Brake-by-wire method and system based on personalized feature learning and intention evolution prediction, and vehicle

Abstract

The invention provides a brake-by-wire method and system based on personalized feature learning and intention evolution prediction and a vehicle. The invention discloses a brake-by-wire method based on personalized feature learning and intention evolution prediction. The brake-by-wire method comprises the following steps: acquiring real-time braking operation information of a driver, current state information of a vehicle and driving environment information of the vehicle; personalized braking characteristic parameters of the driver are obtained or updated; predicting a current braking intention type of the driver by utilizing a preset driving intention recognition model integrated with an attention mechanism; predicting the evolution trend of the current braking intention in a preset future time window; determining a group of target braking characteristic parameters according to the predicted current braking intention type and evolution trend in combination with the current state information of the vehicle and the driving environment information of the vehicle; and a brake control instruction of the brake-by-wire system is dynamically adjusted. According to the invention, more intelligent, smoother and predictable line control brake is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle control, and particularly to a brake-by-wire method, system and vehicle based on personalized feature learning and intention evolution prediction. Background Art

[0002] With the development of automotive electrification and intelligence, the brake-by-wire (BBW) system is gradually becoming the mainstream braking technology due to its advantages such as fast response, high control precision, and easy integration of advanced driver assistance system (ADAS) functions (such as automatic emergency braking AEB, regenerative braking, etc.). In traditional mechanical or hydraulic power-assisted braking systems, the driver directly controls the braking force through the brake pedal, and the pedal feel is closely related to the braking effect. In the brake-by-wire system, the brake pedal is usually decoupled from the brake actuator, and after the pedal input signal is processed by the electronic control unit (ECU), it then controls the brake actuator to generate the braking force.

[0003] The existing brake-by-wire systems have made remarkable progress in improving braking performance and integrating new functions. For example, some technologies focus on providing warnings and protection to the driver when the system experiences performance degradation (such as thermal degradation); some other technologies are dedicated to improving braking comfort under specific working conditions, such as reducing braking impact by adjusting pedal feel parameters during low-speed driving; there are also technical solutions that consider the influence of external factors such as vehicle load and road gradient on braking efficiency, and correct the braking force to ensure the consistency of braking effect.

[0004] However, even though some existing technologies have started to pay attention to the dynamics of the driver's braking intention, they often have the following deep-seated limitations: Firstly, the lack of personalized adaptability. Most intention recognition models are trained based on a large amount of "averaged" driving data, and it is difficult to accurately capture and adapt to the unique braking habits, risk preferences, and subtle differences in vehicle response expectations of specific drivers. This results in the system being less intelligent for some drivers, and the braking experience lacking personalization. Secondly, the lack of understanding and prediction ability for the evolution process of braking intention. The driver's braking intention may undergo continuous and rapid evolution during a braking event. Existing systems mainly focus on the judgment of the current intention, and have insufficient ability to finely model this evolution process and predict the short-term future trend of the intention, which may lead to the braking response lagging behind the actual change of the driver's intention, affecting driving smoothness and safety. Summary of the Invention

[0005] Based on the above background, the purpose of the present invention is to provide a wire control braking method, system and vehicle based on personalized feature learning and intention evolution prediction, aiming to deeply integrate the personalized braking habits of drivers, more accurately perceive the dynamic evolution process of braking intentions and conduct short-term trend prediction, so as to achieve more intelligent, smoother and more predictable wire control braking.

[0006] To achieve the above object of the invention, the first aspect of the present invention provides a wire control braking method based on personalized feature learning and intention evolution prediction, the method comprising the following steps: obtaining real-time braking operation information of the driver, current state information of the vehicle, and driving environment information of the vehicle; obtaining or updating personalized braking characteristic parameters of the driver; predicting the current braking intention type of the driver based on the dynamic characteristics of the real-time braking operation information of the driver, the personalized braking characteristic parameters of the driver, and using a preset driving intention recognition model integrated with an attention mechanism;

[0007] Predicting the evolution trend of the current braking intention within a preset future time window based on the output of the driving intention recognition model or an independent intention trend prediction module;

[0008] Determining a set of target braking characteristic parameters according to the predicted current braking intention type and the evolution trend, and combining the current state information of the vehicle and the driving environment information of the vehicle;

[0009] Based on the target braking characteristic parameters, dynamically adjusting the braking control instruction of the wire control braking system to control the vehicle to perform braking.

[0010] Preferably, the personalized braking characteristic parameters of the driver include at least one of the average pedal depression depth, average pedal depression rate, preference coefficient for different deceleration rates, and risk tolerance level statistically obtained from the driver's historical braking behavior.

[0011] Preferably, the driving intention recognition model integrated with an attention mechanism is constructed based on a recurrent neural network, long short-term memory network or gated recurrent unit, and the attention mechanism is used to assign higher weights to the parts more relevant to intention judgment and evolution trend prediction when processing the time series and auxiliary information of the real-time braking operation information.

[0012] Preferably, predicting the current braking intention type of the driver based on the dynamic characteristics of the real-time braking operation information of the driver, the personalized braking characteristic parameters of the driver, and using a preset driving intention recognition model integrated with an attention mechanism specifically includes:

[0013] Take the dynamic characteristics of the real-time braking operation information as the first input, and take the personalized braking characteristic parameters of the driver as the second input, and jointly input them into the driving intention recognition model;

[0014] Alternatively, first use the dynamic characteristics of the real-time braking operation information to obtain a preliminary intention through the basic intention recognition module, and then use the personalized braking characteristic parameters of the driver to correct the preliminary intention.

[0015] Preferably, predicting the evolution trend of the current braking intention within a preset future time window includes predicting the probability that the current braking intention remains unchanged or the probability of changing to other preset braking intention types.

[0016] Preferably, the current braking intention type includes one or a combination of an emergency braking intention, a conventional smooth deceleration intention, a comfortable slow stop intention, a curve pre-deceleration intention, and a micro speed control intention.

[0017] Preferably, the target braking characteristic parameters include at least one of the shape parameters of the target deceleration curve, the upper limit of the target impact degree, the braking force building pressure rate, and the braking energy recovery intensity.

[0018] The second aspect of the present invention provides a wire control braking system based on personalized feature learning and intention evolution prediction, and the system includes:

[0019] An information acquisition unit for acquiring the real-time braking operation information of the driver, the current state information of the vehicle, and the driving environment information of the vehicle;

[0020] A personalized feature management unit for acquiring or updating the personalized braking characteristic parameters of the driver;

[0021] An intention prediction unit configured with a driving intention recognition model and an intention trend prediction module integrated with an attention mechanism, for predicting the current braking intention type of the driver based on the dynamic characteristics of the real-time braking operation information and the personalized braking characteristic parameters, and predicting the short-term evolution trend of the current braking intention;

[0022] A braking characteristic decision unit for determining a set of target braking characteristic parameters according to the predicted current braking intention type and the short-term evolution trend, and combining the vehicle state and the driving environment;

[0023] A braking control execution unit for dynamically adjusting the control instruction of the wire control braking system based on the target braking characteristic parameters.

[0024] Preferably, the driving intention recognition model integrated with the attention mechanism adopted by the intention prediction unit is constructed based on a long short-term memory network and includes an attention layer for evaluating the importance of information at different time steps in the input sequence.

[0025] The third aspect of the present invention provides a vehicle, including a wire control braking system based on personalized feature learning and intention evolution prediction as described above.

[0026] Compared with the prior art, the present invention has the following advantages:

[0027] A wire control braking method based on personalized feature learning and intention evolution prediction of the present invention can more accurately predict the real braking intention of the driver by analyzing and learning the dynamic features of the driver's braking operation. According to the predicted intention, the braking characteristics are dynamically adjusted, which can make the braking process smoother and more natural, reduce unnecessary braking jerks or response lags. Especially in certain specific situations, the accurate prediction of the driver's intention can identify the initial characteristics of potential emergency braking intention, which helps the system enter the standby state faster or assist the driver to complete more effective braking operations, thereby indirectly improving driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the provided drawings.

[0029] Figure 1 is a flowchart of a wire control braking method based on personalized feature learning and intention evolution prediction of the present invention;

[0030] Figure 2 is an overall functional block diagram of a wire control braking system based on personalized feature learning and intention evolution prediction of the present invention;

[0031] Figure 3 is a detailed architecture diagram of a driving intention recognition model (Att-LSTM) integrated with an attention mechanism in the present invention;

[0032] Figure 4 is a relationship diagram between an intention evolution trend prediction module and a main intention recognition model in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The technical solution of the present invention will be further specifically described below through specific embodiments in combination with the accompanying drawings. It should be understood that the implementation of the present invention is not limited to the following embodiments, and any formal modification and / or change made to the present invention will fall within the protection scope of the present invention.

[0034] In the present invention, unless otherwise specified, all parts and percentages are in weight units, and the equipment and raw materials used can be purchased from the market or are commonly used in the art. The methods in the following embodiments are all conventional methods in the art unless otherwise specified. The components or equipment in the following embodiments are all general standard parts or components known to those skilled in the art, and their structures and principles can all be learned through technical manuals or obtained through conventional experimental methods by those skilled in the art.

[0035] The following will make a detailed description of the embodiments of the present invention in combination with the accompanying drawings. In the following detailed description, for the convenience of explanation, many specific details are elaborated to provide a comprehensive understanding of the embodiments of the present invention. However, one or more embodiments can also be implemented by those skilled in the art without these specific details.

[0036] The embodiments of the present invention disclose a wire control braking method based on personalized feature learning and intention evolution prediction, aiming to optimize the response characteristics of the wire control braking system by deeply understanding the personalized habits and dynamically changing braking intentions of drivers, and improve the riding experience and safety. This method is mainly executed by an in-vehicle controller, and typical in-vehicle controllers include a brake control ECU, an intelligent driving domain controller, or a central computing platform. As Figure 1 shown, this method includes the following steps: obtaining the real-time braking operation information of the driver, the current state information of the vehicle, and the driving environment information of the vehicle; obtaining or updating the personalized braking feature parameters of the driver; based on the dynamic features of the driver's real-time braking operation information, the driver's personalized braking feature parameters, and using a preset driving intention recognition model integrated with an attention mechanism, predicting the current braking intention type of the driver;

[0037] predicting the evolution trend of the current braking intention within a preset future time window based on the output of the driving intention recognition model or an independent intention trend prediction module;

[0038] determining a set of target braking characteristic parameters according to the predicted current braking intention type and evolution trend, and combining the current state information of the vehicle and the driving environment information of the vehicle;

[0039] dynamically adjusting the braking control command of the wire control braking system based on the target braking characteristic parameters to control the vehicle to execute braking.

[0040] An embodiment of the present invention also discloses a wire braking system based on personalized feature learning and intention evolution prediction, as Figure 2 shown. The system includes:

[0041] An information acquisition unit, configured to acquire real-time braking operation information of the driver, current state information of the vehicle, and driving environment information of the vehicle;

[0042] A personalized feature management unit, configured to acquire or update personalized braking feature parameters of the driver;

[0043] An intention prediction unit, configured with a driving intention recognition model and an intention trend prediction module integrated with an attention mechanism, for predicting the current braking intention type of the driver based on the dynamic features of the real-time braking operation information and the personalized braking feature parameters, and predicting the short-term evolution trend of the current braking intention;

[0044] A braking characteristic decision unit, configured to determine a set of target braking characteristic parameters according to the predicted current braking intention type and the short-term evolution trend, and in combination with the vehicle state and driving environment;

[0045] A braking control execution unit, configured to dynamically adjust the control instruction of the wire braking system based on the target braking characteristic parameters.

[0046] The following makes a detailed description of each step of this method.

[0047] The first step: Information acquisition and preprocessing

[0048] When a vehicle start or braking event is triggered, the information acquisition unit collects information related to braking intention judgment from multiple sources in real time:

[0049] 1. Real-time braking operation information of the driver

[0050] This information is obtained by collecting instantaneous data of the brake pedal through a high-precision brake pedal sensor at a preset sampling frequency. These data include pedal displacement, pedal stepping speed, pedal stepping acceleration, and pedal holding duration. The pedal displacement represents the current depression depth or angle of the pedal. The pedal stepping speed is obtained by taking the first-order difference of the pedal displacement or directly from a speed sensor, indicating how fast the pedal is depressed or released. The pedal stepping acceleration is obtained by taking the first-order difference of the pedal speed or the second-order difference of the displacement, indicating the degree of change in the pedal movement state. The pedal holding duration records the duration of the pedal in a certain displacement interval. These raw data are preprocessed through filtering and normalization to form time series features for subsequent model input. For example, the pedal displacement, speed, and acceleration sequences of the past N sampling points (such as N = 20, corresponding to 0.2 seconds) are used as dynamic feature inputs.

[0051] 2. Vehicle current state information

[0052] This information is obtained from the vehicle CAN bus or IMU, wheel speed sensors, and suspension sensors. The vehicle current state information includes vehicle speed, vehicle longitudinal acceleration and lateral acceleration, vehicle yaw rate, and vehicle load state. The vehicle load state is estimated through suspension height sensor data and the relationship between drive system torque and acceleration.

[0053] 3. Vehicle driving environment information

[0054] This information is obtained through in-vehicle ADAS sensors and the navigation system, and includes information on obstacles ahead, road gradient, road curvature, and navigation path information. Information on obstacles ahead, road gradient, and road curvature is obtained through in-vehicle ADAS sensors. Navigation path information is obtained through the navigation system.

[0055] Step 2: Acquisition and update of driver's personalized braking characteristic parameters

[0056] The personalized feature management unit is responsible for maintaining and updating the personalized braking characteristic parameters of the current driver. The personalized braking characteristic parameters are designed to quantify the driver's long-term braking habits and preferences, including average pedal depression depth, average pedal depression rate, deceleration preference coefficient, risk tolerance level, and comfort sensitivity. The average pedal depression depth represents the pedal displacement curve or characteristic points that the driver is accustomed to under different deceleration requirements. The average pedal depression rate represents the characteristic of the pedal depression speed curve that the driver is accustomed to. The deceleration preference coefficient characterizes the driver's preference for the establishment speed and maximum deceleration value of braking deceleration. For example, by analyzing the time and pedal input amount used by the driver to reach the target deceleration at different initial vehicle speeds in historical braking events. The risk tolerance level is evaluated by analyzing the TTC distribution that the driver is accustomed to maintaining in a following scenario and the decision-making behavior when facing a yellow light. The comfort sensitivity is analyzed by observing the fineness of the driver's pedal operation and the response to vehicle impact when braking at low speed or when smooth parking is required.

[0057] When the vehicle starts, if the system identifies the current driver, the personalized parameters previously stored by the driver are loaded. If it is a new driver or cannot be identified, a set of general default parameters is loaded. During the vehicle driving process, the system continuously monitors and records the driver's braking behavior and the corresponding vehicle state and environment information. The data collected regularly is statistically analyzed to update the mean, variance, or distribution characteristics of the above personalized parameters. A clustering algorithm is used to cluster the driver's braking segments to identify several typical braking patterns of the driver, and the characteristics of these patterns are used as part of the personalized parameters. These updated parameters are stored in non-volatile memory for subsequent use.

[0058] Step 3: Prediction of Current Braking Intention Type and Evolution Trend

[0059] The intention prediction unit receives real-time data from the information acquisition unit and personalized parameters from the personalized feature management unit, and outputs the current braking intention type and short-term evolution trend.

[0060] In this embodiment, a driving intention recognition model integrating an attention mechanism (abbreviated as Att-LSTM model) and an intention trend prediction module are deployed inside the intention prediction unit.

[0061] The Att-LSTM model aims to extract key information from complex input time series and combine driver personalized features to accurately judge the current braking intention. As Figure 3 shown, the architecture of the Att-LSTM model includes the following parts: input layer, feature embedding layer, sequence encoding layer, attention mechanism layer, feature fusion layer, and classification output layer.

[0062] The input layer includes dynamic feature input and static feature input. The dynamic feature input receives the time series of the preprocessed real-time braking operation information of the driver, the time series of the current vehicle state information, and the time series of the vehicle driving environment information. The static feature input receives the personalized braking feature parameter vector of the current driver.

[0063] The feature embedding layer maps some categorical or high-dimensional sparse input features to low-dimensional dense vector representations.

[0064] The sequence encoding layer uses one or more stacked LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit) networks to process the time series data of the dynamic feature input. Due to the internal gating mechanism of the LSTM / GRU unit, it can effectively capture long-term dependencies and context information in the time series.

[0065] The purpose of the attention mechanism layer is to enable the model to automatically learn and focus on the most relevant parts of the input sequence for more accurate intention judgment. In this embodiment, the self-attention mechanism is applied to the output sequence of the LSTM layer. The self-attention mechanism calculates the correlation score between each hidden state hi in the output sequence [h1, h2, ..., hT1] and all other hidden states hj, then obtains the attention weight ai through the Softmax function, and finally sums all the hidden states weighted by the weight to obtain a context vector C.

[0066] The feature fusion layer fuses the context vector C output by the attention layer with the static personalized braking feature parameter vector P. There are two fusion methods. The first is concatenation, followed by further non-linear transformation through one or more fully connected layers. The second is to use the personalized feature P to adjust the weights or biases of the subsequent classification layer, or directly correct the probability distribution of the basic intention.

[0067] The classification output layer receives the fused features, which consists of one or more fully connected layers. The last layer uses the Softmax activation function to output the probability distribution for each predefined current braking intention type, including emergency braking intention, conventional smooth deceleration intention, comfortable slow stop intention, curve pre-deceleration intention, and micro speed control intention. The category with the highest probability is selected as the prediction result of the current braking intention.

[0068] The intention trend prediction module shares some underlying networks with the Att-LSTM model (such as the context vector C output by the LSTM layer or the attention layer). The structure of the intention trend prediction module is a small feed-forward neural network (FFN), which is directly added as a branch network after the feature fusion layer.

[0069] As Figure 4 shown, the inputs of the intention trend prediction module include: the probability distribution or determined category of the current braking intention type output by the Att-LSTM model, the key feature representation inside the Att-LSTM model, the historical sequence of intention recognition in the recent period, key vehicle state and environmental information. Its outputs include: the probability of predicting that the current braking intention remains unchanged within a preset future time window, and the probability of transitioning to other preset braking intention types. Its output is a probability vector, which is used to represent the distribution of future intention states.

[0070] The driving intention recognition model integrated with the attention mechanism uses labeled data for supervised learning. The labeled data includes driver operation sequences, vehicle states, environmental information, personalized features, as well as corresponding true braking intention labels and intention evolution labels. Its loss function is designed as a multi-task loss. The main intention classification task uses cross-entropy loss, the predicted probability distribution also uses cross-entropy loss, and the predicted stability index uses mean squared error.

[0071] Step 4: Determination of target braking characteristic parameters

[0072] The braking characteristic decision unit determines a set of optimal target braking characteristic parameters according to the current braking intention type and evolution trend output by the intention prediction unit, and combines the current vehicle state information and driving environment information.

[0073] The target braking characteristic parameters include the shape parameters of the target deceleration curve, the upper limit of the target jerk, the pressure build-up rate of the braking force, and the intensity of braking energy recovery. The shape parameters of the target deceleration curve define the variation law of the expected deceleration with time or pedal displacement. For comfortable braking, it is an S-shaped or trapezoidal curve that rises gently at the beginning, is stable in the middle, and gradually decreases at the end. For emergency braking, it is a step-shaped curve that reaches the maximum value as soon as possible. The upper limit of the target jerk is the change rate of the deceleration, and limiting the jerk can improve comfort. Different intentions correspond to different upper limits of the jerk. The pressure build-up rate of the braking force refers to the speed of building the wheel cylinder pressure for a hydraulic braking system and the speed of building the clamping force for an electric motor-driven brake. The intensity of braking energy recovery is the proportion of regenerative braking in the total braking force or the maximum recovery torque.

[0074] The decision-making logic is based on the current intention type, considering the evolution trend and combining the vehicle state and environment.

[0075] The system will pre-store or calculate through functions the set of basic target braking characteristic parameters corresponding to different intention types internally. The indicators for the emergency braking intention are: the target deceleration curve reaches the maximum value as soon as possible, the upper limit of the jerk is relatively high, the pressure build-up rate is the fastest, and the intensity of energy recovery may be reduced to prioritize the frictional braking force, and the front-to-rear axle distribution is close to the EBD optimization limit. The indicators for the comfortable slow-stop intention are: the target deceleration curve is very gentle, the upper limit of the jerk is extremely low, the pressure build-up rate is slow, and the intensity of energy recovery is as large as possible to decelerate smoothly.

[0076] When considering the evolution trend, if it is predicted that the current intention will remain stable or tend to ease, more emphasis can be placed on smoothness and energy recovery efficiency. If it is predicted that the current intention will tend to stronger braking, even if the currently recognized intention has not been fully changed, the system can adjust some parameters in advance to make preparations. For example, slightly increase the upper limit of the pressure build-up rate, or reduce the response delay of regenerative braking, so that when the true intention is confirmed to change, the braking system can reach the required braking force faster.

[0077] Combining the vehicle state and environment specifically includes: at high vehicle speeds, even for comfortable braking, a relatively large initial deceleration may be required; at low vehicle speeds, more emphasis is placed on smoothness. When the vehicle is heavily loaded, a greater braking force is required to achieve the same deceleration, and the relevant parameters will be adjusted accordingly. When going uphill, gravity assists braking and the required braking force decreases; when going downhill, additional gravity needs to be overcome. On low-adhesion road surfaces, the maximum available deceleration is limited, and the jerk also needs to be more strictly controlled, and the intervention thresholds of ABS / TCS will be adjusted.

[0078] Step 5: Dynamically adjust the control commands of the electronically controlled braking system

[0079] The braking control execution unit receives the target braking characteristic parameters output by the braking characteristic decision unit and converts them into specific and real-time control instructions for each actuator of the line control braking system.

[0080] For an electro-hydraulic braking system (EHB), according to the target deceleration curve, the upper limit of the jerk, and the target pressure build-up rate, the target wheel cylinder pressure of each wheel is calculated through a closed-loop control algorithm, and components such as pumps and valves in the hydraulic control unit (HCU) are controlled to precisely adjust the flow rate and pressure of the brake fluid entering each wheel cylinder. At the same time, according to the regenerative braking coordination strategy, a target negative torque request is sent to the motor controller.

[0081] For an electro-mechanical braking system (EMB), according to the target deceleration curve, etc., the target clamping force or motor drive torque required for each wheel brake is calculated and directly controlled by the motor in the brake to execute precisely.

[0082] According to the determined energy recovery intensity and coordination strategy, the drive motor is controlled to enter the power generation state to recover braking energy. Ensure the smooth transition and synergistic effect of the regenerative braking force and the frictional braking force to achieve the total target deceleration.

[0083] The entire process composed of the above five steps is closed-loop and dynamic. As long as the driver's braking operation, vehicle state, or environmental information changes, the above steps will be re-executed, the intention will be re-evaluated, and the target braking characteristics and control instructions will also be updated in real time, so as to achieve continuous tracking and adaptive response to the driver's intention.

[0084] Specific examples are used in this article to elaborate on the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention. It should be noted that for those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A wire braking method based on personalized feature learning and intention evolution prediction, characterized in that: The method includes the following steps: obtaining real-time braking operation information of the driver, current state information of the vehicle, and driving environment information of the vehicle; obtaining or updating personalized braking characteristic parameters of the driver; predicting the current braking intention type of the driver based on the dynamic characteristics of the driver's real-time braking operation information, the driver's personalized braking characteristic parameters, and using a preset driving intention recognition model integrated with an attention mechanism; predicting the evolution trend of the current braking intention within a preset future time window based on the output of the driving intention recognition model or an independent intention trend prediction module; determining a set of target braking characteristic parameters according to the predicted current braking intention type and the evolution trend, and combining the current state information of the vehicle and the driving environment information of the vehicle; dynamically adjusting the braking control instruction of the electronic stability program based on the target braking characteristic parameters to control the vehicle to perform braking.

2. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, wherein: The personalized braking characteristic parameters of the driver include at least one of the average pedal depression depth, average pedal depression rate, preference coefficient for different deceleration rates, and risk tolerance level statistically obtained from the driver's historical braking behavior.

3. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, wherein: The driving intention recognition model integrated with the attention mechanism is constructed based on a recurrent neural network, a long short-term memory network, or a gated recurrent unit, and the attention mechanism is used to assign higher weights to the parts more relevant to intention judgment and evolution trend prediction when processing the time series and auxiliary information of the real-time braking operation information.

4. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, characterized in that: Predicting the current braking intention type of the driver based on the dynamic characteristics of the driver's real-time braking operation information, the driver's personalized braking characteristic parameters, and using a preset driving intention recognition model integrated with an attention mechanism specifically includes: taking the dynamic characteristics of the real-time braking operation information as the first input and the driver's personalized braking characteristic parameters as the second input, and jointly inputting them into the driving intention recognition model; alternatively, first obtaining a preliminary intention using the dynamic characteristics of the real-time braking operation information through a basic intention recognition module, and then correcting the preliminary intention using the driver's personalized braking characteristic parameters.

5. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, characterized in that: Predicting the evolution trend of the current braking intention within a preset future time window includes predicting the probability that the current braking intention remains unchanged or the probability of changing to other preset braking intention types.

6. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, wherein: The current braking intention type includes one or a combination of an emergency braking intention, a conventional smooth deceleration intention, a comfortable slow stop intention, a corner pre-deceleration intention, and a micro speed control intention.

7. The wire braking method based on personalized feature learning and intention evolution prediction according to claim 1, wherein: The target braking characteristic parameters include at least one of the morphological parameters of the target deceleration curve, the upper limit of the target jerk, the braking force build-up rate, and the braking energy recovery intensity.

8. A wire braking system based on personalized feature learning and intention evolution prediction, characterized in that: The system includes: an information acquisition unit for obtaining real-time braking operation information of the driver, current state information of the vehicle, and driving environment information of the vehicle; a personalized feature management unit for obtaining or updating personalized braking characteristic parameters of the driver; An intention prediction unit, configured with a driving intention recognition model integrating an attention mechanism and an intention trend prediction module, is used to predict the current braking intention type of the driver based on the dynamic features of the real-time braking operation information and the personalized braking characteristic parameters, and predict the short-term evolution trend of the current braking intention; A braking characteristic decision unit is used to determine a set of target braking characteristic parameters according to the predicted current braking intention type and the short-term evolution trend, and in combination with the vehicle state and the driving environment; A braking control execution unit is used to dynamically adjust the control instruction of the wire braking system based on the target braking characteristic parameters.

9. The wire control braking system based on personalized feature learning and intention evolution prediction according to claim 8, characterized in that: The driving intention recognition model integrating the attention mechanism adopted by the intention prediction unit is constructed based on a long short-term memory network and includes an attention layer for evaluating the importance of information at different time steps in the input sequence.

10. A vehicle, characterized in that: The vehicle includes a wire braking system based on personalized feature learning and intention evolution prediction as described in claim 8 or 9.

Images