A method, system and vehicle for robust control of cornering roll pre-conditioning

Abstract

The application discloses a curved road roll-pre-adjustment robustness control method, a system and a vehicle, relates to the field of automobile safety, and the curved road roll-pre-adjustment robustness control method comprises the following steps: collecting multi-source heterogeneous information, and constructing a four-dimensional input matrix; evaluating the confidence of the multi-source heterogeneous information, fusing output multi-source fusion information; based on the multi-source fusion information, self-adapting calibration of a full scene; real-time monitoring of a full-link failure state; and fault-tolerant control based on the full-link failure state grading. The application can construct a four-dimensional multi-source heterogeneous information input matrix, a multi-source information fusion method with confidence evaluation, and a deep learning self-adaptive calibration model, and solves the problems of single input, poor working condition adaptability and insufficient robustness of the prior art from the root, and realizes self-adaptive adaptation of a full-scene working condition.

Description

Technical Field

[0001] This invention belongs to the field of automotive suspension technology, and specifically relates to a method, system, and vehicle for robust control of cornering roll pre-adjustment. Background Technology

[0002] With the mass production and widespread adoption of high-precision maps and intelligent chassis technology, vehicle cornering roll pre-adjustment technology has been upgraded from traditional feedback-based passive control to feedforward-based active control based on road preview information. This effectively solves the lag problem of traditional feedback control and has become a core technology for improving vehicle cornering comfort and handling stability.

[0003] Current mainstream roll pre-adjustment technologies all employ a dual-input core framework of "high-precision map curve curvature + current vehicle speed." Based on a pre-calibrated parameter map, they adjust suspension stiffness and damping before entering a curve to suppress roll. However, existing technologies are always focused on "basic pre-adjustment function implementation" and have not been deeply designed for the robustness of the function across all scenarios and the functional safety requirements of mass production, thus failing to meet the compliance and reliability requirements for mass production of passenger vehicles. Summary of the Invention

[0004] The purpose of this invention is to provide a robust control method, system, and vehicle for cornering roll pre-adjustment, in order to solve one or more technical problems existing in the prior art.

[0005] The technical solution adopted to solve the above-mentioned technical problems is as follows: This invention discloses a robust control method for pre-adjustment of cornering roll, comprising the following steps: Collect heterogeneous information from multiple sources and construct a four-dimensional input matrix; Assess the confidence level of multi-source heterogeneous information and fuse and output multi-source fused information; Based on multi-source fusion information, adaptive calibration is performed across all scenarios; Real-time monitoring of end-to-end failure status; Fault-tolerant control based on end-to-end failure state classification.

[0006] The beneficial effects of this invention are at least as follows: by collecting multi-source heterogeneous information, a four-dimensional input matrix is ​​constructed, covering all core dimensions affecting roll pre-adjustment; the multi-source information fusion method with confidence effectively filters data noise; the accuracy of control decisions under complex working conditions is improved by more than 90%; it can be adapted to all real driving scenarios such as no-load / full-load, dry / slippery road surfaces, and comfortable / aggressive driving styles; it achieves a significant improvement in adaptability and robustness across all scenarios, and completely solves the problem of poor adaptability of existing technologies.

[0007] The adaptive calibration model based on multi-source fusion information can significantly reduce calibration costs, has extremely strong generalization ability, gets rid of the existing dependence on massive manual calibration, greatly shortens the calibration development cycle, and significantly reduces calibration costs.

[0008] The end-to-end failure status monitoring covers all potential systemic failures and random hardware failure scenarios, and can meet the functional safety requirements of ISO26262ASIL-B and above. It completely solves the core pain point of existing products that cannot pass mass production functional safety certification, and has the compliance to be directly implemented in mass production.

[0009] The robust control method for cornering roll pre-adjustment improves the robustness of the vehicle's cornering roll pre-adjustment function across all scenarios and provides full-link functional safety protection.

[0010] As a further improvement to the above technical solution, the evaluation of the confidence level of multi-source heterogeneous information and the fusion output of multi-source fused information includes the following steps: Preprocessing multi-source heterogeneous information; Evaluate the confidence level of preprocessed multi-source heterogeneous information; Dynamic weight allocation; Output multi-source fused information.

[0011] As a further improvement to the above technical solution, the adaptive calibration of the entire scene based on multi-source fusion information includes the following steps: Pre-trained adaptive calibration model; Solve for online dynamic parameters; Adaptive calibration model learns online.

[0012] As a further improvement to the above technical solution, the real-time monitoring of the entire link failure status includes the following steps: Failure to monitor multi-source heterogeneous information; Monitoring and adaptive calibration model failure; The suspension actuator is malfunctioning.

[0013] As a further improvement to the above technical solution, the step of classifying the failure status based on the entire link includes the following steps: When there are no failures in the entire link, the confidence level of all multi-source heterogeneous information is normal, and all suspension actuators are working normally, the failure state of the entire link is judged as a level one state. When a partial failure occurs and does not affect core driving safety, the entire link failure state is judged as a level two state. When the entire link fails severely, the failure status is determined to be Level 3.

[0014] As a further improvement to the above technical solution, the fault-tolerant control includes the following steps: When the failure state is the first level, the optimal control parameters are output based on multi-source fusion information and adaptive calibration model, and the roll pre-adjustment function is executed. When the failure status is level 2, the redundancy backup or strategy adjustment maintains the operation of the tilt pre-adjustment function. When the failure state is level three, the suspension stiffness and damping are controlled to smoothly recover to the default straight-line driving state within a preset time period according to the preset low gradient, and the function is prohibited from being activated.

[0015] As a further improvement to the above technical solution, when the failure state is a level two state, the redundancy backup or strategy adjustment to maintain the roll pre-adjustment function includes the following steps: If the local failure area contains multi-source heterogeneous information, the redundant backup maintains the tilt pre-adjustment function. If the local failure area is the suspension actuator, the strategy adjustment maintains the roll pre-adjustment function.

[0016] As a further improvement to the above technical solution, the curve roll pre-adjustment robust control method further includes the following steps: Based on the aforementioned fault-tolerant control, human-computer interaction prompts are output.

[0017] This invention discloses a system comprising: The acquisition module acquires multi-source heterogeneous information and constructs a four-dimensional input matrix; The fusion module evaluates the confidence level of multi-source heterogeneous information and fuses and outputs multi-source fused information. An adaptive calibration module, which adaptively calibrates across the entire scenario based on multi-source fusion information; The monitoring module monitors the failure status of the entire link in real time. A hierarchical fault-tolerant module, which performs hierarchical fault-tolerant control based on the failure state of the entire link.

[0018] The present invention has at least the following beneficial effects: it enables robust control and monitoring of multi-source heterogeneous information fusion and hierarchical fault-tolerant control of full-link functional safety redundancy.

[0019] This invention discloses a vehicle that applies the cornering roll pre-adjustment robust control method described in any of the above claims. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments; Figure 1 This is a flowchart of the robust control method for pre-adjustment of cornering roll provided in an embodiment of the present invention; Figure 2 yes Figure 1Detailed flowchart of step S200; Figure 3 yes Figure 1 Detailed flowchart of step S300; Figure 4 yes Figure 1 A detailed flowchart of step S400; Figure 5 yes Figure 1 A detailed flowchart of step S500 being classified into first-level states; Figure 6 yes Figure 1 A detailed flowchart of step S500 being divided into two states; Figure 7 yes Figure 1 A detailed flowchart of step S500 being classified into three states. Detailed Implementation

[0021] This section will describe in detail specific embodiments of the present invention. Preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the drawings is to supplement the textual description with graphics, so that people can intuitively and vividly understand each technical feature and overall technical solution of the present invention, but they should not be construed as limiting the scope of protection of the present invention.

[0022] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0023] In the description of this invention, the use of terms such as "a number" means one or more, with "more than" meaning two or more. Terms like "greater than," "less than," and "exceeding" are understood to exclude the stated number, while terms like "above," "below," and "within" are understood to include the stated number. The use of terms like "first," "second," and "third" is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, the number of indicated technical features, or the sequential relationship between indicated technical features.

[0024] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0025] Among related technologies, the main pain points of the mainstream roll pre-adjustment technology are all focused on two dimensions: operational robustness and functional safety redundancy, which are completely decoupled from other technical directions such as suspension parameter solving, timing control, and cross-domain collaboration.

[0026] Pain Point 1: The input dimension is extremely limited, the adaptability to working conditions is extremely poor, and the robustness is seriously insufficient.

[0027] Current technologies rely solely on two fixed input dimensions: high-precision map curvature and current vehicle speed, completely ignoring other core factors affecting roll pre-adjustment effectiveness. A vehicle's roll characteristics during cornering are not only related to curve curvature and vehicle speed, but also strongly correlated with dynamic conditions such as vehicle load, center of gravity height, road surface adhesion coefficient, weather conditions, and driver style. Fixed dual-input control logic can only cover calibrated ideal conditions. In complex scenarios such as fully loaded / unloaded conditions, wet / slippery roads, continuous curves in mountainous areas, and aggressive / comfortable driving styles, the control effect significantly decreases, even exhibiting over-adjustment or under-adjustment issues, failing to adapt to all real-world driving conditions.

[0028] Pain Point 2: Lack of multi-source information fusion and confidence assessment makes it susceptible to data noise interference and results in low decision-making accuracy.

[0029] Existing technologies directly use raw data from high-precision maps and vehicle speed sensors for control decisions without validating the input data, filtering noise, or assessing confidence. In actual driving, high-precision maps suffer from positioning errors, curvature data deviations, and signal fluctuations, while vehicle speed sensors experience numerical deviations due to wheel slippage. These noises and deviations are directly transmitted to the control decision layer, leading to erroneous or missed function triggers, or even the output of incorrect control parameters, resulting in abnormal vehicle posture and extremely poor control reliability and anti-interference capabilities.

[0030] Pain Point 3: The perception layer lacks a redundant backup mechanism, and the failure of a single perception source directly leads to functional paralysis.

[0031] Current technologies exhibit absolute path dependence on high-precision maps and lack any perceptual redundancy design. When vehicles travel in scenarios where high-precision map signals are lost or data becomes invalid, such as in tunnels, mountain valleys, or remote road sections, the function loses its core input source, is forcibly shut down, and becomes completely unusable. This not only significantly limits the applicable scenarios of the function but also leads to scenarios where "high-precision maps suddenly fail during cornering, causing sudden interruptions in the function," resulting in safety risks such as sudden changes in suspension parameters and loss of vehicle attitude control.

[0032] Pain Point 4: The execution layer lacks fault-tolerant control strategies, and the failure of a single actuator can lead to serious safety hazards.

[0033] Current technology assumes that all four air springs and damper actuators in the suspension are in normal working condition, without designing any fault tolerance mechanism for actuator failure scenarios. In actual use, if a single air spring or damper experiences hardware failure, communication interruption, or response failure, the current technology will directly shut down the function, or even continue to output the original control parameters. This results in severe asymmetry in the parameters of the left and right suspensions and the front and rear suspensions, causing sudden changes in vehicle roll and deviation from the driving trajectory. In high-speed cornering scenarios, this can easily lead to serious safety accidents such as skidding, fishtailing, or even rollover.

[0034] Pain Point 5: Lack of end-to-end hierarchical fault-tolerant control, which does not meet the ISO26262 functional safety mass production requirements.

[0035] Current technologies only offer two binary operating states: "function on" and "function off," failing to construct a comprehensive failure monitoring and hierarchical fault-tolerance system covering the entire perception-decision-execution chain. On one hand, they cannot proactively identify and warn of potential failures across the entire chain, only able to forcibly shut down the function after a failure occurs. On the other hand, there is no smooth transition mechanism when the function is shut down, resulting in a sudden jump in suspension parameters, which can easily trigger secondary vehicle posture disturbances. Furthermore, current technologies do not meet the ISO 26262 road vehicle functional safety standard's protection requirements against random hardware failures and systemic failures, making them unsuitable for mass production and lacking compliance for mass production.

[0036] Pain Point 6: It relies on massive manual calibration, has extremely poor generalization ability, and has extremely high development costs.

[0037] Existing technologies rely entirely on manual calibration of control parameters through bench tests and real-vehicle road tests. To cover different operating conditions, tens of thousands of calibrations are required for various combinations of curvature, vehicle speed, load, and road surface conditions, resulting in long calibration cycles and extremely high development costs. Furthermore, manual calibration cannot cover all extreme operating conditions, and the generalization ability of the calibration parameters is extremely poor. If an uncalibrated condition occurs, the control effect will drop sharply, making it unsuitable for complex and ever-changing real-world driving scenarios.

[0038] Reference Figures 1 to 7 The following are several embodiments of a robust control method, system and vehicle for cornering roll pre-adjustment according to the present invention.

[0039] like Figure 1 As shown, the cornering roll pre-adjustment robust control method of this invention includes steps S100 to S500.

[0040] Step S100: Collect multi-source heterogeneous information and construct a four-dimensional input matrix.

[0041] Understandably, step S100 forms the basis of robust control by inputting multi-source heterogeneous information to break the limitations of the existing dual-input technology and construct a four-dimensional input matrix that covers all influencing factors.

[0042] Step S200: Evaluate the confidence level of the multi-source heterogeneous information and fuse and output the multi-source fused information.

[0043] Understandably, step S200 is the core of improving control robustness. Thus, step S200 effectively filters out data noise and invalid information through a confidence-weighted fusion method, ensuring the accuracy and anti-interference capability of control decisions, and completely solving the problem of poor adaptability to operating conditions in existing technologies.

[0044] Step S300: Adaptively calibrate the entire scenario based on multi-source fusion information.

[0045] Understandably, step S300 is used to address the problems of existing technologies relying on manual calibration and having poor generalization ability.

[0046] Step S400: Monitor the failure status of the entire link in real time.

[0047] Understandably, step S400 forms the basis for functional safety protection in the robust control method for cornering roll pre-adjustment, enabling the construction of a full-link failure monitoring system to achieve real-time identification and early warning of failures.

[0048] Step S500: Based on the full-link failure state classification, fault-tolerant control.

[0049] Understandably, step S500 is the core of functional safety protection. It breaks through the limitations of the existing binary switch technology and designs a graded fault-tolerant control strategy based on the severity of failure, so as to achieve smooth transition and safety protection in failure scenarios.

[0050] It is understandable that the multi-source heterogeneous information includes static map information, vehicle status information, environmental perception information, and driver characteristic information. Thus, step S100 constructs a four-dimensional input matrix that includes static map information, vehicle status information, environmental perception information, and driver characteristic information.

[0051] In this embodiment, all multi-source heterogeneous information can be directly obtained by the vehicle's existing sensors and controllers without the need for additional hardware in the vehicle.

[0052] In some embodiments, static map information is collected through the vehicle's high-precision map module. Specifically, static map information includes the curvature of the road ahead curve, the rate of change of curvature, the curve length, the road cross slope / superelevation, the road surface type, the road elevation, tunnel / mountain section markings, map positioning accuracy, and signal strength.

[0053] In some embodiments, vehicle status information is collected through body sensors and chassis domain controllers. Specifically, vehicle status information includes real-time vehicle mass / load, sprung / unsprung mass distribution, center of gravity height, tire pressure / wear status, current operating status of suspension actuators, real-time vehicle speed, lateral / longitudinal acceleration, body roll angle / roll rate, and wheel vertical load.

[0054] In some embodiments, environmental perception information is collected through a forward-facing camera, millimeter-wave radar, and a vehicle domain controller. Specifically, this includes lane curvature recognized by vision, vehicle trajectory estimated by radar, road surface adhesion coefficient, weather conditions (rain / snow / ice / dry), crosswind intensity, road surface unevenness, and oncoming traffic information.

[0055] In some embodiments, driver characteristic information is collected through the intelligent driving domain controller and the body domain controller. Specifically, driver characteristic information includes driving style classification (comfort / standard / sport / aggressive), historical cornering habits (steering angle rate, acceleration and deceleration characteristics), tolerance for roll suppression, and driver attention status.

[0056] Thus, step S100, by covering the four-dimensional input matrix encompassing the road, vehicle, environment, and driver, fully covers all the core factors affecting the roll pre-adjustment effect, solving the problem of single input in existing technologies.

[0057] It is understandable that, such as Figure 2 As shown, step S200 includes steps S210 to S240.

[0058] Step S210 involves preprocessing multi-source heterogeneous information. Specifically, this involves filtering, denoising, time synchronization, and normalization of the various data sources of the multi-source heterogeneous information in the four-dimensional input matrix to eliminate differences in sampling periods and numerical dimensions between different data sources, and to filter out high-frequency noise. Each data source includes high-precision map information, vehicle status information, etc.

[0059] Step S220: Evaluate the confidence level of the preprocessed multi-source heterogeneous information. Specifically, for each data source, a confidence score (0-100 points) is calculated in real time based on data validity, signal stability, and scene adaptability. For example, the confidence level of static map information collected by the high-precision map module is calculated based on positioning accuracy, signal strength, and map freshness; the confidence level of visual recognition of lane line curvature is calculated based on lane line clarity, recognition accuracy, and lighting conditions.

[0060] Step S230: Dynamically allocate weights. Specifically, based on the confidence scores of each data source, dynamically allocate fusion weights. Data sources with higher confidence scores have a larger weight ratio, while data sources with confidence scores below a preset threshold are automatically weighted down or removed to avoid invalid data interfering with decision-making.

[0061] Step S240: Output multi-source fusion information. Specifically, based on a weighted fusion algorithm, multiple data sources of the four-dimensional input matrix are fused, and the fused effective control parameters are output. At the same time, the comprehensive confidence level of the control parameters is also output, providing a basis for subsequent decision-making and fault-tolerant control. That is, the multi-source fusion information includes control parameters and the comprehensive confidence level of the parameters.

[0062] With this setup, step S200 is used to address the issues of noise, bias, and validity of multi-source heterogeneous information.

[0063] In other embodiments, any multi-source information fusion algorithm, such as Kalman filtering, DS evidence theory, Bayesian estimation, or deep learning fusion model, can be used. As long as it is based on multi-source information fusion of a four-dimensional input matrix to improve the accuracy of input parameters, it can replace the confidence-based weighted fusion method used in step S200.

[0064] In this embodiment, as Figure 3 As shown, step S300 includes steps S310 to S330.

[0065] Step S310: Pre-train the adaptive calibration model.

[0066] Understandably, an adaptive calibration model based on deep learning is pre-built and trained offline using massive amounts of real vehicle road test data and bench test data. The input of the adaptive calibration model is the fused four-dimensional control parameters, and the output is the suspension stiffness and damping control parameters for roll pre-adjustment. The training process covers all normal and extreme working conditions to ensure the model's generalization ability.

[0067] Step S320: Solve for the online dynamic parameters.

[0068] Understandably, the fused effective control parameters are input into the pre-trained adaptive calibration model to solve for the optimal suspension control parameters under the current operating conditions in real time, without relying on the pre-calibrated MAP diagram.

[0069] Step S330: Adaptive calibration model online self-learning.

[0070] Understandably, based on the feedback of real-time control performance (i.e., the deviation between the actual roll angle and the target value), the control parameters are fine-tuned and optimized online to continuously improve the adaptability of the adaptive calibration model under different vehicles and operating conditions.

[0071] With this setup, step S300 dynamically outputs control parameters based on multi-source fusion information and an adaptive calibration model. Based on a pre-trained deep learning model, it achieves dynamic solution of control parameters across all operating conditions and scenarios, eliminating the need for massive manual calibration. Furthermore, it continuously optimizes the solution through online self-learning, significantly improving the generalization ability and development efficiency of the solution.

[0072] It is understandable that the suspension actuator executes control parameters.

[0073] In other embodiments, a lookup table method can be used in combination with multidimensional interpolation algorithms, fuzzy control, PID adaptive adjustment, model predictive control (MPC), and other arbitrary parameter adaptive solution methods. As long as the adaptive calibration of parameters under all operating conditions is based on multi-source fusion information, it can replace the deep learning adaptive calibration model in step S300.

[0074] In this embodiment, as Figure 4 As shown, step S400 includes steps S410 to S430.

[0075] Step S410: Monitor the failure of multi-source heterogeneous information.

[0076] Understandably, step S410 monitors whether multi-source heterogeneous information has failed, thereby monitoring the failure of the perception layer. Specifically, it involves real-time monitoring of the high-precision map signal status, positioning accuracy, curvature data confidence level, the recognition effectiveness of the environmental perception system, and the communication status and fault codes of each sensor.

[0077] Understandably, the failure criteria for multi-source heterogeneous information include: high-precision map signal loss exceeding a preset time, positioning error exceeding a safety threshold, all sensing sources failing to output effective curve curvature information, and core sensor communication interruption.

[0078] Step S420: Monitor the failure of the adaptive calibration model.

[0079] It is understandable that monitoring the failure of the decision-making layer is achieved by monitoring whether the adaptive calibration model has failed in step S420. Specifically, this involves real-time monitoring of the output validity of the adaptive calibration model, the rationality of the control parameters, the conflict of control commands, and the computing power and communication status of the decision-making unit.

[0080] It is understandable that the failure judgment rules of the adaptive calibration model include: the control parameters output by the model exceed the physical limits of the actuator, the control commands are seriously inconsistent with the current operating conditions, and the decision unit has hardware / software failures.

[0081] Step S430: Monitor suspension actuator failure.

[0082] Understandably, step S430 monitors whether the suspension actuators have failed, thereby monitoring failures at the execution layer. Specifically, it involves real-time monitoring of the response status, deviation between actual output and target value, actuator fault codes, and communication status of each suspension actuator.

[0083] Understandably, the failure criteria for suspension actuators include: no response from one or more actuators, deviation between actual output and target value exceeding a preset threshold, and actuator reporting a hardware fault.

[0084] It is understandable that by combining full-link failure monitoring and hierarchical fault tolerance through steps S410 to S430, functional safety compliance can be achieved: full-link failure monitoring covering multi-source heterogeneous information, adaptive calibration modules and suspension actuators can identify all potential failure scenarios in real time.

[0085] In this embodiment, step S500, which is based on the classification of failure states across the entire link, includes steps S510 to S530, such as... Figures 5 to 7 As shown.

[0086] Understandably, end-to-end failure states are categorized into Level 1, Level 2, and Level 3 states, from least to most severe.

[0087] Step S510: When there are no failures in the entire link, the confidence of all data sources is normal, and all suspension actuators are working normally, the failure status of the entire link is determined to be a level one failure status.

[0088] Step S520: When a local failure occurs and does not affect core driving safety, the entire link failure state is determined to be a level 2 state.

[0089] Step S530: When the entire link is severely failed, the failure status of the entire link is determined to be a level three state.

[0090] Understandably, the three-level state cannot maintain the normal operation of the roll pre-adjustment function and poses a clear driving safety risk. The three-level state includes the failure of all multi-source heterogeneous information, the core failure of the adaptive calibration model, and the failure of two or more actuators.

[0091] It is understandable that when the end-to-end failure state is at level one, level one is the normal operation mode. At this time, fault tolerance control includes step S511, such as... Figure 5 As shown.

[0092] Step S511: Based on multi-source fusion information and the adaptive calibration model, output optimal control parameters and execute the roll pre-adjustment function. The optimal control parameters are based on the control parameters output by the adaptive calibration model, and the executed roll pre-adjustment function has no restrictions.

[0093] It is understandable that when the end-to-end failure state is at level two, level two is a functional degradation mode. At this time, fault tolerance control includes step S521, such as... Figure 6 As shown.

[0094] Step S521: Redundancy backup or strategy adjustment to maintain the roll pre-adjustment function operation.

[0095] It is understandable that step S521 is divided into step S522 and step S523 based on the local failure area.

[0096] Step S522: If the local failure area is multi-source heterogeneous information, the redundant backup maintains the tilt pre-adjustment function.

[0097] In some embodiments, if the end-to-end failure state is a level 2 state, and the monitored local failure area is multi-source heterogeneous information. Taking the degraded scenario of high-precision map failure / insufficient confidence, but normal redundant perception of vision and millimeter-wave radar as an example, in this case, step S522 maintains the roll pre-adjustment function for redundant backup, that is, immediately performs seamless switching, takes the lane line curvature and driving trajectory prediction identified by vision and radar as the core input source of the four-dimensional input matrix, and at the same time slightly reduces the adjustment range of control parameters to increase the safety margin, and continues to execute the roll pre-adjustment function. There are no parameter jumps or functional interruptions during the switching process.

[0098] In other embodiments, any perception method that can obtain the curvature of a curve can be used, such as LiDAR lane line recognition, road curvature information obtained by V2X vehicle-road cooperation, or inertial navigation combined with historical trajectory prediction. As long as it achieves redundant perception backup after the failure of the high-precision map, it can replace the redundant backup of vision and millimeter-wave radar used in step S522.

[0099] Step S523: If the local failure area is the suspension actuator, the strategy adjustment maintains the roll pre-adjustment function.

[0100] In some embodiments, if the overall failure state is a level two state, and the monitored local failure area is the suspension actuator, taking a degraded scenario where a single suspension actuator fails while the remaining actuators are functioning normally as an example, step S523 is to adjust the strategy to maintain the normal operation of the roll pre-adjustment function. That is, the torque reconstruction control strategy is immediately activated, and the optimal stiffness and damping parameters are recalculated based on the physical limits of the remaining normal actuators. Through differentiated adjustment of the normal actuators, the roll suppression capability of the failed actuator is maximized, while ensuring the balance of left and right suspension torques, avoiding sudden changes in vehicle attitude, and simultaneously limiting the vehicle's maximum speed to reduce roll risk.

[0101] In other embodiments, the roll suppression capability can be compensated after failure by combining active stabilizer bars, differential braking of the braking system, and auxiliary compensation of the steering system. As long as torque reconstruction and roll suppression are achieved by normal actuators after actuator failure, they can replace the parameter redistribution based on the remaining suspension actuators in step S523.

[0102] It is understandable that when the end-to-end failure state is at level three, level three is a safe shutdown mode. At this time, fault tolerance control includes step S531, such as... Figure 7 As shown.

[0103] Step S531: Control the suspension stiffness and damping to smoothly return to the default straight-line driving state within a preset time period according to the preset low gradient, and disable the function activation.

[0104] Understandably, step S531 immediately executes a smooth shutdown strategy, strictly prohibiting sudden changes in control parameters. Simultaneously, it prevents the function from reactivating until the fault is completely eliminated, and records fault codes and failure data for subsequent fault diagnosis.

[0105] Thus, the three-level fault-tolerant control strategy breaks the binary switching logic of existing technologies. It achieves uninterrupted functional degradation through sensing redundancy and actuator torque reconstruction, and avoids secondary attitude disturbances through a smooth shutdown mechanism. It fully meets the functional safety requirements of ISO26262 and solves the core compliance pain point that existing technologies cannot be mass-produced and implemented.

[0106] In other embodiments, the failure scenario can be divided into more levels or simplified to two levels, depending on its complexity. As long as differentiated fault-tolerant control is achieved based on the failure level, it can replace the three-level classification in steps S510 to S530.

[0107] Understandably, the robust control method for cornering roll pre-adjustment includes step S600, such as... Figure 1 As shown.

[0108] Step S600: Output human-computer interaction prompts based on fault-tolerant control.

[0109] It is understandable that the grading and fault-tolerant control of step S500, and step S600, meet the functional safety requirements for the driver's right to know, and output clear and unambiguous human-machine interaction prompts to the driver based on different functional failure levels.

[0110] Understandably, when the end-to-end failure status is at level one, the instrument panel and HUD display green function icons, along with the text message "Tilting pre-adjustment function activated," without any additional interference.

[0111] Understandably, when the end-to-end failure status is at level two, the instrument panel and HUD display yellow function icons, along with corresponding text prompts (such as "tilt pre-adjustment function is degraded" or "tilt pre-adjustment function is partially restricted"), without voice warnings, to avoid interfering with the driver.

[0112] Understandably, when the end-to-end failure status is at level three, the instrument panel and HUD display red function icons, along with the text prompt "Tilting pre-adjustment function is off, please drive with caution," and simultaneously trigger a single voice warning to inform the driver of the function status, while also recording the fault message.

[0113] With this setup, through steps S100 to S600, the robust control method for pre-adjusting cornering roll fully covers the entire process of robustness improvement and functional safety protection.

[0114] Understandably, the cornering roll pre-adjustment robust control method is applied to vehicles equipped with adjustable suspension systems (air springs / continuously adjustable dampers), road preview modules, environmental perception systems, vehicle status sensors, and functional safety domain controllers. The cornering roll pre-adjustment robust control method is completely decoupled from the existing basic roll pre-adjustment control logic and can be independently superimposed on any roll pre-adjustment basic scheme, possessing complete implementation independence.

[0115] In other embodiments, the cornering roll pre-adjustment robust control method is applied to the vehicle's cornering roll pre-adjustment function, and can also be applied to other vehicle attitude control functions such as vehicle straight-line driving anti-crosswind, braking pitch suppression, and acceleration pitch suppression. As long as it is based on multi-source information fusion robust control and full-link redundancy fault-tolerant architecture, it can replace the application scenario.

[0116] This invention also provides a system comprising a data acquisition module, a fusion module, an adaptive calibration module, a monitoring module, and a hierarchical fault-tolerant module. The data acquisition module acquires multi-source heterogeneous information. The fusion module evaluates the confidence level of the multi-source heterogeneous information and outputs fused multi-source information. The adaptive calibration module adaptively calibrates the entire scenario based on the fused multi-source information. The monitoring module monitors the end-to-end failure state in real time. The hierarchical fault-tolerant module performs hierarchical fault-tolerant control based on the end-to-end failure state.

[0117] Furthermore, the system also includes a human-computer interaction module, which is used to output human-computer interaction prompts based on fault-tolerant control.

[0118] Thus, the core control logic of the system revolves around "robust control based on the fusion of multi-source heterogeneous information" and "functional safety redundancy across the entire perception-decision-execution chain".

[0119] Understandably, the system features a fully decoupled modular design, ensuring its independence from other vehicle components. As a robust and safety-redundant module, the system can be independently superimposed on any basic roll pre-adjustment control logic, without involving suspension parameter solving, timing control, cross-domain coordination, or other related aspects, thus enabling the system to be implemented completely independently.

[0120] This invention also provides a vehicle that applies the cornering roll pre-adjustment robust control method described above.

[0121] Understandably, the vehicle is a mass-produced passenger car equipped with a cornering roll pre-adjustment function. Its core configurations include: a high-precision map module, a forward-looking monocular camera and millimeter-wave radar environmental perception system, four-wheel independent adjustable air springs and continuously adjustable damping suspension system, a body status sensor matrix, a chassis domain controller, and a functional safety domain controller that meets the ISO26262ASIL-B standard.

[0122] Example 1 It is understandable that Example 1 is a scenario for redundancy backup switching when the high-precision map fails.

[0123] Scenario setting: The vehicle is traveling at a preset speed on a mountain road. The tilt pre-adjustment function is activated normally, and normal control is performed based on the curvature information of the high-precision map. When the vehicle reaches the tunnel entrance, the high-precision map signal is lost, the positioning accuracy decreases, and the multi-source heterogeneous information failure judgment is triggered.

[0124] Implementation process: I. Level 1 Operation Phase: When the vehicle is traveling outside the tunnel, the high-precision map signal is normal with a confidence level of 95%. The fusion module uses the high-precision map as the core weight and combines it with other information from the four-dimensional input matrix to output the optimal control parameters through the adaptive calibration model. The function is running normally, and the instrument displays a green activation icon.

[0125] II. Failure Monitoring and Judgment: After the vehicle enters the tunnel, the high-precision map signal is lost for more than 500ms, the positioning error exceeds the preset safety threshold, and the confidence of the curvature data drops below 20%. The monitoring module judges it as a level two state of partial failure of multi-source heterogeneous information.

[0126] 3. Redundant backup and seamless switching: The fusion module immediately and dynamically adjusts the weight allocation, raising the weight of the lane curvature information identified by forward vision and millimeter-wave radar to the highest level. Combined with the vehicle driving trajectory prediction, it outputs the fused effective curve curvature information. The entire switching process is without functional interruption or control parameter jump.

[0127] IV. Level 2 Operation: Based on the input information of redundant sensing, the adaptive calibration model slightly reduces the adjustment range of control parameters to increase the safety margin and continues to execute the roll pre-adjustment function; the instrument switches to a yellow icon, displaying "Roll pre-adjustment function degraded operation", with no voice interference.

[0128] V. Functional Recovery: After the vehicle exits the tunnel, the high-precision map signal is restored, the confidence level rises to over 90%, the system seamlessly switches back to Level 1 status, the instrument panel returns to the green icon, there is no fluctuation in vehicle posture throughout the entire process, and the function is continuous and uninterrupted.

[0129] Thus, through the dual redundancy backup mechanism of multi-source heterogeneous information, Implementation 1 achieves seamless switching after the failure of high-precision map, with no interruption or failure of function and stable vehicle posture. It completely solves the problem of functional paralysis caused by the failure of a single sensing source in the existing technology, and can be adapted to all driving scenarios without high-precision map signals, with strong universality.

[0130] Example 2 It is understandable that Example 2 is a fault-tolerant scenario when the suspension actuator fails (single suspension actuator failure).

[0131] Scenario setting: The vehicle is cornering on a highway ramp at a preset speed, and the roll pre-adjustment function is operating normally; during the cornering process, the left front air spring actuator experiences a hardware failure and becomes unresponsive, triggering a single suspension actuator failure judgment.

[0132] Implementation process: Level 1 Operation Phase: When the vehicle enters the highway ramp, the functions are activated normally. Based on multi-source fusion information, the four-wheel control parameters are output, the vehicle body rolls steadily, and the functions operate normally.

[0133] II. Failure Monitoring and Judgment: The monitoring module detects that the actual stiffness of the left front air spring deviates from the target value by more than the preset threshold for a duration of more than 200ms. At the same time, it receives a hardware fault code of the suspension actuator, which determines that a single suspension actuator has failed and triggers the secondary function degradation mode.

[0134] III. Torque Reconfiguration Control: The system immediately initiates the torque reconfiguration control strategy. Based on the physical limits of the remaining three normal suspension actuators, the adaptive calibration module re-solves for the optimal stiffness and damping parameters. By adjusting the parameters of the right front, left rear, and right rear actuators, the system maximizes the compensation for the loss of roll suppression capability caused by the failure of the left front actuator, while ensuring the torque balance of the left and right suspensions and avoiding sudden changes in vehicle body posture.

[0135] IV. Safety Assist Control: The system synchronously limits the vehicle's maximum speed to prevent excessive cornering speed from exacerbating the risk of body roll. The instrument panel displays a yellow icon and the message "Partially limited roll pre-adjustment function, please drive with caution," while also recording a fault code.

[0136] 5. Safety shutdown: After the vehicle exits the ramp and enters the straight road, the system detects no risk of roll and smoothly restores the suspension parameters to the default state according to the preset low gradient. It then switches to the third-level state and prompts the driver that the function has been turned off and recommends maintenance.

[0137] Thus, through the suspension actuator torque reconstruction control strategy, the second embodiment can still control the body roll angle within the safe threshold when a single actuator fails, without sudden attitude changes or loss of control risk. This completely solves the safety hazards caused by actuator failure in the prior art and can be adapted to any extreme scenario of single or multiple actuator failure.

[0138] Example 3 It is understandable that Example 3 is a full-scenario adaptive calibration scenario that dynamically adapts to complex working conditions.

[0139] Scenario setting: When the vehicle is unloaded, it is driving on a dry asphalt road in standard driving mode, and the roll pre-adjustment function is working normally; later, the vehicle is fully loaded with 5 people and luggage, and is driving on a wet and slippery road. The driver switches to sport driving mode, and the working conditions change drastically.

[0140] Implementation process: 1. No-load and dry road surface condition: The acquisition module collects information such as no-load, dry road surface, and standard driving mode. The adaptive calibration model outputs the optimal control parameters for the corresponding condition to achieve a balance between roll suppression and ride comfort. The function operates normally.

[0141] II. Real-time data collection of operating conditions: When the vehicle is fully loaded, the vehicle status information dimension collects data in real time, including the increase in vehicle weight and the rise in center of gravity; when driving in the rain, the environmental perception dimension collects data, including the decrease in road surface adhesion coefficient and the slippery road surface; after the driver switches to sport mode, the driver characteristic dimension collects data, including the preference for sport driving style.

[0142] III. Adaptive Calibration Dynamic Update: The fusion module updates the four-dimensional input matrix in real time. The adaptive calibration model dynamically solves the control parameters that are adapted to full load, wet and slippery road surface, and motion mode based on the new working condition information. It automatically increases the adjustment range of suspension stiffness and damping, while strictly limiting the wheel load transfer rate to avoid the risk of sideslip on wet and slippery road surface.

[0143] IV. Full-condition adaptability: Throughout the entire process of changing operating conditions, the control parameters are dynamically and smoothly adjusted without sudden changes or impacts. Regardless of whether it is unloaded or fully loaded, or on dry or wet roads, it can achieve the best roll suppression effect without control failure or reduced comfort.

[0144] Thus, in embodiment three, through four-dimensional multi-source input and adaptive calibration model, dynamic adaptation to all working conditions is achieved. Without manual calibration, it can cover driving scenarios with different loads, road surfaces, and driving modes, completely solving the problem of poor adaptability of fixed parameter working conditions in existing technologies, and has extremely strong generalization ability and universality.

[0145] The vehicle can be a private car, such as a sedan, SUV, MPV, or pickup truck. It can also be a commercial vehicle, such as a van, bus, small truck, or large semi-trailer. The vehicle must have an electric motor capable of outputting power or acting as a generator to store mechanical energy. When the vehicle is a new energy vehicle, it can be a hybrid or a pure electric vehicle.

[0146] Since the vehicle applies all the technical solutions of the above-described cornering roll pre-adjustment robust control method or system, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0147] The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A robust control method for pre-adjustment of curve roll, characterized in that, Includes the following steps: Collect heterogeneous information from multiple sources and construct a four-dimensional input matrix; Assess the confidence level of multi-source heterogeneous information and fuse and output multi-source fused information; Based on multi-source fusion information, adaptive calibration is performed across all scenarios; Real-time monitoring of end-to-end failure status; Fault-tolerant control based on end-to-end failure state classification.

2. The robust control method for pre-adjustment of cornering roll according to claim 1, characterized in that, The process of evaluating the confidence level of multi-source heterogeneous information and fusing and outputting multi-source fused information includes the following steps: Preprocessing multi-source heterogeneous information; Evaluate the confidence level of preprocessed multi-source heterogeneous information; Dynamic weight allocation; Output multi-source fused information.

3. The robust control method for pre-adjustment of cornering roll according to claim 1, characterized in that, The adaptive calibration of the entire scene based on multi-source fusion information includes the following steps: Pre-trained adaptive calibration model; Solve for online dynamic parameters; Adaptive calibration model learns online.

4. The robust control method for pre-adjustment of curve roll according to claim 1, characterized in that, The real-time monitoring of the entire link failure status includes the following steps: Failure to monitor multi-source heterogeneous information; Monitoring and adaptive calibration model failure; The suspension actuator is malfunctioning.

5. The robust control method for pre-adjustment of cornering roll according to claim 1, characterized in that, The classification based on the end-to-end failure status includes the following steps: When there are no failures in the entire link, the confidence level of all multi-source heterogeneous information is normal, and all suspension actuators are working normally, the failure state of the entire link is judged as a level one state. When a partial failure occurs and does not affect core driving safety, the entire link failure state is judged as a level two state. When the entire link fails severely, the failure status is determined to be Level 3.

6. The robust control method for pre-adjustment of cornering roll according to claim 5, characterized in that, The fault-tolerant control includes the following steps: When the failure state is the first level, the optimal control parameters are output based on multi-source fusion information and adaptive calibration model, and the roll pre-adjustment function is executed. When the failure status is level 2, the redundancy backup or strategy adjustment maintains the operation of the tilt pre-adjustment function. When the failure state is level three, the suspension stiffness and damping are controlled to smoothly recover to the default straight-line driving state within a preset time period according to the preset low gradient, and the function is prohibited from being activated.

7. The robust control method for pre-adjustment of cornering roll according to claim 6, characterized in that, When the failure state is level two, the redundancy backup or strategy adjustment to maintain the roll pre-adjustment function includes the following steps: If the local failure area contains multi-source heterogeneous information, the redundant backup maintains the tilt pre-adjustment function. If the local failure area is the suspension actuator, the strategy adjustment maintains the roll pre-adjustment function.

8. The robust control method for pre-adjustment of cornering roll according to claim 1, characterized in that, It also includes the following steps: Based on the aforementioned fault-tolerant control, human-computer interaction prompts are output.

9. A system, characterized in that, Including: The acquisition module acquires multi-source heterogeneous information and constructs a four-dimensional input matrix; The fusion module evaluates the confidence level of multi-source heterogeneous information and fuses and outputs multi-source fused information. An adaptive calibration module, which adaptively calibrates across the entire scenario based on multi-source fusion information; The monitoring module monitors the failure status of the entire link in real time. A hierarchical fault-tolerant module, which performs hierarchical fault-tolerant control based on the failure state of the entire link.

10. A vehicle, characterized in that, The robust control method for pre-adjustment of cornering roll as described in any one of claims 1 to 8 is applied.

Images