Active lane changing obstacle avoidance control method based on road surface friction coefficient prediction

By combining visual and dynamic data to predict the road surface adhesion coefficient and actively intervene to avoid obstacles, the problem of difficult-to-estimate adhesion coefficient during high-speed driving is solved, and driving safety and comfort under complex road conditions are improved.

CN119078817APending Publication Date: 2024-12-06JILIN UNIVERSITY +1
View PDF 0 Cites 14 Cited by

Patent Information

Application Number
CN202411356657.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-27
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

When driving at high speed, especially on roads with low adhesion coefficient, it is difficult for vehicles to estimate the safe distance from the vehicle in front, resulting in an increase in collision risk. The existing active lane change and obstacle avoidance technology is safe and comfortable when driving under complex road conditions. Inadequate degree.

Method used

By combining vehicle vision and dynamic data, image processing and Kalman filtering technology are used to predict the road surface attachment coefficient in real time, and combined with vehicle dynamics model and driver's intentions, we predict the future trajectory of the vehicle. If there is potential safety risk, we actively intervene and execute it if there is a potential safety risk. Steer the action to avoid obstacles.

Benefits of technology

It significantly improves the driving safety of the vehicle and the comfort of the driving experience under complex road conditions, improves the accuracy and driving safety of active lane change and obstacle avoidance, is suitable for a variety of vehicle types, and enhances braking efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119078817A_ABST
    Figure CN119078817A_ABST
Patent Text Reader

Abstract

The invention relates to an active lane changing obstacle avoidance control method based on pavement friction coefficient prediction. Comprising the steps of 1, obtaining a road adhesion coefficient based on vehicle vision; 2, obtaining a road adhesion coefficient based on vehicle dynamics; 3, fusing the road adhesion coefficient based on the vehicle vision and the road adhesion coefficient based on the vehicle dynamics through a weighted average method to obtain a road adhesion coefficient; 4, sensing the environment; 5, determining a lane changing obstacle avoidance decision based on a safe distance model; and 6, controlling and executing a lane changing obstacle avoidance decision. According to the method, sensor data and environment information collected during vehicle driving are analyzed, the road surface visual adhesion coefficient is estimated by using an image processing technology, and the road surface visual adhesion coefficient is combined with a dynamic road surface adhesion coefficient measured by a vehicle sensor, so that a more accurate comprehensive road surface adhesion coefficient is synthesized; the comprehensive road adhesion coefficient is combined with a vehicle dynamics model and a driver intention to predict a possible advancing track of the vehicle in a period of time in the future.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of automobile technology, and in particular to an active lane-changing obstacle avoidance control method based on road friction coefficient prediction. Background Art

[0002] With the rapid development of my country's economy, people's living standards have been greatly improved, especially in terms of travel. Cars are increasingly used as the main means of transportation for the people. The number of vehicles in my country has increased year by year. However, with the increase in the number of cars, traffic accidents often occur. Therefore, more and more car companies are developing active obstacle avoidance systems to reduce accidents. The active obstacle avoidance technology of automobiles mainly includes longitudinal braking obstacle avoidance and active lane change obstacle avoidance. The more mature longitudinal collision avoidance control uses radar sensors to detect the distance from the vehicle and pedestrian in front to control the throttle and brake system to decelerate and avoid collisions. The more mature active lane change obstacle avoidance uses sensors to detect the distance from the front. When the longitudinal braking distance is insufficient, the vehicle is given a steering wheel angle. However, when driving at high speed, the distance required for vehicle braking increases, and a large safe distance needs to be maintained from the vehicle in front, resulting in a decrease in road utilization. When driving on a road with a low adhesion coefficient, it is impossible to predict the safe distance from the vehicle in front and cause a collision. Summary of the invention

[0003] To solve the above problems, the present invention provides an active lane change obstacle avoidance control method based on road friction coefficient prediction. Based on the real-time prediction of the road friction coefficient, the sensor data and environmental information (such as road conditions and meteorological conditions) collected during vehicle driving are analyzed, and the visual road adhesion coefficient is estimated using image processing technology, and combined with the dynamic road adhesion coefficient measured by the vehicle sensor, a more accurate comprehensive road adhesion coefficient is synthesized. The comprehensive road adhesion coefficient is combined with the vehicle dynamics model and the driver's intention to predict the possible trajectory of the vehicle in the future; if there is a deviation between the predicted path and the actual path, and there is a potential safety risk or a high possibility of collision, the system will actively intervene and perform steering operations to avoid obstacles, thereby significantly improving the driving safety of the vehicle under complex road conditions and the comfort of the driving experience.

[0004] The technical solution of the present invention is described as follows in conjunction with the accompanying drawings:

[0005] An active lane-changing obstacle avoidance control method based on road friction coefficient prediction includes the following steps:

[0006] Step 1, obtaining a first road adhesion coefficient μ1 based on vehicle vision;

[0007] Step 2, obtaining a second road adhesion coefficient μ2 based on vehicle dynamics;

[0008] Step 3: The first road adhesion coefficient μ1 based on vehicle vision and the second road adhesion coefficient μ2 based on vehicle dynamics are integrated by a weighted average method to obtain a third road adhesion coefficient μ;

[0009] Step 4: Perceive the environment to obtain the position, speed and size of obstacles;

[0010] Step 5: Determine the lane change obstacle avoidance decision based on the safety distance model;

[0011] Step 6: Control and execute lane change and obstacle avoidance decision.

[0012] Furthermore, the specific method of step one is as follows:

[0013] 11) Use the on-board camera to capture real-time images of the road surface and mark the adhesion coefficient of the road surface in dry and wet conditions;

[0014] 12) Perform normalization on the collected images to perform image denoising, image enhancement, and image correction preprocessing;

[0015] 13) Extract the roughness of the road surface texture and the depth of the color from the preprocessed image, and divide the collected data set into a training set and a test set;

[0016] 14) The preprocessed images and the corresponding friction coefficient labels are used to train the BP neural network, and the trained model is verified and tested using the test set. According to the results of the performance evaluation, the network structure and training parameters are adjusted to optimize the performance of the model;

[0017] 15) The data collected by the camera is input into the model to predict the road friction coefficient in real time, and obtain the first road adhesion coefficient μ1 based on visual prediction.

[0018] Furthermore, the images include several different types of road surfaces in dry and wet conditions, including gravel road surfaces, cement road surfaces, icy and snowy road surfaces, and asphalt road surfaces.

[0019] Furthermore, 70% of the data are used as training sets and 30% of the data are used as test sets, and each data set contains all working conditions.

[0020] Furthermore, the specific method of step 2 is as follows:

[0021] 21) Use vehicle gyroscopes, steering wheel angle sensors, accelerometers, and wheel speed sensors to collect dynamic data during vehicle driving;

[0022] 22) Preprocessing the collected kinetic data using the median filter method;

[0023] 23) According to the dynamic characteristics of the vehicle, a dynamic model of a two-degree-of-freedom vehicle and a Dugoff tire model are established;

[0024] 24) The unscented Kalman filter (UKF) is used for estimation, and the estimated road adhesion coefficient provided by the UKF algorithm is compared in detail with the field measurement value to evaluate its ability to reflect the actual road conditions. According to the comparison and experimental results, the parameters of the unscented Kalman filter and the accuracy of the vehicle dynamics model are adjusted. The actual driving data is used as feedback to continuously optimize and iterate the estimation strategy, predict the road friction coefficient in real time, and obtain the second road adhesion coefficient μ2 based on dynamics.

[0025] Furthermore, the dynamic data includes vehicle acceleration, speed, steering wheel angle, lateral acceleration, yaw rate, wheel speed and sideslip angle of the center of mass.

[0026] Furthermore, the specific method of step three is as follows:

[0027] 31) Allocate the weight ratio of the visual road adhesion coefficient μ1 and the dynamic road adhesion coefficient μ2 according to different environments, and assign different weight ratio relationships;

[0028] 32) The third road adhesion coefficient after the fusion of the visual road adhesion coefficient and the estimated dynamic road adhesion coefficient is:

[0029] μ=(μ1×weight1+μ2×weight1) / (weight1+weight2).

[0030] Furthermore, the specific method of step 4 is as follows:

[0031] 41) Collect and detect obstacles in front and on the sides, remove noise from the collected data through Gaussian filter, and pre-process the radar data using Kalman filter;

[0032] 42) The SLAM technology is used to create an environmental model around the vehicle on a high-precision map. The grid map construction method divides the environment into discrete grids. Each grid unit represents an area in the environment and contains the probability that the area is occupied. Computer vision and image processing technology are used to identify and classify perceived obstacles, including distinguishing between static obstacles and dynamic obstacles. The point cloud data collected by the lidar is processed through point cloud processing technology to estimate the position, speed and size of the obstacles.

[0033] Furthermore, the specific method of step five is as follows:

[0034] 51) The safety distance model is established as follows: based on two-degree-of-freedom vehicle dynamics, the vehicle is regarded as a mass point moving on the road plane; the actual size of the vehicle and the safety space requirements are considered, and the model is further refined;

[0035] The safety distance model is optimized through simulation calculations. Various road adhesion coefficients and different driving scenarios are taken into account during the simulation process. The safety distance model combines the vehicle's speed and acceleration to calculate the longitudinal safety distance under various road adhesion coefficients. The system monitors the surrounding environment in real time through sensors, identifies potential dangers and triggers early warning mechanisms.

[0036] The safety distance model formula is:

[0037]

[0038] a 1 =μ·g·cos(θ)

[0039] a 2 =μ·g·cos(θ)

[0040] Where, d br is the braking distance; v is the vehicle speed; v rel is the relative vehicle speed; a 1 is the maximum deceleration of the vehicle; a 2 is the maximum deceleration of the target vehicle; t 1 is the driver's reaction delay time; t 2 is the brake delay time; d 0 is the minimum stopping distance; g is the acceleration of gravity; θ is the road inclination angle; μ is the adhesion coefficient of the third road surface after fusion;

[0041] 52) Whether to make longitudinal control or lateral control obstacle avoidance decisions based on longitudinal distance, that is, to select obstacle avoidance strategy decisions by real-time monitoring of the vehicle's motion state and surrounding environment: when the distance between the vehicle and the vehicle in front is less than the safe distance but greater than the preset emergency braking threshold, the system gives priority to longitudinal braking obstacle avoidance; when the distance between the two vehicles is lower than the emergency braking threshold and the system determines that longitudinal braking cannot avoid collision, it will give priority to lateral steering obstacle avoidance; when avoiding obstacles, the distribution of steering and braking forces will be dynamically adjusted, and the specific distribution needs to be dynamically adjusted according to the warning level: the algorithm is adjusted to achieve the best obstacle avoidance effect; according to the preset urgency of the danger, the system will issue different levels of warnings to give the driver or the autonomous driving system more time to react;

[0042] Among them, when lateral steering to avoid obstacles, a lateral displacement tracker needs to be designed to calculate the steering angle required for steering to avoid obstacles and calculate the safe distance for lateral steering;

[0043] The lateral displacement tracker is designed with a feedforward-feedback structure; the feedforward part is responsible for pre-calculating the control input according to the obstacle avoidance trajectory, while the feedback part is adjusted according to the actual state of the vehicle to reduce the tracking error;

[0044] The feedforward-feedback control formula is:

[0045] u(t)=G_{ff}(s)r(t)+G_{fb}(s)(r(t)-y(t))

[0046] Where u(t) is the control input signal; G_{ff}(s) is the feedforward gain; r(t) is the reference input signal, G_{fb}(s) is the feedback gain, r(t) is the reference input signal, and y(t) is the system output signal;

[0047] The lateral safety distance model is as follows: a lateral safety distance model is established based on the width, speed, steering characteristics and predicted road adhesion coefficient of the target vehicle. Through this model, the active obstacle avoidance algorithm can dynamically calculate the required safety distance in different situations.

[0048] The formula for lateral safety distance is:

[0049]

[0050] Where D y is the total safety distance required by the vehicle in the lateral direction; V is the speed of the vehicle; t is the driver's reaction time; W car is the width of the vehicle itself; W margin is the additional safety margin; g is the acceleration of gravity; μ is the adhesion coefficient of the third road surface after fusion;

[0051] The longitudinal braking obstacle avoidance warning is graded as follows: Level 1 warning: the system will perform emergency braking or remind the driver to perform evasive maneuvers to avoid collision; Level 2 warning: the system will prepare for emergency braking and remind the driver to slow down or maintain a safe distance; Level 3 warning: the system will prompt the driver to slow down and maintain a safe following distance; Level 4 warning: the system will remind the driver to maintain the current speed or slightly adjust the speed to ensure driving safety;

[0052] The warning levels for lateral steering obstacle avoidance are as follows: Level 1 warning: the driver will receive a preliminary warning, with a flashing indicator light on the dashboard and a gentle voice prompt, reminding him to pay attention to possible obstacles ahead and suggesting that he remain vigilant; Level 2 warning: the system sends an enhanced warning signal to the driver, with continuous voice warnings and more eye-catching visual prompts, requiring the driver to reduce speed and be prepared to intervene in control at any time; Level 3 warning: the driver will receive an emergency warning informing him that the system is implementing obstacle avoidance measures and requiring the driver to be on standby immediately so as to take over control when necessary; Level 4 warning: the driver will receive an emergency alert, with continuous sharp alarm sounds and accompanying emergency text / icon prompts, and the system will start the automatic obstacle avoidance program;

[0053] 53) Trajectory planning: applying convolutional neural network method to perform trajectory planning and generate obstacle avoidance trajectory adapted to current road and traffic conditions; the safety distance model is dynamically adjusted according to the real-time status and environmental information of the vehicle to ensure the optimality of the obstacle avoidance path;

[0054] 54) Trajectory tracking is: applying the MPC algorithm to track the trajectory, predicting the future motion state of the vehicle, and adjusting the control input accordingly, that is, adjusting the acceleration, steering angle and other control parameters of the vehicle in real time to achieve trajectory tracking; MPC will re-optimize the control strategy in each control cycle to adapt to the real-time changes in the environment; within a limited time domain, the motion state of the obstacle vehicle and surrounding traffic vehicles is modeled on a high-precision map through SLAM technology; including their position, speed and predicted motion trajectory;

[0055] 55) By simulating various driving scenarios and abnormal situations, the performance of the obstacle avoidance strategy is comprehensively evaluated and optimized according to the actual situation.

[0056] Furthermore, the specific method of step six is ​​as follows:

[0057] 61) According to the lane change trajectory and the current state of the vehicle, the required turn signal, accelerator and brake instructions are calculated, specifically:

[0058] Furthermore, when changing lanes is required, the system will first turn on the turn signal, adjust the throttle to smoothly accelerate or decelerate to a suitable speed, and adjust the vehicle speed by braking when necessary to ensure a safe distance. The collision distance between the vehicle and the dangerous target is compared with the set threshold to determine the current degree of collision risk, where the warning stage threshold is Lw, the light braking stage threshold is L1, the medium braking stage threshold is L2, the emergency braking stage threshold is L3, and the emergency braking minimum distance threshold is Lmin;

[0059] The braking distance L_2 of the current vehicle is:

[0060]

[0061] In the formula, v 1 The current vehicle speed is m / s; g is the acceleration due to gravity, which is 9.8 m / S 2 ; μ is the road adhesion coefficient after fusion;

[0062] The safe driving distance d between the two vehicles is:

[0063] d=L_1-L_2

[0064] Where, L_1 is the current distance between two vehicles;

[0065] The current collision risk level is determined as follows:

[0066] When the actual distance between the vehicle and the dangerous target vehicle is L>Lw, the automatic emergency braking system does not work;

[0067] When the actual distance L between the vehicle and the dangerous target vehicle satisfies Lw<L≤L1, the vehicle's longitudinal automatic emergency braking system enters the early warning stage and issues an alarm to prompt the driver to pay attention to the road vehicle situation and take corresponding measures;

[0068] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L1<L≤L2, the vehicle's longitudinal automatic emergency braking system enters the light braking stage, provides 25% braking force and sends a slight braking signal and lights up the brake lights;

[0069] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L2<L≤L3, the vehicle provides 50% longitudinal braking force and sends a moderate braking signal and lights up the brake lights;

[0070] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤L3, the automatic emergency braking system enters the emergency braking stage, provides 100% braking force and lights up the brake lights;

[0071] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤Lmin, a collision will occur. The vehicle will implement a lateral obstacle avoidance control strategy, turn on the turn signal, adjust the throttle, give the vehicle a steering angle, provide braking force, send a brake signal, and light up the brake lights.

[0072] The beneficial effects of the present invention are:

[0073] 1) The present invention analyzes the sensor data and environmental information (such as road conditions and weather conditions) collected during vehicle driving, uses image processing technology to estimate the visual road adhesion coefficient, and combines it with the dynamic road adhesion coefficient measured by the vehicle sensor to synthesize a more accurate comprehensive road adhesion coefficient. The comprehensive road adhesion coefficient is combined with the vehicle dynamics model and the driver's intention to predict the possible trajectory of the vehicle in the future. If there is a deviation between the predicted path and the actual path, and there is a potential safety risk or a high possibility of collision, the system will actively intervene and perform steering operations to avoid obstacles, thereby significantly improving the vehicle's driving safety and driving comfort in complex road conditions;

[0074] 2) The present invention analyzes the sensor data and environmental information (such as road conditions and weather conditions) collected during vehicle driving, uses image processing technology to estimate the visual road adhesion coefficient, and combines it with the dynamic road adhesion coefficient measured by the vehicle sensor to synthesize a more accurate comprehensive road adhesion coefficient. The comprehensive road adhesion coefficient is combined with the vehicle dynamics model and the driver's intention to predict the possible trajectory of the vehicle in the future. If there is a deviation between the predicted path and the actual path, and there is a potential safety risk or a high possibility of collision, the system will actively intervene and perform steering operations to avoid obstacles, thereby significantly improving the vehicle's driving safety and driving comfort under complex road conditions;

[0075] 3) The active lane change and obstacle avoidance control strategy based on the prediction of the road friction coefficient can more accurately evaluate the actual adhesion conditions of the road surface, thereby improving the accuracy of active lane change and driving safety. This strategy is applicable to a variety of vehicle types and aims to enhance the braking efficiency and driving safety of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0077] Figure 1 It is a schematic diagram of the process of the present invention;

[0078] Figure 2 To obtain the visual road adhesion coefficient flow chart of the vehicle;

[0079] Figure 3 Flow chart for obtaining the dynamic road adhesion coefficient of the vehicle;

[0080] Figure 4 It is a flowchart for environmental perception;

[0081] Figure 5 It is a decision flow chart for lane change and obstacle avoidance;

[0082] Figure 6 This is the first set of tests of the lane change obstacle avoidance braking force / steering angle and the distance between the two vehicles without adhesion coefficient estimation;

[0083] Figure 7 This is the first set of tests with estimated adhesion coefficients for lane change and obstacle avoidance braking force / steering angle versus the distance between the two vehicles;

[0084] Figure 8This is the second group of tests without adhesion coefficient estimation for lane change and obstacle avoidance braking force / steering angle vs. distance between two vehicles;

[0085] Fig. 9 This is the second set of tests on the lane change obstacle avoidance braking force / steering angle and the distance between the two vehicles without adhesion coefficient estimation. DETAILED DESCRIPTION

[0086] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0087] Embodiment 1

[0088] See also Figure 1 The present invention provides an active lane change obstacle avoidance control method based on road friction coefficient prediction, which specifically includes the following steps:

[0089] Step 1: Obtain the first road adhesion coefficient μ1 based on vehicle vision, as follows:

[0090] 11) Use the on-board camera to capture real-time images of the road surface; the images include several different types of road surfaces in dry and wet conditions, such as gravel road surface, cement road surface, ice and snow road surface, and asphalt road surface, with no less than 500 images of each type of road surface, and the adhesion coefficients of the road surfaces in dry and wet conditions are marked;

[0091] 12) Perform normalization on the collected images to perform image denoising, image enhancement, and image correction preprocessing, and extract the roughness of the road surface texture and the depth of color;

[0092] The specific methods of data preprocessing are as follows:

[0093] The collected images are normalized to perform image denoising, image enhancement, and image correction preprocessing, specifically:

[0094] The Z-score normalization method is:

[0095]

[0096] Among them, X is the original data; μ is the mean of the data; δ is the standard deviation of the data; X non is the normalized data;

[0097] The specific method for extracting the roughness and color depth characteristics of road surface texture is as follows:

[0098] Color features are extracted from the processed image using color histogram.

[0099] 13) Divide the collected data set into a training set and a test set; use 70% of the data as the training set and 30% of the data as the test set, and each data set contains all working conditions;

[0100] 14) Use the BP neural network method to compare the divided training set with the known friction coefficient and train a model that can accurately predict the friction coefficient; use the deep learning algorithm to evaluate the trained model using the test set (collected road photos) based on the extracted features (roughness of road texture, depth of color); optimize the model based on the evaluation results, adjust model parameters, increase training data, etc.; update the database with the fused road adhesion coefficient and collected photos.

[0101] 15) Input the data collected by the camera into the model, predict the road friction coefficient in real time, obtain the first road adhesion coefficient μ1 based on visual prediction, and output the result of the first road adhesion coefficient μ1 based on vision.

[0102] Step 2: See Figure 2 , obtain the second road adhesion coefficient μ2 based on vehicle dynamics, as follows:

[0103] 21) Using a vehicle gyroscope, a steering wheel angle sensor, an accelerometer, and a wheel speed sensor to collect dynamic data of the vehicle during driving; the dynamic data includes vehicle acceleration, speed, steering wheel angle, lateral acceleration, yaw rate, wheel speed, and sideslip angle of the center of mass;

[0104] 22) Preprocessing the collected kinetic data using the median filter method;

[0105] The preprocessing method is as follows:

[0106] The collected data is subjected to median filtering for noise removal and filtering preprocessing to improve the quality and accuracy of the data.

[0107] The median filter is:

[0108] out(i,j)=P(in(i-1:i+1,j-1,j+1))

[0109] Among them, in(i-1:i+1,j-1,j+1) represents the pixel value in the 3×3 neighborhood around the pixel point (i,j), and P() represents the calculation of the median of a group of numbers.

[0110] 23) According to the dynamic characteristics of the vehicle, a dynamic model of a two-degree-of-freedom vehicle and a Dugoff tire model are established;

[0111] The specific method of establishing the vehicle dynamics model is as follows:

[0112] Longitudinal motion equation:

[0113]

[0114] Lateral motion equation:

[0115]

[0116] In the formula, a y is the lateral acceleration; r is the lateral deviation angle; V is the vehicle speed; F f,max is the maximum longitudinal force of the front wheel; F r,max is the maximum longitudinal force of the rear wheel; L f is the distance from the front wheel to the center of mass; L r is the distance from the rear wheel to the center of mass; I is the moment of inertia of the vehicle; F X is the driving force; m is the vehicle mass.

[0117] 23) The unscented Kalman filter (UKF) is used for estimation, and the estimated road adhesion coefficient provided by the UKF algorithm is compared in detail with the field measurement value to evaluate its ability to reflect the actual road conditions. According to the comparison and experimental results, the parameters of the unscented Kalman filter and the accuracy of the vehicle dynamics model are adjusted to improve the estimation precision and accuracy of the road adhesion coefficient. The actual driving data is used as feedback to continuously optimize and iterate the estimation strategy, predict the road friction coefficient in real time, and obtain the second road adhesion coefficient μ2 based on dynamics.

[0118] The specific method of unscented Kalman filter (UKF) is as follows:

[0119] Nonlinear Systems:

[0120]

[0121] Where f is the nonlinear state equation function; h is the nonlinear observation equation function; it is assumed that W(k) has a covariance matrix Q and V(k) has a covariance matrix R;

[0122] Step 3: Read Figure 3 , the first road adhesion coefficient μ1 based on vehicle vision and the second road adhesion coefficient μ2 based on vehicle dynamics are integrated by weighted average method to obtain the third road adhesion coefficient μ, as follows:

[0123] 31) According to different environments, the weight ratio of the visual road adhesion coefficient μ1 and the dynamic road adhesion coefficient μ2 is allocated, and different weight ratios are allocated; sunny day μ1:μ2=0.7:0.3, sunny night μ1:μ2=0.3:0.7, rainy day μ1:μ2=0.25:0.75, rainy night μ1:μ2=0.2:0.8, snowy day μ1:μ2=0.4:0.6, snowy night μ1:μ2=0.2:0.8;

[0124] 32) The third road adhesion coefficient after the fusion of the visual road adhesion coefficient and the estimated dynamic road adhesion coefficient is:

[0125] μ=(μ1×weight1+μ2×weight1) / (weight1+weight2).

[0126] Step 4: Read Figure 4 , perceive the environment to obtain the position, speed and size of obstacles, as follows:

[0127] 41) Use GPS, gyroscope, accelerometer, camera, millimeter wave radar and laser radar, and vehicle-mounted data acquisition system to detect obstacles in front and on the side, remove noise from the image through Gaussian filter, and use Kalman filter to pre-process radar data;

[0128] 42) The SLAM technology is used to create an environmental model around the vehicle on a high-precision map. The grid map construction method divides the environment into discrete grids. Each grid unit represents an area in the environment and contains the probability that the area is occupied. The perceived obstacles are identified and classified through computer vision and image processing technology, including distinguishing between static obstacles (roadside trees, buildings) and dynamic obstacles (pedestrians, vehicles). The point cloud data collected by the lidar is processed through point cloud technology to estimate the position, speed and size of the obstacles.

[0129] Step 5: Read Figure 5 , the lane change obstacle avoidance decision is determined based on the safety distance model, as follows:

[0130] 51) The safety distance model is established as follows: Based on two-degree-of-freedom vehicle dynamics, the vehicle is regarded as a mass point moving on the road plane; in order to more accurately simulate the actual driving situation, the model is further refined, taking into account the actual size of the vehicle (vehicle length, vehicle width) and the safety space requirements (driver's reaction time, vehicle braking performance, vehicle steering performance); the vehicle can accurately calculate the safety distance from the vehicle in front in various complex driving situations, thereby effectively preventing accidents such as rear-end collisions; the safety distance model is optimized through a large number of simulation calculations; various road adhesion coefficients and different driving situations (different vehicle speeds, different road conditions, different weather conditions) are considered during the simulation process. The model combines the vehicle's speed and acceleration to accurately calculate the longitudinal safety distance under various road adhesion coefficients. The model is dynamically adjusted according to the real-time vehicle status and environmental information to adapt to the ever-changing driving environment. The system monitors the surrounding environment in real time through sensors, identifies potential dangers and triggers early warning mechanisms.

[0131] The safety distance formula is:

[0132]

[0133] a 1 =μ·g·cos(θ)

[0134] a 2 =μ·g·cos(θ)

[0135] Where: d br represents the braking distance, v is the vehicle speed, and v rel is the relative vehicle speed, a 1 is the maximum deceleration of the vehicle, a 2 is the maximum deceleration of the target vehicle, t 1 is the driver’s reaction delay time, t 2 is the brake delay time, d 0 is the minimum stopping distance, g is the acceleration of gravity, θ is the road inclination angle, and μ is the fused third road adhesion coefficient.

[0136] 52) Whether to make longitudinal control or lateral control obstacle avoidance decisions based on longitudinal distance, that is, to select obstacle avoidance strategy decisions by real-time monitoring of the vehicle's motion state and surrounding environment: When the distance between the vehicle and the vehicle in front is less than the safety distance but greater than the preset emergency braking threshold, the system gives priority to longitudinal braking obstacle avoidance. When the distance between the two vehicles is lower than the emergency braking threshold, and the system determines that longitudinal braking cannot avoid collision, lateral steering obstacle avoidance will be given priority. When avoiding obstacles, the distribution of steering and braking forces will be dynamically adjusted, and the specific distribution needs to be based on the warning level: dynamic adjustment is performed through existing algorithms, such as outputting braking percentages according to the safety distance level PID, and outputting different percentages at different distances to achieve the best obstacle avoidance effect. According to the urgency of the danger, the system will issue different levels of warnings to give the driver or the autonomous driving system more time to respond. The urgency of the danger is divided according to the safety distance set by the controller, which is set in advance in the controller and can correspond to the threshold in 61).

[0137] Among them, when lateral steering to avoid obstacles, a lateral displacement tracker needs to be designed to calculate the steering angle required for steering to avoid obstacles and calculate the safe distance for lateral steering;

[0138] The lateral displacement tracker is designed with a feedforward-feedback structure; the feedforward part is responsible for pre-calculating the control input according to the obstacle avoidance trajectory, while the feedback part is adjusted according to the actual state of the vehicle to reduce the tracking error; by real-time monitoring and analysis of the lateral speed and lateral acceleration of the target vehicle, the active obstacle avoidance algorithm can more accurately predict the future position of the target vehicle, thereby making more appropriate obstacle avoidance decisions;

[0139] The feedforward-feedback control formula is:

[0140] u(t)=G_{ff}(s)r(t)+G_{fb}(s)(r(t)-y(t))

[0141] Where u(t) is the control input signal; G_{ff}(s) is the feedforward gain; r(t) is the reference input signal, G_{fb}(s) is the feedback gain, r(t) is the reference input signal, and y(t) is the system output signal;

[0142] The lateral safety distance model is as follows: a lateral safety distance model is established based on the width, speed, steering characteristics and predicted road adhesion coefficient of the target vehicle. Through this model, the active obstacle avoidance algorithm can dynamically calculate the required safety distance in different situations.

[0143] The formula for lateral safety distance is:

[0144]

[0145] Where D yis the total safety distance required by the vehicle in the lateral direction; V is the speed of the vehicle; t is the driver's reaction time; W car is the width of the vehicle itself; W margin is the additional safety margin; g is the acceleration of gravity; μ is the adhesion coefficient of the third road surface after fusion;

[0146] The longitudinal braking obstacle avoidance warning is graded as follows: Level 1 warning: the system will perform emergency braking or remind the driver to perform evasive maneuvers to avoid collision; Level 2 warning: the system will prepare for emergency braking and remind the driver to slow down or maintain a safe distance; Level 3 warning: the system will prompt the driver to slow down and maintain a safe following distance; Level 4 warning: the system will remind the driver to maintain the current speed or slightly adjust the speed to ensure driving safety;

[0147] The warning levels for lateral steering obstacle avoidance are as follows: Level 1 warning: the driver will receive a preliminary warning, with a flashing indicator light on the dashboard and a gentle voice prompt, reminding him to pay attention to possible obstacles ahead and suggesting that he remain vigilant; Level 2 warning: the system sends an enhanced warning signal to the driver, with continuous voice warnings and more eye-catching visual prompts, requiring the driver to reduce speed and be prepared to intervene in control at any time; Level 3 warning: the driver will receive an emergency warning informing him that the system is implementing obstacle avoidance measures and requiring the driver to be on standby immediately so as to take over control when necessary; Level 4 warning: the driver will receive an emergency alert, with continuous sharp alarm sounds and accompanying emergency text / icon prompts, and the system will start the automatic obstacle avoidance program;

[0148] 53) Trajectory planning is as follows: applying the convolutional neural network method for trajectory planning to ensure the smoothness of the trajectory, avoid sudden changes during the steering process, improve the driving stability of the vehicle, quickly generate an obstacle avoidance trajectory that adapts to the current road and traffic conditions, and improve the success rate of obstacle avoidance; the model is dynamically adjusted according to the real-time status of the vehicle and environmental information to ensure the optimality of the obstacle avoidance path.

[0149] 54) Trajectory tracking: The MPC algorithm performs trajectory tracking, predicts the future motion state of the vehicle, and adjusts the control input accordingly to achieve accurate trajectory tracking. MPC will re-optimize the control strategy in each control cycle to adapt to real-time changes in the environment and improve the adaptability and robustness of the control system. In a limited time domain, SLAM technology models the motion state of the obstacle vehicle and surrounding traffic vehicles in a high-precision map, including their position, speed and predicted motion trajectory, the same as step 42).

[0150] In summary, the active obstacle avoidance algorithm develops corresponding obstacle avoidance strategies based on the calculation results of the lateral safety distance model. These strategies may include adjusting the vehicle speed, changing lanes, or taking emergency braking. In order to ensure the effectiveness and reliability of the emergency obstacle avoidance strategy, a large number of simulation tests and real vehicle verification are required. By simulating various driving scenarios (different vehicle speeds, different road conditions, different weather conditions) and abnormal situations, the performance of the obstacle avoidance strategy is comprehensively evaluated, and optimized and adjusted according to the actual situation.

[0151] Step 6: Control and execute the lane change and obstacle avoidance decision, as follows:

[0152] 61) Ensure that the vehicle changes lanes according to the planned path based on the designed vehicle lateral and longitudinal control algorithms. Monitor the actual motion state of the vehicle (such as speed, acceleration, steering angle, etc.) in real time and make appropriate corrections. During the lane change process, monitor the vehicle's motion state in real time and compare it with the planned path. If there is a deviation, the control system will immediately adjust to ensure driving safety. Design the system's redundancy and fault tolerance mechanism to deal with sensor failures or abnormal situations. Enable emergency braking and intervention measures when the system cannot determine a safe lane change.

[0153] Specifically: when changing lanes, the system will first turn on the turn signal, adjust the throttle to smoothly accelerate or decelerate to the appropriate speed, and adjust the vehicle speed by braking when necessary to ensure a safe distance. The collision distance between the vehicle and the dangerous target is compared with the set threshold to determine the current degree of collision risk, where the warning stage threshold is Lw, the light braking stage threshold is L1, the moderate braking stage threshold is L2, the emergency braking stage threshold is L3, and the emergency braking minimum distance threshold is Lmin;

[0154] The braking distance L_2 of the current vehicle is:

[0155]

[0156] In the formula, v 1 The current vehicle speed is m / s; g is the acceleration due to gravity, which is 9.8 m / S 2 ; μ is the road adhesion coefficient after fusion;

[0157] The safe driving distance d between the two vehicles is:

[0158] d=L_1-L_2

[0159] Where, L_1 is the current distance between two vehicles;

[0160] The current collision risk level is determined as follows:

[0161] When the actual distance between the vehicle and the dangerous target vehicle is L>Lw, the automatic emergency braking system does not work;

[0162] When the actual distance L between the vehicle and the dangerous target vehicle satisfies Lw<L≤L1, the vehicle's longitudinal automatic emergency braking system enters the early warning stage and issues an alarm to prompt the driver to pay attention to the road vehicle situation and take corresponding measures;

[0163] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L1<L≤L2, the vehicle's longitudinal automatic emergency braking system enters the light braking stage, provides 25% braking force and sends a slight braking signal and lights up the brake lights;

[0164] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L2<L≤L3, the vehicle provides 50% longitudinal braking force and sends a moderate braking signal and lights up the brake lights;

[0165] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤L3, the automatic emergency braking system enters the emergency braking stage, provides 100% braking force and lights up the brake lights;

[0166] When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤Lmin, a collision will occur. The vehicle will implement a lateral obstacle avoidance control strategy, turn on the turn signal, adjust the throttle, give the vehicle a steering angle, provide braking force, send a brake signal, and light up the brake lights.

[0167] Embodiment 2

[0168] This example verifies the multi-sensor road adhesion coefficient fusion and braking strategy:

[0169] In order to verify the superiority of the present invention, the active lane change obstacle avoidance control algorithm without adhesion coefficient estimation and the active lane change obstacle avoidance control algorithm with fusion adhesion coefficient are simulated respectively as follows:

[0170] This group of simulation experiments is divided into two groups. In order to highlight the superiority of the algorithm of the present invention, the control variable method is adopted to conduct the experiment. Except for the different control algorithms, the other simulation conditions of the two groups of experiments are exactly the same. When conducting the simulation, the experimental environment is first set. The first group is an ice and snow road with a preset adhesion coefficient of 0.2, a two-way lane and no vehicles in the adjacent lane, the speed of the target vehicle is 20Km / h, the speed of the vehicle is 60Km / h, and the distance between the target vehicle and the vehicle is 100m. The second group is a wet cement road with a preset adhesion coefficient of 0.4, the speed of the target vehicle is 15Km / h, the speed of the vehicle is 50Km / h, and the distance between the target vehicle and the vehicle is 50m.

[0171] See also Figure 6 and Figure 7, the first group, firstly, conducts simulation experiments on the active lane change obstacle avoidance control algorithm without adhesion coefficient estimation. It can be seen from the experimental results that the car generates 25% braking force when the distance is 23.9m, and then provides a steering angle to the vehicle. Due to the small adhesion coefficient of the ice and snow road, the steering time of the car becomes longer, and it is impossible to avoid obstacles by steering at the preset safe distance. The final result is that the vehicle collides with the target vehicle. Then the second simulation experiment of this group of experiments was carried out to verify the active lane change obstacle avoidance control algorithm with fusion adhesion coefficient of the present invention. The environmental conditions in the second experiment were the same as the first. When verifying, the collected images and dynamic data were denoised and filtered for preprocessing. The road adhesion coefficient predicted by the image decision tree model was fused with the road adhesion coefficient predicted by the dynamic model. The fused adhesion coefficient was transferred to the simulation model through the CAN network protocol. It can be seen from the simulation results that the car generates 25% braking force when the distance is 76.7m, and provides a steering angle to the vehicle in this process. The car is in a state of continuous deceleration and lane change. The minimum longitudinal distance between the vehicle and the leading vehicle after changing lanes is 10 meters.

[0172] See also Figure 8 and Fig. 9 , the second group, firstly, conducted simulation experiments on the algorithm without adhesion coefficient estimation. The experimental results show that the car generates 25% braking force when the distance is 20 meters, and then provides a steering angle to the vehicle. Due to the small adhesion coefficient of the wet cement road surface, the braking and steering distance of the vehicle is greater than the distance between the two vehicles, and the vehicle cannot be braked within the preset safety time. The final result is that the vehicle collides with the target vehicle. Then the second simulation experiment of this group of experiments was carried out to verify the AEB strategy with fusion adhesion system of the present invention. The environmental conditions in the second experiment were the same as those in the first experiment. When verifying, the collected images and dynamic data were denoised and filtered for preprocessing. The road adhesion coefficient predicted by the image decision tree model was fused with the road adhesion coefficient predicted by the dynamic model. The fused adhesion coefficient was transferred to the simulation model through the CAN network protocol. The simulation results show that the car generates 25% braking force when the distance is 33 meters, and gives a steering angle in this process. The car is in a state of continuous deceleration and lane change. The minimum longitudinal distance between the vehicle and the front vehicle after braking is 10 meters.

[0173] In summary, the method based on the fusion of visual road adhesion coefficient and dynamic road adhesion coefficient can more accurately reflect the actual adhesion of the road surface, improve the accuracy and safety of emergency braking. The method can be applied to various types of vehicles to improve the braking performance and driving safety of the vehicles.

[0174] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0175] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An active lane-changing obstacle avoidance control method based on road friction coefficient prediction, characterized in that: The following steps are involved: Step 1, obtaining a first road adhesion coefficient μ1 based on vehicle vision; Step 2, obtaining a second road adhesion coefficient μ2 based on vehicle dynamics; Step 3: The first road adhesion coefficient μ1 based on vehicle vision and the second road adhesion coefficient μ2 based on vehicle dynamics are integrated by a weighted average method to obtain a third road adhesion coefficient μ; Step 4: Perceive the environment to obtain the position, speed and size of obstacles; Step 5: Determine the lane change obstacle avoidance decision based on the longitudinal safety distance model; Step 6: Control and execute lane change and obstacle avoidance decision.

2. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step one is as follows: 11) Use the on-board camera to capture real-time images of the road surface and mark the adhesion coefficient of the road surface in dry and wet conditions; 12) Perform normalization on the collected images to perform image denoising, image enhancement, and image correction preprocessing; 13) Extract the roughness of the road surface texture and the depth of the color from the preprocessed image, and divide the collected data set into a training set and a test set; 14) The preprocessed images and the corresponding friction coefficient labels are used to train the BP neural network, and the trained model is verified and tested using the test set. According to the results of the performance evaluation, the network structure and training parameters are adjusted to optimize the performance of the model; 15) The data collected by the camera is input into the model to predict the road friction coefficient in real time, and obtain the first road adhesion coefficient μ1 based on visual prediction.

3. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 2, characterized in that: The images include several different types of road surfaces, including gravel road surface, cement road surface, icy and snowy road surface, and asphalt road surface, in dry and wet conditions.

4. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 2, characterized in that: 70% of the data are used as training sets and 30% of the data are used as test sets. Each data set contains all working conditions.

5. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step 2 is as follows: 21) Use vehicle gyroscopes, steering wheel angle sensors, accelerometers, and wheel speed sensors to collect dynamic data during vehicle driving; 22) Preprocessing the collected kinetic data using the median filter method; 23) According to the dynamic characteristics of the vehicle, a dynamic model of a two-degree-of-freedom vehicle and a Dugoff tire model are established; 24) The unscented Kalman filter (UKF) is used for estimation, and the estimated road adhesion coefficient provided by the UKF algorithm is compared in detail with the field measurement value to evaluate its ability to reflect the actual road conditions. According to the comparison and experimental results, the parameters of the unscented Kalman filter and the accuracy of the vehicle dynamics model are adjusted. The actual driving data is used as feedback to continuously optimize and iterate the estimation strategy, predict the road friction coefficient in real time, and obtain the second road adhesion coefficient μ2 based on dynamics.

6. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 5, characterized in that: The dynamic data includes vehicle acceleration, speed, steering wheel angle, lateral acceleration, yaw rate, wheel speed and center of mass sideslip angle.

7. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step three is as follows: 31) Allocate the weight ratio of the visual road adhesion coefficient μ1 and the dynamic road adhesion coefficient μ2 according to different environments, and assign different weight ratio relationships; 32) The third road adhesion coefficient after the fusion of the visual road adhesion coefficient and the estimated dynamic road adhesion coefficient is: μ=(μ1×weight1+μ2×weight1) / (weight1+weight2).

8. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step 4 is as follows: 41) Collect and detect obstacles in front and on the sides, remove noise from the collected data through Gaussian filter, and pre-process the radar data using Kalman filter; 42) The SLAM technology is used to create an environmental model around the vehicle on a high-precision map. The grid map construction method divides the environment into discrete grids. Each grid unit represents an area in the environment and contains the probability that the area is occupied. Computer vision and image processing technology are used to identify and classify perceived obstacles, including distinguishing between static obstacles and dynamic obstacles. The point cloud data collected by the lidar is processed through point cloud processing technology to estimate the position, speed and size of the obstacles.

9. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step five is as follows: 51) The safety distance model is established as follows: based on two-degree-of-freedom vehicle dynamics, the vehicle is regarded as a mass point moving on the road plane; the actual size of the vehicle and the safety space requirements are considered, and the model is further refined; The safety distance model is optimized through simulation calculations. Various road adhesion coefficients and different driving scenarios are taken into account during the simulation process. The safety distance model combines the vehicle's speed and acceleration to calculate the longitudinal safety distance under various road adhesion coefficients. The system monitors the surrounding environment in real time through sensors, identifies potential dangers and triggers early warning mechanisms. The safety distance model formula is: a1=μ·g·cos(θ) a2=μ·g·cos(θ) Where, d br is the braking distance; v is the vehicle speed; v rel is the relative vehicle speed; a1 is the maximum deceleration of the vehicle; a2 is the maximum deceleration of the target vehicle; t1 is the driver's reaction delay time; t2 is the brake delay time; d0 is the minimum stopping distance; g is the gravity acceleration; θ is the road inclination angle; μ is the third road adhesion coefficient after fusion; 52) Whether to make longitudinal control or lateral control obstacle avoidance decisions based on longitudinal distance, that is, to select obstacle avoidance strategy decisions by real-time monitoring of the vehicle's motion state and surrounding environment: when the distance between the vehicle and the vehicle in front is less than the safe distance but greater than the preset emergency braking threshold, the system gives priority to longitudinal braking obstacle avoidance; when the distance between the two vehicles is lower than the emergency braking threshold and the system determines that longitudinal braking cannot avoid collision, it will give priority to lateral steering obstacle avoidance; when avoiding obstacles, the distribution of steering and braking forces will be dynamically adjusted, and the specific distribution needs to be dynamically adjusted according to the warning level: the algorithm is adjusted to achieve the best obstacle avoidance effect; according to the preset urgency of the danger, the system will issue different levels of warnings to give the driver or the autonomous driving system more time to react; Among them, when lateral steering to avoid obstacles, a lateral displacement tracker needs to be designed to calculate the steering angle required for steering to avoid obstacles and calculate the safe distance for lateral steering; The lateral displacement tracker is designed with a feedforward-feedback structure; the feedforward part is responsible for pre-calculating the control input according to the obstacle avoidance trajectory, while the feedback part is adjusted according to the actual state of the vehicle to reduce the tracking error; The feedforward-feedback control formula is: u(t)=G_{ff}(s)r(t)+G_{fb}(s)(r(t)-y(t)) Where u(t) is the control input signal; G_{ff}(s) is the feedforward gain; r(t) is the reference input signal, G_{fb}(s) is the feedback gain, r(t) is the reference input signal, and y(t) is the system output signal; The lateral safety distance model is as follows: a lateral safety distance model is established based on the width, speed, steering characteristics and predicted road adhesion coefficient of the target vehicle. Through this model, the active obstacle avoidance algorithm can dynamically calculate the required safety distance in different situations. The formula for lateral safety distance is: Where D y is the total safety distance required by the vehicle in the lateral direction; V is the speed of the vehicle; t is the driver's reaction time; W car is the width of the vehicle itself; W margin is the additional safety margin; g is the acceleration of gravity; μ is the adhesion coefficient of the third road surface after fusion; The longitudinal braking obstacle avoidance warning is graded as follows: Level 1 warning: the system will perform emergency braking or remind the driver to perform evasive maneuvers to avoid collision; Level 2 warning: the system will prepare for emergency braking and remind the driver to slow down or maintain a safe distance; Level 3 warning: the system will prompt the driver to slow down and maintain a safe following distance; Level 4 warning: the system will remind the driver to maintain the current speed or slightly adjust the speed to ensure driving safety; The warning levels for lateral steering obstacle avoidance are as follows: Level 1 warning: the driver will receive a preliminary warning, with a flashing indicator light on the dashboard and a gentle voice prompt, reminding him to pay attention to possible obstacles ahead and suggesting that he remain vigilant; Level 2 warning: the system sends an enhanced warning signal to the driver, with continuous voice warnings and more eye-catching visual prompts, requiring the driver to reduce speed and be prepared to intervene in control at any time; Level 3 warning: the driver will receive an emergency warning informing him that the system is implementing obstacle avoidance measures and requiring the driver to be on standby immediately so as to take over control when necessary; Level 4 warning: the driver will receive an emergency alert, with continuous sharp alarm sounds and accompanying emergency text / icon prompts, and the system will start the automatic obstacle avoidance program; 53) Trajectory planning: applying convolutional neural network method to perform trajectory planning and generate obstacle avoidance trajectory adapted to current road and traffic conditions; the safety distance model is dynamically adjusted according to the real-time status and environmental information of the vehicle to ensure the optimality of the obstacle avoidance path; 54) Trajectory tracking is: applying the MPC algorithm to track the trajectory, predicting the future motion state of the vehicle, and adjusting the control input accordingly, that is, adjusting the acceleration, steering angle and other control parameters of the vehicle in real time to achieve trajectory tracking; MPC will re-optimize the control strategy in each control cycle to adapt to the real-time changes in the environment; within a limited time domain, the motion state of the obstacle vehicle and surrounding traffic vehicles is modeled on a high-precision map through SLAM technology; including their position, speed and predicted motion trajectory; 55) By simulating various driving scenarios and abnormal situations, the performance of the obstacle avoidance strategy is comprehensively evaluated and optimized according to the actual situation.

10. The active lane change obstacle avoidance control method based on road friction coefficient prediction according to claim 1, characterized in that: The specific method of step six is ​​as follows: 61) According to the lane change trajectory and the current state of the vehicle, the required turn signal, accelerator and brake instructions are calculated; specifically: When changing lanes, the system will first turn on the turn signal, adjust the throttle to smoothly accelerate or decelerate to the appropriate speed, and adjust the speed through braking when necessary to ensure a safe distance. The collision distance between the vehicle and the dangerous target is compared with the set threshold to determine the current collision risk level, where the warning stage threshold is Lw, the light braking stage threshold is L1, the moderate braking stage threshold is L2, the emergency braking stage threshold is L3, and the emergency braking minimum distance threshold is Lmin; The braking distance L_2 of the current vehicle is: Where v1 is the current vehicle speed, m / s; g is the acceleration due to gravity, 9.8 m / s 2 ; μ is the road adhesion coefficient after fusion; The safe driving distance d between the two vehicles is: d=L_1-L_2 Where, L_1 is the current distance between two vehicles; The current collision risk level is determined as follows: When the actual distance between the vehicle and the dangerous target vehicle is L>Lw, the automatic emergency braking system does not work; When the actual distance L between the vehicle and the dangerous target vehicle satisfies Lw<L≤L1, the vehicle's longitudinal automatic emergency braking system enters the early warning stage and issues an alarm to prompt the driver to pay attention to the road vehicle situation and take corresponding measures; When the actual distance L between the vehicle and the dangerous target vehicle satisfies L1<L≤L2, the vehicle's longitudinal automatic emergency braking system enters the light braking stage, provides 25% braking force and sends a slight braking signal and lights up the brake lights; When the actual distance L between the vehicle and the dangerous target vehicle satisfies L2<L≤L3, the vehicle provides 50% longitudinal braking force and sends a moderate braking signal and lights up the brake lights; When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤L3, the automatic emergency braking system enters the emergency braking stage, provides 100% braking force and lights up the brake lights; When the actual distance L between the vehicle and the dangerous target vehicle satisfies L≤Lmin, a collision will occur. The vehicle will implement a lateral obstacle avoidance control strategy, turn on the turn signal, adjust the throttle, give the vehicle a steering angle, provide braking force, send a brake signal, and light up the brake lights.

Citation Information

Cited By

  • Vehicle control method, device and equipment, readable storage medium and program product

    CN119329548A

  • Intelligent vehicle attachment coefficient online estimation method fusing real vehicle data

    CN119659636A

  • Vehicle brake-by-wire method and system

    CN119821341A

  • Vehicle brake-by-wire method and system

    CN119821341B

  • Vehicle driving optimization method based on road surface friction coefficient intelligent prediction model

    CN119975409A