Multi-objective optimization method and system for wire-controlled brake system based on digital twin

Abstract

The present invention relates to the field of automobile braking technology, and in particular to a multi-objective optimization method and system for a wire-controlled brake system based on digital twins, comprising: establishing a virtual platform including a physical model of the wire-controlled brake system and a digital twin model, wherein the virtual platform integrates a computational model, simulation data, and a physical entity on the same platform; constructing a virtual-reality interactive interface on the virtual platform to realize real-time interaction between the physical model data and the digital twin model data; performing multi-objective optimization on the wire-controlled brake system based on a deep deterministic policy gradient algorithm, and coordinating the distribution of friction braking and electric motor force according to the driver's driving intention while ensuring braking safety; using a physical engine to simulate hydraulic pipeline pressure fluctuations in real time, and optimizing the brake energy recovery amount and system response characteristics under ABS activation conditions, and using a deep deterministic policy gradient algorithm to perform multi-objective optimization on the wire-controlled brake system, which can better balance multiple objectives such as energy recovery efficiency.

Description

Technical Field

[0001] The present invention relates to the field of automobile braking technology, and in particular to a multi-objective optimization method and system for a wire-controlled brake system based on digital twins, which is applicable to brake energy recovery and brake control of electric vehicles, hybrid vehicles and extended-range electric vehicles. Background Art

[0002] In recent years, with the rapid development of new energy vehicle technology, brake-by-wire systems have been widely used in electric vehicles due to their high degree of integration and strong controllability. These systems intelligently distribute braking force through an electronic control unit (ECU) controlling the motor and hydraulic system. Compared to traditional braking systems, these systems offer higher energy recovery efficiency and more precise braking force control.

[0003] However, existing brake-by-wire systems still face the following technical challenges in practical applications: First, there is a contradiction between brake energy recovery efficiency and braking safety. Maximizing energy recovery while ensuring braking safety is a complex multi-objective optimization problem. Second, the dynamic characteristics of brake-by-wire systems under different operating conditions are complex, especially when the ABS (anti-lock braking system) is activated, making brake force distribution and control more difficult. Third, traditional linear control strategies are difficult to adapt to complex and changing driving scenarios and lack adaptive adjustment capabilities.

[0004] Currently, conventional brake-by-wire systems mostly employ rule-based control strategies or simple PID controllers, lacking precise modeling and optimization of the system's dynamic characteristics. Furthermore, regenerative braking strategies often employ fixed allocation ratios, failing to dynamically adjust to real-time operating conditions, resulting in low regenerative efficiency. Furthermore, existing systems lack sufficient research on coordinated control of braking force under ABS activation conditions, making it difficult to maintain adequate regenerative efficiency while ensuring safety.

[0005] Therefore, how to construct an accurate wire-controlled brake system model and, on this basis, achieve multi-objective optimization of brake energy recovery and braking safety is a technical problem that needs to be solved urgently. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-objective optimization method and system for a wire-controlled brake system based on digital twins, aiming to resolve the contradiction between braking energy recovery efficiency and braking safety in existing wire-controlled brake systems. A high-precision system model is constructed through digital twin technology, and a deep deterministic policy gradient algorithm is combined to achieve multi-objective optimization, thereby maximizing energy recovery efficiency while ensuring braking safety, while maintaining good braking comfort.

[0007] The present invention discloses a multi-objective optimization method for a wire-controlled brake system based on digital twins, comprising:

[0008] Establishing a virtual platform that includes a physical model and a digital twin model of the brake-by-wire system, wherein the virtual platform integrates the computational model, simulation data, and physical entities on the same platform;

[0009] Constructing a virtual-reality interaction interface on the virtual platform to achieve real-time interaction between physical model data and digital twin model data;

[0010] A multi-objective optimization approach is used to optimize the brake-by-wire system based on a deep deterministic policy gradient algorithm. Based on the driver's driving intention, the system coordinates the distribution of friction braking and electric motor force while ensuring braking safety.

[0011] A physics engine is used to simulate hydraulic line pressure fluctuations in real time, and to optimize brake energy recovery and system response characteristics under ABS activation conditions.

[0012] Preferably, the establishing of a virtual platform including a physical model of the brake-by-wire system and a digital twin model comprises:

[0013] Establishing a physical model of a brake-by-wire system, wherein the physical model includes a vehicle kinematic model, a hydraulic model, and a wheel brake dynamics model;

[0014] Establishing a digital twin model of the brake-by-wire system, the digital twin model including a digital environment, a virtual hydraulic system, and a virtual motor system;

[0015] Importing the physical model into a virtual laboratory built based on MATLAB and Simulink, and obtaining brake pedal displacement and hydraulic line pressure data according to the physical model;

[0016] The brake pedal displacement and hydraulic line pressure data are stored in a physical engine to obtain a training sample set for updating the digital twin model parameters.

[0017] Preferably, the vehicle kinematic model is used to obtain the total deceleration of the target vehicle; the hydraulic model is used to obtain the pressure value of the hydraulic cylinder; and the wheel braking dynamics model is used to obtain the slip rate of the target vehicle's wheel and determine whether the wheel enters the ABS safety area.

[0018] Preferably, the construction of the virtual-reality interaction interface includes:

[0019] Equipped with a physical-to-digital data acquisition interface to collect real-time sensor data such as brake pedal displacement, wheel speed, hydraulic pressure, motor speed, motor current, and vehicle acceleration;

[0020] Implement the data preprocessing module to perform noise filtering, outlier detection, data compensation and data synchronization on the collected sensor data;

[0021] Configure the digital-to-physical control execution interface to convert the control instructions into motor torque instructions, brake master cylinder pressure instructions and ABS regulating valve control instructions, and realize control through the corresponding execution unit.

[0022] Preferably, the multi-objective optimization of the brake-by-wire system based on the deep deterministic policy gradient algorithm includes:

[0023] Design the control state space, including states related to braking demand, vehicle state, system state, and environment;

[0024] Design the control action space, including motor braking torque command, hydraulic brake pressure command and state switching flag;

[0025] Design a multi-objective reward function, including energy recovery efficiency, braking performance, comfort, safety, and system response terms;

[0026] Build an experience replay pool to store control experience and improve sample utilization efficiency based on priority sampling strategy;

[0027] The dual-network architecture is used to separate strategy evaluation and improvement, ensuring the stability of the learning process.

[0028] Preferably, the weight of the multi-objective reward function adopts an adaptive adjustment mechanism to dynamically adjust the weight ratio of each indicator according to different braking conditions:

[0029] In emergency braking conditions, the safety weight is increased and the energy recovery weight is reduced;

[0030] Under normal braking conditions, the energy recovery weight is increased and the comfort weight is maintained at a moderate level;

[0031] Under ABS activation conditions, safety is weighted the highest and energy recovery is weighted the lowest.

[0032] Under low-speed conditions, increase the comfort weighting and system response weighting.

[0033] Preferably, the experience replay pool adopts a multi-level structure design, including:

[0034] The short-term pool is used to store the most recently generated experience samples, with a capacity of 100;

[0035] The medium-term pool is used to store experience samples with a medium time span and has a capacity of 500;

[0036] Long-term pool, used to store long-term valid experience samples, with a capacity of 2000;

[0037] A priority calculation strategy based on time difference error is used to improve the sampling probability of rare working condition samples and samples with rich gradient information.

[0038] Preferably, the real-time simulation of hydraulic pipeline pressure fluctuations using a physical engine includes:

[0039] Use the PX-OSG plug-in for simulation and implement real-time data exchange between the physical engine and the digital twin model based on the WebSocket protocol;

[0040] Real-time simulation of hydraulic pipeline pressure fluctuations based on physical model data and digital twin model data;

[0041] The physical model data includes ABS pressure, friction brake wheel cylinder pressure, motor force and system temperature;

[0042] The digital twin data includes brake pedal displacement, motor torque, regenerative current and line pressure.

[0043] Preferably, the optimizing of the braking energy recovery amount and the system response characteristics under the ABS activation condition includes:

[0044] Coordinates friction braking and electric motor force distribution under ABS activation conditions, adjusting the solenoid valve state and hydraulic cylinder drive according to changes in wheel cylinder pressure;

[0045] When the wheel cylinder pressure rises and approaches the target value, the solenoid valve opens to connect the hydraulic cylinder control reservoir to the brake master cylinder;

[0046] When the wheel cylinder pressure drops and the brake pressure is low, the solenoid valve is closed to disconnect the oil circuit between the reservoir and the brake master cylinder, and at the same time the hydraulic cylinder is driven to adjust the oil pressure;

[0047] The above control strategy can increase the braking energy recovery per 100 kilometers by 19%, while keeping the deceleration fluctuation rate within 0.3g.

[0048] A multi-objective optimization system for a brake-by-wire system based on digital twins, used to implement the method, includes:

[0049] Physical model building module, used to build the physical model of the wire control brake system, including vehicle kinematic model, hydraulic model and wheel brake dynamics model;

[0050] A digital model building module, used to establish a digital twin model of the brake-by-wire system, including a virtual laboratory construction unit, a training sample collection unit, a neural network model training unit, and a digital twin mapping unit;

[0051] The virtual-reality interaction interface module is used to realize real-time interaction between physical model data and digital twin model data;

[0052] Optimization target setting module, used to set and manage multiple optimization indicators for brake energy recovery and braking safety;

[0053] Weight optimization module, used to determine and optimize the weight ratio of each optimization indicator to achieve multi-objective balance;

[0054] Real-time optimization module, used to search for the optimal brake pedal displacement and brake motor torque in real time under actual working conditions;

[0055] The brake control module is used to control the brake energy recovery and braking safety of the vehicle's wire control brake system based on the optimization results, and coordinate the distribution of friction braking and electric machine power under ABS activation conditions.

[0056] The present invention has the following beneficial effects:

[0057] 1. A high-precision virtual platform for the brake-by-wire system was built using digital twin technology, enabling real-time mapping between the physical entity and the digital model, providing a reliable foundation for system optimization.

[0058] 2. A deep deterministic policy gradient algorithm is used to perform multi-objective optimization on the brake-by-wire system. Compared with traditional control methods, it can better balance multiple objectives such as energy recovery efficiency, braking safety, and comfort.

[0059] 3. A specialized braking force distribution strategy has been designed for ABS activation conditions, resolving the existing system's issue of insufficient braking force coordination under these conditions. This strategy improves braking energy recovery by 19% per 100 kilometers while maintaining a deceleration fluctuation rate of no more than 0.3g.

[0060] 4. An efficient sample learning strategy was designed based on the experience replay mechanism, which significantly improved the convergence speed and control effect of the algorithm;

[0061] 5. Through the adaptive weight adjustment mechanism, the system can intelligently adapt to different braking conditions, further improving overall performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of the multi-objective optimization method for the wire control brake system based on digital twins of the present invention;

[0063] Figure 2 Schematic diagram of the physical model structure of the wire control brake system of the present invention;

[0064] Figure 3 Schematic diagram of data interaction between the digital twin model and the physical model of the present invention;

[0065] Figure 4 This is a schematic diagram of the network structure of the deep deterministic policy gradient algorithm of the present invention;

[0066] Figure 5 Schematic diagram of the braking force distribution strategy under ABS activation conditions of the present invention;

[0067] Figure 6 Schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION

[0068] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0069] like Figure 1 As shown, the present invention provides a multi-objective optimization method for a wire control brake system based on digital twins, comprising the following steps:

[0070] Step S1: Establishing a virtual platform 30 including a physical model 10 of a brake-by-wire system and a digital twin model 20 , wherein the virtual platform 30 integrates computational models, simulation data, and physical entities on the same platform.

[0071] Step S2: Construct a virtual-reality interaction interface 40 on the virtual platform 30 to achieve real-time interaction between the physical model data and the digital twin model data.

[0072] Step S3: Based on the deep deterministic policy gradient algorithm, a multi-objective optimization is performed on the wire control brake system. According to the driver's driving intention, the friction braking and electric machine force distribution are coordinated under the premise of ensuring braking safety.

[0073] Step S4: Using a physical engine to simulate hydraulic line pressure fluctuations in real time, and optimizing the braking energy recovery amount and system response characteristics under ABS activation conditions.

[0074] like Figure 2 As shown, in step S1, establishing a virtual platform 30 including a physical model 10 and a digital twin model 20 of a wire control brake system includes: establishing a physical model 10 of a wire control brake system, wherein the physical model 10 includes a vehicle kinematic model 11, a hydraulic model 12, and a wheel brake dynamics model 13; establishing a digital twin model 20 of a wire control brake system, wherein the digital twin model 20 includes a digital environment 21, a virtual hydraulic system 22, and a virtual motor system 23; importing the physical model 10 into a virtual laboratory built based on MATLAB and Simulink, and obtaining brake pedal displacement and hydraulic line pressure data according to the physical model 10; and storing the brake pedal displacement and hydraulic line pressure data in a physical engine to obtain a training sample set for updating the parameters of the digital twin model 20.

[0075] In one embodiment of the present invention, the vehicle kinematic model 11 is used to obtain the total deceleration of the target vehicle; the hydraulic model 12 is used to obtain the pressure value of the hydraulic cylinder; and the wheel braking dynamics model 13 is used to obtain the slip rate of the target vehicle's wheel and determine whether the wheel enters the ABS safety area.

[0076] Preferably, the vehicle kinematic model 11 can be expressed as:

[0077] ,

[0078] in, is the vehicle speed in m / s; is time, unit is s; is the braking force, in N; The unit of rolling resistance is N; is the air resistance, in N; is the slope resistance, in N; is the vehicle mass in kg; It is the rotational mass conversion coefficient, dimensionless, usually 1.1-1.5, and can be 1.2 for ordinary passenger cars.

[0079] In the actual wire control brake system, taking a certain type of pure electric SUV as an example, its quality is 1850kg, the rotational mass conversion factor is 1.25, when the vehicle speed When braking at 60km / h on a dry asphalt road, the total braking force It is about 6000N, which can achieve a deceleration of about 0.3g, meeting the daily comfortable braking needs. It can be calculated as:

[0080] ,

[0081] in, is the rolling resistance coefficient, dimensionless, related to tire type, tire pressure and road surface type, usually taken as 0.013-0.018; g is the acceleration due to gravity, taken as 9.8m / s²; is the road inclination angle, in rad. For ordinary passenger cars driving on standard asphalt roads, the rolling resistance coefficient f is generally taken as 0.015. When the vehicle mass m is 1850kg, on a flat road ( =0) About 270N.

[0082] air resistance It can be calculated as:

[0083] ,

[0084] in, is the air resistance coefficient, dimensionless, determined by the vehicle shape, usually 0.3-0.4; is the frontal area of ​​the vehicle, in m 2 is the air density, usually 1.225kg / m 3 is the vehicle speed in m / s. For the electric SUV involved in the optimization of the wire control brake system, its air resistance coefficient is 0.35, windward area 2.4m 2 , when the vehicle speed is 60km / h (16.7m / s), the air resistance About 290N.

[0085] Slope resistance It can be calculated as:

[0086] ,

[0087] in, is the vehicle mass in kg; is the acceleration due to gravity, take 9.8m / s 2 , is the road inclination angle, in rad. When driving on a 3% slope (about 1.7°), the slope resistance This has a significant impact on the braking energy recovery strategy. The energy recovery potential is higher in downhill conditions and needs to be taken into account in the control strategy.

[0088] The hydraulic model 12 is used to describe the dynamic characteristics of the brake hydraulic system and can be expressed as:

[0089] ,

[0090] in, is the hydraulic cylinder pressure, in Pa; The bulk elastic modulus of the oil is in Pa, which depends on the type of brake fluid and temperature. Usually, DOT4 brake fluid is taken at room temperature. Pa; is the pipeline volume, in units of ; is the flow rate entering the pipeline, in units of ; is the flow rate leaving the pipeline, in units of ; is the volume change rate, in units of , taking into account factors such as pipe expansion. In a wire-controlled brake system, the hydraulic model is crucial for accurately controlling brake pressure. For a typical mid-sized passenger car brake system, the master cylinder diameter is 20mm, the stroke is 30mm, the brake pipe inner diameter is 3.2mm, and the total length is approximately 8m. These parameters directly affect the system response characteristics. When the brake pedal is pressed at a speed of 20mm / s, the flow generated by the master cylinder is Approximately , considering the pipeline volume for , system pressure rise rate Approximately Pa / s (1.17 bar / s), this parameter has a direct impact on the responsiveness and comfort of the braking system. The wheel brake dynamics model 13 mainly describes the tire slip characteristics and can be expressed as:

[0091] ,

[0092] in, is the tire slip ratio, dimensionless; is the vehicle speed in m / s; is the wheel angular velocity in rad / s; is the wheel radius in m.

[0093] Slip rate There is an important relationship between braking force and road adhesion, which is usually expressed as an adhesion coefficient-slip rate curve (μ-λ curve). For dry asphalt roads, the maximum adhesion coefficient Typically, the peak adhesion coefficient reaches approximately 0.9 at a slip ratio of λ = 0.15. On slippery roads, the peak adhesion coefficient drops to around 0.6, occurring around a slip ratio of λ = 0.1. ABS control typically maintains a slip ratio between 0.1 and 0.3 for optimal braking and directional stability. When the slip ratio exceeds 0.3, the wheels enter an unstable zone, potentially leading to wheel lock. At this point, ABS safety control is considered necessary.

[0094] In the optimization of wire-controlled braking systems, slip ratio control accuracy significantly impacts braking distance during emergency braking at high speeds (e.g., 100 km / h). Experimental data shows that optimal slip ratio control (λ=0.15) can shorten braking distance by approximately 15% compared to inadequate control (λ=0.4), while also improving energy recovery efficiency by approximately 12%.

[0095] like Figure 3As shown, in step S2, the construction of the virtual-reality interaction interface 40 includes: configuring a physical-to-digital data acquisition interface 41 to collect sensor data such as brake pedal displacement, wheel speed, hydraulic pressure, motor speed, motor current and vehicle acceleration in real time; implementing a data preprocessing module 42 to perform noise filtering, outlier detection, data compensation and data synchronization on the collected sensor data; configuring a digital-to-physical control execution interface 43 to convert control instructions into motor torque instructions, brake master cylinder pressure instructions and ABS regulating valve control instructions, and realize control through corresponding execution units.

[0096] In an embodiment of the present invention, the physical-to-digital data acquisition interface 41 is implemented using a high-speed data acquisition card, with sampling frequencies designed based on the signal characteristics of different sensors. Specifically, the brake pedal displacement sensor has a sampling frequency of 100 Hz and a signal bandwidth of 0-20 Hz; the wheel speed sensor has a sampling frequency of 200 Hz and a signal bandwidth of 0-50 Hz; the hydraulic pressure sensor has a sampling frequency of 200 Hz and a signal bandwidth of 0-80 Hz; the motor speed sensor has a sampling frequency of 100 Hz and a signal bandwidth of 0-30 Hz; the motor current sensor has a sampling frequency of 200 Hz and a signal bandwidth of 0-60 Hz; and the vehicle acceleration sensor has a sampling frequency of 100 Hz and a signal bandwidth of 0-20 Hz. The sampling frequency design adheres to the Nyquist sampling theorem, ensuring that the sampling frequency is at least twice the signal bandwidth to avoid signal aliasing.

[0097] Data preprocessing module 42 uses an adaptive window sliding average filtering algorithm, with the window size automatically adjusted based on the data change rate. Specifically, when the signal change rate is low (such as during the initial braking phase and during constant-speed braking), a larger window size (7 sampling points) is used to improve filtering effectiveness. When the signal change rate is high (such as during emergency braking and ABS activation), a smaller window size (3 sampling points) is used to ensure faster response. The filtering algorithm can be expressed as:

[0098] ,

[0099] in, is the filtered data at the nth moment; is the original data; is the filter window size, which is adaptively adjusted according to the signal change rate and ranges from 3 to 7. The signal change rate is calculated as:

[0100] ,

[0101] in, is the normalized signal change rate; is the sampling period, in seconds; is the full-scale value of the signal. Adaptive adjustment according to the rate of change: hour, when hour, when hour, . Outlier detection uses 3 The principle is that when data deviates from the mean by more than three standard deviations, it is considered an outlier and replaced with the nearest valid value. The normal operating range for brake pressure signals is 0-8 MPa. When a sudden change, such as 12 MPa, is detected, the system automatically identifies it as an outlier and makes corrections to ensure the stability of the control algorithm.

[0102] Data compensation primarily compensates for sensor delays. Different sensors have different response times. The brake pedal displacement sensor has a response time of approximately 0.8ms, the hydraulic pressure sensor has a response time of approximately 15ms, and the wheel speed sensor has a response time of approximately 5ms. Time compensation ensures data consistency. The compensation formula is:

[0103] ,

[0104] in, is the data after compensation; is the original data; The sensor response delay can be obtained through system identification. For the hydraulic pressure sensor, its compensation value It is about -15ms, and the Lagrange interpolation method is used to obtain the values ​​of non-sampling points.

[0105] Data synchronization is achieved through timestamps, using a master clock to trigger sampling, with synchronization accuracy within ±2ms. All sensor data is timestamped to ensure temporal consistency during data fusion processing. This is crucial for accurately assessing system status and formulating control strategies, especially during emergency braking at high speeds. A 2ms time synchronization error is equivalent to only approximately 6cm of vehicle movement, ensuring highly accurate system status assessment and control.

[0106] The digital-to-physical control execution interface 43 converts control commands into signals recognizable by the hardware in a specified format. The motor torque command range is ±300 Nm with a resolution of 0.5 Nm, with positive values ​​representing driving torque and negative values ​​representing braking torque. The master cylinder pressure command range is 0-8 MPa with a resolution of 0.05 MPa. The ABS control valve control commands include open, closed, and adjustment states. The solenoid valve opening is controlled by a PWM signal with a 100 Hz frequency and a duty cycle range of 0% to 100%. In practice, to avoid system shock caused by sudden changes in control commands, a ramp transition strategy is implemented, limiting the motor torque change rate to within 1000 Nm / s and the brake pressure change rate to within 20 MPa / s, ensuring smooth and rapid braking system response.

[0107] like Figure 4 As shown, in step S3, the multi-objective optimization of the wire control brake system based on the deep deterministic policy gradient algorithm includes: designing the control state space, including braking demand-related states, vehicle state-related states, system state-related states and environment-related states; designing the control action space, including motor braking torque instructions, hydraulic braking pressure instructions and state switching flags; designing a multi-objective reward function, including energy recovery efficiency items, braking performance items, comfort items, safety items and system response items; constructing an experience replay pool to store control experience, and improving sample utilization efficiency based on a priority sampling strategy; and realizing the separation of strategy evaluation and improvement through a dual network architecture to ensure the stability of the learning process.

[0108] In a preferred embodiment of the present invention, the control state space is designed as follows:

[0109] Braking demand related states include brake pedal travel and pedal displacement change rate .in, The range is 0-100mm, indicating the intensity of the driver's braking intention; The range is usually -200mm / s to 200mm / s, indicating the changing trend of braking intention, which is important for distinguishing between sudden and slow pedaling. , wheel angular velocity and slip rate .in, The range is 0-180km / h and is a key input for brake control; is the set of angular velocities of the four wheels , the unit is rad / s, subscript 、 、 、 Respectively represent the left front wheel, right front wheel, left rear wheel, and right rear wheel; is the set of slip rates of the four wheels , which is the core parameter of ABS control.

[0110] System status related status includes motor speed 、 and ,in, The range is usually 0-12000rpm, affecting the motor's power and energy recovery efficiency; The range is -20°C to 120°C. Under high temperature conditions, it is necessary to limit the motor's braking force to protect the motor; The brake pressure of the four wheels , in MPa, directly determines the size of the mechanical braking force. Environmental related conditions include the estimated road adhesion coefficient and estimated road slope .in, The range is 0.1-1.0, estimated from the slip ratio and wheel angular acceleration; The range is -20° to 20° and is estimated by the vehicle's longitudinal acceleration and speed changes. These parameters have a significant impact on the choice of braking strategy, such as requiring a more conservative braking strategy on low-adhesion roads and increasing the energy recovery ratio on downhill sections.

[0111] The historical state memory includes the key state variables of the previous N steps (N=10), forming a time-series state vector, providing information on the dynamic changes of the system, and helping to improve the stability and predictability of the control strategy.

[0112] In summary, the complete state vector can be expressed as:

[0113] ,

[0114] in, Represents the historical state sequence, including the key state information of the previous N steps. The state vector dimension is 10+4×4+2+10=38, which is an acceptable input size for deep reinforcement learning. The control action space design includes: motor braking torque command , continuous value, range is [-300Nm, 300Nm], negative value indicates braking, positive value indicates driving; hydraulic brake pressure command , continuous value, range is [0MPa, 8MPa], uniformly controls the pressure of the four wheel brakes, and then distributes it to each wheel according to the vehicle state through the brake force distributor; state switching flag , a discrete value, represents the ABS activation / deactivation control, where a value of 0 indicates deactivation and a value of 1 indicates activation. The action vector can be expressed as:

[0115] ,

[0116] The action vector dimension is 3. This low-dimensional action space design is conducive to improving learning efficiency and control accuracy. In actual control, and The coordination of the two is the key to ensuring energy recovery efficiency and braking safety. For example, in high-speed light braking conditions (such as 80km / h, deceleration 0.1g), motor braking is used first. About -120Nm, Close to 0; in medium speed and moderate braking conditions (such as 50km / h, deceleration 0.3g), the motor brake and hydraulic brake work together. About -280Nm, About 2MPa; in low-speed emergency braking conditions (such as 30km / h, deceleration 0.8g), it mainly relies on hydraulic braking, About -100Nm, About 6MPa.

[0117] The core of the Deep Deterministic Policy Gradient algorithm is to approximate the policy function and value function through a deep neural network, combining deterministic policy gradients and experience replay to achieve efficient optimization of the continuous action space. The algorithm consists of four networks: the actor network, the critic network, the target actor network, and the target critic network.

[0118] The Actor network generates deterministic actions based on the current state. It is a four-layer neural network consisting of an input layer, two hidden layers, and an output layer. The input layer has 38 neurons, matching the state dimensions. The first hidden layer contains 32 neurons, and the second hidden layer contains 16 neurons, both using the ReLU activation function. The output layer has three neurons, matching the action dimensions. The first two use the tanh activation function, mapping the output to the [-1, 1] range before scaling it to the actual control range. The final layer uses a sigmoid activation function, mapping the output to the [0, 1] range for ABS state control.

[0119] The ReLU activation function is defined as:

[0120] ,

[0121] in, is the neuron input; is the output of the activation function. The ReLU function has the advantages of simple calculation and stable gradient, and is suitable for deep network training. The tanh activation function is defined as:

[0122] ,

[0123] in, is the neuron input; is the output of the activation function, ranging from [-1, 1]. The zero-centering property of the tanh function is beneficial to gradient updates. The sigmoid activation function is defined as:

[0124] ,

[0125] in, is the neuron input; is the activation function output, ranging from [0, 1]. The sigmoid function is suitable for binary classification problems and is used to determine the activation of the ABS state in this case.

[0126] The Actor network can be represented as:

[0127] ,

[0128] in, is the action vector, including 、 and ; is the state vector; The parameters are Actor network. Parameters Including all network weights and biases, the total number of parameters is approximately The Critic network is used to evaluate the value of state-action pairs. Its structure is a four-layer neural network: the input layer receives the connection vector of the state and action, and the number of neurons is the state dimension plus the action dimension, that is, 41 neurons; the first hidden layer contains 64 neurons, and the second hidden layer contains 32 neurons, both using the ReLU activation function; the output layer has 1 neuron, using the linear activation function, to represent the Q value of the state-action pair. The Critic network can be expressed as:

[0129] ,

[0130] in, is the value of the state-action pair; is the state vector; is the action vector; The parameters are Critic network. Parameters Including all network weights and biases, the total number of parameters is approximately The target actor network and target critic network structures are the same as the corresponding main network, but the parameter update frequency is low, which is used to provide a stable learning target and prevent training instability. The parameter update adopts a soft update strategy with an update frequency of every 20 steps and an update coefficient of .

[0131] The multi-objective reward function is designed as the weighted sum of various performance indicators:

[0132] ,

[0133] in, is the total reward; It is the energy recovery efficiency reward, which is proportional to the motor recovery current; It is a braking performance reward, which is inversely proportional to the deviation between the actual deceleration and the target deceleration; It is a comfort bonus and is inversely proportional to the deceleration fluctuation rate; It is a safety reward, inversely proportional to the degree to which the tire slip rate deviates from the optimal slip rate; It is the system response reward, which is inversely proportional to the rate of change of the control command; to is the weight coefficient of each reward, and satisfies The energy recovery efficiency bonus can be calculated as:

[0134] ,

[0135] in: is the recovery current, in A; is the motor terminal voltage, in V; is the total braking power, in W; The energy recovery efficiency coefficient is usually 2-5, which is used to adjust the weight of the reward. Taking a certain type of pure electric vehicle as an example, when the vehicle speed is 50km / h and the deceleration braking is 0.2g, the motor speed is about 3000rpm, and the motor terminal voltage is About 350V, recovery current About 80A, total braking power The power is about 38kW, the energy recovery power is about 28kW, and the recovery efficiency is about 74%. The calculated value is 3.7 (assuming ), which is a positive incentive.

[0136] The braking performance bonus can be calculated as:

[0137] ,

[0138] in, is the actual deceleration, in units of is the target deceleration, in units of The braking performance coefficient is usually 1-3, which is used to adjust the weight of the reward. Taking the comfortable braking condition as an example, when the driver presses the brake pedal 40mm, the corresponding target deceleration is Approximately , if the deceleration achieved by the actual control system 0.32g(3.14 ), then the deviation is , assuming ,Braking Performance Reward The calculated value is -0.4, which is a slight penalty. The comfort bonus can be calculated as:

[0139] ,

[0140] in, is the deceleration standard deviation, indicating the degree of deceleration fluctuation, in m / s It is the comfort coefficient, usually 2-4, used to adjust the weight of the reward.

[0141] Deceleration standard deviation Calculated as:

[0142] ,

[0143] in, is the deceleration at the i-th moment, in units of is the average deceleration at N moments, in units of is the number of sampling points, usually 10-20 moments of data are taken. Under ideal braking conditions, the deceleration fluctuation is small. About 0.05g (0.49m / s2), assuming , then the comfort reward The calculated value is -1.47, which is a medium penalty. Under the ABS activation condition, the deceleration fluctuates greatly. reachable ,Under the same conditions, the calculated value of the comfort reward is -5.88, which is a stronger penalty, ,which will prompt the control strategy to seek a smoother braking method.

[0144] The security bonus can be calculated as:

[0145] ,

[0146] in, is the actual slip rate of the i-th wheel, dimensionless; is the optimal slip ratio, usually 0.15-0.2, dimensionless; The safety factor, usually between 3 and 5, is used to adjust the weight of this reward. The summation symbol indicates the accumulation of the slip rate deviations of the four wheels.

[0147] Under normal braking conditions, the slip ratios of the four wheels are close to the optimal values, such as Both are 0.16, is 0.18, the total slip rate deviation is 0.08, assuming ,Safety Rewards The calculated value is -0.32, which is a slight penalty. In emergency braking conditions on slippery roads, if the slip rate of a wheel reaches 0.4, far away from the optimal value of 0.18, the safety reward will be greatly reduced, prompting the control strategy to activate ABS for intervention.

[0148] The system response reward can be calculated as:

[0149] ,

[0150] in, is the change in motor torque between two adjacent control cycles, in Nm; is the hydraulic pressure change between two adjacent control cycles, in MPa; The response coefficient is usually 0.5-1.5, which is used to adjust the weight of the reward; is the pressure change weight coefficient, usually ranging from 2 to 5, considering that the impact of pressure change on comfort is usually greater than the motor torque change. The hydraulic pressure change is 10Nm is 0.2MPa, assuming ,System Response Reward The calculated value is -10-3 , is a medium penalty; under the braking force sudden change condition, such as 50Nm, The reward value is -53 under the same conditions, which is a strong penalty. This will prompt the control strategy to seek a smoother braking transition method.

[0151] In this embodiment of the present invention, the weighting of the reward function utilizes an adaptive adjustment mechanism, dynamically adjusting the weighting ratios of various indicators based on different braking conditions. Specifically, in emergency braking conditions, safety is prioritized while energy regeneration is prioritized; in normal braking conditions, energy regeneration is prioritized while comfort is prioritized moderately; in ABS-activated conditions, safety is prioritized while energy regeneration is prioritized minimally; and in low-speed conditions, comfort and system response are prioritized.

[0152] For example, the basic working condition weight configuration is: Energy recovery , security , comfort , responsiveness =0.1); Emergency braking weight configuration is: safety ( , responsiveness , comfort 、Energy recovery( The general braking weight configuration is: Energy recovery , comfort , security , responsiveness Low-speed braking weight configuration: Comfort , energy recovery , responsiveness , security The working condition judgment standard is: when the brake pedal is displaced or pedal speed When the vehicle speed is When the slip rate of any wheel is When the braking angle is 0.3, it is determined as the ABS activation condition; the rest of the conditions are determined as normal braking conditions.

[0153] To improve the convergence efficiency and stability of the algorithm, the present invention uses an experience replay mechanism to store and utilize control experience. In a preferred embodiment of the present invention, the experience replay pool adopts a multi-level structure design, including: a short-term pool for storing recently generated experience samples, with a capacity of 100; a medium-term pool for storing experience samples with a medium time span, with a capacity of 500; and a long-term pool for storing long-term effective experience samples, with a capacity of 2000.

[0154] The structure of an experience sample is <current state, action performed, reward received, next state, operating condition label>. The operating condition label identifies the type of operating condition to which the sample belongs, such as emergency braking, general braking, low-adhesion road conditions, and ABS activation. To ensure a balanced distribution of operating condition samples during the initial training phase, a condition balancing strategy is employed. For example, the constructed training scenarios are configured with 30% general braking conditions, 30% emergency braking conditions, 20% low-adhesion road conditions, and 20% ABS activation conditions. This ensures that the algorithm learns the optimal control strategy for each condition.

[0155] In order to improve the efficiency of sample utilization, the present invention adopts a priority calculation strategy based on time difference error to increase the sampling probability of rare working condition samples and samples with rich gradient information. The time difference error is calculated as:

[0156] ,

[0157] in, is the time difference error; For immediate rewards; is the discount factor, usually 0.95-0.99; Target Critic network; is the target Actor network; The main critic network; is the current state; To perform an action; In the ABS activation condition, due to the rapid change of state and large change of instant reward, the time difference error It is usually larger, such as up to 3-5, indicating that these samples contain richer gradient information; while in steady-state braking, the state and reward changes are small, and the time difference error Usually small, such as 0.1-0.5. The priority sampling strategy will give the ABS activation condition samples a higher sampling probability, accelerating the learning process. The priority of the sample is proportional to the time difference error and is calculated as:

[0158] ,

[0159] in, is the sample priority; A small positive number, usually 0.01, to prevent the priority from being zero; It is the priority index, usually ranging from 0.6 to 0.8, and is used to control the impact of priority. When the value is small (such as 0.6), the priority difference between samples is small and the sampling is more uniform; When the value is large (such as 0.8), high error samples are more likely to be sampled, but it may lead to excessive attention to some samples.

[0160] At the same time, for rare working condition samples, such as ABS activation working condition, low adhesion road working condition, etc., additional priority coefficient is added , usually 1.5-3 is taken to further increase the sampling probability of these samples. For example, for the ABS activation sample under slippery road conditions, its original priority may be 2.5, and the rare operating condition coefficient is applied. After that, the priority is raised to 5.0, which significantly increases the probability of being sampled. Based on the above priorities, a priority sampling strategy is used to sample training samples from the experience replay pool. In order to prevent high-priority samples from over-dominantly dominating the training process, importance sampling weights are used for correction:

[0161] ,

[0162] in, is the importance sampling weight; is the size of the experience replay pool; is the probability that sample i is sampled, calculated as is the importance sampling index, the initial value is usually 0.4, and then increases linearly to 1. When the value is small, the correction effect of importance sampling is weaker; When the value is close to 1, the correction effect is stronger and closer to uniform sampling. The network parameters are updated using temporal difference (TD) learning and policy gradient algorithm. Specifically, the loss function of the critic network is:

[0163] ,

[0164] in: is the loss function of the Critic network; The batch size is usually 32-64; is the importance sampling weight of sample i; is the TD target of sample i, calculated as ; is the predicted value of the Critic network for sample i. The sum symbol indicates that all samples in the batch are accumulated and the average is taken as the loss function. The parameter update of the Actor network uses the Deterministic Policy Gradient (DPG) algorithm:

[0165] ,

[0166] in: is the policy gradient of the Actor network; is the gradient of Q value with respect to action a, in action Calculation at is the policy function About parameters The sum symbol indicates that the policy gradient is calculated for all samples in the batch, and then the average is taken to update the parameters. The intuitive understanding of the policy gradient is: adjust the policy parameters , so that the strategy In state The action generated by Able to maximize Q value Taking motor torque control as an example, if the motor braking torque is increased The energy recovery efficiency can be improved, thereby increasing the Q value. The policy gradient will adjust the parameters so that the network outputs a larger motor braking torque under similar conditions. To ensure the stability of learning, the target network parameters adopt a soft update strategy:

[0167] ,

[0168] ,

[0169] in, and are the parameters of the target Actor network and the target Critic network respectively; and These are the parameters of the main Actor network and the main Critic network respectively; It is the soft update coefficient, usually 0.001-0.01, which controls the smoothness of the update. When the value is small (such as 0.001), the target network parameters change slowly, the learning stability is good, but the convergence speed is slow; When the value is large (such as 0.01), the target network parameters are updated faster and the convergence speed is fast, but it may cause unstable learning.

[0170] In actual applications, dynamic adjustment is made according to the system convergence situation. Value: Initially, keep it small, such as 0.001, and gradually increase it to 0.01 as training progresses, balancing stability and convergence speed. The target network update frequency is set to every 20 steps, rather than every step, to further enhance learning stability.

[0171] like Figure 5 As shown, in step S4, the real-time simulation of hydraulic pipeline pressure fluctuations using a physical engine includes: using a PX-OSG plug-in for simulation, and realizing real-time data exchange between the physical engine and the digital twin model based on the WebSocket protocol; simulating hydraulic pipeline pressure fluctuations in real time according to the physical model data and the digital twin model data; the physical model data includes ABS pressure, friction brake wheel cylinder pressure, motor force and system temperature; the digital twin data includes brake pedal displacement, motor torque, recovery current and pipeline pressure.

[0172] In one embodiment of the present invention, the physics engine utilizes the open-source physics simulation engine PhysX combined with the OSG (OpenSceneGraph) visualization library to construct both the physical and visual models of the virtual hydraulic system. The PX-OSG plug-in, based on the WebSocket protocol and using TCP / IP communication, enables real-time data exchange between the physics engine and the digital twin model. The communication frequency is 100Hz, with data latency controlled to less than 10ms. The communication protocol consists of two parts: a request message and a response message.

[0173] The simulation of hydraulic pipeline pressure fluctuations is based on a one-dimensional fluid dynamics model, taking into account factors such as pipeline friction, oil compressibility, and pipe wall elasticity. The wave equation can be expressed as:

[0174] ,

[0175] ,

[0176] Where P is pressure, unit is Pa; v is flow velocity, unit is m / s; t is time, unit is s; x is position, unit is m; is the oil density, usually 850-900kg / m³; c is the pressure wave velocity, usually 1200-1500m / s; f is the friction coefficient, dimensionless, usually 0.02-0.04; D is the pipe diameter, in meters. It represents the partial derivative of pressure with respect to time, and characterizes the rate of change of pressure with time; It represents the partial derivative of pressure with respect to position, and characterizes the distribution of pressure along the pipeline; It represents the partial derivative of flow velocity with respect to time, and characterizes the rate of change of flow velocity with time; It represents the partial derivative of flow velocity with respect to position, and characterizes the distribution of flow velocity along the pipeline. For a typical hydraulic pipeline of a brake system, the pipe diameter is 3.2mm, length 1-2m, brake fluid density is 880kg / m³, and the pressure wave speed is 1350m / s, friction coefficient The pressure in the master cylinder can rise from 0 to 8 MPa in just 0.1 seconds during emergency braking. This hydraulic shock can cause line pressure fluctuations, with amplitudes of up to 1-2 MPa and a period of approximately 2-5 ms. This pressure fluctuation propagates through the hydraulic lines, ultimately causing caliper pressure fluctuations, impacting braking stability and comfort.

[0177] A key innovation of this invention is optimizing brake energy recovery and system response characteristics under ABS activation. Specifically, this involves coordinating friction braking and motor force distribution under ABS activation, adjusting the solenoid valve state and hydraulic cylinder drive according to changes in wheel cylinder pressure. When wheel cylinder pressure rises and approaches the target value, the solenoid valve opens, connecting the hydraulic cylinder control reservoir to the brake master cylinder. When wheel cylinder pressure drops and brake pressure is low, the solenoid valve closes, disconnecting the hydraulic line between the reservoir and the brake master cylinder while simultaneously actuating the hydraulic cylinder to adjust oil pressure. This control strategy improves brake energy recovery by 19% per 100 kilometers, while maintaining a deceleration fluctuation rate of no more than 0.3g.

[0178] ABS activation is determined by the following criteria: ABS control is triggered when the slip ratio of any wheel, lambda_i, exceeds 0.3. The goal of ABS control is to maintain the slip ratio within the optimal range (0.1-0.2), ensuring high braking force while maintaining directional stability.

[0179] Under normal operating conditions, the hydraulic cylinder controls the oil flow between the fluid reservoir and the master cylinder, driving the pistons of the secondary and master cylinders to move and output brake pressure. Under ABS activation, based on braking system parameters, the hydraulic cylinder closes the oil flow between the fluid reservoir and the master cylinder, and the pressure control motor adjusts the brake pressure, directing it to the wheel cylinders. The specific control strategy is as follows:

[0180] 1. When wheel cylinder pressure rises and reaches 90% of the target value (the threshold is set based on actual system response characteristics), the solenoid valve opens to connect the hydraulic cylinder control reservoir to the brake master cylinder, preventing excessive wheel cylinder pressure from causing wheel lock.

[0181] 2. When wheel cylinder pressure drops below 60% of the target value (the threshold is set based on system stability and response speed), the solenoid valve closes, disconnecting the oil line between the reservoir and the master cylinder. Simultaneously, the hydraulic cylinder is actuated to adjust the oil pressure to prevent insufficient braking force and increased braking distance.

[0182] 3. When the vehicle speed is lower than 10km / h, the ABS control mode will be gradually exited and control will be returned to the normal braking control mode to prevent instability caused by frequent system switching at low speeds.

[0183] Conventional ABS systems typically significantly reduce or completely disable motor braking when activated, which is detrimental to energy recovery. The innovation of this invention is that even when ABS is activated, it still maintains appropriate motor braking force to achieve energy recovery. The specific strategy is:

[0184] 1. When ABS is not activated, the electric motor can account for up to 90% of the power, maximizing energy recovery;

[0185] 2. During the initial activation of ABS, the motor force is quickly reduced to 40%, ensuring the system can quickly respond to slip conditions;

[0186] 3. As ABS control stabilizes, the motor's braking force gradually recovers to 60% to 70%, improving energy recovery while ensuring braking safety.

[0187] 4. When ABS is frequently activated under special road conditions (such as icy or snowy roads), the motor braking force is further reduced to below 20%, prioritizing vehicle safety.

[0188] The dynamic distribution of electric motor force and hydraulic braking force is achieved through the following formula:

[0189] ,

[0190] in: is the ratio of the electric braking force to the total braking force; is the slip rate of the i-th wheel; The time after ABS activation, in seconds; The time coefficient is usually 0.5-1.5, which controls the speed of motor power recovery; is the estimated road adhesion coefficient. In the conditional judgment statement, "if" represents the condition and "max" represents the maximum value. Represents a negative exponential decay function, which gradually approaches 0 as time increases.

[0191] For a certain pure electric SUV, during emergency braking at 100 km / h on a dry asphalt road (adhesion coefficient μ = 0.8), using the aforementioned control strategy, the electric braking force initially accounted for 0.4% of the braking force. As the system stabilized, the electric braking force percentage recovered to 0.67% after three seconds, and the braking energy regenerated reached 0.65 kWh. In contrast, under the traditional control strategy, the electric braking was completely disabled after ABS activation, resulting in only 0.24 kWh of braking energy regenerated. Comprehensive testing under various operating conditions showed that the control strategy of the present invention increased the braking energy regenerated per 100 km from 82 kWh with the traditional system to 97.6 kWh, a 19% improvement.

[0192] At the same time, by precisely controlling the hydraulic pressure and motor torque, the deceleration fluctuation rate is controlled within 0.3g, significantly improving braking comfort. The deceleration fluctuation rate is defined as the ratio of the deceleration standard deviation to the average deceleration, and the calculation formula is:

[0193]

[0194] in, is the deceleration fluctuation rate, dimensionless; is the average deceleration, in units of .

[0195] Under emergency braking conditions (average deceleration of approximately 0.8g or 7.84m / s²), the standard deviation of deceleration under the traditional control strategy is About 0.4g (3.92m / s²), deceleration fluctuation rate is 0.5; however, by adopting the control strategy of the present invention, the standard deviation of deceleration is reduced to 0.24g (2.35m / s²), the deceleration fluctuation rate is 0.3, and the fluctuation is reduced by 40%, which significantly improves the braking comfort.

[0196] like Figure 6As shown, the present invention also provides a multi-objective optimization system for a wire control brake system based on digital twins, including: a physical model establishment module 100 for establishing a physical model of the wire control brake system, including a vehicle kinematic model, a hydraulic model, and a wheel brake dynamics model; a digital model construction module 200 for establishing a digital twin model of the wire control brake system, including a virtual laboratory construction unit, a training sample acquisition unit, a neural network model training unit, and a digital twin mapping unit; a virtual-reality interaction interface module 300 for realizing real-time interaction between physical model data and digital twin model data; an optimization target setting module 40 0, used to set and manage multiple optimization indicators for brake energy recovery and brake safety; weight optimization module 500, used to determine and optimize the weight ratio of each optimization indicator to achieve multi-objective balance; real-time optimization module 600, used to search for the optimal brake pedal displacement and brake motor torque in real time under actual working conditions; brake control module 700, used to control the brake energy recovery and brake safety of the vehicle's wire control brake system according to the optimization results, coordinate the distribution of friction braking and electric machine power under ABS activation conditions, and increase the brake energy recovery per 100 kilometers by 19%, while keeping the deceleration fluctuation rate no more than 0.3g.

[0197] The implementation of the physical model building module 100 is the same as the physical model building steps described above, including the establishment of a vehicle kinematic model, a hydraulic model, and a wheel brake dynamics model. These models together form the physical foundation of the brake-by-wire system and provide a basis for subsequent optimization.

[0198] The digital model building module 200 is responsible for establishing the digital twin model of the wire control brake system. It includes a virtual laboratory construction unit, a training sample acquisition unit, a neural network model training unit, and a digital twin mapping unit. The virtual laboratory construction unit builds a virtual simulation platform based on MATLAB and Simulink, constructs a virtual vehicle model, and introduces real-world driving scenario data to simulate vehicle behavior under different operating conditions. The training sample acquisition unit is responsible for collecting data such as brake pedal displacement, vehicle speed, wheel speed, and brake pressure. The sampling frequency is 100Hz, and each test acquisition lasts 30 to 60 seconds. It covers various operating conditions such as conventional braking, emergency braking, and low-adhesion road braking. The neural network model training unit trains a deep deterministic policy gradient model based on the collected data, using the aforementioned network structure and training parameters. The digital twin mapping unit establishes a mapping relationship between the physical model and the digital model, achieving real-time conversion from physical state to digital state, with an error control of less than 5%.

[0199] The virtual-reality interaction interface module 300 realizes real-time interaction between physical model data and digital twin model data. Its implementation method is the same as the above-mentioned virtual-reality interaction interface construction steps, mainly including physical to digital data acquisition interface, data preprocessing module and digital to physical control execution interface.

[0200] The optimization target setting module 400 is responsible for setting and managing multiple optimization indicators for brake energy recovery and braking safety. It includes a control variable setting unit, an objective function construction unit, a constraint setting unit, and an indicator evaluation unit. The control variable setting unit uses the control variable method to set optimization indicators, such as fixed vehicle speed and slope conditions, and analyzes the impact of different braking strategies on energy recovery efficiency. The objective function construction unit constructs an objective function for comprehensively evaluating various indicators, such as the aforementioned multi-objective reward function. The constraint setting unit sets hard constraints such as safety and comfort, such as a slip ratio not exceeding 0.4 and a braking distance not exceeding 110% of the reference value. The indicator evaluation unit evaluates the achievement of optimization indicators in real time, calculates indicators such as energy recovery efficiency, braking distance, and comfort, and provides feedback.

[0201] The weight optimization module 500 determines and optimizes the weight ratios of various optimization indicators to achieve multi-objective balance. It includes an initial weight setting unit, a differential evolution algorithm unit, a weight evaluation unit, and a weight adaptation unit. The initial weight setting unit sets the initial weights of each optimization indicator, such as the aforementioned basic operating condition weight configuration. The differential evolution algorithm unit adjusts the weight ratios based on the differential evolution algorithm, with a population size of 30, a maximum number of iterations of 50, a mutation factor F of 0.5, and a crossover probability CR of 0.7. The weight evaluation unit evaluates the impact of weight adjustments on system performance and calculates comprehensive performance indicators under different weight combinations. The weight adaptation unit adaptively adjusts the weight ratios according to different operating conditions, such as dynamically increasing the safety weight under ABS activation conditions.

[0202] The real-time optimization module 600 searches for the optimal brake pedal displacement and brake motor torque in real time under actual operating conditions. It includes a target speed setting unit, an optimization algorithm execution unit, a fitness evaluation unit, and a digital twin update unit. The target speed setting unit sets the acquisition target speed, such as 0.2g for smooth deceleration conditions and 0.8g for emergency braking conditions. The optimization algorithm execution unit executes an optimization algorithm based on deep deterministic policy gradients with a control cycle of 10ms, updating the policy network output in real time. The fitness evaluation unit evaluates the fitness value of the optimization result, calculating indicators such as the deviation between the actual deceleration and the target deceleration and the energy recovery efficiency. The digital twin update unit updates the digital twin model based on the fitness value, adjusting the model parameters to ensure that the predicted results are consistent with the actual performance.

[0203] Braking control module 700 controls the vehicle's brake-by-wire system's brake energy recovery and braking safety based on the optimization results. It includes a control strategy application unit, a motor brake control unit, a hydraulic brake control unit, and a coordination control unit. The control strategy application unit applies the optimized control strategy, converting the strategy network's output into specific control instructions. The motor brake control unit controls the motor's output braking torque, with a response time of no more than 30ms and a torque control accuracy of ±5Nm. The hydraulic brake control unit controls the hydraulic system's output braking pressure, with a response time of no more than 50ms and a pressure control accuracy of ±0.2MPa. The coordination control unit coordinates the smooth transition between motor braking and hydraulic braking, dynamically adjusting the output of each actuator based on the aforementioned braking force distribution strategy to ensure smooth braking.

[0204] Based on this system architecture, the present invention achieves multi-objective optimization of the brake-by-wire system, maximizing energy recovery efficiency while ensuring braking safety and maintaining good braking comfort. Specifically, under ABS-activated conditions, by precisely coordinating friction braking and electric motor force distribution, brake energy recovery per 100 kilometers is increased by 19%, while maintaining deceleration fluctuations below 0.3g, significantly improving overall system performance.

[0205] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various obvious changes and modifications can be made to the present invention without departing from the scope of the present invention, and these changes and modifications all fall within the scope of protection of the present invention.

Claims

1. A multi-objective optimization method for a wire control brake system based on digital twins, characterized by: include: Establishing a virtual platform that includes a physical model and a digital twin model of the brake-by-wire system, wherein the virtual platform integrates the computational model, simulation data, and physical entities on the same platform; Constructing a virtual-reality interaction interface on the virtual platform to achieve real-time interaction between physical model data and digital twin model data; A multi-objective optimization approach is used to optimize the brake-by-wire system based on a deep deterministic policy gradient algorithm. Based on the driver's driving intention, the system coordinates the distribution of friction braking and electric motor force while ensuring braking safety. A physics engine is used to simulate hydraulic line pressure fluctuations in real time, and to optimize brake energy recovery and system response characteristics under ABS activation conditions: Optimizing the braking energy recovery amount and system response characteristics under the ABS activation condition includes: Coordinates friction braking and electric motor force distribution under ABS activation conditions, adjusting the solenoid valve state and hydraulic cylinder drive according to changes in wheel cylinder pressure; When the wheel cylinder pressure rises and approaches the target value, the solenoid valve opens to connect the hydraulic cylinder control reservoir to the brake master cylinder; When the wheel cylinder pressure drops and the brake pressure decreases, the solenoid valve is closed to disconnect the oil circuit between the reservoir and the brake master cylinder, and at the same time the hydraulic cylinder is driven to adjust the oil pressure; The dynamic distribution of electric motor force and hydraulic braking force is achieved through the following formula: , in: is the ratio of the electric braking force to the total braking force; is the slip rate of the i-th wheel; The time after ABS activation, in seconds; is the time coefficient, ranging from 0.5 to 1.5, which controls the motor power recovery speed; is the road adhesion coefficient. In the conditional judgment statement, "if" indicates the condition and "max" indicates the maximum value. Represents a negative exponential decay function, which gradually approaches 0 as time increases.

2. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 1, characterized in that: The virtual platform including the physical model and digital twin model of the wire control brake system includes: Establishing a physical model of a brake-by-wire system, wherein the physical model includes a vehicle kinematic model, a hydraulic model, and a wheel brake dynamics model; Establishing a digital twin model of the brake-by-wire system, the digital twin model including a digital environment, a virtual hydraulic system, and a virtual motor system; Importing the physical model into a virtual laboratory built based on MATLAB and Simulink, and obtaining brake pedal displacement and hydraulic line pressure data according to the physical model; The brake pedal displacement and hydraulic line pressure data are stored in a physical engine to obtain a training sample set for updating the digital twin model parameters.

3. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 2, characterized in that: The vehicle kinematic model is used to obtain the total deceleration of the target vehicle; the hydraulic model is used to obtain the pressure value of the hydraulic cylinder; and the wheel braking dynamics model is used to obtain the slip rate of the target vehicle's wheel and determine whether the wheel enters the ABS safety area.

4. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 1, characterized in that: The construction of the virtual-reality interaction interface includes: Equipped with a physical-to-digital data acquisition interface to collect real-time sensor data such as brake pedal displacement, wheel speed, hydraulic pressure, motor speed, motor current, and vehicle acceleration; Implement the data preprocessing module to perform noise filtering, outlier detection, data compensation and data synchronization on the collected sensor data; Configure the digital-to-physical control execution interface to convert the control instructions into motor torque instructions, brake master cylinder pressure instructions and ABS regulating valve control instructions, and realize control through the corresponding execution unit.

5. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 1, characterized in that: The multi-objective optimization of the wire control brake system based on the deep deterministic policy gradient algorithm includes: Design the control state space, including states related to braking demand, vehicle state, system state, and environment; Design the control action space, including motor braking torque command, hydraulic brake pressure command and state switching flag; Design a multi-objective reward function, including energy recovery efficiency, braking performance, comfort, safety, and system response terms; Build an experience replay pool to store control experience and improve sample utilization efficiency based on priority sampling strategy; The dual-network architecture is used to separate strategy evaluation and improvement, ensuring the stability of the learning process.

6. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 5, characterized in that: The weight of the multi-objective reward function adopts an adaptive adjustment mechanism to dynamically adjust the weight ratio of each indicator according to different braking conditions: In emergency braking conditions, the safety weight is increased and the energy recovery weight is reduced; Under normal braking conditions, the energy recovery weight is increased and the comfort weight is maintained at a moderate level; Under ABS activation conditions, safety is weighted the highest and energy recovery is weighted the lowest. Under low-speed conditions, increase the comfort weighting and system response weighting.

7. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 5, characterized in that: The experience replay pool adopts a multi-level structure design, including: The short-term pool is used to store the most recently generated experience samples, with a capacity of 100; The medium-term pool is used to store experience samples with a medium time span and has a capacity of 500; Long-term pool, used to store long-term valid experience samples, with a capacity of 2000; A priority calculation strategy based on time difference error is used to improve the sampling probability of rare working condition samples and samples with rich gradient information.

8. The multi-objective optimization method for a wire control brake system based on digital twin according to claim 1, characterized in that: The real-time simulation of hydraulic pipeline pressure fluctuations using a physical engine includes: Use the PX-OSG plug-in for simulation and implement real-time data exchange between the physical engine and the digital twin model based on the WebSocket protocol; Real-time simulation of hydraulic pipeline pressure fluctuations based on physical model data and digital twin model data; The physical model data includes ABS pressure, friction brake wheel cylinder pressure, motor force and system temperature; The digital twin model data includes brake pedal displacement, motor torque, regenerative current and line pressure.

9. A multi-objective optimization system for a wire control brake system based on digital twins, used to implement the method according to any one of claims 1 to 8, characterized in that: include: Physical model building module, used to build the physical model of the wire control brake system, including vehicle kinematic model, hydraulic model and wheel brake dynamics model; A digital model building module, used to establish a digital twin model of the brake-by-wire system, including a virtual laboratory construction unit, a training sample collection unit, a neural network model training unit, and a digital twin mapping unit; The virtual-reality interaction interface module is used to realize real-time interaction between physical model data and digital twin model data; Optimization target setting module, used to set and manage multiple optimization indicators for brake energy recovery and braking safety; Weight optimization module, used to determine and optimize the weight ratio of each optimization indicator to achieve multi-objective balance; Real-time optimization module, used to search for the optimal brake pedal displacement and brake motor torque in real time under actual working conditions; The brake control module is used to control the brake energy recovery and braking safety of the vehicle's wire control brake system based on the optimization results, and coordinate the distribution of friction braking and electric machine power under ABS activation conditions.

Images