A braking control optimization method and system combining vehicle multi-state perception

Abstract

This invention discloses a braking control optimization method and system combining vehicle multi-state perception, belonging to the field of vehicle intelligent control technology. The method includes five steps: multi-source data acquisition, data preprocessing, multi-state risk modeling, multi-objective braking force distribution, and braking control optimization. Based on a tire dynamics model and a semantic-physical dual-source weighting mechanism, high-precision vehicle multi-state perception feature data is generated. A multi-state risk modeling method based on path prediction and physical constraint models, combined with a dynamic risk weight adjustment mechanism, calculates the comprehensive target deceleration. A multi-objective optimization function including braking distance, stability, and tire reliability is constructed to perform four-wheel braking force distribution and coaxial coordinated control. Pressure control is executed based on the optimization results to achieve braking control optimization. This invention can achieve real-time adaptive adjustment of braking control strategies under complex environments and multiple operating conditions, improving braking safety, stability, and response accuracy.

Description

Technical Field

[0001] This invention relates to the field of vehicle intelligent control technology, specifically to a braking control optimization method and system that combines multi-state perception of the vehicle. Background Technology

[0002] The braking control optimization method and system combining multi-state vehicle perception refers to an intelligent control method that uses multi-source sensors to collect real-time vehicle dynamic parameters (such as vehicle speed, wheel speed, acceleration, and braking pressure) and external environmental information (such as road adhesion coefficient, slope, obstacle distance, and weather conditions). After unified spatiotemporal synchronization and physical modeling processing, it dynamically calculates braking risk and performs multi-objective braking force distribution. This system comprehensively utilizes various algorithms, including vehicle dynamics models, environmental semantic fusion, path prediction, and optimized control, enabling braking control to adaptively adjust under different road conditions and driving states. Its fundamental function is to optimize braking force distribution and execution strategies through joint perception and intelligent analysis of vehicle operating status and the external environment, thereby shortening braking distance, improving vehicle stability, reducing tire slippage or yaw risks, and achieving synergistic optimization of active safety braking control and intelligent driving systems. It is one of the core foundational technologies in the field of intelligent connected vehicles and autonomous driving control.

[0003] However, existing braking control optimization methods suffer from technical problems such as poor vehicle environmental adaptability, lack of dynamic risk perception in braking force distribution, and difficulty in achieving stable braking control under different operating conditions.

[0004] In the existing process of vehicle multi-state pre-processing, there are technical problems such as asynchronous sensor data, large noise interference, and low state estimation accuracy.

[0005] Existing multi-state risk modeling methods suffer from technical problems such as path prediction relying on a single model, fixed environmental risk weights, and insufficient flexibility in calculating braking demand.

[0006] Existing multi-objective braking force distribution methods suffer from technical problems such as a single optimization objective, coarse distribution of braking torque, and uncoordinated control of the left and right wheels. Summary of the Invention

[0007] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides a braking control optimization method and system that integrates vehicle multi-state perception. Addressing the technical problems of poor vehicle environmental adaptability, lack of dynamic risk perception in braking force distribution, and difficulty in achieving stable braking control under different operating conditions in existing braking control optimization methods, this solution creatively adopts a comprehensive braking control optimization method combining multi-state risk modeling and multi-objective braking force distribution. By jointly modeling vehicle multi-source state and environmental information, it achieves real-time adaptive optimization of the braking control strategy under different adhesion coefficients, slopes, and weather conditions, significantly improving the safety and robustness of the braking system. Furthermore, addressing the technical problems of asynchronous sensor data, high noise interference, and low state estimation accuracy in existing vehicle multi-state pre-processing, this solution creatively employs environmental information fusion and a multi-observation joint state estimation method based on tire dynamics models. Through semantic-physical dual-source adaptive weighted adhesion coefficient estimation and a spatiotemporal synchronization mechanism, it improves the accuracy of state perception and environmental... The ability to understand the underlying principles provides highly reliable state input for subsequent braking force distribution. Addressing the technical issues of existing multi-state risk modeling methods, such as path prediction relying on a single model, fixed environmental risk weights, and insufficient flexibility in braking demand calculation, this solution creatively employs risk assessment calculation based on path prediction and a physical model. By introducing a dynamic risk weight adjustment mechanism, it can adjust the risk factor weights in real time according to environmental variables such as adhesion coefficient, slope, and visibility, thereby generating a comprehensive target deceleration, making braking demand more in line with the dynamic safety requirements of complex road scenarios. Furthermore, addressing the technical issues of existing multi-objective braking force distribution methods, such as a single optimization objective, coarse braking torque distribution, and uncoordinated left and right wheel control, this solution creatively uses a multi-objective braking optimization function to calculate four-wheel braking force distribution. It combines tire adhesion constraints and front-rear axle ratio constraints for real-time solution and introduces a co-axle dual-wheel micro-distribution strategy, achieving integrated distribution of independent braking force for four wheels and co-axle coordinated control, significantly improving vehicle braking balance and handling stability.

[0008] The technical solution adopted by this invention is as follows: This invention provides a braking control optimization method combining vehicle multi-state perception, the method comprising the following steps:

[0009] Step S1: Multi-source data acquisition;

[0010] Step S2: Data preprocessing;

[0011] Step S3: Multi-state risk modeling;

[0012] Step S4: Multi-target braking force distribution;

[0013] Step S5: Braking control optimization.

[0014] Further, in step S1, the multi-source data acquisition is used to collect multi-dimensional perception data of the vehicle and the environment. Specifically, it is obtained by using vehicle dynamics sensors and environmental perception sensors to perform high-precision time synchronization and data aggregation to obtain a synchronized multi-source dataset. The synchronized multi-source dataset specifically includes vehicle dynamics data, environmental perception data, control execution feedback data, and synchronization annotation information.

[0015] Further, in step S2, the data preprocessing, used to optimize the data and generate semantic features and key state features, specifically involves obtaining vehicle multi-state perception feature data based on the synchronous multi-source dataset through data filtering, environmental information fusion, and joint state estimation based on the tire dynamics model, including the following steps:

[0016] Step S21: Multi-source signal consistency filtering, specifically, based on the synchronous multi-source dataset, the sampling frequency and noise characteristics of different sensors are processed by channel filtering and resampling, and exponential smoothing filtering is applied to vehicle dynamics data, timestamp interpolation is applied to environmental perception data, and time compensation is applied to lost frame data to obtain a smoothed filtered dataset.

[0017] Step S22: Environmental information fusion, specifically, by defining an environmental semantic set, based on the smoothed filtering dataset, generating prior values ​​of adhesion coefficient and slope, which are used for initial value correction in subsequent dynamic estimation, to obtain environmental semantic information data;

[0018] The environmental semantic set specifically includes road type, weather conditions, road surface type, and slope grade;

[0019] Step S23: Construction of the state observation model, specifically, based on the vehicle's longitudinal dynamic balance and tire mechanical characteristics, an observation model for estimating the adhesion coefficient and the vertical load of the four wheels is established to obtain the tire dynamic state model;

[0020] Step S24: Joint state estimation of multiple observations, specifically by using a semantic-physical dual-source adaptive weighting method to jointly estimate the adhesion coefficients, obtain the joint adhesion coefficients, and construct a multi-observation state vector by combining the joint adhesion coefficients;

[0021] The multi-observation state vector includes longitudinal acceleration, four-wheel wheel speed, braking pressure, semantic adhesion coefficient prior, and slope prior;

[0022] Step S25: Vehicle multi-state perception feature generation, specifically by structuring the multi-observation state vectors to obtain structured state vectors, and by fusing the structured state vectors with the environmental semantic information data to obtain vehicle multi-state perception feature data;

[0023] The vehicle multi-state perception feature data specifically includes vehicle travel state data, environmental semantic information, real-time road surface adhesion coefficient, road slope, and four-wheel vertical load.

[0024] Furthermore, in step S3, the multi-state risk modeling is used to transform the environment and vehicle state into a quantitative braking demand. Specifically, based on the vehicle's multi-state perception feature data, a comprehensive target deceleration is obtained by performing a risk assessment calculation based on path prediction and a physical model.

[0025] In the risk assessment calculation stage, a physical constraint model based on vehicle longitudinal dynamics is used to quantify braking risk. On this basis, a dynamic risk weight adjustment mechanism is introduced to adaptively adjust the risk factor weights according to real-time environmental semantics and vehicle state to form a dynamic comprehensive assessment value, thereby obtaining the final comprehensive target deceleration.

[0026] The comprehensive target deceleration reflects the vehicle's current stability, braking distance, gradient impact, and environmental risks, and serves as the input parameter for multi-target braking force distribution in the subsequent step S4.

[0027] Further, in step S4, the multi-objective braking force distribution is used to convert the comprehensive target deceleration into an optimal coaxial control command that can be executed on the actual braking system. Specifically, based on the comprehensive target deceleration and the vehicle multi-state perception feature data, a real-time optimization solution with stability, braking distance, and tire reliability as objectives is performed to obtain the four-wheel independent target braking torque. Based on the four-wheel independent target braking torque, an actual axle braking torque for coaxial dual-wheel cooperative control is generated through an actuator constraint mapping strategy. This includes the following steps:

[0028] Step S41: Constructing a multi-objective braking optimization function. Based on the comprehensive target deceleration and vehicle multi-state perception feature data, a multi-objective optimization function containing three types of indicators—braking distance, vehicle stability, and tire reliability—is constructed to model the braking force distribution target and obtain a quantifiable braking optimization function.

[0029] Among them, by converting the comprehensive target deceleration into the braking force required by the whole vehicle, and combining the vehicle's real-time adhesion coefficient, vertical load distribution and attitude parameters, a multi-objective optimization function is established.

[0030] Step S42: Optimize the four-wheel braking force distribution. Specifically, based on the multi-objective optimization function, by introducing tire adhesion constraints and front and rear axle braking ratio constraints, perform constrained optimization to obtain the independent target braking torque of the four wheels.

[0031] Step S43: Actuator constraint mapping, specifically, based on the four independent target braking torques, by establishing an equivalent relationship model between wheel-end braking force and actuator output pressure, actuator capability mapping and optimization solution are performed to obtain the front and rear axle target pressures;

[0032] The equivalent relationship model solves for the front and rear axle pressure parameters by minimizing the error between the ideal braking force and the braking force that the actuator can output.

[0033] Step S44: Coaxial dual-wheel control strategy generation, specifically, based on the target pressure of the front and rear axles, by analyzing the real-time vertical load distribution of the left and right wheels of the coaxial axle, a coaxial dual-wheel micro-distribution strategy is designed to obtain the coaxial cooperative control command;

[0034] Step S45: Braking force distribution, specifically, based on the target pressure of the front and rear axles, the braking control unit sends pressure and provides execution feedback to distribute and execute the braking torque, thereby obtaining the actual axle braking torque.

[0035] Furthermore, in step S5, the brake control optimization is used to execute the distribution command, specifically by obtaining the brake actuator pressure control command based on the actual axle braking torque through pressure tracking control and parameter adaptation.

[0036] The present invention provides a braking control optimization system that combines vehicle multi-state perception, including a multi-source data acquisition module, a data preprocessing module, a multi-state risk modeling module, a multi-objective braking force distribution module, and a braking control optimization module;

[0037] The multi-source data acquisition module is used for multi-source data acquisition, obtaining a synchronous multi-source dataset through multi-source data acquisition, and sending the synchronous multi-source dataset to the data preprocessing module;

[0038] The data preprocessing module is used for data preprocessing. Through data preprocessing, vehicle multi-state perception feature data is obtained, and the vehicle multi-state perception feature data is sent to the multi-state risk modeling module and the multi-target braking force distribution module.

[0039] The multi-state risk modeling module is used for multi-state risk modeling. Through multi-state risk modeling, a comprehensive target deceleration is obtained, and the comprehensive target deceleration is sent to the multi-target braking force distribution module.

[0040] The multi-target braking force distribution module is used for multi-target braking force distribution. Through multi-target braking force distribution, the actual axle braking torque is obtained, and the actual axle braking torque is sent to the braking control optimization module.

[0041] The braking control optimization module is used for braking control optimization, and through braking control optimization, it obtains the brake actuator pressure control command.

[0042] The beneficial effects achieved by the present invention using the above solution are as follows:

[0043] (1) In view of the technical problems that exist in the existing braking control optimization methods, such as poor vehicle environmental adaptability, lack of dynamic risk perception in braking force distribution, and difficulty in achieving stable braking control under different working conditions, this solution creatively adopts a comprehensive braking control optimization method that combines multi-state risk modeling and multi-objective braking force distribution. By jointly modeling the multi-source state of the vehicle and environmental information, the braking control strategy can be optimized in real time under different adhesion coefficients, slopes and weather conditions, which significantly improves the safety and robustness of the braking system.

[0044] (2) In view of the technical problems of asynchronous sensor data, large noise interference and low state estimation accuracy in the existing vehicle multi-state pre-processing process, this solution creatively adopts environmental information fusion and multi-observation joint state estimation method based on tire dynamics model. Through semantic and physical dual-source adaptive weighted adhesion coefficient estimation and spatiotemporal synchronization mechanism, the state perception accuracy and environmental understanding ability are improved, providing a highly reliable state input for subsequent braking force distribution.

[0045] (3) In view of the technical problems in the existing multi-state risk modeling methods, such as path prediction relying on a single model, fixed environmental risk weights, and insufficient flexibility in braking demand calculation, this solution creatively adopts risk assessment calculation based on path prediction and physical model. By introducing a dynamic risk weight adjustment mechanism, the risk factor weights can be adjusted in real time according to environmental variables such as adhesion coefficient, slope, and visibility, thereby generating a comprehensive target deceleration, making the braking demand more in line with the dynamic safety requirements under complex road scenarios.

[0046] (4) In view of the technical problems of single optimization target, coarse distribution of braking torque and uncoordinated control of left and right wheels in the existing multi-objective braking force distribution method, this scheme creatively adopts a multi-objective braking optimization function to calculate the four-wheel braking force distribution, combines tire adhesion constraint and front and rear axle ratio constraint for real-time solution, and introduces a coaxial dual-wheel micro-distribution strategy to realize the integrated distribution of independent braking force of four wheels and coaxial coordinated control, which significantly improves the vehicle braking balance and handling stability. Attached Figure Description

[0047] Figure 1 A flowchart illustrating a braking control optimization method combining vehicle multi-state perception provided by the present invention;

[0048] Figure 2A schematic diagram of a braking control optimization system that combines vehicle multi-state perception provided by the present invention;

[0049] Figure 3 This is a flowchart illustrating the multi-source data acquisition process in step S1.

[0050] Figure 4 This is a flowchart illustrating the data preprocessing process in step S2.

[0051] Figure 5 This is a flowchart illustrating the multi-target braking force allocation process in step S4.

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0054] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship 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 limitations on this invention.

[0055] Example 1, see Figure 1 The present invention provides a braking control optimization method combining vehicle multi-state perception, the method comprising the following steps:

[0056] Step S1: Multi-source data acquisition;

[0057] Step S2: Data preprocessing;

[0058] Step S3: Multi-state risk modeling;

[0059] Step S4: Multi-target braking force distribution;

[0060] Step S5: Braking control optimization.

[0061] By performing the above operations, this solution addresses the technical problems of poor vehicle environmental adaptability, lack of dynamic risk perception in braking force distribution, and difficulty in achieving stable braking control under different operating conditions in existing braking control optimization methods. It creatively adopts a comprehensive braking control optimization method that combines multi-state risk modeling and multi-objective braking force distribution. By jointly modeling the multi-source state of the vehicle and environmental information, it realizes real-time adaptive optimization of the braking control strategy under different adhesion coefficients, slopes, and weather conditions, significantly improving the safety and robustness of the braking system.

[0062] Example 2, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S1, the multi-source data acquisition is used to collect multi-dimensional perception data of the vehicle and the environment. Specifically, it obtains a synchronized multi-source dataset by performing high-precision time synchronization and data aggregation through vehicle dynamics sensors and environmental perception sensors, including the following steps:

[0063] Step S11: Vehicle status signal acquisition, used to acquire the vehicle's own dynamic parameter signals, specifically through synchronous reading via the vehicle bus and sensor nodes to obtain raw dynamic parameter signal data;

[0064] The raw signal data of the dynamic parameters include vehicle speed, braking force, wheel speed, yaw rate, longitudinal acceleration, lateral acceleration, steering wheel angle, brake pedal travel, and brake hydraulic pressure data; the sensor nodes include IMU, wheel speed sensor, and brake pressure sensor;

[0065] Step S12: Environmental perception information acquisition, used to collect external environmental information, specifically by constructing a multimodal sensing unit including millimeter-wave radar, camera, lidar, and environmental temperature and humidity sensor, and performing joint acquisition to obtain environmental perception information data;

[0066] The environmental perception information data includes road type, road surface adhesion coefficient, weather conditions, distance and relative speed to obstacles ahead, and road slope;

[0067] Step S13: Spatiotemporal synchronization processing, used to achieve a unified spatiotemporal reference for different data sources. Specifically, a high-precision clock synchronization protocol is used to align the time of each sensor channel, and spatial coordinate synchronization is performed based on the GPS / IMU joint positioning system to form a unified spatiotemporal reference system and obtain spatiotemporal synchronized data.

[0068] Step S14: Multi-source fusion and aggregation, used to cache and align the spatiotemporal synchronization data according to timestamps and data source identifiers, and achieve unified formatting and fusion through the vehicle edge computing unit to obtain intermediate fused data;

[0069] Step S15: Data quality enhancement to ensure data reliability. Specifically, this involves frame loss detection, abnormal peak detection, and noise ratio evaluation through signal integrity detection, and combining time interpolation or Kalman filtering to repair abnormal signals, resulting in a synchronized multi-source dataset. The synchronized multi-source dataset specifically includes vehicle dynamics data, environmental perception data, control execution feedback data, and synchronization annotation information.

[0070] The vehicle dynamics data include vehicle speed, wheel speed, acceleration, yaw rate, pitch angle, roll angle, braking torque, brake hydraulic pressure, and tire slip ratio.

[0071] The environmental perception data includes road slope, road surface adhesion coefficient, road curvature, obstacle distance and relative speed, weather conditions, visibility and road type labels;

[0072] The control execution feedback data includes braking execution pressure feedback, actuator response signal, and control deviation parameters;

[0073] The synchronized annotation information includes timestamps, spatial coordinates, data source identifiers, and environmental tags;

[0074] Preferably, the signal integrity detection specifically includes frame loss detection, abnormal peak detection, and noise ratio evaluation.

[0075] Example 3, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to optimize the data and generate semantic features and key state features. Specifically, based on the synchronous multi-source dataset, vehicle multi-state perception feature data is obtained by performing data filtering, environmental information fusion, and joint state estimation based on the tire dynamics model. The steps include:

[0076] Step S21: Multi-source signal consistency filtering, specifically, based on the synchronous multi-source dataset, the sampling frequency and noise characteristics of different sensors are processed by channel filtering and resampling, and exponential smoothing filtering is applied to vehicle dynamics data, timestamp interpolation is applied to environmental perception data, and time compensation is applied to lost frame data to obtain a smoothed filtered dataset.

[0077] Step S22: Environmental information fusion, specifically, by defining an environmental semantic set, based on the smoothed filtering dataset, generating prior values ​​of adhesion coefficient and slope, which are used for initial value correction in subsequent dynamic estimation, to obtain environmental semantic information data;

[0078] The environmental semantic set includes semantic tags that reflect the state of roads and the external environment, specifically including road type, weather conditions, road surface type, and slope grade;

[0079] Preferably, the semantic tags can be obtained by parsing multimodal environmental perception data, specifically by the camera recognition model performing semantic segmentation and scene classification on image frames to obtain road types (e.g., highways, urban roads, and rural roads) and weather categories (e.g., sunny, rainy, snowy, and foggy).

[0080] The reflection intensity and point cloud characteristics of millimeter-wave radar and lidar can be used to determine the road surface type (dry, slippery, waterlogged, snowy, or icy) and the changes in the road slope ahead;

[0081] The slope prior value can be calculated based on the height difference of the lidar point cloud, the visual horizon offset angle, or the longitudinal attitude angle of the IMU, and the final slope level prior is formed after confirmation by semantic tags.

[0082] In addition, the vehicle-mounted embedded environmental recognition algorithm can perform secondary confirmation of weather conditions and visibility levels by fusing temperature and humidity sensor data with visual recognition confidence information.

[0083] To improve the robustness and consistency of environmental semantic labels, this embodiment adopts a cross-modal consistency verification mechanism, which cross-verifies the camera recognition results with the scene features of millimeter-wave radar / LiDAR. When the visual recognition confidence is lower than a preset threshold (e.g., 0.7), the road surface category is determined first by radar reflection features. When the radar intensity is abnormal or the point cloud noise is large, the visual semantic recognition is used as a compensation reference to obtain stable environmental semantic information.

[0084] Through the above-mentioned environmental information fusion processing, the system can obtain environmental semantic information data reflecting road semantics, weather conditions and slope status, providing stable and interpretable semantic prior input for subsequent dynamic state estimation;

[0085] Based on the combination of road type, pavement category, and weather conditions, the system can use a pre-defined semantic-to-adhesion coefficient mapping table to generate prior values ​​for the adhesion coefficient, which are then used as the semantic-mapped adhesion coefficients. Preferably, the following empirical mapping model can be used:

[0086] For dry asphalt roads, the value range is 0.75 to 0.85;

[0087] Slippery asphalt roads or waterlogged roads The value range is 0.45 to 0.60;

[0088] Snow covered the road surface. The value range is 0.25 to 0.35;

[0089] Icy or very low adhesion road surfaces The value range is 0.10 to 0.25;

[0090] gravel and sand roads The value range is 0.55 to 0.70;

[0091] The above range can be adjusted based on vehicle calibration tests and statistical priors to ensure the stability and reproducibility of the semantic source of the adhesion coefficient;

[0092] Step S23: Construction of the state observation model, specifically, based on the vehicle's longitudinal dynamic balance and tire mechanical characteristics, an observation model for estimating the adhesion coefficient and the vertical load of the four wheels is established to obtain the tire dynamic state model;

[0093] The calculation formula for the tire dynamics state model is as follows:

[0094] ;

[0095] In the formula, m is the total weight of the vehicle, and a x It is the longitudinal acceleration, i is the tire index, and F is the longitudinal acceleration. x,i R is the longitudinal force of the i-th wheel. res This is the resistance term; the longitudinal force of the wheel is calculated using the adhesion coefficient, vertical load, and slip ratio function.

[0096] Preferably, the tire longitudinal force F x,i Modeling can be performed using the standard Pacejka tire model or a simplified slip ratio function model, specifically including calculations based on the nonlinear relationship between slip ratio, vertical load, and adhesion coefficient;

[0097] Step S24: Joint state estimation of multiple observations, specifically by using a semantic-physical dual-source adaptive weighting method to jointly estimate the adhesion coefficients, obtain the joint adhesion coefficients, and construct a multi-observation state vector by combining the joint adhesion coefficients;

[0098] The formula for calculating the combined adhesion coefficient is as follows:

[0099] ;

[0100] In the formula, It is the combined adhesion coefficient, w dyn It is to reverse-engineer adaptive weights. The adhesion coefficient, w, is derived from vehicle dynamics. sem It is a mapping of adaptive weights. It is the semantic mapping attachment coefficient;

[0101] The adhesion coefficient based on vehicle dynamics backpropagation The measured longitudinal acceleration, wheel speed, and braking pressure can be substituted into the tire dynamics state model and estimated by inversion using the slip ratio function.

[0102] Preferably, w dyn with w sem The adhesion coefficient can be adjusted in real time based on the observational consistency between the adhesion coefficient derived from vehicle dynamics and the adhesion coefficient obtained from semantic mapping. The variance is denoted as Semantic mapping attachment coefficient The variance is denoted as By calculating the weights based on the inverse uncertainty principle, the reverse adaptive weights and the mapped adaptive weights are obtained. The calculation formula is as follows:

[0103] ;

[0104] In the formula, Based on the short time window The volatility estimate was obtained. It can be obtained based on semantic label stability, visual recognition confidence, or statistical prior; this implementation makes the adhesion coefficient estimation robust under conditions of low adhesion, high noise, or semantic uncertainty;

[0105] The multi-observation state vector includes longitudinal acceleration, four-wheel wheel speed, braking pressure, semantic adhesion coefficient prior, and slope prior;

[0106] Step S25: Vehicle multi-state perception feature generation, specifically by structuring the multi-observation state vectors to obtain structured state vectors, and by fusing the structured state vectors with the environmental semantic information data to obtain vehicle multi-state perception feature data;

[0107] The vehicle multi-state perception feature data specifically includes vehicle travel state data, environmental semantic information, real-time road surface adhesion coefficient, road slope, and four-wheel vertical load.

[0108] By performing the above operations, this solution addresses the technical problems of asynchronous sensor data, high noise interference, and low state estimation accuracy in existing vehicle multi-state pre-processing. It creatively adopts an environmental information fusion and a multi-observation joint state estimation method based on tire dynamics models. Through semantic and physical dual-source adaptive weighted adhesion coefficient estimation and spatiotemporal synchronization mechanism, it improves the state perception accuracy and environmental understanding capabilities, providing highly reliable state input for subsequent braking force distribution.

[0109] Example 4, see Figure 1 , Figure 2This embodiment is based on the above embodiment. In step S3, the multi-state risk modeling is used to transform the environment and vehicle state into a quantitative braking demand. Specifically, based on the vehicle multi-state perception feature data, a comprehensive target deceleration is obtained by performing a risk assessment calculation based on path prediction and a physical model.

[0110] Specifically, based on the vehicle's multi-state perception feature data, a trajectory prediction model based on kinematic constraints is used to predict the vehicle's short-term trajectory. The path prediction model can be a quadratic trajectory extrapolation model based on the vehicle's longitudinal speed, steering wheel angle, and current road geometry information. Its input parameters come from the vehicle's speed sensor, steering angle sensor, and road boundary and obstacle information extracted from the environmental perception system. Through this model, the predicted displacement and speed changes of the vehicle in the next 0.5 to 1.0 seconds can be obtained, forming path prediction data.

[0111] In the risk assessment calculation phase, a physical constraint model based on vehicle longitudinal dynamics is used to quantitatively analyze braking risk. The model combines vehicle longitudinal acceleration, road adhesion coefficient, and slope information to calculate the vehicle's basic braking requirements under current operating conditions. The assessment parameters include braking distance, adhesion limit, stability margin, and slope compensation value. Preferably, the braking safety margin is set to 10%~20%, and the slope correction factor is set to 0.05~0.15. By comprehensively analyzing the above parameters, the basic deceleration for safe braking of the vehicle under different environmental conditions is obtained.

[0112] Based on this, in order to improve the real-time performance and environmental adaptability of risk assessment, this embodiment introduces a dynamic risk weight adjustment mechanism. This mechanism adaptively adjusts the risk factor weights according to the real-time environmental semantics and vehicle status to form a dynamic comprehensive assessment value, thereby obtaining the final comprehensive target deceleration.

[0113] The dynamic risk weight adjustment mechanism is specifically as follows:

[0114] When the adhesion coefficient is detected to be below 0.45, the system determines it to be a low adhesion or slippery scenario, and automatically increases the calculation weight of braking distance and slope risk factor, so that the deceleration calculation is more biased towards safe braking requirements.

[0115] When the adhesion coefficient is detected to be higher than 0.7, the system determines it to be a high adhesion or dry scenario, and automatically increases the weight of vehicle stability and response time factors, so that the deceleration calculation focuses more on response efficiency and stable control.

[0116] When nighttime driving or visibility below 50 meters is detected, the system introduces environmental uncertainty parameters as additional risk factors and assigns them an extra weight of 5% to 10% to compensate for the potential risks caused by perception lag.

[0117] Preferably, the dynamic risk weight adjustment period is set to 100~200 milliseconds, and the risk factor weight change rate does not exceed ±20% to avoid control oscillations; through the above weight adjustment mechanism, the system can achieve adaptive dynamic balance of risk factors under different environments and conditions;

[0118] Finally, the system calculates a comprehensive risk index based on the dynamically weighted risk factors, and integrates this index with the basic braking demand under physical constraints to generate a comprehensive target deceleration; the comprehensive risk index can be obtained by weighting and summing the risk factors according to their weight coefficients.

[0119] The comprehensive target deceleration reflects the vehicle's current stability, braking distance, gradient impact, and environmental risks, and serves as the input parameter for multi-target braking force distribution in the subsequent step S4.

[0120] By performing the above operations, this solution addresses the technical problems of existing multi-state risk modeling methods, such as path prediction relying on a single model, fixed environmental risk weights, and insufficient flexibility in braking demand calculation. It creatively adopts risk assessment calculation based on path prediction and physical models, and by introducing a dynamic risk weight adjustment mechanism, it can adjust the risk factor weights in real time according to environmental variables such as adhesion coefficient, slope, and visibility, thereby generating a comprehensive target deceleration, making the braking demand more in line with the dynamic safety requirements of complex road scenarios.

[0121] Example 5, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment. In step S4, the multi-target braking force distribution is used to convert the comprehensive target deceleration into an optimal coaxial control command that can be executed on the actual braking system. Specifically, based on the comprehensive target deceleration and the vehicle multi-state perception feature data, a real-time optimization solution is performed with stability, braking distance, and tire reliability as objectives to obtain the four-wheel independent target braking torque. Based on the four-wheel independent target braking torque, an actual axle braking torque for coaxial dual-wheel cooperative control is generated through an actuator constraint mapping strategy. This includes the following steps:

[0122] Step S41: Constructing a multi-objective braking optimization function. Based on the comprehensive target deceleration and vehicle multi-state perception feature data, a multi-objective optimization function containing three types of indicators—braking distance, vehicle stability, and tire reliability—is constructed to model the braking force distribution target and obtain a quantifiable braking optimization function.

[0123] Specifically, by converting the comprehensive target deceleration into the braking force required by the entire vehicle, and combining the vehicle's real-time adhesion coefficient, vertical load distribution, and attitude parameters, a multi-objective optimization function is established, and the calculation formula is as follows:

[0124] ;

[0125] In the formula, J is the multi-objective optimization function, i is the tire index, and F... b,i F is the target braking force of the i-th wheel. req It is the total braking force required for the entire vehicle. This is the braking efficiency weight, with a default value range of [0.4, 0.6]. It is an additional yaw moment used to reflect deviations in vehicle braking stability. This is the vehicle yaw weight, with a default value range of [0.2, 0.4]. It is the combined adhesion coefficient, F z,i It is the real-time vertical load of the i-th tire. This is the attachment limit weight, with a default value range of [0.1, 0.3].

[0126] As a further optimization of this embodiment, The yaw imbalance during vehicle braking can be calculated based on the difference in braking force between the left and right wheels. The preferred calculation method is as follows:

[0127] ;

[0128] In the formula, It is the additional yaw moment, F b,fl It is the target braking force of the front left wheel, F b,fr It is the target braking force of the front right wheel, t f It is half the front axle wheel diameter, F b,rl It is the target braking force of the rear left wheel, F b, rr It is the target braking force of the rear right wheel, t r It is half the rear axle track.

[0129] Step S42: Optimize the four-wheel braking force distribution. Specifically, based on the multi-objective optimization function, by introducing tire adhesion constraints and front and rear axle braking ratio constraints, perform constrained optimization to obtain the independent target braking torque of the four wheels.

[0130] Step S43: Actuator constraint mapping, specifically, based on the four independent target braking torques, by establishing an equivalent relationship model between wheel-end braking force and actuator output pressure, actuator capability mapping and optimization solution are performed to obtain the front and rear axle target pressures;

[0131] The equivalent relationship model solves for the front and rear axle pressure parameters by minimizing the error between the ideal braking force and the actuator's output braking force. The calculation formula is as follows:

[0132] ;

[0133] In the formula, J P It is to minimize the objective function, by applying the objective function J P The gradient descent method is used to minimize the solution to obtain the target front axle pressure P. f Rear axle target pressure P r This results in the formation of the target pressure F between the front and rear axles. B ={P f ,P r}, P f It is the target pressure on the front axle, P r It is the target pressure on the rear axle, i front It is the front wheel index, k i It is the equivalent braking gain coefficient of the i-th tire, representing the conversion ratio between wheel-end braking force and hydraulic pressure. It is the independent target braking torque for the four wheels corresponding to the i-th tire, i rear It's the rear wheel index;

[0134] Step S44: Coaxial dual-wheel control strategy generation, specifically, based on the target pressure of the front and rear axles, by analyzing the real-time vertical load distribution of the left and right wheels of the coaxial axle, a coaxial dual-wheel micro-distribution strategy is designed to obtain the coaxial cooperative control command;

[0135] The formula for calculating the real-time vertical load distribution of the coaxial left and right wheels is as follows:

[0136] ;

[0137] In the formula, It is the command braking force of the left front axle wheel. It is the real-time load proportionality factor of the left and right wheels of the front axle, used to compensate for minor braking forces based on the load distribution between the left and right wheels, k f It is the equivalent braking gain coefficient of the braking system, specifically determined by the equivalent braking gain coefficient k of the i-th tire. i Perform conversion calculations. It is the command braking force of the right wheel on the front axle;

[0138] The specific calculation formula for the real-time load ratio factor of the left and right front axle wheels is as follows:

[0139] ;

[0140] In the formula, It is the real-time load ratio factor for the left and right wheels of the front axle. It is the real-time vertical load on the left front axle wheel. It is the real-time vertical load on the right wheel of the front axle;

[0141] Similarly, the commanded braking force of the left and right rear wheels can be based on the rear axle load ratio factor. The allocation is performed while satisfying the following computational relationships:

[0142] ;

[0143] In the formula, It is the command braking force of the left rear wheel. It is the real-time load ratio factor of the left and right wheels of the rear axle, used to compensate for minor braking forces based on the load distribution between the left and right wheels, k r It is the equivalent braking gain coefficient of the braking system, specifically determined by the equivalent braking gain coefficient k of the i-th tire. i Perform conversion calculations. It is the command braking force of the right wheel on the rear axle;

[0144] The specific calculation formula for the real-time load ratio factor of the left and right rear axle wheels is as follows:

[0145] ;

[0146] In the formula, It is the real-time load ratio factor for the left and right wheels of the rear axle. It is the real-time vertical load on the left rear axle wheel. It is the real-time vertical load on the right rear wheel;

[0147] Step S45: Braking force distribution, specifically, based on the target pressure of the front and rear axles, the braking control unit sends pressure and provides execution feedback to distribute and execute the braking torque, thereby obtaining the actual axle braking torque.

[0148] By performing the above operations, this solution addresses the technical problems of existing multi-objective braking force distribution methods, such as a single optimization objective, coarse braking torque distribution, and uncoordinated control of the left and right wheels. It creatively adopts a multi-objective braking optimization function to calculate the four-wheel braking force distribution, combines tire adhesion constraints and front and rear axle ratio constraints for real-time solution, and introduces a coaxial dual-wheel micro-distribution strategy to achieve integrated distribution of independent braking force of four wheels and coaxial coordinated control, which significantly improves vehicle braking balance and handling stability.

[0149] Example 6, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S5, the brake control optimization is used to execute the distribution command. Specifically, based on the actual axle braking torque, the brake actuator pressure control command is obtained by performing pressure tracking control and parameter adaptation.

[0150] Preferably, the pressure tracking control can adopt proportional-integral-derivative (PID) control, feedforward compensation control, or model predictive control (MPC). By comparing the target pressure with the actual actuator pressure in real time, the pressure tracking error is calculated, and the drive signal of the hydraulic actuator is adjusted according to the error to make the pressure error converge to a preset threshold (such as ±2% to ±5%), thereby ensuring the precise application of braking torque.

[0151] Meanwhile, the control gain can be compensated for under conditions such as brake pad wear, temperature changes, or pressure hysteresis through a parameter adaptive mechanism, making the pressure control process more stable and robust.

[0152] Example 7, see Figure 1 and Figure 2 Based on the above embodiments, the present invention provides a braking control optimization system that combines vehicle multi-state perception, including a multi-source data acquisition module, a data preprocessing module, a multi-state risk modeling module, a multi-target braking force distribution module, and a braking control optimization module.

[0153] The multi-source data acquisition module is used for multi-source data acquisition, obtaining a synchronous multi-source dataset through multi-source data acquisition, and sending the synchronous multi-source dataset to the data preprocessing module;

[0154] The data preprocessing module is used for data preprocessing. Through data preprocessing, vehicle multi-state perception feature data is obtained, and the vehicle multi-state perception feature data is sent to the multi-state risk modeling module and the multi-target braking force distribution module.

[0155] The multi-state risk modeling module is used for multi-state risk modeling. Through multi-state risk modeling, a comprehensive target deceleration is obtained, and the comprehensive target deceleration is sent to the multi-target braking force distribution module.

[0156] The multi-target braking force distribution module is used for multi-target braking force distribution. Through multi-target braking force distribution, the actual axle braking torque is obtained, and the actual axle braking torque is sent to the braking control optimization module.

[0157] The braking control optimization module is used for braking control optimization, and through braking control optimization, it obtains the brake actuator pressure control command.

[0158] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0159] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0160] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A braking control optimization method combining vehicle multi-state perception, characterized in that: The method includes the following steps: Step S1: Collect multi-source data to obtain a synchronized multi-source dataset; Step S2: Data preprocessing. Based on the synchronous multi-source dataset, vehicle multi-state perception feature data is obtained by performing data filtering, environmental information fusion, and joint state estimation based on the tire dynamics model. This includes the following steps: multi-source signal consistency filtering, environmental information fusion, state observation model construction, multi-observation joint state estimation, and vehicle multi-state perception feature generation. The multi-observation joint state estimation specifically involves using a semantic-physical dual-source adaptive weighting method to jointly estimate the adhesion coefficients, obtaining joint adhesion coefficients, and then combining the joint adhesion coefficients to construct a multi-observation state vector. The formula for calculating the combined adhesion coefficient is as follows: ； In the formula, It is the combined adhesion coefficient, w dyn It is to reverse-engineer adaptive weights. The adhesion coefficient, w, is derived from vehicle dynamics. sem It is a mapping of adaptive weights. It is the semantic mapping attachment coefficient; w dyn With w sem The adhesion coefficient can be adjusted in real time based on the observational consistency between the adhesion coefficient derived from vehicle dynamics and the adhesion coefficient obtained from semantic mapping. The variance is denoted as Semantic mapping attachment coefficient The variance is denoted as By calculating the weights based on the inverse uncertainty principle, the reverse adaptive weights and the mapped adaptive weights are obtained. The calculation formula is as follows: ； In the formula, Based on the short time window The volatility estimate was obtained. It can be obtained based on semantic label stability, visual recognition confidence, or statistical prior. Step S3: Multi-state risk modeling. Based on the vehicle's multi-state perception feature data, a comprehensive target deceleration is obtained by performing risk assessment calculations based on path prediction and physical models. In step S3, the multi-state risk modeling, in the risk assessment calculation stage, adopts a physical constraint model based on vehicle longitudinal dynamics to quantitatively analyze braking risk; on this basis, a dynamic risk weight adjustment mechanism is introduced to adaptively adjust the risk factor weights according to real-time environmental semantics and vehicle state to form a dynamic comprehensive evaluation value and obtain the comprehensive target deceleration. Step S4: Multi-target braking force distribution. Based on the comprehensive target deceleration and the vehicle multi-state perception feature data, the four-wheel independent target braking torque is obtained by real-time optimization with stability, braking distance and tire reliability as objectives. Based on the four-wheel independent target braking torque, the actual axle braking torque for coaxial dual-wheel cooperative control is generated through the actuator constraint mapping strategy. In step S4, the multi-target braking force distribution includes the following steps: Step S41: Constructing a multi-objective braking optimization function. Based on the comprehensive target deceleration and vehicle multi-state perception feature data, a multi-objective optimization function containing three types of indicators—braking distance, vehicle stability, and tire reliability—is constructed to model the braking force distribution target and obtain a quantifiable braking optimization function. Specifically, by converting the comprehensive target deceleration into the braking force required by the entire vehicle, and combining the vehicle's real-time adhesion coefficient, vertical load distribution, and attitude parameters, a multi-objective optimization function is established, and the calculation formula is as follows: ； In the formula, J is the multi-objective optimization function, i is the tire index, and F... b,i F is the target braking force of the i-th wheel. req It is the total braking force required for the entire vehicle. It is the braking energy efficiency weight. It is an additional yaw moment used to reflect deviations in vehicle braking stability. It is the vehicle yaw weight. It is the combined adhesion coefficient, F z,i It is the real-time vertical load of the i-th tire. It is the attachment limit weight; Step S42: Optimize the distribution of braking force among four wheels; Step S43: Map actuator constraints; Step S44: Generate the coaxial dual-wheel control strategy; Step S45: Distribute braking force. Step S5: Braking control optimization, obtaining brake actuator pressure control commands.

2. The braking control optimization method combining vehicle multi-state perception according to claim 1, characterized in that: In step S1, the multi-source data acquisition is used to collect multi-dimensional perception data of the vehicle and the environment. Specifically, it is obtained by using vehicle dynamics sensors and environmental perception sensors to perform high-precision time synchronization and data aggregation to obtain a synchronized multi-source dataset. The synchronized multi-source dataset specifically includes vehicle dynamics data, environmental perception data, control execution feedback data, and synchronization annotation information.

3. The braking control optimization method combining vehicle multi-state perception according to claim 2, characterized in that: In step S2, the data preprocessing includes the following steps: Step S21: Multi-source signal consistency filtering, specifically, based on the synchronous multi-source dataset, the sampling frequency and noise characteristics of different sensors are processed by channel filtering and resampling, and exponential smoothing filtering is applied to vehicle dynamics data, timestamp interpolation is applied to environmental perception data, and time compensation is applied to lost frame data to obtain a smoothed filtered dataset. Step S22: Environmental information fusion, specifically, by defining an environmental semantic set, based on the smoothed filtering dataset, generating prior values ​​of adhesion coefficient and slope, which are used for initial value correction in subsequent dynamic estimation, to obtain environmental semantic information data; The environmental semantic set specifically includes road type, weather conditions, road surface type, and slope grade; Step S23: Construction of the state observation model, specifically, based on the vehicle's longitudinal dynamic balance and tire mechanical characteristics, an observation model for estimating the adhesion coefficient and the vertical load of the four wheels is established to obtain the tire dynamic state model; The multi-observation state vector includes longitudinal acceleration, four-wheel wheel speed, braking pressure, semantic adhesion coefficient prior, and slope prior; Step S25: Vehicle multi-state perception feature generation, specifically, by structuring the multi-observation state vectors to obtain structured state vectors, and by fusing the structured state vectors with the environmental semantic information data to obtain vehicle multi-state perception feature data.

4. The braking control optimization method combining vehicle multi-state perception according to claim 3, characterized in that: In step S2, the vehicle multi-state perception feature data specifically includes vehicle travel state data, environmental semantic information, real-time road surface adhesion coefficient, road slope, and four-wheel vertical load.

5. The braking control optimization method combining vehicle multi-state perception according to claim 4, characterized in that: In step S42, the optimization solution for the four-wheel braking force distribution is specifically obtained by performing a constrained optimization solution based on the multi-objective optimization function by introducing tire adhesion constraints and front and rear axle braking ratio constraints, thereby obtaining the independent target braking torque of the four wheels. In step S43, the actuator constraint mapping specifically involves, based on the four independent target braking torques, establishing an equivalent relationship model between wheel-end braking force and actuator output pressure, performing actuator capability mapping and optimization solution, and obtaining the front and rear axle target pressures. The equivalent relationship model solves for the front and rear axle pressure parameters by minimizing the error between the ideal braking force and the braking force that the actuator can output. In step S44, the coaxial dual-wheel control strategy is generated by analyzing the real-time vertical load distribution of the left and right wheels of the coaxial axle based on the target pressure of the front and rear axles, and designing a coaxial dual-wheel micro-distribution strategy to obtain the coaxial cooperative control command. In step S45, the braking force distribution specifically involves distributing and executing the braking torque based on the target pressures of the front and rear axles through the brake control unit, thereby obtaining the actual axle braking torque.

6. The braking control optimization method combining vehicle multi-state perception according to claim 5, characterized in that: In step S5, the brake control optimization is used to execute the distribution command. Specifically, based on the actual axle braking torque, pressure tracking control and parameter adaptation are performed to obtain the brake actuator pressure control command.

7. A braking control optimization system incorporating vehicle multi-state perception, used to implement the braking control optimization method incorporating vehicle multi-state perception as described in any one of claims 1-6, characterized in that: It includes a multi-source data acquisition module, a data preprocessing module, a multi-state risk modeling module, a multi-target braking force distribution module, and a braking control optimization module.

8. The braking control optimization system combining vehicle multi-state perception according to claim 7, characterized in that: The multi-source data acquisition module is used for multi-source data acquisition, obtaining a synchronous multi-source dataset through multi-source data acquisition, and sending the synchronous multi-source dataset to the data preprocessing module; The data preprocessing module is used for data preprocessing. Through data preprocessing, vehicle multi-state perception feature data is obtained, and the vehicle multi-state perception feature data is sent to the multi-state risk modeling module and the multi-target braking force distribution module. The multi-state risk modeling module is used for multi-state risk modeling. Through multi-state risk modeling, a comprehensive target deceleration is obtained, and the comprehensive target deceleration is sent to the multi-target braking force distribution module. The multi-target braking force distribution module is used for multi-target braking force distribution. Through multi-target braking force distribution, the actual axle braking torque is obtained, and the actual axle braking torque is sent to the braking control optimization module. The braking control optimization module is used for braking control optimization, and through braking control optimization, it obtains the brake actuator pressure control command.

Images