A method for active adjustment of brake pedal feel

Abstract

The application provides a kind of line control brake pedal feeling active adjustment method, comprising the following steps: S1, the brake operation data of driver and the dynamic response data of vehicle are acquired in real time;S2, based on the brake operation data and the dynamic response data, a multi-dimensional driving habit image is constructed for representing driver's braking preference, the multi-dimensional driving habit image includes response preference dimension and intensity preference dimension;S3, based on the multi-dimensional driving habit image, adjust the pedal feeling parameter of line control brake system, the pedal feeling parameter includes the damping characteristic of pedal force simulator and the pedal stroke-brake force gain curve in brake controller.The application changes the fixed passive pedal feeling setting in prior art into intelligent adjustment mode of active learning and adaptation, and realizes the high personalization of pedal feeling.

Description

Technical Field

[0001] This invention relates to a method for use in a brake-by-wire system, specifically a method for actively adjusting the feel of the brake-by-wire pedal, belonging to the field of automotive intelligent control technology. Background Technology

[0002] With the evolution of automotive electronic and electrical architecture, brake-by-wire systems are gradually becoming standard equipment in modern cars, especially new energy vehicles. In brake-by-wire systems, there is no rigid mechanical or hydraulic connection between the driver's brake pedal and the brake actuator; the pedal displacement or pressing force is interpreted as an electrical signal by sensors. This decoupled structure brings significant advantages to precise control of braking force and energy recovery, while also allowing for flexible design of the brake pedal feedback force using a pedal force simulator.

[0003] However, existing pedal feel designs typically employ fixed or limited modular strategies. At the factory, the pedal feel characteristics (such as the relationship between damping, travel, and braking force) are set to one or more preset modes, such as comfort mode and sport mode, for the driver to manually select. This design ignores the differences in driving style, height, weight, and handling preferences among different drivers. For example, some drivers prefer gentle, linear braking, while others prefer a shorter travel and faster response. This design requires the driver to passively adapt to the vehicle's preset pedal feel, rather than allowing the vehicle to actively adapt to the driver's personalized needs. This "driver adapts to car" approach, to some extent, affects driving comfort. Summary of the Invention

[0004] Based on the above background, the purpose of this invention is to provide a method for actively adjusting the brake pedal feel of drive-by-wire, providing drivers with a personalized, comfortable and controllable braking experience.

[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0006] A method for actively adjusting the feel of a brake-by-wire pedal includes the following steps:

[0007] S1. Real-time acquisition of driver braking operation data and vehicle dynamic response data;

[0008] S2. Based on the braking operation data and the dynamic response data, construct a multi-dimensional driving habit profile to characterize the driver's braking preferences. The multi-dimensional driving habit profile includes a response preference dimension and a force preference dimension.

[0009] S3. Based on the multi-dimensional driving habit profile, adjust the pedal feel parameters of the brake-by-wire system. The pedal feel parameters include the damping characteristics of the pedal force simulator and the pedal travel-braking force gain curve in the brake controller.

[0010] Preferably, step S1 further includes: dividing the continuous braking operation data stream into independent braking event segments, and performing feature engineering on each braking event segment to extract braking features for constructing the multi-dimensional driving habit profile.

[0011] Preferably, the braking characteristics include the maximum pedal angular velocity and the time to reach the maximum braking force for characterizing the response preference dimension, and the peak pedal force and average pedal force for characterizing the force preference dimension.

[0012] Preferably, in step S2, the braking event is divided into different driving conditions by using a fuzzy clustering algorithm, and the braking features are statistically analyzed under each driving condition to generate the multi-dimensional driving habit profile.

[0013] Preferably, the fuzzy clustering algorithm clusters the driving conditions based on the vehicle's speed, longitudinal acceleration, and following distance data.

[0014] Preferably, in step S3, the multi-dimensional driving habit profile is mapped to target pedal feel parameters using a pre-trained neural network model, and the pedal force simulator and the pedal travel-braking force gain curve are controlled to smoothly transition to the target pedal feel parameters.

[0015] Preferably, the neural network model is a multilayer perceptron model with a parallel processing structure. The multilayer perceptron model includes a conditional embedding branch, a preference processing branch, and a fusion output module. The conditional embedding branch is used to receive the driving conditions and convert them into conditional feature vectors. The preference processing branch is used to receive the multi-dimensional driving habit profile and process it into preference feature vectors. The fusion output module is used to fuse the conditional feature vectors and the preference feature vectors, and map the fused result to the target pedal feel parameter.

[0016] The multilayer perceptron model with the aforementioned specific parallel processing structure processes preference and operating condition information separately and then intelligently fuses them, enabling the model to learn how to differentiate and satisfy the same driving preference under different operating conditions, thereby improving the intelligence level and robustness of the adjustment strategy.

[0017] Preferably, the smooth transition is achieved through a PID controller, which uses the target pedal feel parameter output by the neural network model as the set value and controls the current pedal feel parameter as a process variable.

[0018] Preferably, the target pedal feel parameters include a set of pedal damping coefficient values ​​for defining the damping characteristics, and a set of key points for defining the pedal travel-braking force gain curve.

[0019] Compared with the prior art, the present invention has the following advantages:

[0020] The present invention provides a method for actively adjusting the pedal feel of brake-by-wire. By collecting driver operation data in real time and combining it with vehicle dynamic conditions to construct a multi-dimensional profile of the driver's braking habits, a specially designed neural network model is used to map the profile into pedal feel control parameters. This transforms the fixed passive pedal feel setting in the prior art into an intelligent adjustment mode that actively learns and adapts, thereby achieving a highly personalized pedal feel. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating an active adjustment method for the feel of a brake-by-wire pedal according to the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of the multilayer perceptron model with parallel processing in this invention;

[0024] Figure 3 These are the target pedal feel parameters output by the multilayer perceptron model in this invention for driver A and driver B respectively under urban congestion conditions;

[0025] Figure 4 These are the target pedal feel parameters output by the multilayer perceptron model in this invention to driver A and driver B respectively under high-speed cruise conditions. Detailed Implementation

[0026] The technical solution of the present invention will be further described in detail below through specific embodiments and in conjunction with the accompanying drawings. It should be understood that the implementation of the present invention is not limited to the following embodiments, and any modifications and / or alterations made to the present invention will fall within the protection scope of the present invention.

[0027] In this invention, unless otherwise specified, all parts and percentages are by weight, and the equipment and raw materials used are commercially available or commonly used in the art. Unless otherwise specified, the methods in the following embodiments are conventional methods in the art. Unless otherwise specified, the components or equipment in the following embodiments are general standard parts or components known to those skilled in the art, and their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0028] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. In this detailed description, numerous specific details are set forth to facilitate explanation and provide a thorough understanding of the embodiments of the present invention. However, one or more embodiments may be practiced by those skilled in the art without these specific details.

[0029] An embodiment of the present invention discloses a method for actively adjusting the feel of a brake-by-wire pedal, comprising the following steps:

[0030] S1. Real-time acquisition of driver braking operation data and vehicle dynamic response data;

[0031] S2. Based on braking operation data and the dynamic response data, construct a multi-dimensional driving habit profile to characterize the driver's braking preferences. The multi-dimensional driving habit profile includes a response preference dimension and a force preference dimension.

[0032] S3. Based on a multi-dimensional driving habit profile, adjust the pedal feel parameters of the brake-by-wire system. The pedal feel parameters include the damping characteristics of the pedal force simulator and the pedal travel-braking force gain curve in the brake controller.

[0033] The following is for reference. Figure 1 Each step is explained in detail.

[0034] Step S1: Real-time data acquisition and preprocessing

[0035] During vehicle operation, the controller collects a series of data reflecting the driver's intentions and vehicle status in real time through the vehicle's CAN bus and high-frequency sensor network.

[0036] The data collection types are as follows:

[0037] Braking operation data: high-precision pedal travel (directly reflects the depth of the driver's pedal), pedal force / pressure (reflects the force applied by the driver), and pedal angular velocity (obtained from the differential of the travel, reflecting how fast the driver presses the pedal).

[0038] Dynamic response data: vehicle speed, longitudinal / lateral acceleration, vehicle impact (rate of change of acceleration, reflecting braking smoothness), and following distance to the vehicle in front (provided by radar / camera).

[0039] The controller segments the continuous data stream into independent braking event segments. For example, all data from the moment the driver's foot touches the pedal (travel > 0.1 mm) until it is fully released (travel < 0.1 mm) constitutes an independent braking event. The advantage of this is that it transforms ambiguous continuous behavior into discrete samples with clear start and end points that can be analyzed.

[0040] For each braking event segment, the controller calculates a series of quantified braking characteristics as follows.

[0041] Response preference-related characteristics:

[0042] Maximum pedal angular velocity: reflects the explosive force of the driver's pedaling action.

[0043] Time to reach maximum braking force: reflects the speed at which the driver expects the braking force to be established.

[0044] Strength preference related characteristics:

[0045] Peak pedal force: Reflects the maximum force that a driver is accustomed to applying during a single braking action.

[0046] Average pedal force: reflects the driver's average force application habits throughout the braking process.

[0047] Step S2: Construct a multi-dimensional driving habit profile

[0048] The purpose of this step is to model the driver's long-term behavior based on an understanding of the current driving scenario.

[0049] The controller employs a fuzzy clustering algorithm to categorize the current driving condition into several preset classes based on data such as vehicle speed, longitudinal acceleration, and following distance. Unlike hard clustering, fuzzy clustering allows an event to belong to multiple conditions simultaneously with varying degrees of membership. For example, a deceleration process might resemble 70% ordinary road conditions and 30% urban traffic congestion, which better reflects the complexity and transitions of real-world driving scenarios.

[0050] Typical operating conditions include:

[0051] Condition A: Urban congestion, characterized by vehicle speed <20km / h, high-frequency acceleration and deceleration, and following distance <10m.

[0052] Operating Condition B: High-speed cruising, characterized by vehicle speed >80km / h, stable speed, and following distance >50m.

[0053] Condition C: Ordinary road, between A and B.

[0054] After identifying the current operating condition, the controller performs a long-term weighted average of the feature values ​​of the braking events that occur under that condition to form a stable profile.

[0055] For example, suppose the system currently identifies the vehicle as being in urban congestion. Driver A habitually uses intermittent braking to maintain distance. The system records 100 braking events under this condition, calculating the average time to reach maximum braking force as 0.25 seconds and the average peak pedal force as 80 N. Driver B habitually brakes and accelerates abruptly. The system records 100 braking events under this condition, calculating the average time to reach maximum braking force as 0.5 seconds and the average peak pedal force as 150 N.

[0056] The system normalizes these statistical values, mapping them to the 0-1 interval, ultimately forming a profile vector. Driver A's profile is [Response Preference: 0.9, Intensity Preference: 0.2], while Driver B's profile is [Response Preference: 0.4, Intensity Preference: 0.8]. These profiles are dynamically updated to reflect long-term style changes in drivers.

[0057] Step S3: Based on the image, adaptively adjust the pedal feel parameters.

[0058] The purpose of this step is to translate the abstract image into control over the hardware.

[0059] The mapping from image to parameters is accomplished through a specially designed neural network model. Conventional MLP models, which simply flatten all inputs and process them into a single network, essentially attempt to fit all situations with a single, overarching function. This structure struggles to reflect, at the model level, the moderating or constraining effects of driving conditions on driving preferences. Therefore, it fails to learn the distinctly different physical implications of high-response preference under urban congestion and highway cruising conditions.

[0060] To overcome the above limitations, this invention employs a multilayer perceptron model with a parallel processing structure. (Reference) Figure 2 The multilayer perceptron model comprises the following core components.

[0061] The input to the conditional embedding branch is a discrete working condition ID, which contains an embedding layer that transforms a simple numeric ID into a high-dimensional conditional feature vector (V_cond) containing rich working condition semantics.

[0062] The input to the preference processing branch is a multi-dimensional driving habit profile, which consists of multiple fully connected layers. It is used to learn the basic mapping relationship from the driver's general preferences to general pedal feel features, and finally outputs a preference feature vector (V_pref).

[0063] The fusion output module receives V_cond and V_pref from the two parallel branches mentioned above. It first passes the vector through a fusion layer, then feeds the fused vector into a series of subsequent fully connected layers. Finally, the output layer maps the processed information into a set of hardware-executable target pedal feel parameters.

[0064] Based on the aforementioned profiles of driver A and driver B, the working mechanism of this step will be further explained.

[0065] In urban congestion conditions (condition ID=1), the conditional embedding branch outputs V_cond representing congestion, and the preference processing branch outputs V_pref representing the driver's general preference. For driver A, the fusion output module combines the information of the condition requiring sensitive response and the driver's preference for fast response, and the final output target pedal feel parameters are: a pedal travel-braking force gain curve with a very steep initial slope, and damping characteristics with low initial damping. Correspondingly, the system outputs another set of target pedal feel parameters for driver B. (Reference) Figure 3 In the graph, the darker curves represent the pedal travel-braking force gain curve and damping characteristic curve for driver A, while the lighter curves represent those for driver B. It can be seen that in this scenario, the system provides driver A with a highly sensitive initial response to handle frequent starts and stops, while providing driver B with a slightly slower but more forceful curve in the later stages, and increasing overall damping to suppress sudden braking tendency and improve comfort.

[0066] Under high-speed cruising conditions (condition ID=2), the conditional embedding branch outputs V_cond, representing high speed. For driver A, the fusion output module combines the information of a stable and linear condition and a driver who prefers a fast response. The final output target pedal feel parameters are: a more linear pedal travel-braking force gain curve and stronger, more stable overall damping characteristics. Correspondingly, the system outputs another set of target pedal feel parameters for driver B. (Reference) Figure 4 In the graph, the darker curves represent driver A's pedal travel-braking force gain curve and damping characteristic curve, while the lighter curves represent driver B's. It can be seen that in this scenario, the system prioritizes stability and predictability. For both drivers, the system outputs a very linear braking force gain curve for accurate control. Simultaneously, the system significantly increases the overall pedal damping, providing stronger damping, especially for driver B who is accustomed to sudden braking and acceleration.

[0067] Therefore, this method transforms the fixed passive pedal feel setting in the existing technology into an intelligent adjustment mode that actively learns and adapts, thereby achieving a highly personalized pedal feel.

[0068] Furthermore, to ensure a seamless and comfortable driving experience, a PID controller is used to achieve smooth parameter transitions. For example, when the target damping coefficient output by the multilayer perceptron model is 1.2 N·m / s and the current value is 1.0 N·m / s, the PID controller generates a series of smooth control commands to the motor of the pedal force simulator, so that the target damping coefficient increases to 1.2 N·m / s imperceptibly in a short time, rather than a sudden jump.

[0069] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A method for actively adjusting the feel of a brake-by-wire pedal, characterized in that: The active adjustment method for the feel of the brake pedal includes the following steps: S1. Real-time acquisition of driver's braking operation data and vehicle dynamic response data; segmenting the continuous braking operation data stream into independent braking event segments, and performing feature engineering on each braking event segment to extract braking features for constructing a multi-dimensional driving habit profile. S2. Based on the braking operation data and the dynamic response data, a multi-dimensional driving habit profile is constructed to characterize the driver's braking preference. The multi-dimensional driving habit profile includes a response preference dimension and a force preference dimension. The braking events are divided into different driving conditions by a fuzzy clustering algorithm, and the braking features are statistically analyzed under each driving condition to generate the multi-dimensional driving habit profile. S3. Based on the multi-dimensional driving habit profile, adjust the pedal feel parameters of the brake-by-wire system. The pedal feel parameters include the damping characteristics of the pedal force simulator and the pedal travel-braking force gain curve in the brake controller. In step S3, the multi-dimensional driving habit profile is mapped to a target pedal feel parameter using a pre-trained neural network model. The pedal force simulator and the pedal travel-braking force gain curve are then smoothly transitioned to the target pedal feel parameter. The neural network model is a multilayer perceptron model with a parallel processing structure. The multilayer perceptron model includes a conditional embedding branch, a preference processing branch, and a fusion output module. The conditional embedding branch receives the driving conditions and converts them into conditional feature vectors. The preference processing branch receives the multi-dimensional driving habit profile and processes it into preference feature vectors. The fusion output module fuses the conditional feature vectors and the preference feature vectors, and maps the fused result to the target pedal feel parameter. The target pedal feel parameter includes a set of pedal damping coefficient values ​​for defining the damping characteristics, and a set of braking force gain key points for defining the pedal travel-braking force gain curve. The smooth transition is achieved through a PID controller. The PID controller uses the target pedal feel parameter output by the neural network model as a setpoint and controls the current pedal feel parameter as a process variable.

2. The method for actively adjusting the feel of a brake-by-wire pedal according to claim 1, characterized in that: The braking characteristics include the maximum pedal angular velocity and the time to reach maximum braking force for characterizing the response preference dimension, and the peak pedal force and average pedal force for characterizing the force preference dimension.

3. The method for actively adjusting the feel of a brake-by-wire pedal according to claim 1, characterized in that: The fuzzy clustering algorithm clusters the driving conditions based on vehicle speed, longitudinal acceleration, and following distance data.

Images