An electro-hydraulic wire-controlled braking system based on CPS architecture and its energy optimization method

Abstract

This invention discloses an electro-hydraulic drive-by-wire braking system based on a CPS architecture and an energy optimization method. The method includes: acquiring driver input data and vehicle driving condition data; processing the driver input data and vehicle driving condition data; performing driving condition prediction based on the processed data to obtain a rolling time-domain driving condition prediction result; solving the overall objective function of the electro-hydraulic hybrid braking system based on the obtained driving condition prediction result to obtain the optimal braking strategy; and performing vehicle braking according to the optimal braking strategy to complete the energy optimization of the braking system. This invention performs braking based on the current vehicle state and driving condition prediction, improves the energy optimization strategy, achieves reasonable braking torque distribution, further reduces the energy consumption of the electro-hydraulic hybrid braking system, and achieves energy saving and emission reduction throughout the vehicle's entire life cycle.

Description

Technical Field

[0001] This invention belongs to the field of vehicle brake-by-wire technology, specifically relating to an electro-hydraulic brake-by-wire system based on CPS architecture and an energy optimization method. Background Technology

[0002] As the automotive industry moves towards intelligent and electronic systems, electro-hydraulic hybrid braking systems are widely used in modern vehicles. These systems eliminate mechanical connections, employing electronic controllers, integrated sensors, and actuators. The vehicle's brake pedal signal is transmitted to the electronic controller, and the control signal is then transmitted to the actuator as an electrical signal.

[0003] Traditional braking systems, limited by their inherent mechanical-hydraulic transmission methods, have significant shortcomings in dynamic response and energy recovery efficiency. Compared to traditional braking systems, electro-hydraulic hybrid braking systems improve the braking system's response speed and achieve effective braking energy recovery. During low-intensity braking, the electronic control unit controls the inverter to recover braking energy; while during high-intensity braking, the hydraulic braking system intervenes to ensure braking response and safety requirements.

[0004] Due to various physical factors involved in vehicle braking, the braking decisions made by the controller are affected, resulting in low braking energy recovery efficiency and poor vehicle ride comfort. The brake-by-wire system, combined with a CPS (Cyber-Physical System) architecture, collects physical information such as driver input data, vehicle driving data, and road condition data through sensors. This collected data is then uploaded to the control system, which combines the driver input with the vehicle's driving conditions to make reasonable braking decisions, thereby improving the braking system's energy recovery efficiency and enhancing vehicle ride comfort.

[0005] Chinese invention patent application CN202410154776.2, entitled "A Braking Energy Recovery System and its Working Method for a Pure Electric Vehicle," proposes connecting a motor module to a transmission and a high-frequency switching circuit. During braking, the braking torque is transmitted to the motor module via the transmission, and the motor module connects to the high-frequency switching circuit to generate an electromotive force to achieve energy recovery. However, due to the long energy transmission path and the presence of friction, the energy recovery efficiency is low, and the motor braking effect is incomplete. Chinese invention patent application CN202210721810.0, entitled "An Electro-hydraulic Composite Braking Force Distribution Method for Regenerative Braking Energy Recovery in Electric Vehicles," proposes collecting vehicle braking intensity and battery SOC data, and distributing the braking force of the electro-hydraulic composite braking system based on these data. However, this method only collects partial data from the braking system and does not incorporate a CPS architecture to collect key physical information such as vehicle driving information, road condition information, and driver input. The existing braking strategies have low braking energy recovery efficiency and poor vehicle ride comfort.

[0006] Currently, there is no publicly available technical information on energy optimization in the field of electro-hydraulic braking by combining cyber-physical systems with driving condition prediction. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, the present invention aims to provide an electro-hydraulic brake-by-wire system and energy optimization method based on a CPS architecture. This system achieves a reasonable braking torque distribution strategy, solving the problems of excessive braking energy consumption and the inability to integrate road traffic information physics for energy prediction and optimization in the prior art. The present invention can comprehensively consider driving intention, battery status, and vehicle status to achieve energy optimization of the braking system.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] The present invention provides an electro-hydraulic brake-by-wire system based on a CPS architecture, comprising: a brake-by-wire unit and a brake optimization management unit, which are communicatively connected;

[0010] The brake-by-wire unit is used to collect vehicle information and realize the vehicle's braking state according to the braking control command sent by the braking optimization management unit. It includes: a driver input module, a braking execution module, and an environmental perception module.

[0011] The driver input module is used to collect braking, driving, and steering data input by the driver.

[0012] The environmental perception module is used to collect operating condition data of the vehicle.

[0013] The braking execution module is used to receive braking control commands sent by the braking optimization management unit to realize vehicle braking;

[0014] The braking optimization management unit is used to predict driving conditions based on the driver input data and vehicle driving condition data sent by the brake-by-wire unit, generate braking control commands, and send the braking control commands to the brake-by-wire unit.

[0015] Furthermore, the driver input module includes: a brake pedal sensor, an accelerator pedal sensor, a steering wheel angle sensor, and a torque sensor.

[0016] Furthermore, the environmental perception module includes: a vehicle speed sensor, a vehicle acceleration sensor, a yaw rate sensor, a road condition detection device, and a battery status sensor; the vehicle's operating condition data includes: vehicle speed, vehicle lateral acceleration, vehicle longitudinal acceleration, vehicle yaw rate, battery SOC status, and road curvature.

[0017] Furthermore, the braking execution module is used to implement vehicle braking, and it receives braking control commands sent by the braking optimization management unit via the CAN bus.

[0018] Furthermore, the braking execution module includes: a hydraulic wheel cylinder, a solenoid valve, an electric hydraulic pump, a coordinating controller, an electronic booster, a hub motor, an inverter, a power battery, and a brake disc; the hydraulic wheel cylinder is equipped with a solenoid valve at the hydraulic oil inlet for hydraulic pressure control; the hydraulic wheel cylinder and the brake disc are mechanically connected; the coordinating controller is connected to the electric hydraulic pump, the solenoid valve, and the hub motor via a control bus; the electronic booster is connected to the brake pedal via a wiring harness; the hub motor is connected to the wheel via a coupling to drive the wheel to rotate; when the vehicle brakes, the kinetic energy of the hub motor is converted into electrical energy by the inverter and stored in the power battery to achieve regenerative braking; at the same time, the electric hydraulic pump moves according to the control command, and the solenoid valve changes its opening to build up pressure in the hydraulic wheel cylinder, causing the brake pads to press against the brake disc for friction braking.

[0019] Furthermore, the braking optimization management unit includes: a data processing module and a decision optimization module;

[0020] The data processing module is used to process the driver input data and vehicle driving condition data sent by the brake-by-wire unit;

[0021] The decision optimization module receives the data processed by the data processing module, calculates the operating condition prediction results, and obtains the optimal braking strategy based on the operating condition prediction results.

[0022] Furthermore, the data processing module uses smoothing filtering and correlation analysis to process the data.

[0023] Furthermore, the optimal braking strategy includes: optimal braking energy recovery strategy, optimal vehicle ride comfort strategy, optimal tracking braking torque strategy, and optimal regenerative braking power strategy;

[0024] The optimal braking energy recovery strategy is to maximize the energy recovery value achieved by the hub motor through the inverter;

[0025] The optimal vehicle ride comfort strategy is to minimize the change in braking torque and the vehicle bumps.

[0026] The optimal tracking braking torque strategy achieves the best vehicle controllability and the shortest time to reach the target torque.

[0027] The optimal regenerative braking power strategy is to ensure that the braking torque of the hub motor does not exceed the maximum torque that the motor can generate, and that the power battery is in the best condition.

[0028] Furthermore, the driving condition prediction uses an encoder decoupled prediction network (ED-TFN) based on a temporal Transformer model, which is implemented based on a self-attention mechanism. It calculates the vehicle driving condition at future times based on the driver input data sent by the data processing module and the vehicle driving condition data, thereby realizing the prediction of the vehicle driving condition.

[0029] This invention also provides an energy optimization method for an electro-hydraulic wire-controlled braking system based on a CPS architecture. Based on the aforementioned system, the method steps are as follows:

[0030] 1) Obtain the driver's input data and the vehicle's operating condition data;

[0031] 2) Perform data processing on the driver's input data and the vehicle's operating condition data;

[0032] 3) Based on the processed data, predict the operating conditions to obtain the rolling time-domain driving condition prediction results;

[0033] 4) Solve the overall objective function of the electro-hydraulic composite braking based on the obtained working condition prediction results to obtain the optimal braking strategy;

[0034] 5) Perform vehicle braking according to the optimal braking strategy to complete the energy optimization of the braking system.

[0035] Furthermore, the driver input data in step 1) includes: the current steering wheel angle θ and torque T, brake pedal travel, and accelerator pedal travel; the vehicle's operating condition data includes vehicle speed v and vehicle lateral acceleration a. x Vehicle longitudinal acceleration a y yaw rate ω of the vehicle r Battery SOC status, road curvature ρ.

[0036] Furthermore, the data processing method in step 2) is smoothing filtering and correlation analysis;

[0037] The smoothing filter process removes noise from the dataset, and the expression is as follows:

[0038]

[0039] Among them, y t-n To smooth out the driver input data and vehicle operating condition data before processing, y t The driver input dataset and the vehicle driving condition dataset are obtained after smoothing, where n is the number of filter terms;

[0040] The correlation analysis is a Pearson correlation coefficient analysis, which screens out key parameters that are highly correlated with braking torque, simplifying the calculation of driving condition prediction. The smoothed driver input dataset and the vehicle driving condition dataset are used to perform correlation analysis, as shown in the following expression:

[0041]

[0042] Where X represents the smoothed driver input dataset and the vehicle driving condition dataset, Y represents the braking torque, and ρ represents the braking torque. X,Y Let σ be the Pearson correlation coefficient, cov(X,Y) be the covariance of X and Y, and σ be the Pearson correlation coefficient. X and σ Y Let μ be the standard deviation of X and Y, respectively. X and μ Y X and Y are the average values, respectively, and E is the expected value. After correlation analysis, it was found that the vehicle speed v has the strongest correlation with the braking torque. The dataset of vehicle speed v is used as the input data for the working condition prediction.

[0043] Furthermore, the condition prediction in step 3) uses an encoder decoupling prediction network based on a temporal Transformer model, which embeds each variable into a variable token and takes the time step as a feature of a single variable as input; the prediction network includes an embedding layer, a linear layer, a normalization layer, and a feedforward neural network.

[0044] Furthermore, the specific steps for predicting operating conditions in step 3) are as follows:

[0045] 31) The embedding layer maps the processed vehicle speed dataset to obtain the mapping vector corresponding to the vehicle speed v dataset.

[0046] 32) The linear layer captures the correlation between the vehicle speed v dataset mapping vector after the embedding layer and the driving conditions at future times, thus obtaining the driving condition prediction feature vector.

[0047] 33) Normalize the working condition prediction feature vector obtained from the linear layer through a normalization layer;

[0048] 34) The normalized operating condition prediction feature vector is decoded by a feedforward neural network to obtain the vehicle driving condition prediction result.

[0049] Furthermore, step 31) is implemented based on a multilayer perceptron (MLP), performing feature mapping on the entire historical observation sequence for each variable, as shown in the following expression:

[0050] H 0 =MLP(V 0 )

[0051] Among them, H 0 V is a mapping vector. 0 This is the vehicle speed dataset obtained after smoothing.

[0052] Furthermore, step 32) is implemented based on a multi-head attention mechanism, which consists of multiple self-attention modules. Each self-attention module converts the input vector into a query vector, a key vector, and a value vector, and uses the Softmax function to generate a probability distribution based on the pairwise similarity scores between vectors. The expression is:

[0053]

[0054] Here, Attention(Q,K,V) is the comprehensive information extracted from the driver input dataset and the vehicle driving condition dataset. Q is the query vector, K is the key vector, and V is the value vector. The query vector represents the direction or focus of information search, the key vector determines the information that the query vector should focus on, and the value vector contains the actual information to be output. The final output is obtained by weighted summation.

[0055] Furthermore, in step 33), the mean and standard deviation of all features are calculated to improve network convergence and training stability. The process expression is as follows:

[0056]

[0057]

[0058]

[0059] Output = γ × H norm +β

[0060] Where, μ l The mean, Let n be the i-th neuron node of the l-th encoding module in the input. l Let l represent the number of neuron nodes in the l-th encoding module, where l∈{1,2,...,L} represents the l-th encoding module. For variance, H norm H represents the normalized value of the vehicle velocity v mapping vector, where γ and β are the learnable scaling and offset parameters, respectively. l is the input variable of the l-th encoding module, and Output is the output of the normalization layer, which includes the vehicle speed v mapping vector.

[0061] Furthermore, the feedforward neural network in step 34) has a two-layer fully connected network with an activation function sandwiched in between. The activation function is a Gaussian error linear unit (GELU), and its expression is:

[0062]

[0063] In the formula, GELU is the Gaussian error linear unit activation function, and x is the function input variable.

[0064] Furthermore, the feedforward neural network encodes the observed time series through stacked Transformer modules and decodes the representation of future sequences using dense nonlinear connections to obtain the output, expressed as:

[0065] H l =LayerNorm(H l-1 +Feed-Forward(H l-1 ))

[0066] Among them, H l H is the output of the current layer. l-1 The output is the previous layer's output. LayerNorm is the layer normalization function, and Forward is the feedforward layer function, which receives the input and transforms it through a series of transformations to generate the output.

[0067] Furthermore, the future sequence is decoded by inputting the output of the feedforward neural network into a multilayer perceptron for mapping, as expressed by:

[0068]

[0069] in, For the operating condition prediction results, H L This is the output of the feedforward neural network.

[0070] Furthermore, the optimal braking strategy in step 4) includes: optimal braking energy recovery strategy, optimal vehicle ride comfort strategy, optimal tracking braking torque strategy, and optimal regenerative braking power strategy.

[0071] The objective of the optimal braking energy recovery strategy is to maximize the energy recovered during vehicle deceleration. The objective function for the braking energy recovery rate is expressed as follows:

[0072]

[0073] Among them, J bra_rec For total recovered energy, T mot_ij ω is the motor torque. ij v is the angular velocity of the motor. i0 v is the vehicle speed at the start of braking. ik η is the vehicle speed at the end of braking. d For braking system efficiency, η m_reg Where f is the regenerative braking efficiency of the motor, m is the rolling resistance coefficient, and C is the vehicle mass.D Where A is the air resistance coefficient, A is the vehicle's frontal area, and t is the air resistance coefficient. ik t is the braking end time. i0 This is the braking start time;

[0074] The objective of the optimal vehicle ride comfort strategy is to achieve the best ride comfort during braking. The objective function for ride comfort is expressed as follows:

[0075]

[0076] Among them, J bra_f Let u(k+i|k) be the target value for vehicle ride comfort, where u(k+i|k) represents the vehicle speed at the current time step k as the initial value and u(k+i-1 ...

[0077] The objective of the optimal tracking braking torque strategy is to ensure that the actually applied braking torque can accurately track and meet the desired braking torque setpoint. The objective function expression for the tracking braking torque is:

[0078]

[0079] Among them, J bra_T For tracking purposes, T if (k+i|k) represents the front axle braking torque at the current time step k, with the current time step k as the initial value, and T representing the front axle braking torque at the i-th future time step. jr (k+i|k) represents the rear axle braking torque at the current time step k as the initial value, and T represents the rear axle braking torque at the i-th future time step. req (k+i|k) represents the expected braking torque at the current time step k as the initial value, and the expected braking torque at the i-th future step.

[0080] The optimal regenerative braking power strategy needs to consider the regenerative braking torque of the hub motor and the battery SOC value.

[0081] The regenerative braking torque of the hub motor is less than or equal to the maximum braking torque that the motor can generate, expressed as:

[0082]

[0083] Where n is the motor speed, η m For motor efficiency, P chg_max T is the maximum charging power of the motor. m_max T is the maximum torque of the motor. m For regenerative braking torque;

[0084] The expression for a battery's SOC value being less than 90% is:

[0085]

[0086] Among them, SOC initial The initial charge state of the battery, C bat I0 represents the rated capacity of the battery and the current value.

[0087] Further, in step 4), the overall objective function of the electro-hydraulic composite braking is obtained by multiplying the braking energy recovery rate objective function, the ride comfort objective function, and the tracking braking torque objective function by their respective weighting coefficients and then summing them. The expression is as follows:

[0088] J = min(w rec ·J bra_rec +w f ·J bra_f +w T ·J bra_T )

[0089] Where J is the target value of the optimal braking strategy, w rec For energy recovery weighting coefficient, w f For smoothness weighting coefficient, w T The weighting coefficient for the braking force tracking target.

[0090] Furthermore, in step 4), when solving the overall objective function of electro-hydraulic composite braking, a deep Q-learning algorithm is needed to recalculate the weight coefficients of energy recovery rate, ride comfort, and tracking braking torque under different braking conditions to ensure that the benefits are maximized for each braking operation.

[0091] The beneficial effects of this invention are:

[0092] This invention integrates hub motors with a hydraulic braking system, eliminating the original mechanical connection, simplifying the structure of the braking system, and applying braking torque to different wheels through four hub motors, satisfying independent wheel control and improving active safety and operational stability; during regenerative braking, energy can be recovered directly through the inverter, increasing the energy recovery efficiency of the braking system; and the hub motors can cooperate with the hydraulic control unit to brake by applying reverse torque, quickly compensating for the required braking torque and improving braking reliability.

[0093] This invention combines the CPS architecture with the electro-hydraulic brake-by-wire system, inputting the driving conditions collected by the brake-by-wire unit to the brake optimization management unit for decision optimization and control, thus realizing a closed loop of "perception-analysis-decision-execution".

[0094] The Transformer network architecture in this invention improves upon traditional deep learning neural networks, enabling it to adapt to operating conditions with varying sensor signal frequencies. It is suitable for application in electro-hydraulic hybrid braking systems. Using this network for operating condition prediction can improve the stability of the algorithm and ensure the correctness of the braking system control strategy.

[0095] This invention performs braking based on the current vehicle status and operating condition prediction, improves the energy optimization strategy, achieves reasonable braking torque distribution, further reduces the energy consumption of the electro-hydraulic hybrid braking system, and realizes energy saving and emission reduction throughout the vehicle's entire life cycle. Attached Figure Description

[0096] Figure 1 This is a schematic diagram of the system of the present invention.

[0097] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0098] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0099] Reference Figure 1 As shown, an electro-hydraulic brake-by-wire system based on a CPS architecture according to the present invention includes: a brake-by-wire unit and a brake optimization management unit, which are communicatively connected;

[0100] The brake-by-wire unit is used to collect vehicle information and realize the vehicle's braking state according to the braking control command sent by the braking optimization management unit. It includes: a driver input module, a braking execution module, and an environmental perception module.

[0101] The driver input module is used to collect braking, driving, and steering data input by the driver. It includes a brake pedal sensor, an accelerator pedal sensor, a steering wheel angle sensor, and a torque sensor.

[0102] The environmental perception module is used to collect vehicle driving condition data, which includes: vehicle speed sensor, vehicle acceleration sensor, yaw rate sensor, road condition detection equipment, and battery status sensor; the vehicle driving condition data includes: vehicle speed, vehicle lateral acceleration, vehicle longitudinal acceleration, vehicle yaw rate, battery SOC status, and road condition information including road curvature.

[0103] The braking execution module is used to receive braking control commands sent by the braking optimization management unit to realize vehicle braking; it receives braking control commands sent by the braking optimization management unit via the CAN bus.

[0104] The braking execution module includes: a hydraulic wheel cylinder, a solenoid valve, an electric hydraulic pump, a coordinating controller, an electronic booster, a hub motor, an inverter, a power battery, and a brake disc. The hydraulic wheel cylinder has a solenoid valve at its hydraulic oil inlet for hydraulic pressure control. The hydraulic wheel cylinder and brake disc are mechanically connected. The coordinating controller is connected to the electric hydraulic pump, solenoid valve, and hub motor via a control bus. The electronic booster is connected to the brake pedal via a wiring harness. The hub motor is connected to the wheel via a coupling, driving the wheel to rotate. When the vehicle brakes, the kinetic energy of the hub motor is converted into electrical energy by the inverter and stored in the power battery, achieving regenerative braking. Simultaneously, the electric hydraulic pump moves according to control commands, and the solenoid valve changes its opening to build pressure in the hydraulic wheel cylinder, pressing the brake pads against the brake disc for friction braking.

[0105] The braking optimization management unit is used to predict driving conditions based on the driver input data and vehicle driving condition data sent by the brake-by-wire unit, generate braking control commands, and send the braking control commands to the brake-by-wire unit; it includes: a data processing module and a decision optimization module;

[0106] The data processing module is used to process the driver input data and vehicle driving condition data sent by the brake-by-wire unit;

[0107] The decision optimization module receives the data processed by the data processing module, calculates the operating condition prediction results, and obtains the optimal braking strategy based on the operating condition prediction results.

[0108] The data processing module uses smoothing filtering and correlation analysis to process the data.

[0109] The optimal braking strategy includes: optimal braking energy recovery strategy, optimal vehicle ride comfort strategy, optimal tracking braking torque strategy, and optimal regenerative braking power strategy.

[0110] The optimal braking energy recovery strategy is to maximize the energy recovery value achieved by the hub motor through the inverter;

[0111] The optimal vehicle ride comfort strategy is to minimize the change in braking torque and the vehicle bumps.

[0112] The optimal tracking braking torque strategy achieves the best vehicle controllability and the shortest time to reach the target torque.

[0113] The optimal regenerative braking power strategy is to ensure that the braking torque of the hub motor does not exceed the maximum torque that the motor can generate, and that the power battery is in the best condition.

[0114] The driving condition prediction uses an encoder decoupled prediction network (ED-TFN) based on a temporal Transformer model. It is based on a self-attention mechanism and calculates the vehicle driving condition at future times based on the driver input data sent by the data processing module and the vehicle driving condition data, thereby realizing the prediction of the vehicle driving condition.

[0115] Reference Figure 2 As shown, the present invention also provides an energy optimization method for an electro-hydraulic brake-by-wire system based on a CPS architecture. Based on the above system, the method steps are as follows:

[0116] 1) Obtain the driver's input data and the vehicle's operating condition data;

[0117] The driver input data in step 1) includes: current steering wheel angle θ and torque T, brake pedal travel, and accelerator pedal travel; the vehicle's operating condition data includes vehicle speed v and vehicle lateral acceleration a. x Vehicle longitudinal acceleration a y yaw rate ω of the vehicle r Battery SOC status, road curvature ρ.

[0118] 2) Perform data processing on the driver's input data and the vehicle's operating condition data;

[0119] The data processing methods include smoothing filtering and correlation analysis;

[0120] The smoothing filter process removes noise from the dataset, and the expression is as follows:

[0121]

[0122] Among them, y t-n To smooth out the driver input data and vehicle operating condition data before processing, y t The driver input dataset and the vehicle driving condition dataset are obtained after smoothing, where n is the number of filter terms;

[0123] The correlation analysis is a Pearson correlation coefficient analysis, which screens out key parameters that are highly correlated with braking torque, simplifying the calculation of driving condition prediction. The smoothed driver input dataset and the vehicle driving condition dataset are used to perform correlation analysis, as shown in the following expression:

[0124]

[0125] Where X represents the smoothed driver input dataset and the vehicle driving condition dataset, Y represents the braking torque, and ρ represents the braking torque. X,Y Let σ be the Pearson correlation coefficient, cov(X,Y) be the covariance of X and Y, and σ be the Pearson correlation coefficient.X and σ Y Let μ be the standard deviation of X and Y, respectively. X and μ Y X and Y are the average values, respectively, and E is the expected value. After correlation analysis, it was found that the vehicle speed v has the strongest correlation with the braking torque. The dataset of vehicle speed v is used as the input data for the working condition prediction.

[0126] 3) Based on the processed data, predict the operating conditions to obtain the rolling time-domain driving condition prediction results;

[0127] The operating condition prediction uses an encoder-decoupled prediction network (ED-TFN) based on a temporal Transformer model, which embeds each variable into a variable token and takes the time step as a feature of a single variable as input; the prediction network includes an embedding layer, a linear layer, a normalization layer, and a feedforward neural network.

[0128] The specific steps for performing load condition prediction are as follows:

[0129] 31) The embedding layer maps the processed vehicle speed dataset to obtain the mapping vector corresponding to the vehicle speed v dataset.

[0130] 32) The linear layer captures the correlation between the vehicle speed v dataset mapping vector after the embedding layer and the driving conditions at future times, thus obtaining the driving condition prediction feature vector.

[0131] 33) Normalize the working condition prediction feature vector obtained from the linear layer through a normalization layer;

[0132] 34) The normalized operating condition prediction feature vector is decoded by a feedforward neural network to obtain the vehicle driving condition prediction result.

[0133] Step 31) is implemented based on a multilayer perceptron (MLP), which performs feature mapping on the entire historical observation sequence for each variable, as shown in the following expression:

[0134] H 0 =MLP(V 0 )

[0135] Among them, H 0 V is a mapping vector. 0 The vehicle speed dataset obtained after slippage processing.

[0136] Step 32) is implemented based on a multi-head attention mechanism, which consists of multiple self-attention modules. Each self-attention module converts the input vector into a query vector, a key vector, and a value vector. The Softmax function is used to generate a probability distribution based on the pairwise similarity scores between vectors, expressed as:

[0137]

[0138] Here, Attention(Q,K,V) is the comprehensive information extracted from the driver input dataset and the vehicle driving condition dataset. Q is the query vector, K is the key vector, and V is the value vector. The query vector represents the direction or focus of information search, the key vector determines the information that the query vector should focus on, and the value vector contains the actual information to be output. The final output is obtained by weighted summation.

[0139] In step 33), the mean and standard deviation of all features are calculated to improve network convergence and training stability. The process expression is as follows:

[0140]

[0141]

[0142]

[0143] Output = γ × H norm +β

[0144] Where, μ l The mean, Let n be the i-th neuron node of the l-th encoding module in the input. l Let l represent the number of neuron nodes in the l-th encoding module, where l∈{1,2,...,L} represents the l-th encoding module. For variance, H norm H represents the normalized value of the vehicle velocity v mapping vector, where γ and β are the learnable scaling and offset parameters, respectively. l is the input variable of the l-th encoding module, and Output is the output of the normalization layer, which includes the vehicle speed v mapping vector.

[0145] In step 34), the feedforward neural network has a two-layer fully connected network with an activation function sandwiched in between. The activation function is a Gaussian error linear unit (GELU), and its expression is:

[0146]

[0147] In the formula, GELU is the Gaussian error linear unit activation function, and x is the function input variable.

[0148] The feedforward neural network encodes the observed time series through stacked Transformer modules and decodes the representation of future sequences using dense nonlinear connections to obtain the output, expressed as:

[0149] Hl =LayerNorm(H l-1 +Feed-Forward(H l-1 ))

[0150] Among them, H l H is the output of the current layer. l-1 The output is the previous layer's output. LayerNorm is the layer normalization function, and Forward is the feedforward layer function, which receives the input and transforms it through a series of transformations to generate the output.

[0151] The future sequence is decoded by inputting the output of the feedforward neural network into a multilayer perceptron for mapping, as expressed by:

[0152]

[0153] in, For the operating condition prediction results, H L This is the output of the feedforward neural network.

[0154] 4) Solve the overall objective function of the electro-hydraulic composite braking based on the obtained working condition prediction results to obtain the optimal braking strategy;

[0155] The optimal braking strategy includes: optimal braking energy recovery strategy, optimal vehicle ride comfort strategy, optimal tracking braking torque strategy, and optimal regenerative braking power strategy.

[0156] The objective of the optimal braking energy recovery strategy is to maximize the energy recovered during vehicle deceleration. The objective function for the braking energy recovery rate is expressed as follows:

[0157]

[0158] Among them, J bra_rec For total recovered energy, T mot_ij ω is the motor torque. ij v is the angular velocity of the motor. i0 v is the vehicle speed at the start of braking. ik η is the vehicle speed at the end of braking. d For braking system efficiency, η m_reg Where f is the regenerative braking efficiency of the motor, m is the rolling resistance coefficient, and C is the vehicle mass. D Where A is the air resistance coefficient, A is the vehicle's frontal area, and t is the air resistance coefficient. ik t is the braking end time. i0 This is the braking start time;

[0159] The objective of the optimal vehicle ride comfort strategy is to achieve the best ride comfort during braking. The objective function for ride comfort is expressed as follows:

[0160]

[0161] Among them, J bra_f Let u(k+i|k) be the target value for vehicle ride comfort, and u(k+i-1|k) be the vehicle speed at the i-th future step, with the current vehicle speed at time step k as the initial value. Let u(k+i-1|k) be the vehicle speed at the (i-1)-th future step, with the current vehicle speed at time step k as the initial value.

[0162] The objective of the optimal tracking braking torque strategy is to ensure that the actually applied braking torque can accurately track and meet the desired braking torque setpoint. The objective function expression for the tracking braking torque is:

[0163]

[0164] Among them, J bra_T For tracking purposes, T if (k+i|k) represents the front axle braking torque at the current time step k, with the current time step k as the initial value, and T representing the front axle braking torque at the i-th future time step. jr (k+i|k) represents the rear axle braking torque at the current time step k as the initial value, and T represents the rear axle braking torque at the i-th future time step. req (k+i|k) represents the expected braking torque at the current time step k as the initial value, and the expected braking torque at the i-th future step.

[0165] The optimal regenerative braking power strategy needs to consider the regenerative braking torque of the hub motor and the battery SOC value.

[0166] The regenerative braking torque of the hub motor is less than or equal to the maximum braking torque that the motor can generate, expressed as:

[0167]

[0168] Where n is the motor speed, η m For motor efficiency, P chg_max T is the maximum charging power of the motor. m_max T is the maximum torque of the motor. m For regenerative braking torque;

[0169] The expression for a battery's SOC value being less than 90% is:

[0170]

[0171] Among them, SOC initial The initial charge state of the battery, C bat I0 represents the rated capacity of the battery and the current value.

[0172] The overall objective function of electro-hydraulic hybrid braking is obtained by multiplying the objective functions of braking energy recovery rate, ride comfort, and tracking braking torque by their respective weighting coefficients and then summing them. The expression is as follows:

[0173] J = min(w rec ·J bra_rec +w f ·J bra_f +w T ·J bra_T )

[0174] Where J is the target value of the optimal braking strategy, w rec For energy recovery weighting coefficient, w f For smoothness weighting coefficient, w T The weighting coefficient for the braking force tracking target.

[0175] Specifically, in step 4), when solving the overall objective function of electro-hydraulic composite braking, a deep Q-learning algorithm is needed to recalculate the weight coefficients of energy recovery rate, ride comfort, and tracking braking torque under different braking conditions to ensure that the benefits are maximized for each braking operation.

[0176] 5) Perform vehicle braking according to the optimal braking strategy to complete the energy optimization of the braking system.

[0177] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. An energy optimization method for an electro-hydraulic brake-by-wire system based on a CPS architecture, wherein the electro-hydraulic brake-by-wire system based on a CPS architecture includes: The brake-by-wire unit and the brake optimization management unit are communicatively connected. The brake-by-wire unit is used to collect vehicle information and realize the vehicle's braking state according to the braking control command sent by the braking optimization management unit. It includes: a driver input module, a braking execution module, and an environmental perception module. The driver input module is used to collect braking, driving, and steering data input by the driver. The environmental perception module is used to collect operating condition data of the vehicle. The braking execution module is used to receive braking control commands sent by the braking optimization management unit to realize vehicle braking; The braking optimization management unit is used to predict the driving conditions based on the driver input data and vehicle driving condition data sent by the brake-by-wire unit, generate braking control commands, and send the braking control commands to the brake-by-wire unit. The method is characterized by the following steps: 1) Obtain the driver's input data and the vehicle's operating condition data; 2) Perform data processing on the driver's input data and the vehicle's operating condition data; 3) Based on the processed data, predict the operating conditions to obtain the rolling time-domain driving condition prediction results; 4) Solve the overall objective function of the electro-hydraulic composite braking based on the obtained working condition prediction results to obtain the optimal braking strategy; 5) Perform vehicle braking according to the optimal braking strategy to optimize braking system energy; The data processing method in step 2) is smoothing filtering and correlation analysis; The smoothing filter process removes noise from the dataset, and the expression is as follows: Among them, y t-n To smooth out the driver input data and vehicle operating condition data before processing, y t The driver input dataset and the vehicle driving condition dataset are obtained after smoothing, where n is the number of filter terms; The correlation analysis is a Pearson correlation coefficient analysis, which screens out key parameters that are highly correlated with braking torque, simplifying the calculation of driving condition prediction. The smoothed driver input dataset and the vehicle driving condition dataset are used to perform correlation analysis, as shown in the following expression: Where X represents the smoothed driver input dataset and the vehicle driving condition dataset, Y represents the braking torque, and ρ represents the braking torque. X,Y Let σ be the Pearson correlation coefficient, cov(X,Y) be the covariance of X and Y, and σ be the Pearson correlation coefficient. X and σ Y Let μ be the standard deviation of X and Y, respectively. X and μ Y X and Y are the average values, respectively, and E is the expected value. After correlation analysis, it was found that the vehicle speed v has the strongest correlation with the braking torque. The dataset of vehicle speed v is used as the input data for the working condition prediction.

2. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, The braking execution module includes: a hydraulic wheel cylinder, a solenoid valve, an electric hydraulic pump, a coordinating controller, an electronic booster, a hub motor, an inverter, a power battery, and a brake disc. The hydraulic wheel cylinder has a solenoid valve at its hydraulic oil inlet for hydraulic pressure control. The hydraulic wheel cylinder and brake disc are mechanically connected. The coordinating controller is connected to the electric hydraulic pump, solenoid valve, and hub motor via a control bus. The electronic booster is connected to the brake pedal via a wiring harness. The hub motor is connected to the wheel via a coupling, driving the wheel to rotate. When the vehicle brakes, the kinetic energy of the hub motor is converted into electrical energy by the inverter and stored in the power battery, achieving regenerative braking. Simultaneously, the electric hydraulic pump moves according to control commands, and the solenoid valve changes its opening to build pressure in the hydraulic wheel cylinder, pressing the brake pads against the brake disc for friction braking.

3. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, The braking optimization management unit includes: a data processing module and a decision optimization module; The data processing module is used to process the driver input data and vehicle driving condition data sent by the brake-by-wire unit; The decision optimization module receives the data processed by the data processing module, calculates the operating condition prediction results, and obtains the optimal braking strategy based on the operating condition prediction results.

4. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, The driver input data in step 1) includes: current steering wheel angle and torque, brake pedal travel, and accelerator pedal travel; the vehicle's operating condition data includes vehicle speed v and vehicle lateral acceleration a. x Vehicle longitudinal acceleration a y yaw rate ω of the vehicle r Battery SOC status, road curvature ρ.

5. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, The condition prediction in step 3) uses an encoder-decoupled prediction network based on the temporal Transformer model, which embeds each variable into a variable token and takes the time step as a feature of a single variable as input; the prediction network includes an embedding layer, a linear layer, a normalization layer, and a feedforward neural network.

6. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 5, characterized in that, The specific steps for predicting the operating conditions in step 3) are as follows: 31) The embedding layer maps the processed vehicle speed dataset to obtain the mapping vector corresponding to the vehicle speed v dataset. 32) The linear layer captures the correlation between the vehicle speed v dataset mapping vector after the embedding layer and the driving conditions at future times, thus obtaining the driving condition prediction feature vector. 33) Normalize the working condition prediction feature vector obtained from the linear layer through a normalization layer; 34) The normalized operating condition prediction feature vector is decoded by a feedforward neural network to obtain the vehicle driving condition prediction result.

7. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, The optimal braking strategy in step 4) includes: optimal braking energy recovery strategy, optimal vehicle ride comfort strategy, optimal tracking braking torque strategy, and optimal regenerative braking power strategy. The objective of the optimal braking energy recovery strategy is to maximize the energy recovered during vehicle deceleration. The objective function for the braking energy recovery rate is expressed as follows: Among them, J bra_rec For total recovered energy, T mot_ij ω is the motor torque. ij v is the angular velocity of the motor. i0 v is the vehicle speed at the start of braking. ik η is the vehicle speed at the end of braking. d For braking system efficiency, η m_reg Where f is the regenerative braking efficiency of the motor, m is the rolling resistance coefficient, and C is the vehicle mass. D Where A is the air resistance coefficient, A is the vehicle's frontal area, and t is the air resistance coefficient. ik t is the braking end time. i0 This is the braking start time; The objective of the optimal vehicle ride comfort strategy is to achieve the best ride comfort during braking. The objective function for ride comfort is expressed as follows: Among them, J bra_f Let u(k+i|k) be the target value for vehicle ride comfort, where u(k+i|k) represents the vehicle speed at the current time step k as the initial value and u(k+i-1 ... The objective of the optimal tracking braking torque strategy is to ensure that the actually applied braking torque can accurately track and meet the desired braking torque setpoint. The objective function expression for the tracking braking torque is: Among them, J bra_T For tracking purposes, T if (k+i|k) represents the front axle braking torque at the current time step k, with the current time step k as the initial value, and T representing the front axle braking torque at the i-th future time step. jr (k+i|k) represents the rear axle braking torque at the current time step k as the initial value, and T represents the rear axle braking torque at the i-th future time step. req (k+i|k) represents the expected braking torque at the current time step k as the initial value, and the expected braking torque at the i-th future step. The optimal regenerative braking power strategy needs to consider the regenerative braking torque of the hub motor and the battery SOC value. The regenerative braking torque of the hub motor is less than or equal to the maximum braking torque that the motor can generate, expressed as: Where n is the motor speed, η m For motor efficiency, P chg_max T is the maximum charging power of the motor. m_max T is the maximum torque of the motor. m This is regenerative braking torque; The expression for a battery's SOC value being less than 90% is: Among them, SOC initial The initial charge state of the battery, C bat I0 represents the rated capacity of the battery and the current value.

8. The energy optimization method for an electro-hydraulic brake system based on a CPS architecture according to claim 1, characterized in that, In step 4), the overall objective function of the electro-hydraulic composite braking system is obtained by multiplying the braking energy recovery rate objective function, the ride comfort objective function, and the tracking braking torque objective function by their respective weighting coefficients and then summing them. The expression is as follows: J=min(w rec ·J bra_rec +in f ·J bra_f +in T ·J bra_T ) Where J is the target value of the optimal braking strategy, w rec For energy recovery weighting coefficient, w f For smoothness weighting coefficient, w T The weighting coefficient for the braking force tracking target.

Images