A path planning and control method considering driving habits in a static obstacle avoidance scenario

Abstract

The application discloses a path planning and control method considering driving habits in a static obstacle avoidance scene, and relates to the technical field of safe driving. The method comprises the following steps: collecting simulation driving data of different drivers through a driving experiment platform; building a driving style classifier and classifying driving styles of the drivers by using a kmeans-SVM algorithm, selecting a proper speed as a driving expected speed of different style drivers, and taking the expected speeds of different type drivers as speeds during vehicle obstacle avoidance planning; a vehicle-mounted sensor transmits information of obstacles to a trajectory re-planning controller, and re-plans a trajectory in combination with environmental information and characteristic parameters of different styles; and a trajectory tracking controller is used to realize tracking of the planned path by the vehicle. The application establishes a vehicle dynamics model based on a model prediction method, converts a path planning problem into an optimization problem, considers environmental information and dynamics constraints, and directly optimizes vehicle state information as a reference for tracking control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of safe driving, and particularly relates to a path planning and control method considering driving habits in a static obstacle avoidance scene. BACKGROUND

[0002] With the development of intelligent driving technology, the application in vehicles is more and more widely used, and the most basic application is the application of active safety auxiliary system in the obstacle avoidance scene. The most important thing in the obstacle avoidance process is path planning, which needs to consider the self-vehicle state, the state of the surrounding vehicles, the state of the traffic signal and other environmental information to decide the safe driving area. Different drivers have high consistency in decision-making behavior in the emergency obstacle avoidance scene, but for ordinary obstacle avoidance scenes, they will be influenced by individual driving characteristics and show great differences, for example, experienced drivers and drivers with insufficient driving experience usually have great differences in the judgment of obstacle distance and speed and the avoidance decision behavior. Therefore, it is necessary to study the path planning suitable for the driving style of the driver.

[0003] Most of the existing obstacle avoidance ideas adopt the way of directly planning the driving path and then tracking the path to avoid obstacles. Commonly used algorithms include A* algorithm, D* algorithm, ant colony algorithm, artificial potential field method, RRT algorithm, etc. Among them, A*, D* and ant colony algorithm are mainly applied to global path planning. In local path planning, RRT and artificial potential field method are more widely used. The main problems of current obstacle avoidance planning are the problems of insufficient speed and insufficient accuracy of identification, which correspond to RRT algorithm and artificial potential field method respectively. When the dimension of input data increases, the convergence speed of RRT will be greatly reduced. Similarly, the artificial potential field method also has the problem of falling into local optimum and being unable to continue. These methods are difficult to consider the problem of vehicle dynamics in the planning stage, which may affect the tracking effect. Model predictive control (MPC) is an optimization control algorithm generated by modern industrial process control, which takes prediction model, rolling optimization and feedback correction as the control strategy. It has the advantages of strong robustness, great flexibility, good dynamic control effect, etc. It can overcome the uncertainty of external environment, controlled object parameters and structure, and is suitable for control process with complex controlled object and low accuracy requirement for established mathematical model. In recent years, model predictive control has gradually been applied to trajectory planning and tracking control of unmanned vehicles due to its ability to handle multiple constraints in the control process. For example, MPC algorithm is used in active steering of unmanned vehicles. The vehicle model is processed online linearly, and various constraint conditions are added. The front wheel steering angle is calculated by a nonlinear MPC controller. Real vehicle test shows that the algorithm has good stability at high speed and on ice and snow road; or Laguerre function is introduced into the model predictive controller, which can optimize a large number of control parameters in the MPC controller, reduce the calculation difficulty and improve the calculation efficiency while meeting the tracking accuracy. In addition, most of the current obstacle avoidance focuses on how to avoid obstacles, ignoring the consideration of the driving personality of the current driver.

[0004] The traditional path planning method mainly takes safety as the design point, and rarely considers the driving characteristics of the driver, so for a manned vehicle, the difference between the controller and the human decision increases the difficulty of human-vehicle shared control. For unmanned driving, ignoring the difference between different driving styles also ignores the comfort of the passengers to a certain extent. In addition, when the intelligent vehicle completes automatic driving in the actual environment, since the external environment of the vehicle is unknown and changing, trajectory tracking under a given predetermined trajectory may not guarantee that the vehicle accurately completes any trajectory tracking problem. When there are obstacles on the given trajectory, the vehicle cannot complete trajectory tracking according to the predetermined trajectory, and the vehicle needs to identify obstacle information to judge the obstacle and continue the trajectory tracking process after avoiding the obstacle. At present, the obstacle avoidance is mostly carried out by directly planning the driving path and then tracking the path, such as A* algorithm, dynamic window method, artificial potential field method and the like. These methods are difficult to consider the problem of vehicle dynamics in the planning stage, which may affect the tracking effect. SUMMARY

[0005] The purpose of the present application is to provide a path planning and control method considering driving habits in a static obstacle avoidance scenario to solve the problems in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] A path planning and control method considering driving habits in a static obstacle avoidance scenario, comprising the following steps: step S1 data collection and processing: collecting simulation driving data of different drivers through a driving experiment platform; step S2: classifying the driving styles of the drivers by building a driving style classifier and using a kmeans-SVM algorithm, selecting an appropriate speed as the driving expected speed of different style drivers, and taking the expected speed of different types of drivers as the speed during vehicle obstacle avoidance planning; step S3: taking the front wheel steering angle range of different style drivers as a constraint reference in the vehicle planning process; step S4: the vehicle-mounted sensor transmits the information of the obstacle to the trajectory re-planning controller, and the trajectory re-planning controller re-plans the trajectory in combination with the environmental information and the characteristic parameters of different styles; and the vehicle tracks the planned path by the trajectory tracking controller.

[0008] On the basis of the above technical scheme, the present application further provides the following optional technical scheme:

[0009] In an optional scheme, in the step S1 data collection and processing, the specific steps of data collection are as follows: collecting driving data of drivers of different genders, different ages and different driving ages and pre-processing and feature extraction, and establishing a feature data set representing driving style; selecting lateral and longitudinal acceleration, lateral and longitudinal speed, and yaw angular velocity as the basis for classifying the driving style of the driver.

[0010] In an alternative, in the step S1 data collection and processing, the specific steps of data preprocessing are as follows: the selected driving data is subjected to abnormal data elimination and missing data supplement, and then subjected to smoothing treatment by using Gaussian filtering.

[0011] In an alternative, the specific steps of building the driving style classifier are as follows: the unsupervised classification result provided by the k-means algorithm is used as the training data of the support vector machine (SVM) algorithm; and the labeled data is transmitted to the SVM algorithm training as the training data, and the trained kmeans-SVM model is used as the driving style recognizer.

[0012] In an alternative, in the calculation of the support vector machine (SVM) model, a certain amount of training data is given, and the data set in the space that cannot be grouped is projected into a certain high-dimensional space by using the kernel variable, and an optimal hyperplane that meets the type condition is found, so that the distance between the data points in the space structure and the hyperplane is optimized, and finally the classification of the data set is completed.

[0013] In an alternative, the trajectory re-planning controller and the trajectory tracking controller constitute a trajectory tracking control system with a re-planning layer, which establishes a trajectory re-planning layer above the trajectory tracking layer from the perspective of trajectory re-planning. The planning layer can re-plan a local expected route that can bypass obstacles according to the obstacle information and reference path information obtained by the vehicle-mounted sensor, and then input the local expected trajectory information to the tracking control layer, so as to realize the tracking of the global reference path while avoiding obstacles.

[0014] In an alternative, in the trajectory re-planning process of the trajectory re-planning controller, the front wheel turning angle range of different style drivers is used as a constraint reference in the vehicle planning process, and different obstacle avoidance function functions are designed. Wherein, S is a weight coefficient, v des is the different expected speed under different driving styles selected for extracting driving feature parameters, v des By affecting the obstacle avoidance function function, (x i ,y i ) are the coordinates of the obstacle points, (x0, y0) are the coordinates of the mass center of the vehicle, and β is a positive number to prevent the denominator of the function from being zero.

[0015] In an alternative, the model predictive controller of the re-planning layer has the following specific form:

[0016]

[0017] U pmin ≤U p ≤U pmax

[0018] wherein η p is the output value, Y p is the lateral position of the re-planned path η pref is the desired output value, U p is the control variable, J obs is the obstacle avoidance function, and ε, z are the output weight matrix and the control variable weight matrix, respectively.

[0019] In an alternative, the trajectory tracking controller is designed using a model predictive control algorithm, and the model prediction uses a linear time-varying model obtained by linearizing a nonlinear model, wherein the nonlinear vehicle dynamics model is as follows:

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] wherein m is the vehicle mass, is the body roll angle, is the vehicle roll rate, I z is the vehicle roll inertia about the z axis, X and Y are the positions of the vehicle in the X and Y directions of the inertial coordinate system, are the velocities of the vehicle in the X and Y directions of the inertial coordinate system, are the velocities of the vehicle in the x and y directions of the vehicle coordinate system, are the accelerations of the vehicle in the x and y directions of the vehicle coordinate system, C cf , C cr are the lateral stiffnesses of the front and rear tires, respectively, C lf , C lr are the longitudinal stiffnesses of the front and rear tires, respectively, s f , s r are the slip ratios of the front and rear wheels, respectively.

[0026] In an alternative, the specific steps of linearizing the nonlinear model are as follows: a state space equation for vehicle lateral control is constructed:

[0027]

[0028] η c = C * M

[0029] State variables Output control quantity u c =δ f C is the coefficient matrix, where δ f For the steering angle of the vehicle's front wheels, Y c These are the yaw angle and lateral position of the vehicle's trajectory, respectively.

[0030] Linearizing the nonlinear dynamic model, we obtain:

[0031]

[0032] The system is discretized using the forward Euler method, and the discretized system is as follows:

[0033] M(k+1)=A k M(k)+B k u c (k)+d(k)

[0034] η c (k)=C*M(k).

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] This invention classifies driving styles using the kmeans-SVM algorithm based on collected driver driving data. The average longitudinal maximum speed of different types of drivers is selected as the expected speed of the corresponding driving style, and the front wheel steering angle range of different driving styles is used as the constraint in the planning process. Then, by combining a trajectory replanning module and a trajectory tracking module based on model predictive control algorithm, obstacle avoidance function functions for different styles are designed. When avoiding obstacles, obstacle avoidance reference paths that meet different driving styles are planned. After avoiding obstacles, the expected path is continued to be tracked. The differences in vehicle behavior affected by driving style during the obstacle avoidance process are verified by path tracking.

[0037] This invention outlines obstacle avoidance paths that cater to different driving styles. It has practical significance for enhancing the understanding of human driver behavior in intelligent vehicles during obstacle avoidance. It is not only a necessary condition for ensuring safety and sufficient performance, but also a necessary condition for adjusting driver needs, improving driver acceptance, and ultimately satisfying driver preferences in a safe environment. Attached Figure Description

[0038] Figure 1 This is a flowchart illustrating the obstacle avoidance planning and design process for the driving style in this invention.

[0039] Figure 2 This is a schematic diagram of the driver data collection process in this invention.

[0040] Figure 3 The figure is a schematic diagram of the Gaussian filtering process in the application.

[0041] Figure 4 The figure is a flow chart of driving style classification in the application.

[0042] Figure 5 The figure is a schematic diagram of the obstacle avoidance module scheme in the application.

[0043] Figure 6 The figure is a schematic diagram of the k-means algorithm clustering result in the application.

[0044] Figure 7 The figure is a schematic diagram of the SVM algorithm clustering result in the application.

[0045] Figure 8 The figure is a schematic diagram of the obstacle avoidance trajectory tracking verification in the application.

[0046] Figure 9 The figure is a schematic diagram of the comparison of obstacle avoidance paths of different styles in the application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with embodiments. The embodiments listed in the application are only used to illustrate the application, and are not used to limit the scope of the application. Any obvious modification or change made to the application does not deviate from the spirit and scope of the application.

[0048] As shown in Figure 1 The application provides a path planning and control method considering driving habits in a static obstacle avoidance scenario. The steps of the method are as follows:

[0049] First, the simulation driving data of different drivers is collected through a driving experiment platform. The drivers are divided into three types through the establishment of a driving style classifier. The expected speed of drivers of different types is used as the speed in the vehicle obstacle avoidance planning. The front wheel angle range of drivers of different styles is used as a constraint reference in the vehicle planning process. Different obstacle avoidance penalty functions are designed in the subsequent trajectory re-planning process. The environmental information, i.e. the information of obstacles, is transmitted to the trajectory re-planning controller by the vehicle-mounted sensor. The controller re-plans the trajectory in combination with the environmental information and the characteristic parameters of different styles. The trajectory tracking controller realizes the tracking of the planned path by the vehicle.

[0050] The specific steps of the method are as follows:

[0051] 1. Driving style classification

[0052] 1.1 Driving data collection and data processing

[0053] Firstly, the experimental road model is designed based on SCANeRTM studio software. The experimental road is a basic driving environment. In order to make the simulated driving road closer to the real life driving road, static things such as trees, buildings and dynamic things such as pedestrians, cars, electric vehicles and bicycles are added in the driving environment. The factors such as traffic participants and road conditions are comprehensively considered. Through collecting the driving data of 28 drivers with different genders, ages and driving ages, the feature data set representing the driving style is established by preprocessing and feature extraction, so as to provide data support for the quantification and classification of driving style. According to the principles of revealing the attribute characteristics of driving style and distinguishing different style types to a large extent, finally, the five dimensions of lateral acceleration, longitudinal acceleration, lateral speed, longitudinal speed and yaw angular velocity are selected to classify the driving style of drivers. The data collection scheme is as shown in Figure 2

[0054] Data processing is to remove the abnormal data of the selected driving data, supplement the missing data, and then use Gaussian filtering for smoothing processing. The present application defines the elements with mean value and three times standard deviation as outliers in the data set and removes them. Finally, the processed information is smoothed by using Gaussian filtering. The data completes the preprocessing to obtain the input data which can be directly used for driving style recognizer. The data processing flow is as shown in Figure 3 .

[0055] 1.2 Building driving style recognizer

[0056] The present application adopts semi-supervised support vector machine algorithm, uses the unsupervised classification result provided by k-means algorithm as the training data of support vector machine SVM (support vector machines) algorithm, then transmits the labeled data to SVM algorithm training as training data, and the trained kmeans-SVM model is used as driving style recognizer. The flow is as shown in Figure 4 .

[0057] As Figure 6 and Figure 7 ​As shown, the k-means algorithm uses time as the similarity index. Given k types in a dataset, the core of each type is derived by averaging the information in each type; therefore, the core of each type can be represented by the cluster center. The Support Vector Machine (SVM) model, during computation, uses a certain amount of training data and kernel variables to project datasets that cannot be grouped in the spatial data onto a high-dimensional space. It then finds an optimal hyperplane that meets the type conditions, optimizing the distance between data points and the hyperplane within the spatial structure, ultimately completing the dataset classification. After clustering, the K-means-SVM algorithm classifies drivers 4, 19, 20, 21, and 27 as aggressive drivers, drivers 1, 2, 3, 5, 7, 8, 9, 12, 13, 17, 18, 22, 24, and 26 as average drivers, and drivers 6, 10, 11, 14, 15, 16, 23, and 25 as robust drivers.

[0058] 2. Add a trajectory tracking control system with a replanning layer.

[0059] When a vehicle is autonomously driving in a real-world environment, tracking control under a given desired trajectory cannot guarantee that the vehicle will accurately handle every trajectory tracking problem. When a static obstacle exists on the given desired trajectory, the vehicle must replan based on the obstacle information, avoid the obstacle, and then continue tracking the desired trajectory. From the perspective of trajectory replanning, a trajectory replanning layer, or local planning layer, is established above the trajectory tracking layer. This planning layer can replan a local desired route that can avoid obstacles based on obstacle information and reference path information obtained from sensors. The local desired trajectory information is then input into the tracking control layer, achieving obstacle avoidance while tracking the global reference path. The control system, as shown below, can then achieve this. Figure 5 As shown, a complete vehicle model was first built on CarSim, a multidisciplinary complex system modeling and simulation platform, and then co-simulated with Simulink to verify the feasibility of the research method.

[0060] 2.1 Trajectory Replanning Controller Design

[0061] The purpose of designing a trajectory replanning controller using model predictive control (MMC) is to improve tracking stability by ensuring that the planning results meet the vehicle's dynamic constraints. However, the planning algorithm itself is computationally intensive. To reduce computational load, overly precise models should not be used. Therefore, a point mass model that ignores vehicle dimensions is chosen for the trajectory replanning controller. The most important part of this section is designing a reasonable evaluation function. Under various constraints, combined with the expected speed and parameter constraints of different driving styles, obstacle avoidance is achieved while minimizing the deviation between the vehicle and the global reference path. Finally, the function is output to the tracking control layer in a reasonable manner.

[0062] The point mass model of the vehicle is as follows:

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] where x, y are the position of the vehicle in x, y axis in the vehicle coordinate system, is the velocity of the vehicle in x axis in the vehicle coordinate system, is the acceleration of the vehicle in x, y axis in the vehicle coordinate system, X, Y are the position of the vehicle in X, Y axis in the inertial coordinate system, is the velocity of the vehicle in X, Y axis in the inertial coordinate system, is the yaw angle of the vehicle, is the yaw rate of the vehicle, a y is the lateral acceleration of the vehicle.

[0069] The state space equation is constructed as:

[0070]

[0071] The state variable of the system is μ(k) represents the control variable, which is the lateral acceleration a y ; the output variable where Y p are the yaw angle and lateral position of the re-planned path, respectively.

[0072] Discretization is performed on it, and the discretized system is as follows: for each sampling time, there are

[0073] I(k+1) = I(k) + T * f(I(k), μ(k))

[0074] In the process of vehicle driving, the information of obstacles is mainly transmitted through on-board sensors. Therefore, the obstacle avoidance function needs to be designed for each obstacle. The penalty function mainly adjusts the function value according to the difference between the obstacle point and the target point. The farther the distance, the smaller the function value. Combined with the expected speed of the vehicle and the penalty function, the obstacle avoidance function is selected as follows:

[0075]

[0076] where S is a weight coefficient, and the increase of the weight coefficient makes the planning result tend to be conservative. The aggressive driver drives faster and usually perceives the obstacle in advance to avoid it; the conservative driver drives slower and perceives the obstacle slower to avoid it; and the ordinary driver is between the two. Therefore, different weight coefficients can be added for different styles of drivers to reflect the influence of the difference of driving styles on the planning curve. des The different expected speeds under different driving styles are selected for extracting the driving feature parameters. des The different reactions of different styles of drivers in the scene of obstacle avoidance are reflected by affecting the obstacle avoidance function. In the formula, (x i ,y i ) are the coordinates of the obstacle point, (x0,y0) are the coordinates of the mass center of the vehicle, and β is a positive number to prevent the denominator of the function from being zero.

[0077] The model predictive controller of the re-planning layer is specifically as follows:

[0078]

[0079] U pmin ≤U p ≤U pmax

[0080] In the objective function, η p is an output value, Y p is the lateral position of the re-planning path, and η pref is an expected output value, and the first item is to avoid the obstacle while minimizing the deviation between the vehicle trajectory and the expected path; U p is a control quantity, which is the lateral acceleration of the vehicle, and the second item is to make the vehicle travel as smoothly as possible to ensure the safety of travel; the third item J obs is an obstacle avoidance function, and the purpose is to avoid the obstacle; and ε and z are output weight matrix and control quantity weight matrix respectively.

[0081] Different driving styles have different ways of handling the steering of the vehicle and the steering force, so the driving style can also be reflected in the front wheel steering angle. Aggressive drivers usually tend to turn the steering wheel more quickly and more sharply to quickly change the direction of the vehicle, and may tend to use a large steering angle when turning; moderate drivers usually handle steering more cautiously, they may turn the steering wheel more gently to ensure the stability and safety of the vehicle, and may use a smaller steering angle when turning to drive smoothly; ordinary drivers are between the two. Therefore, in the process of trajectory re-planning, in order to plan an obstacle avoidance curve that reflects different driving styles, the front wheel steering angle in the feature parameters extracted by different driving styles needs to be converted into a control quantity, that is, the constraint range of the lateral acceleration of the vehicle, through the conversion relationship.

[0082] According to the conversion relationship between the front wheel steering angle and the lateral acceleration, the constraint range of the front wheel steering angle is converted into the constraint range of the control quantity lateral acceleration, R is the turning radius, L is the vehicle wheelbase, and δ is the front wheel steering angle, and there is the following relationship:

[0083]

[0084]

[0085] The new trajectory re-planned is given in the form of discrete points in the prediction time domain, and the number of discrete points containing trajectory information increases with the increase of the prediction time domain. Directly inputting these points into the control layer will occupy many data input interfaces, thereby causing the complexity of the controller design. Therefore, the local trajectory information needs to be processed to realize the docking of the planning layer and the control layer. The processing of discrete points mostly uses curve fitting, and different curves are used, such as polynomial fitting, polynomial fitting and power exponential fitting.

[0086] The vehicle designed by the application has constraints on vehicle dynamics, the vehicle position curve is required to be continuous, the yaw angle curve is required to be first-order continuous, and the acceleration curve is required to be second-order continuous, and a quintic polynomial method is selected for fitting the discrete points, and the form is as follows:

[0087] Y p = a0t 5 + a1t 4 + a2t 3 + a3t 2 + a4t + a5

[0088]

[0089] Since the real-time requirement of planning layer is lower than that of control layer, the point mass model is also greatly simplified compared with the nonlinear dynamics model. The nonlinear model predictive control algorithm with higher solving accuracy can fully meet the requirements of trajectory re-planning, and the use of nonlinear objective function can also bring convenience to the design of subsequent penalty function. Therefore, the linearization of equation (2.5) is no longer performed, but the nonlinear model is directly solved. The lateral position Y p and the yaw angle of the fitted re-planning path are output as the input of the trajectory tracking controller to achieve tracking of the new trajectory.

[0090] 2.2 Trajectory tracking controller design

[0091] The trajectory tracking controller is designed using the model predictive control algorithm, and the classic three-degree-of-freedom vehicle dynamics model is selected. Since the real-time requirement of the vehicle to the controller is relatively strict during collision avoidance, nonlinear model predictive control is difficult to meet, so the nonlinear model is linearized to obtain a linear time-varying model.

[0092] The nonlinear dynamics model of the vehicle is as follows

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] In the formula, m is the mass of the vehicle, is the body yaw angle, is the vehicle yaw rate, I z is the moment of inertia of the vehicle around the z axis, X and Y are the positions of the vehicle in the X and Y axis directions in the inertial coordinate system, are the velocities of the vehicle in the X and Y axis directions in the inertial coordinate system, are the velocities of the vehicle in the x and y axis directions in the vehicle coordinate system, are the accelerations of the vehicle in the x and y axis directions in the vehicle coordinate system, C cf and C cr are the lateral stiffness of the front and rear tires of the vehicle, C lf and C lr are the longitudinal stiffness of the front and rear tires of the vehicle, s f and s r are the slip rates of the front and rear wheels of the vehicle.

[0099] First, the state-space equations for vehicle lateral control are constructed as follows:

[0100]

[0101] η c =C*M

[0102] State variables Output control quantity u c =δ f C is the coefficient matrix. Where δ f For the steering angle of the vehicle's front wheels, Y c These represent the yaw angle and lateral position of the vehicle's trajectory, respectively.

[0103] Linearizing the nonlinear dynamic model, we obtain:

[0104]

[0105] The system is discretized using the forward Euler method. The discretized system is as follows, and for each sampling time step:

[0106] M(k+1)=A k M(k)+B k u c (k)+d(k)

[0107] η c (k)=C*M(k)

[0108] The control variable for the system shown in the above equation is the front wheel steering angle. If the above system is controlled, only the front wheel steering angle can be constrained, not the increment of the front wheel steering angle. In actual control, it is desirable for the change in the front wheel steering angle to be as gradual as possible to meet the requirements of comfort index. Therefore, a new state space is constructed using the change in the front wheel steering angle as the control variable:

[0109]

[0110]

[0111] Define Δu c (k), that is, the control quantity increment at time k is the front wheel steering angle increment, N P For prediction in the time domain, N c To control the time domain, and N p >N c Iterative derivation yields:

[0112]

[0113] In the formula

[0114]

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] The system not only needs to consider the following performance of the vehicle, but also needs to consider whether it is appropriate to ride, and the performance index target function is defined as follows

[0122]

[0123] The function of the whole expression is to make the system track the desired trajectory quickly and smoothly, where is the predicted output value, that is, the yaw angle and lateral position of the predicted vehicle trajectory, η cref is the desired output value, that is, the yaw angle and lateral position of the re-planned trajectory, the first term of the target function reflects the following ability of the system to the reference trajectory; ΔU c is the control increment in the control time domain, that is, the increment of the front wheel steering angle, the second term reflects the requirement for smooth change of the control quantity; Since the system is a real-time system, the target function may appear no solution at some time, so the third term adds a relaxation factor ε to prevent the system from appearing no solution of the target function, and avoid the system from appearing dead lock and collapse due to no solution. In the formula, Q, R and p are output weight, control weight and relaxation factor weight respectively.

[0124] When the vehicle is running at high speed or turning, the change of the front wheel steering angle can directly affect the curve driving radius of the vehicle, and the front wheel steering angle constraint can limit the steering radius of the vehicle when turning and curve driving, so limiting the front wheel steering angle can effectively avoid the vehicle exceeding the curve driving range to meet the safety and comfort requirements, where u cmin and u cmax represent the minimum and maximum values of the front wheel steering angle respectively.

[0125]

[0126] In addition, in the vehicle trajectory tracking process, the large front wheel angle change of the vehicle needs to be decomposed into small increments to ensure the accuracy and smoothness of the vehicle driving on the trajectory, and to help the vehicle better follow the established trajectory. At the same time, excessive front wheel angle change is easy to cause the vehicle to shake in motion, and the constraint of the front wheel angle increment can effectively prevent this situation from happening and improve the stability of the vehicle control. The control increment constraint is as follows:

[0127]

[0128]

[0129] Where, Δu cmin and Δu cmax represent the minimum and maximum values of the front wheel angle increment, respectively.

[0130] According to Figure 8 and Figure 9 , in the trajectory re-planning controller, the front wheel angle constraint under different styles has been converted into a constraint on the lateral acceleration, and in the trajectory tracking controller, according to different driving styles, the corresponding upper and lower limit constraints of the front wheel angle are added to ensure that the vehicle can track the re-planned route of different styles under the premise of safe and stable driving of the vehicle.

[0131] The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present disclosure, which should be covered within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A path planning and control method considering driving habits in a static obstacle avoidance scenario, characterized in that, Comprise the following steps: Step S1 data collection and processing: collecting simulation driving data of different drivers through a driving experiment platform; step S2: classifying the driving style of the driver by building a driving style classifier and using a kmeans-SVM algorithm, and selecting appropriate speed as the driving expected speed of different style drivers, and taking the expected speed of different types of drivers as the speed during vehicle obstacle avoidance planning; Step S3: taking the front wheel steering angle range of different style drivers as a constraint reference in the vehicle planning process; step S4: the vehicle sensor transmits the information of the obstacle to the trajectory re-planning controller, and the trajectory re-planning controller re-plans the trajectory in combination with the environmental information and the characteristic parameters of different styles; The trajectory tracking controller realizes the tracking of the planned path by the vehicle; The trajectory re-planning controller and the trajectory tracking controller constitute a trajectory tracking control system with a re-planning layer, which establishes a trajectory re-planning layer above the trajectory tracking layer from the perspective of trajectory re-planning, and the planning layer can re-plan a local expected route that can bypass obstacles according to the obstacle information obtained by the vehicle sensor and the reference path information, and then input the local expected trajectory information to the tracking control layer, so as to realize the tracking of the global reference path while avoiding obstacles; In the trajectory replanning controller trajectory replanning process, the front wheel turning angle range of different driving styles is taken as a constraint reference in the vehicle planning process, and different obstacle avoidance function functions are designed; the obstacle avoidance function function is ; wherein S is a weight coefficient, is the different expected speeds under different driving styles selected by extracting driving characteristic parameters, by affecting the obstacle avoidance function function, is the coordinates of the obstacle points, is the centroid coordinates of the vehicle, is a positive number to prevent the denominator of the function from being zero; The model predictive controller of the re-planning layer has the following specific form: wherein, is an output value, a lateral position of a replanned path and a yaw angle , is a desired output value, a desired path, is a control amount, is an obstacle avoidance function function, are an output weight matrix and a control amount weight matrix, respectively.

2. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 1, characterized in that, In the step S1 data collection and processing, the specific steps of data collection are as follows: collecting driving data of drivers of different genders, ages and driving ages and pre-processing and feature extraction, and establishing a characteristic data set representing driving style; selecting lateral and longitudinal acceleration, lateral and longitudinal speed, and yaw rate as the basis for classifying the driving style of the driver.

3. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 2, characterized in that, In the step S1 data collection and processing, the specific steps of data preprocessing are as follows: removing abnormal data and supplementing missing data from the selected driving data, and then performing smoothing processing by using Gaussian filtering.

4. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 1, characterized in that, The specific steps of building a driving style classifier are as follows: using the unsupervised classification result provided by the k-means algorithm as the training data of the support vector machine (SVM) algorithm; the labeled data is used as the training data to train the SVM algorithm, and the trained kmeans-SVM model is used as the driving style recognizer.

5. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 4, characterized in that, The support vector machine (SVM) model calculates by giving a certain amount of training data and using the kernel variable method to project the data set that cannot be grouped in space to a certain high-dimensional space, and find an optimal hyperplane that meets the type condition, so that the distance between the data points in the space structure and the hyperplane is optimized, and finally the classification of the data set is completed.

6. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 1, characterized in that, The trajectory tracking controller is designed using the model predictive control algorithm, and the model prediction uses a linear time-varying model obtained by linearizing a nonlinear model; wherein the nonlinear dynamics model of the vehicle is as follows: where m is the vehicle mass, is the vehicle body roll angle, is the vehicle roll rate, is the vehicle moment of inertia about the z axis, is the vehicle position in the X, Y axis directions in the inertial coordinate system, is the vehicle velocity in the X, Y axis directions in the inertial coordinate system, is the vehicle velocity in the x, y axis directions in the vehicle coordinate system, is the vehicle acceleration in the x, y axis directions in the vehicle coordinate system, are the front and rear tire cornering stiffness, respectively, are the front and rear tire longitudinal stiffness, respectively, are the front and rear wheel slip ratio, respectively, is the vehicle front wheel steering angle.

7. The path planning and control method considering driving habits in a static obstacle avoidance scenario according to claim 6, characterized in that, The specific steps of linearizing the nonlinear model are as follows: constructing the state space equation of the vehicle lateral control: state variable , output variable , control variable , C is a coefficient matrix, wherein is a front wheel steering angle of the vehicle, are a yaw angle and a lateral position of a vehicle trajectory, respectively; Linearize the nonlinear dynamics model to obtain: Discretize it using the forward Euler method, and the discretized system is as follows: 。

Images