A high-efficiency vehicle braking system based on hybrid road surface adhesion recognition

Abstract

This invention discloses a high-efficiency braking system for automobiles based on hybrid recognition of road surface adhesion conditions, comprising: a road surface recognition module, an actuator subsystem, a sensor subsystem, and a vehicle integrated planning controller. The road surface recognition module initially identifies the current road surface adhesion information and obtains an initial estimate of the road surface adhesion coefficient. The vehicle integrated planning controller collects vehicle state information through the sensor subsystem, applies different recognition strategies, and iteratively optimizes the results using an extended Kalman filter in the road surface recognition module, outputting the estimated road surface adhesion information and vehicle operating state. The actuator controls the drive-by-wire braking system to achieve high-efficiency braking. This invention enables advance recognition of road surface adhesion information ahead, thereby achieving high-efficiency braking in intelligent vehicles and solving the expected functional safety problems caused by delayed recognition of road surface adhesion information.

Description

Technical Field

[0001] This invention belongs to the field of automotive electronic control technology, and mainly involves the real-time acquisition and accuracy of road information, and the realization of efficient vehicle braking based on the accurate road information. Background Technology

[0002] With the continuous development and improvement of intelligent vehicle technology, modern cars are gradually gaining the ability to acquire and perceive information about their surrounding environment. Intelligent vehicles use various sensors and devices to obtain information about the surrounding environment, such as road conditions, weather conditions, traffic conditions, pedestrians, and obstacles, thereby making more accurate and safer decisions to improve driving safety and experience. Furthermore, intelligent vehicles can utilize high-precision maps and navigation systems to provide drivers with optimal routes and speeds, avoiding congestion and dangerous driving situations. Simultaneously, intelligent vehicles can also use sensors such as millimeter-wave radar, lidar, and cameras to monitor surrounding vehicles, pedestrians, and obstacles in real time and react promptly, such as slowing down, changing lanes, or stopping, to avoid traffic accidents.

[0003] While intelligent vehicles can achieve autonomous driving and accident avoidance through various sensors and technologies, numerous safety hazards remain in practical applications. For example, when a road obstacle or building suddenly appears on a sidewalk, the intelligent vehicle may react too slowly and fail to brake effectively, leading to an accident. Secondly, intelligent vehicles rely heavily on onboard computers and software systems. In complex road conditions, prolonged processing time for the onboard computer can cause delays in issuing control commands, easily resulting in driving accidents. Therefore, malfunctions or software problems can prevent the vehicle from operating normally, placing the driver and passengers in dangerous situations. Furthermore, the manufacturing and selling costs of sensors such as radar and cameras in intelligent vehicles are very high, making the cost of experiencing the technology and convenience offered by intelligent driving too burdensome for ordinary consumers. This makes intelligent vehicles a luxury for most consumers. Therefore, minimizing the use of sensors and using algorithms to address the cost, safety, and reliability issues of intelligent vehicles, without compromising the consumer's intelligent driving experience and driving safety, will remain a research direction for a long time to come.

[0004] The development of intelligent vehicles has also placed higher demands on the expected functional safety of braking: First, the braking system of intelligent vehicles must possess high reliability to ensure the vehicle can stop safely under any circumstances. Second, the braking system of intelligent vehicles needs to have fault diagnosis and warning functions to promptly detect and handle braking system faults. When a braking system malfunctions, the intelligent vehicle should be able to promptly alert the driver and provide relevant fault diagnosis information. Third, the braking system of intelligent vehicles should be able to adaptively adjust braking force according to different driving scenarios and road conditions to ensure vehicle stability and safety during braking. In rainy or slippery road conditions, the braking system should be able to adaptively adjust braking force to prevent vehicle skidding or loss of control. Finally, the braking system of intelligent vehicles should work in conjunction with other intelligent systems such as anti-lock braking systems (ABS) and electronic stability control systems (ESP) to ensure vehicle stability and safety during braking.

[0005] Current vehicle perception systems in intelligent vehicles primarily focus on acquiring information about the driving environment, while their acquisition and utilization of road surface adhesion information are insufficient. The system lacks adequate advance knowledge of information such as the road surface adhesion coefficient, thus failing to effectively adjust the vehicle's operating state in advance. When road conditions change, the system cannot quickly and effectively adjust the vehicle's driving and braking states based on changes in road parameters. Furthermore, relying solely on visual recognition of road surface information is often inaccurate, resulting in significant lag in the vehicle's active power control. This fails to adequately meet the expected functional safety requirements of the vehicle and reduces passenger comfort. Therefore, a control method that can accurately identify road surface parameters under different road conditions and precisely control the target vehicle's braking is currently lacking. Summary of the Invention

[0006] The present invention aims to address the shortcomings of the existing technology by proposing a high-efficiency vehicle braking system based on hybrid recognition of road surface adhesion. This system utilizes existing vehicle sensing cameras to collect road surface information and refines it through different control strategies under different operating conditions, thereby achieving the goal of high-efficiency vehicle braking.

[0007] To solve the technical problem, the technical solution of the present invention is as follows:

[0008] The present invention provides a high-efficiency vehicle braking system based on hybrid road surface adhesion recognition, which includes: a road surface recognition module, an actuator subsystem, a sensor subsystem, and a vehicle integrated planning controller.

[0009] The road surface recognition module includes: an on-board perception camera module and a road surface type recognition system;

[0010] The sensing camera module is used to collect road surface environment information in real time and transmit it to the road surface type identification system;

[0011] The road surface type identification system compares the collected road surface environment information with various typical original road surfaces to determine the current road surface type and transmits it to the machine learning training and testing system. Each road surface type represents a different estimated range of road surface adhesion coefficient.

[0012] The machine learning training and testing system calculates the road adhesion coefficient estimation interval corresponding to the current road surface type, obtains the initial estimated value of the road adhesion coefficient, and sends it to the vehicle integrated planning controller.

[0013] The actuator subsystem includes a drive motor and a vehicle brake actuator;

[0014] The sensor subsystem is connected to the vehicle integrated planning controller; the sensor subsystem includes: a steering angle sensor, a steering wheel torque sensor, an electromagnetic wheel speed sensor, a yaw rate sensor, a brake pressure sensor, an acceleration sensor, and a wheel speed sensor;

[0015] The steering angle sensor is used to acquire the steering state and steering angle of the steering wheel in real time.

[0016] The steering wheel torque sensor is used to detect the total resistance torque / return torque of the steering system in real time.

[0017] The electromagnetic wheel speed sensor is used to collect the pulse signals emitted when the wheel rotates in real time to obtain the speed of the drive wheel.

[0018] The GPS sensor is used to collect the longitudinal speed of the vehicle in real time.

[0019] The yaw rate sensor is used to collect the yaw rate at the vehicle's center of gravity in real time.

[0020] The brake pressure sensor is used to acquire the vehicle braking status in real time and collect the brake wheel cylinder pressure in the brake-by-wire mechanism; the vehicle braking status is the working state of the brake when the vehicle encounters a sudden situation, and is divided into half braking state and full braking state according to the change of master cylinder pressure.

[0021] The acceleration sensor is used to acquire the vehicle's lateral acceleration and longitudinal deceleration in real time;

[0022] The wheel speed sensor is used to obtain tire angular velocity information in real time;

[0023] The decision layer of the vehicle integrated planning controller monitors the vehicle's steering and braking status in real time based on the information feedback from the steering angle sensor and the brake pressure sensor. It uses an adaptive extended Kalman filter to estimate the vehicle's driving status parameters for the next moment, and iteratively updates the vehicle's driving status parameters at the current moment. The real-time updated driving status parameters are then input into the internal road adhesion coefficient estimator, thereby identifying and updating the road adhesion coefficient based on the initial estimated value.

[0024] The decision layer of the vehicle integrated planning controller calculates the control quantity at the current moment based on the vehicle's current driving state and the road surface adhesion coefficient.

[0025] The execution layer of the vehicle integrated planning controller issues corresponding control commands to the actuator subsystem based on the control quantity at the current moment, thereby realizing real-time braking of the vehicle.

[0026] The high-efficiency braking system for automobiles based on hybrid recognition of road surface adhesion conditions described in this invention is also characterized in that the vehicle integrated planning controller achieves real-time braking of the vehicle in the following steps.

[0027] Step 1: The decision layer of the vehicle integrated planning controller determines whether the vehicle is currently performing a steering operation based on the obtained steering wheel angle. If so, it executes the vehicle state and road adhesion coefficient identification strategy based on steering return torque and then proceeds to Step 2; otherwise, it proceeds to Step 4.

[0028] Step 2: The decision layer of the vehicle integrated planning controller converts the steering wheel angle measured by the steering angle sensor into the front wheel angle, and uses the front wheel angle as the control input, and the measured yaw rate and lateral acceleration as the observations, and inputs them together into the adaptive extended Kalman filter observer for processing to obtain the current driving state parameters of the vehicle after filtering.

[0029] Step 3: Using the estimated initial value of the road surface adhesion coefficient as the initial value of the iteration, the filtered vehicle state parameters at the current moment are used as the observation values ​​and input together into the road surface adhesion coefficient estimator for processing. After obtaining the estimated value of the road surface adhesion coefficient at the next moment, step 9 is executed.

[0030] Step 4: The decision layer of the vehicle integrated planning controller determines whether the vehicle is braking based on the obtained vehicle braking status. If so, it executes the vehicle status and road adhesion coefficient identification strategy based on the longitudinal dynamics model and then executes Step 5; otherwise, it executes Step 8.

[0031] Step 5: The decision layer of the vehicle integrated planning controller calculates the slip ratio based on the vehicle's longitudinal speed obtained from the GPS sensor and the drive wheel speed obtained from the electromagnetic wheel speed sensor. It then determines whether the slip ratio under the current driving state is lower than the set threshold. If it is lower, proceed to step 6; otherwise, proceed to step 7.

[0032] Step 6: The decision layer of the vehicle integrated planning controller takes the slip ratio under the current driving state as the small slip ratio and uses a linear tire model to estimate the road adhesion coefficient in real time. After obtaining the estimated value of the road adhesion coefficient at the next moment, it executes step 9.

[0033] Step 7: The decision layer of the vehicle integrated planning controller uses a steady-state tire model to estimate the road adhesion coefficient in real time, and then executes step 9 after obtaining the estimated value of the road adhesion coefficient at the next moment.

[0034] Step 8: The execution layer of the vehicle integrated planning controller controls the brake pedal according to the initial value of the road surface adhesion coefficient and the driver's operation command to achieve vehicle braking.

[0035] Step 9: The decision layer of the vehicle integrated planning controller calculates the target braking deceleration of the vehicle under the current operating state based on the real-time updated road adhesion coefficient and vehicle operating state; the execution layer of the vehicle integrated planning controller calculates the total braking force required by the actuator subsystem based on the vehicle's target braking deceleration, the longitudinal deceleration measured by the acceleration sensor, and the wheel cylinder pressure measured by the braking pressure sensor, thereby dynamically distributing the output torque of the vehicle's drive motor and the wheel cylinder pressure of the four brake wheel cylinders of the vehicle's brake actuator to achieve real-time braking of the vehicle.

[0036] Compared with existing technologies, the innovation and practicality of this invention are reflected in:

[0037] 1. This invention utilizes existing vehicle-mounted sensors and cameras to effectively capture and accurately obtain road surface adhesion information on the current road surface, based on limited hardware. It then efficiently and accurately calculates the output braking pressure and motor reversing torque required for vehicle braking based on the accurate road surface information, thereby achieving more efficient braking control compared to traditional vehicles.

[0038] 2. This invention fully considers the various working conditions faced by vehicles when recognizing road information, and designs different recognition and control strategies according to different working conditions, ensuring real-time and accurate control of the vehicle's wheel braking force and improving vehicle driving safety.

[0039] 3. The control strategy of this invention caters to the future development trend of intelligent connected vehicles and autonomous vehicles. The multi-camera used is compatible with intelligent driving control algorithms and can be installed in various types of vehicles such as fuel vehicles, pure electric vehicles, hybrid vehicles, and hydrogen fuel cell vehicles, and has broad application prospects. Attached Figure Description

[0040] Figure 1 This is a flowchart illustrating the overall framework of the present invention;

[0041] Figure 2 This is a schematic diagram illustrating the road surface estimation principle of the present invention.

[0042] Figure 3 This is a flowchart of the process for identifying road surface adhesion coefficient based on longitudinal dynamics in this invention. Detailed Implementation

[0043] In this embodiment, a high-efficiency vehicle braking system based on hybrid road surface adhesion recognition includes: a road surface recognition module, an actuator subsystem, a sensor subsystem, and a vehicle integrated planning controller; wherein, the road surface recognition module includes an information acquisition mechanism; the actuator subsystem includes a drive motor and a vehicle brake actuator;

[0044] The sensor subsystem is connected to the vehicle integrated planning controller; the sensor subsystem includes: steering angle sensor, steering wheel torque sensor, electromagnetic wheel speed sensor, yaw rate sensor, brake pressure sensor, acceleration sensor and wheel speed sensor;

[0045] Among them, the steering angle sensor is used to obtain the steering state and steering angle of the steering wheel in real time;

[0046] The steering wheel torque sensor is used to detect the total resistance torque / return torque of the steering system in real time;

[0047] Electromagnetic wheel speed sensors are used to collect pulse signals emitted by the wheels in real time to obtain the speed of the drive wheels;

[0048] GPS sensors are used to collect the longitudinal speed of the vehicle in real time;

[0049] The yaw rate sensor is used to collect the yaw rate at the vehicle's center of gravity in real time;

[0050] Brake pressure sensors are used to acquire the vehicle's braking status in real time and to collect the pressure of the brake wheel cylinders in the brake-by-wire mechanism;

[0051] Among them, the vehicle braking status refers to the working status of the brakes when the vehicle encounters a sudden situation.

[0052] Accelerometers are used to acquire the vehicle’s lateral acceleration and longitudinal deceleration in real time;

[0053] Wheel speed sensors are used to obtain tire angular velocity information in real time;

[0054] like Figure 1 As shown, the perception camera module is used to collect road environment information in real time and transmit it to the road recognition module. The road recognition module calculates the initial value of the road adhesion coefficient based on the road environment information and sends it to the vehicle integrated planning controller.

[0055] The decision layer of the vehicle integrated planning controller monitors the vehicle's steering and braking status in real time based on the information feedback from the steering angle sensor and the brake pressure sensor. It also uses an adaptive extended Kalman filter to estimate the vehicle's driving status parameters for the next moment, and iteratively updates the vehicle's current driving status parameters. The real-time updated driving status parameters are then input into the internal road adhesion coefficient estimator to identify and update the road adhesion coefficient.

[0056] The decision layer of the vehicle integrated planning controller calculates the control quantity at the current moment based on the vehicle's current driving state and the road adhesion coefficient.

[0057] The execution layer of the vehicle integrated planning controller issues corresponding control commands to the actuator subsystem based on the control quantity at the current moment, thereby realizing real-time braking of the vehicle.

[0058] In this embodiment, as Figure 1 As shown, the vehicle integrated planning controller achieves real-time braking of the vehicle in the following steps;

[0059] Step 1, as follows Figure 2 As shown, a multi-view camera is used to perform preliminary identification of typical road surfaces, and an initial estimated value is selected based on the preliminary identification by the perception camera.

[0060] Step 1.1. The sensing camera module collects road surface environment information in real time and transmits it to the road surface type identification system.

[0061] Step 1.2. The road surface type identification system classifies the original road surface into three typical types: dry asphalt road, wet asphalt road, and icy / snowy road. The system compares the collected road environment information with these typical original road surfaces, imports a large number of photographs of these typical road surfaces into the system, and classifies and labels each type of road surface by establishing training and testing datasets. The training dataset distinguishes different road surface environments, assigns different display labels to different road surfaces, and uses different colors to differentiate them.

[0062] Step 1.3. Using the machine learning training and testing system integrated into the vehicle ECU, compare the pixels of the road in the image with the pixels of the training model one by one. By predicting each pixel in the image, a large number of predicted labels are obtained, and each label corresponds to a specific road surface type. Each road surface type represents a different estimation range for the road surface adhesion coefficient.

[0063] Step 1.4. The machine learning training and testing system calculates the estimated interval of the road adhesion coefficient corresponding to the current road surface type; it counts the number of predicted labels for each pixel in the image and uses the predicted label with the most occurrences as the road surface type prediction result for the entire image to obtain the initial estimated value μ of the road adhesion coefficient. 01 And send it to the vehicle integration planning controller.

[0064] Step 2: The decision layer of the vehicle integrated planning controller determines whether the vehicle is currently performing a steering operation based on the obtained steering wheel angle. If so, it executes the vehicle status and road adhesion coefficient identification strategy based on steering return torque and then proceeds to Step 3; otherwise, it proceeds to Step 5.

[0065] Step 3: The decision layer of the vehicle integrated planning controller converts the steering wheel angle measured by the steering angle sensor into the front wheel steering angle, and uses the front wheel steering angle as the control input. The measured yaw rate and lateral acceleration are used as observations, and are input together into the adaptive extended Kalman filter observer for processing to obtain the filtered driving state parameters of the vehicle at the current moment. The specific implementation steps are as follows:

[0066] A two-degree-of-freedom vehicle model is established, and the parameters of the steady-state tire model are corrected through experimental data. The relationship curve between tire self-alignment torque and sideslip angle under typical road conditions is output and stored in the road adhesion coefficient estimator.

[0067] Establish a two-degree-of-freedom kinematic model of the vehicle based on a steady-state tire model:

[0068]

[0069] In equation (1): x is a state variable and x = [β, r] T β is the sideslip angle of the vehicle's center of gravity, r is the yaw rate of the vehicle, and T is the transpose; Let x be the derivative of the state variable; specifically:

[0070]

[0071]

[0072] Each state variable is obtained by sensor measurement and is used as the raw input parameter in the Kalman filter, where: k f k r... z This is the moment of inertia of the car.

[0073] Using the sideslip angle and yaw rate as state variables, and the yaw rate and lateral acceleration as observation variables, and the front wheel steering angle as the control input, these parameters are input into the extended Kalman filter to obtain the vehicle's accurate driving state parameters at the current moment after filtering.

[0074] Step 4: Using the initial value of the road surface adhesion coefficient as the initial value for iteration, the filtered vehicle state parameters at the current moment are used as observations and input together with the road surface adhesion coefficient estimator for processing to obtain the estimated value of the road surface adhesion coefficient at the next moment. Then, step 10 is executed, specifically:

[0075] Step 4.1. Establish a simplified model of the electric power steering system, namely:

[0076]

[0077] Step 4.2. Torque information and power steering motor current are obtained from the steering wheel torque sensor in the EPS system. An interference observer is constructed to obtain the total steering resistance torque of the tires. The steering resistance torque is converted into a tire self-centering torque sequence using a six-component force model. The difference between the real-time estimated tire self-centering torque and the weighted calculated tire self-centering sequence is calculated to obtain the self-centering torque deviation sequence. The self-centering torque deviation is used as a feedback observation term and multiplied by the corresponding weighting coefficients to identify the initial value μ of the road adhesion coefficient. 02 .

[0078] Step 4.3. Apply the estimated initial value μ obtained in Step 1. 01 The estimated initial value μ obtained in step 4.1 is used as the first input variable. 02 As the second input variable, the average value μ0 of the first and second input variables is obtained and used as the original input u. k The Kalman filter principle is used to iteratively estimate the parameters of the road surface.

[0079] Step 4.4. Construct a nonlinear observer based on the vehicle dynamics equations. Use the real-time lateral force and restoring torque calculated from the parameters measured by the sensors as compensation to identify the derivatives of the lateral velocity and the road adhesion coefficient. Then, use an integrator to calculate the lateral velocity and the third road adhesion coefficient μ. 03 The nonlinear observer is:

[0080]

[0081] In equation (5), L 11 L 12 L 21 L 22The system feedback gain has the following values:

[0082]

[0083] Step 4.5. The identification results of the two methods are optimized by weighted fusion of root mean square difference to obtain a more accurate tire road adhesion coefficient.

[0084] Step 5: The decision layer of the vehicle integrated planning controller determines whether the vehicle is braking based on the obtained vehicle braking status. If so, it executes the vehicle status and road adhesion coefficient identification strategy based on the longitudinal dynamics model and then executes step 6; otherwise, it executes step 9.

[0085] Step 6, as follows Figure 3 As shown, the decision layer of the vehicle integrated planning controller calculates the slip ratio based on the vehicle's longitudinal speed obtained from the GPS sensor and the drive wheel speed obtained from the electromagnetic wheel speed sensor. It then determines whether the slip ratio under the current driving state is lower than a set threshold. If it is lower, step 7 is executed; otherwise, step 8 is executed. Specifically:

[0086] The vehicle's longitudinal speed v is obtained from the vehicle's native GPS sensor. x The wheel speed sensor obtains tire angular velocity information and transmits it to the vehicle ECU to calculate the slip ratio. x Calculate, where,

[0087] Step 7: Set the slip ratio threshold to 20%. When the slip ratio s x When the slip ratio is below 20%, the decision layer of the vehicle integrated planning controller takes the slip ratio under the current driving state as the small slip ratio and uses a linear tire model to estimate the road adhesion coefficient in real time, thereby obtaining the estimated value of the road adhesion coefficient at the next moment, and then executes step 10; specifically including:

[0088] When the system determines that the current operating state is a small slip ratio, the tire slip angle is small, and the slip angle and the self-aligning torque are approximately linearly related.

[0089] The linear tire model is represented as:

[0090] F y =K a α (7)

[0091] In equation (7), K a The tire side stiffness is obtained by fitting the experimental data, where α is the tire side slip angle. The tire lateral force F under small side slip angle conditions is obtained using equation (7). y .

[0092] Combined with the above-obtained total tire lateral force F yThe tire-road adhesion coefficient μ is calculated using equation (8) based on the numerical value and the vehicle weight m. 04 .

[0093]

[0094] In equation (7), g is the gravitational acceleration, the magnitude of which is determined according to the local conditions.

[0095] Step 8: The decision layer of the vehicle integrated planning controller uses a steady-state tire model to estimate the road adhesion coefficient in real time, thereby obtaining the estimated value of the road adhesion coefficient at the next moment, and then executes step 10.

[0096] The steady-state tire model outputs the relationship curve between longitudinal force and slip ratio. Using previously obtained parameters, a recursive least squares method is employed to estimate the road adhesion coefficient using a lookup table approach. The real-time estimated road adhesion coefficient μ for the next time step is then output. 05 .

[0097] Step 9: The execution layer of the vehicle integrated planning controller controls the brake pedal according to the initial value of the road adhesion coefficient and the driver's operation command to achieve vehicle braking.

[0098] The prediction model is updated based on the precise road surface parameters obtained above, and the motor reversal torque and brake wheel cylinder pressure are calculated according to the predictive control scheme. Based on the control variables calculated by the integrated planning controller, a vehicle deceleration control method based on the tire inverse model is designed to dynamically and individually control the motor output torque and the wheel cylinder pressure of the four brake wheel cylinders of the target vehicle, ultimately achieving the goal of real-time and efficient braking of the vehicle.

[0099] Step 10: The decision layer of the vehicle integrated planning controller calculates the target braking deceleration of the vehicle under the current operating state based on the real-time updated road adhesion coefficient and vehicle operating state; the execution layer of the vehicle integrated planning controller calculates the total braking force required by the actuator subsystem based on the vehicle's target braking deceleration, the longitudinal deceleration measured by the acceleration sensor, and the wheel cylinder pressure measured by the brake pressure sensor, thereby dynamically distributing the output torque of the vehicle's drive motor and the wheel cylinder pressure of the four brake wheel cylinders of the vehicle's brake actuator to achieve real-time braking of the vehicle.

[0100] Step 11: When the system does not detect braking operation, the control strategy automatically enters the worst-case option. At this time, the system directly uses the preliminary road adhesion coefficient collected by the sensor and inputs it directly into the control strategy. This road adhesion coefficient has not been accurately processed by Kalman filtering. At this time, the driver mainly relies on judgment to control the brake pedal to achieve accurate braking of the vehicle.

[0101] In this embodiment, as Figure 2 As shown, the extended Kalman filter process is as follows:

[0102] S1. Set initial data:

[0103] Set the number of iterations to k, and initialize k = 0; set the initial condition, i.e., the error covariance matrix P. k-1 and initial state input

[0104] S2. Perform state prior prediction:

[0105] The system uses sensors to acquire estimated initial values ​​of vehicle longitudinal velocity, longitudinal acceleration, and front wheel steering angle, which are then transmitted to the state prediction module to obtain prior estimates for that moment. These estimates are then sent to the correction module for refinement. Here, A is the system's state transition matrix, B is the input control matrix, and w... k This represents the process noise at time k. It is the result of the prediction from the previous state. This is the result of the optimal estimate from the previous state, where Q is the system process noise w. k The covariance matrix.

[0106] S3. The prior estimates obtained by the prediction module are combined with the observed values ​​for correction and update, including the weighted update of state variables and error terms. These are then fed back into the prediction module to form a continuously updating and iterative logical closed loop.

[0107] Among them, K k Let C be the Kalman filter gain matrix, C be the observation matrix, and R be the measurement noise v. k The covariance matrix. The state at time k The predicted covariance.

[0108] S4. An adaptive extended Kalman filter-based algorithm for tire-road adhesion coefficient identification is designed. Based on the vehicle dynamics equations and the modified steady-state tire model, an adaptive extended Kalman filter observer is designed to accurately identify the tire-road adhesion coefficient, using lateral acceleration and self-correcting torque as observables. Specifically:

[0109] S5. Set the number of cycles to k, and initialize k = 0. Utilize the lateral acceleration, steering wheel torque, steering wheel angle, longitudinal vehicle speed, and yaw rate signals obtained above. Use the sideslip angle α as the system input, and the lateral velocity v... y The average road surface adhesion coefficient μ0 is taken as the system state variable, and the lateral acceleration a is taken as the system state variable. y and front wheel return positive torque M zf For feedback observations.

[0110] S6. Use equation (9) to obtain the set of state variables x for the kth cycle. k For initial prior state estimation

[0111]

[0112] In equation (9), E[x k ] represents the population mean function of the first and second input variables;

[0113] S7. Use equation (10) to obtain the input variable u of the mean value of the kth cycle. k and initial prior state estimation The prior covariance matrix S0:

[0114]

[0115] In equation (10), T is the transpose symbol;

[0116] S8. Use equation (11) to obtain the set of state variables x for the kth cycle. k Prior estimates

[0117]

[0118] In equation (11), A represents the first coefficient matrix and B represents the second coefficient matrix;

[0119] S9. Obtain the prior estimate using equation (12) The prior covariance matrix P k+1|k :

[0120] P k+1|k =AP k|k A T +Q (12)

[0121] In equation (12), Q represents the system process noise covariance matrix, and A T This represents the transpose of the first coefficient matrix;

[0122] S10. Use equation (13) to obtain the Kalman gain K for the (k+1)th cycle. k+1|k :

[0123] K k+1|k =C T P k+1|k (C T P k+1|k +R) (13)

[0124] In equation (13), R represents the measurement noise covariance matrix of the system; C represents the third coefficient state matrix, C TThis represents the transpose of the third coefficient state matrix;

[0125] S11. Using equation (14), the error covariance matrix S of the (k+1)th cycle is obtained. k+1 :

[0126] S k+1 =[D T R -1 (I-CK k+1 )D] -1 (14)

[0127] In equation (14), D is the fourth coefficient state matrix. T Let I be the transpose of the fourth coefficient state matrix, and let I denote the identity matrix.

[0128] S12. Obtain the input variable u using equation (15) k State estimation in the (k+1)th cycle

[0129]

[0130] In equation (15), y k+1 u represents the output variable of the (k+1)th iteration of the loop. k+1 This represents the second input vector in the (k+1)th iteration.

[0131] S13. Using equation (16), obtain the set of state variables x for the (k+1)th iteration. k+1 :

[0132] x k+1 =Ax k +Bu k +w k (16)

[0133] In equation (16), w k This represents the system process noise vector in the k-th cycle.

[0134] S14. Use equation (17) to obtain the output vector y of the kth cycle. k :

[0135] y k =Cx k +Du k +v k (17)

[0136] In equation (17), v k This is represented as the measurement noise signal in the k-th cycle;

[0137] In this embodiment, the identification of road surface adhesion coefficient involves data calculations from various aspects. Depending on the actual computing capabilities of the on-board equipment, offline calculations should be performed within the equipment, or the data can be uploaded and downloaded for cloud computing through the 5G cloud computing information stream provided by the operator. Finally, the information spectrum with the corresponding road surface adhesion information is uploaded to the cloud database.

[0138] In this embodiment, under ideal conditions (i.e., when a large number of vehicles equipped with the identification and control system of this invention are operating on various parts of the road), the big data system formed by this method can quickly, accurately, and in real time refresh the road surface adhesion information. This allows all passing vehicles to plan ahead based on different road surface information, which has significant reference value for the development of intelligent and connected vehicles. At the same time, when the cloud system suddenly detects a significant change in the road surface adhesion information, an abnormal situation occurs, which can easily lead to traffic accidents. Municipal and traffic management departments can promptly detect and handle the situation, making a positive contribution to intelligent road management.

[0139] In summary, the method of this invention mainly calculates and identifies the road surface adhesion coefficient based on a discrete Kalman filter prediction model, and couples the vehicle-mounted sensing camera, the vehicle state observer based on discrete Kalman filtering, and other information such as the road surface adhesion coefficient in real time. By combining existing sensors and control modules of traditional vehicles to collect vehicle motion and road surface signals, and after predicting, measuring, and analyzing the collected signals, the corresponding accurate road surface adhesion information is obtained. The controller outputs precise control signals to control the vehicle's braking in real time, protecting the lives of occupants and pedestrians, and has strong practicality and feasibility.

Claims

1. A high-efficiency vehicle braking system based on hybrid recognition of road surface adhesion conditions, characterized in that, include: Road surface recognition module, actuator subsystem, sensor subsystem, and vehicle integrated planning controller; The road surface recognition module includes: an on-board perception camera module and a road surface type recognition system; The sensing camera module is used to collect road surface environment information in real time and transmit it to the road surface type identification system; The road surface type identification system compares the collected road surface environment information with various typical original road surfaces to determine the current road surface type and transmits it to the machine learning training and testing system. Each road surface type represents a different estimated range of road surface adhesion coefficient. The machine learning training and testing system calculates the road adhesion coefficient estimation interval corresponding to the current road surface type, obtains the initial estimated value of the road adhesion coefficient, and sends it to the vehicle integrated planning controller. The actuator subsystem includes a drive motor and a vehicle brake actuator; The sensor subsystem is connected to the vehicle integrated planning controller; the sensor subsystem includes: a steering angle sensor, a steering wheel torque sensor, an electromagnetic wheel speed sensor, a yaw rate sensor, a brake pressure sensor, an acceleration sensor, and a wheel speed sensor; The steering angle sensor is used to acquire the steering state and steering angle of the steering wheel in real time. The steering wheel torque sensor is used to detect the total resistance torque / return torque of the steering system in real time. The electromagnetic wheel speed sensor is used to collect the pulse signals emitted when the wheel rotates in real time to obtain the speed of the drive wheel. The GPS sensor is used to collect the longitudinal speed of the vehicle in real time. The yaw rate sensor is used to collect the yaw rate at the vehicle's center of gravity in real time. The brake pressure sensor is used to acquire the vehicle braking status in real time and collect the brake wheel cylinder pressure in the brake-by-wire mechanism; the vehicle braking status is the working state of the brake when the vehicle encounters a sudden situation, and is divided into half braking state and full braking state according to the change of master cylinder pressure. The acceleration sensor is used to acquire the vehicle's lateral acceleration and longitudinal deceleration in real time; The wheel speed sensor is used to obtain tire angular velocity information in real time; The decision layer of the vehicle integrated planning controller monitors the vehicle's steering and braking status in real time based on the information feedback from the steering angle sensor and the brake pressure sensor. It uses an adaptive extended Kalman filter to estimate the vehicle's driving status parameters for the next moment, and iteratively updates the vehicle's driving status parameters at the current moment. The real-time updated driving status parameters are then input into the internal road adhesion coefficient estimator, thereby identifying and updating the road adhesion coefficient based on the initial estimated value. The decision layer of the vehicle integrated planning controller calculates the control quantity at the current moment based on the vehicle's current driving state and the road surface adhesion coefficient. The execution layer of the vehicle integrated planning controller issues corresponding control commands to the actuator subsystem based on the control quantity at the current moment, thereby realizing real-time braking of the vehicle.

2. The high-efficiency vehicle braking system based on hybrid road surface adhesion recognition according to claim 1, characterized in that, The vehicle integrated planning controller achieves real-time braking of the vehicle according to the following steps; Step 1: The decision layer of the vehicle integrated planning controller determines whether the vehicle is currently performing a steering operation based on the obtained steering wheel angle. If so, it executes the vehicle status and road adhesion coefficient identification strategy based on steering return torque, and then executes Step 2. Otherwise, proceed to step 4; Step 2: The decision layer of the vehicle integrated planning controller converts the steering wheel angle measured by the steering angle sensor into the front wheel angle, and uses the front wheel angle as the control input, and the measured yaw rate and lateral acceleration as the observations, and inputs them together into the adaptive extended Kalman filter observer for processing to obtain the current driving state parameters of the vehicle after filtering. Step 3: Using the estimated initial value of the road surface adhesion coefficient as the initial value of the iteration, the filtered vehicle state parameters at the current moment are used as the observation values ​​and input together into the road surface adhesion coefficient estimator for processing. After obtaining the estimated value of the road surface adhesion coefficient at the next moment, step 9 is executed. Step 4: The decision layer of the vehicle integrated planning controller determines whether the vehicle is braking based on the obtained vehicle braking status. If so, it executes the vehicle status and road adhesion coefficient identification strategy based on the longitudinal dynamics model and then executes Step 5; otherwise, it executes Step 8. Step 5: The decision layer of the vehicle integrated planning controller calculates the slip ratio based on the vehicle's longitudinal speed obtained from the GPS sensor and the drive wheel speed obtained from the electromagnetic wheel speed sensor. It then determines whether the slip ratio under the current driving state is lower than the set threshold. If it is lower, then proceed to step 6. Otherwise, proceed to step 7; Step 6: The decision layer of the vehicle integrated planning controller takes the slip ratio under the current driving state as the small slip ratio and uses a linear tire model to estimate the road adhesion coefficient in real time. After obtaining the estimated value of the road adhesion coefficient at the next moment, it executes step 9. Step 7: The decision layer of the vehicle integrated planning controller uses a steady-state tire model to estimate the road adhesion coefficient in real time, and then executes step 9 after obtaining the estimated value of the road adhesion coefficient at the next moment. Step 8: The execution layer of the vehicle integrated planning controller controls the brake pedal according to the initial value of the road surface adhesion coefficient and the driver's operation command to achieve vehicle braking. Step 9: The decision layer of the vehicle integrated planning controller calculates the target braking deceleration of the vehicle under the current operating state based on the real-time updated road adhesion coefficient and vehicle operating state; the execution layer of the vehicle integrated planning controller calculates the total braking force required by the actuator subsystem based on the vehicle's target braking deceleration, the longitudinal deceleration measured by the acceleration sensor, and the wheel cylinder pressure measured by the braking pressure sensor, thereby dynamically distributing the output torque of the vehicle's drive motor and the wheel cylinder pressure of the four brake wheel cylinders of the vehicle's brake actuator to achieve real-time braking of the vehicle.

Images