holding time is determined based on the vehicle's vertical response decay characteristics. With this setting, the controller will not immediately cancel the corresponding control boundary due to partial recovery in a single control cycle after an abnormal event is confirmed.

[0114] During the effective hold phase, the vehicle controller continuously outputs the target regenerative braking upper limit, target hydraulic braking distribution ratio, additional braking force differential upper limit, and steering correction target upper limit, while maintaining the window timing. If the confidence level of the abnormal event decreases briefly during the hold window timing but the exit condition is not triggered, the controller maintains the existing constraints and does not change the control objective. During the suspension assist hold window timing, the vehicle controller maintains the current suspension assist control objective unchanged. This is because short-term recovery signals often appear at the tail end of an abnormal event or in a localized stable phase, but these recovery signals do not necessarily mean that the vehicle has completely left the abnormal influence.

[0115] In some embodiments, after entering the exit hold phase, the vehicle controller first checks whether the brake path hold window meets the exit conditions. These exit conditions include: the hold time of the brake path hold window reaches a predetermined requirement; the current confidence level is lower than the exit threshold; and no new main event has replaced the current main event within the last few cycles. Once the conditions are met, the vehicle controller increases the target regenerative braking upper limit by a preset slope and simultaneously decreases the hydraulic braking additional ratio. The preset slope is the maximum allowable change of the target upper limit per unit time; a larger preset slope results in faster recovery, and a smaller preset slope results in smoother recovery. Its specific value can be preset according to actual conditions.

[0116] The vehicle controller further checks the attitude correction holding window. In addition to the holding time condition, it also requires the yaw rate change to return to the normal range. The normal range means that the yaw rate change has returned to the normal fluctuation range of the vehicle under normal, stable, and non-abnormal intervention conditions. It can be obtained by statistical calibration of ordinary road data, and its upper and lower limits are determined during vehicle development. After the condition is met, the vehicle controller relaxes the upper limit of the additional braking force differential and the upper limit of the steering correction target. The relaxation process adopts a gradual approach to avoid abrupt changes in vehicle response due to the sudden disappearance of the attitude correction boundary at the end of an abnormal event.

[0117] The vehicle controller performs a final check on the suspension assist hold window. If the hold time meets the requirements and the vehicle's vertical response returns to the normal range, it outputs a suspension assist exit command, causing the suspension controller to exit the assist state. Here, the normal range means that the vehicle's vertical response and / or suspension travel change has returned to the normal range of variation under normal vehicle driving conditions, and no longer falls within the abnormal fluctuation range requiring continuous suspension assist control. The vehicle's vertical response can be determined by vehicle vertical acceleration, suspension travel change, or a combination of these indicators. In this implementation, the suspension only performs assist entry and exit functions, does not participate in the confidence calculation of the main abnormal event, and does not change the stage determination rules.

[0118] In some embodiments, if a new abnormal event occurs during the exit hold phase, and the confidence level of this new abnormal event reaches the entry threshold for the preparatory phase or the entry threshold for the effective duration phase, the vehicle controller suspends the current exit process, retains the current braking path constraint parameters and attitude correction constraint parameters, and re-executes the abnormal event determination, confidence level calculation, and phase determination with the new abnormal event as the new primary event. This ensures that when multiple adjacent abnormal events occur consecutively, the controller will not repeatedly switch between exit and re-entry, thereby maintaining the continuity of the control path.

[0119] In summary, the embodiments of this disclosure enable the vehicle controller to sequentially complete wheel position-level segment construction, time difference calculation, reliability assessment, stage determination, constraint parameter generation, hold window management, and exit release around an abnormal event. Because the vehicle controller simultaneously considers the sequential relationship of the front and rear axles and left and right sides entering the abnormal road surface, the stability of the abnormal event in a continuous cycle, and the actual entry state, the switching between braking paths and attitude correction paths is more coherent under conditions such as localized water accumulation, bridge joints, single-sided repair strips, and short, rough road sections. The release process at the end of the event is also smoother, and the changes in the vehicle's longitudinal deceleration, yaw response, and trajectory correction are more stable.

[0120] Please see Figure 6 , Figure 6 This is a schematic diagram of a vehicle control system based on road conditions provided in an embodiment of this application. Figure 6 As shown, the control system includes: The trajectory and segment construction module 601 is configured to generate a vehicle prediction trajectory based on vehicle operating status data and road condition data ahead, and to construct wheel position level road segment sequences corresponding to the left front wheel, right front wheel, left rear wheel and right rear wheel based on the prediction trajectory. The abnormal event determination module 602 is configured to determine abnormal events based on the spatial correspondence of abnormal segments in each wheel position level road segment sequence, and to calculate the inter-axle entry time difference and left and right entry time difference of the abnormal event. The credibility assessment module 603 is configured to input the event characteristics of the abnormal event into the credibility assessment model to obtain the credibility of the abnormal event. The stage determination module 604 is configured to determine whether the vehicle is in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry state of each wheel in response to the abnormal event. The control command output module 605 is configured to generate corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters according to the determined stage. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

[0121] Each processing unit and / or module in the embodiments of this application can be implemented by analog circuits that implement the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0122] Based on the same inventive concept, this application also provides an electronic device, the method corresponding to which can be the method in the foregoing embodiments, and its problem-solving principle is similar to that method. For example... Figure 7 As shown, Figure 7 The device structure diagram provided for the embodiments of this application specifically includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the methods and / or technical solutions of the aforementioned embodiments of this application.

[0123] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above technical features, but should also cover other technical solutions formed by arbitrary combinations of the above technical features or their equivalent features without departing from the inventive concept.

Claims

1. A vehicle control method based on road conditions, characterized in that, include: A vehicle prediction trajectory is generated based on vehicle operating status data and road condition data ahead. Based on this prediction trajectory, wheel position-level road surface segment sequences are constructed for each of the left front wheel, right front wheel, left rear wheel, and right rear wheel. This includes generating a vehicle center of gravity prediction trajectory based on vehicle speed, steering wheel angle, steering wheel angle change rate, and road curvature; converting the vehicle center of gravity prediction trajectory into prediction wheel tracks for each of the four wheels based on the vehicle wheelbase and track width; mapping road segments from the road condition data ahead to each of the prediction wheel tracks to obtain the corresponding wheel position-level road surface segment sequences; and recording the segment start distance, segment end distance, adhesion level, smoothness level, estimated entry time, and estimated exit time for each segment in each wheel position-level road surface segment sequence. Abnormal events are determined based on the spatial correspondence of abnormal segments in each wheel position level road surface segment sequence, and the inter-axle entry time difference and left and right entry time difference of the abnormal events are calculated; wherein the abnormal events are obtained by aggregating the wheel position level road surface segment sequences of different wheels; The credibility of the abnormal event is obtained by inputting the event characteristics of the abnormal event into the credibility evaluation model. The credibility evaluation model is a feedforward neural network model, which sequentially includes an input layer, a feature normalization layer, a first fully connected layer, a first activation layer, a second fully connected layer, a second activation layer, a third fully connected layer, and an output layer. The event characteristics received by the input layer include at least the primary attachment level, primary smoothness level, abnormal event length, the inter-axle entry time difference, the left and right entry time difference, the duration of the abnormal event determined according to the expected entry and exit times of each relevant wheel position, wheel speed change amplitude statistics, longitudinal deceleration rate change statistics, yaw rate change statistics, vehicle speed, braking request quantity, road curvature, and consistency identifier of front and rear periodic events. The first fully connected layer performs a linear transformation on the normalized feature vector and outputs a first hidden feature vector. The first activation layer performs a non-linear mapping on the first hidden feature vector; the second fully connected layer receives the output of the first activation layer and performs another linear transformation to obtain the second hidden feature vector. The second activation layer continues to perform a non-linear mapping on the second hidden feature vector; the third fully connected layer maps the output of the second activation layer to a single real number; the output layer outputs the credibility of the abnormal event; Based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry status of each wheel in response to the abnormal event, the vehicle is determined to be in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage. According to the determined stage, corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters are generated. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

2. The method according to claim 1, characterized in that, Aggregation rules include: When multiple wheel position segments correspond to the same area in the longitudinal position, and at least one of them has a decrease in adhesion level or a decrease in smoothness level, or both the adhesion level and the smoothness level decrease simultaneously, the multiple wheel position segments are merged into the same abnormal event. The primary attachment level and primary flatness level of the aforementioned abnormal event are determined as the most unfavorable attachment level and the most unfavorable flatness level among the merged segments.

3. The method according to claim 2, characterized in that, The determination of whether a vehicle is in the preparation phase, partial entry phase, effective continuous phase, or exit holding phase includes: When the earliest expected entry time of the abnormal event falls within a preset preparation time window and the confidence level reaches the preparation stage entry threshold, it is determined to be in the preparation stage. When at least one relevant wheel has entered the abnormal event and at least one relevant wheel has not yet entered the abnormal event, and the confidence level reaches the partial entry stage threshold, it is determined to be in the partial entry stage. When at least two related wheel positions have entered the abnormal event, or when the same abnormal event persists for multiple consecutive control cycles and the credibility reaches the threshold for entering the effective persistence stage, it is determined to be an effective persistence stage. When the wheel that first entered the abnormal event begins to leave the abnormal event and at least one holding window is in the open state, it is determined to exit the holding phase.

4. The method according to claim 3, characterized in that, The preparation phase generates the braking path constraint parameters and the attitude correction constraint parameters, including: Based on the main adhesion level, the main smoothness level, the reliability, vehicle speed, and abnormal event type, a preset mapping table is queried to obtain the upper limit of regenerative braking growth rate, the upper limit of hydraulic braking additional pressure change rate, the upper limit of additional braking force differential, and the upper limit of steering correction change rate. The upper limit of the regenerative braking growth rate and the upper limit of the hydraulic braking additional pressure change rate are written into the braking path constraint parameters, and the upper limit of the additional braking force differential and the upper limit of the steering correction change rate are written into the attitude correction constraint parameters.

5. The method according to claim 4, characterized in that, During the partial entry phase, braking control commands and steering control commands are generated, including: Determine the set of wheel positions that have entered the abnormal event and the set of wheel positions that have not entered the abnormal event; apply a first upper limit constraint to the braking additional amount corresponding to the set of wheel positions that have entered the abnormal event, and maintain a normal control upper limit for the set of wheel positions that have not entered the abnormal event; A unidirectional gradual constraint is applied to the attitude correction amount, so that the attitude correction amount changes in the same direction in adjacent control cycles until the stage is re-determined.

6. The method according to claim 5, characterized in that, The step of opening the hold window during the effective duration phase includes: Based on the confidence level, the primary adhesion level, the primary smoothness level, the vehicle speed, the braking request amount, and the left and right entry time difference, the target regenerative braking upper limit, the target hydraulic braking distribution ratio, the additional braking force differential upper limit, and the steering correction target upper limit are generated. Open the brake path holding window, attitude correction holding window, and suspension assist holding window, and record the start time and shortest holding time of the corresponding holding window respectively.

7. The method according to claim 6, characterized in that, During the exit hold phase, constraints are gradually released according to the exit conditions of each hold window, including: When the holding time of the braking path holding window meets the requirements and the confidence level is lower than the exit threshold, the upper limit of regenerative braking is increased by a preset slope and the additional ratio of hydraulic braking is reduced. When the holding time of the attitude correction holding window meets the requirements and the yaw rate of change returns to the normal range, the upper limit of the additional braking force differential and the upper limit of the steering correction target are relaxed. When the suspension assist hold window holds for the required duration and the vehicle vertical response and / or suspension travel change return to normal range, a suspension assist exit command is output.

8. A vehicle control system based on road conditions, characterized in that, include: The trajectory and segment construction module is configured to generate a predicted vehicle trajectory based on vehicle operating status data and road condition data ahead, and to construct wheel position-level road surface segment sequences corresponding to the left front wheel, right front wheel, left rear wheel, and right rear wheel based on the predicted trajectory. This includes generating a predicted trajectory of the vehicle's center of gravity based on vehicle speed, steering wheel angle, steering wheel angle change rate, and road curvature; converting the predicted trajectory of the vehicle's center of gravity into predicted wheel tracks for each of the four wheels based on the vehicle's wheelbase and track width; mapping the road segments in the road condition data ahead to each of the predicted wheel tracks to obtain the corresponding wheel position-level road surface segment sequences; and recording the segment start distance, segment end distance, adhesion level, smoothness level, expected entry time, and expected exit time for each segment in each wheel position-level road surface segment sequence. An abnormal event determination module is configured to determine abnormal events based on the spatial correspondence of abnormal segments in each wheel position level road surface segment sequence, and to calculate the inter-axle entry time difference and left and right entry time difference of the abnormal events; wherein the abnormal events are obtained by aggregating wheel position level road surface segment sequences of different wheels; A credibility assessment module is configured to input the event characteristics of the abnormal event into a credibility assessment model to obtain the credibility of the abnormal event. The credibility assessment model is a feedforward neural network model, which sequentially includes an input layer, a feature normalization layer, a first fully connected layer, a first activation layer, a second fully connected layer, a second activation layer, a third fully connected layer, and an output layer. The event characteristics received by the input layer include at least the primary attachment level, primary smoothness level, abnormal event length, the inter-axle entry time difference, the left and right entry time difference, the duration of the abnormal event determined based on the expected entry and exit times of each relevant wheel position, wheel speed change amplitude statistics, longitudinal deceleration rate change statistics, yaw rate change statistics, vehicle speed, braking request quantity, road curvature, and consistency identifier of preceding and following periodic events. The first fully connected layer performs a linear transformation on the normalized feature vector and outputs a first hidden feature vector. The first activation layer performs a non-linear mapping on the first hidden feature vector; the second fully connected layer receives the output of the first activation layer and performs another linear transformation to obtain the second hidden feature vector. The second activation layer continues to perform a non-linear mapping on the second hidden feature vector; the third fully connected layer maps the output of the second activation layer to a single real number; the output layer outputs the credibility of the abnormal event; The stage determination module is configured to determine whether the vehicle is in the preparation stage, the partial entry stage, the effective continuous stage, or the exit holding stage based on the inter-axle entry time difference, the left and right entry time difference, the confidence level, and the actual entry state of each wheel in response to the abnormal event. The control command output module is configured to generate corresponding braking path constraint parameters, attitude correction constraint parameters, and / or suspension auxiliary control parameters according to the determined stage. During the effective continuous stage, a holding window is opened, which includes a braking path holding window, an attitude correction holding window, and a suspension auxiliary holding window. Based on the braking path constraint parameters, the attitude correction constraint parameters, and / or the suspension auxiliary control parameters and the holding window, the corresponding control commands are held and released.

Images