Adaptive MPC Vehicle Trajectory Tracking Control Method for Online Identification of Hazard Levels

Abstract

This solution provides an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels, including the following steps: obtaining the vehicle reference trajectory of the vehicle to be controlled; establishing a three-degree-of-freedom vehicle dynamics model; obtaining the vehicle state information of the vehicle to be controlled; identifying the overall danger level of the current vehicle to be controlled based on the vehicle state information; using the overall danger level and lateral error of the vehicle to be controlled as input variables of the ant algorithm to adaptively adjust the prediction horizon and the state error weight coefficient of the objective function; establishing an MPC trajectory tracking controller according to the three-degree-of-freedom vehicle dynamics model and the reference trajectory, and assigning the prediction horizon and the state error weight coefficient to the MPC trajectory tracking controller. The MPC trajectory tracking controller performs rolling optimization on the vehicle state information of the vehicle to be controlled with the vehicle reference trajectory as the reference information to obtain an adaptive MPC vehicle trajectory, improving the accuracy of MPC vehicle trajectory tracking.

Description

Technical Field

[0001] The present application relates to the field of vehicle trajectory tracking, and particularly to an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels. Background Art

[0002] In recent years, intelligent driving vehicles have been developing rapidly towards electrification and automation. As a key technology for intelligent driving vehicles, trajectory tracking control has become the focus of research in this field. Its core lies in ensuring that the intelligent driving vehicle travels stably and the passengers are comfortable while accurately tracking the desired trajectory. Therefore, designing a trajectory tracking control strategy with accuracy, driving stability, and real-time performance has extremely high research value.

[0003] Currently, the mainstream methods for trajectory tracking include PID control, fuzzy control, linear quadratic regulator (LQR), model predictive control (MPC), etc. Among them, methods such as PID control, fuzzy control, and linear quadratic regulator (LQR) can simplify the control model and shorten the calculation time. However, in the face of the nonlinearity of the vehicle system, it is necessary to frequently adjust the weight parameters to adapt to different working conditions, and the dependence on the weight parameters is too high; MPC has attracted much attention in the field of intelligent driving because it can achieve multi-variable and multi-constraint rolling optimization and has robustness to uncertain parameter perturbations. However, since the prediction horizon in the MPC parameters is closely related to the trajectory tracking accuracy and vehicle body stability, a fixed prediction horizon is difficult to meet the control requirements in complex driving environments, severely restricting the accuracy, stability, and real-time performance of vehicle trajectory tracking in complex driving environments. Therefore, how to improve the trajectory tracking accuracy, stability, and reduce the calculation time is the focus of research.

[0004] In order to solve the above-mentioned problems of MPC trajectory tracking control in practical applications, many researchers have conducted in-depth research on the influence of time domain parameters and weight parameters on trajectory tracking accuracy, computational efficiency and vehicle stability, and produced a series of representative patent results. For example, patent number CN118672251A proposes an online variable time domain model predictive control method based on lateral stability, combining expert scoring method, multiple chain method and matter-element extension analysis method to classify the danger level of vehicle status, and then adaptively adjust the prediction time domain, control time domain and sampling time to achieve vehicle trajectory tracking control: This method can better achieve tracking accuracy, vehicle stability and real-time performance under simple working conditions at medium and low speeds, but under complex working conditions at medium and high speeds, because matter-element extension analysis needs to build a multi-dimensional matter-element model and perform complex transformation calculations, when working in conjunction with the PID regulator, the huge amount of calculation generated by real-time adjustment of the prediction time domain reduces the real-time performance and accuracy of the control, and the vehicle trajectory tracking response lags. Patent No. CN119370127A proposes a trajectory tracking control method for electric vehicles, constructs a vehicle trajectory tracking control model, and uses the obtained state quantity as an evaluation index. The speed-pause particle swarm algorithm is introduced to optimize the state quantity error weight coefficient matrix Q in real time, which can greatly improve the vehicle trajectory tracking accuracy. However, the speed-pause particle swarm algorithm is prone to fall into the local optimum, making it difficult to achieve the optimal trajectory tracking effect and unable to accurately track the expected trajectory. At the same time, it is necessary to perform multiple iterative update calculations on the speed and position of each particle. As the number of particles increases and the dimension of the solution space increases, the amount of calculation will increase exponentially, resulting in reduced real-time performance of trajectory tracking.

[0005] The above two patents attempt to optimize the time domain and weight parameters of MPC from different angles to improve trajectory tracking performance. However, under complex working conditions at medium and high speeds, due to the low computational efficiency of the existing parameter adjustment strategy, it is unable to respond to complex environments in a timely manner, which leads to poor real-time vehicle trajectory tracking, easy shaking and deviation from the trajectory during driving, and driving stability is affected; and the parameter optimization is not accurate, making it difficult for the vehicle to accurately track the desired trajectory, and the trajectory tracking accuracy is reduced. It can be seen that the current strategy faces huge challenges in balancing computational efficiency, tracking accuracy and vehicle stability, and there is a clear gap in achieving the trajectory tracking control goal of combining accuracy, stability and real-time performance. Summary of the invention

[0006] The embodiment of the present application provides an adaptive MPC vehicle trajectory tracking control method with online identification of hazard levels. In view of the technical difficulties that are common in existing adaptive MPC trajectory tracking control methods, which cannot balance accuracy, stability and real-time performance, a trajectory tracking control method with higher accuracy, stronger stability and better real-time performance is provided.

[0007] In a first aspect, an embodiment of the present application provides an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels, including the following steps:

[0008] Step S1: Obtain the vehicle reference trajectory of the controlled vehicle;

[0009] Step S2: Establish a three-degree-of-freedom vehicle dynamics model;

[0010] Step S3: Obtain the vehicle state information of the controlled vehicle;

[0011] Step S4: Identify the overall danger level of the current controlled vehicle based on the vehicle state information;

[0012] Step S5: Use the overall danger level and lateral error of the controlled vehicle as input variables of the ant algorithm to adaptively adjust the prediction horizon and the state error weight coefficient of the objective function;

[0013] Step S6: Establish an MPC trajectory tracking controller according to the three-degree-of-freedom vehicle dynamics model and the reference trajectory, and assign the prediction horizon and the state error weight coefficient to the MPC trajectory tracking controller. The MPC trajectory tracking controller performs rolling optimization on the vehicle state information of the controlled vehicle with the vehicle reference trajectory as the reference information to obtain an adaptive MPC vehicle trajectory.

[0014] In a second aspect, an embodiment of the present application provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the adaptive MPC vehicle trajectory tracking control method for online identification of danger levels.

[0015] The main contributions and innovations of the present invention are as follows:

[0016] This solution uses the k-means clustering algorithm combined with the analytic hierarchy process to offline construct an online identification system for the overall danger level of vehicles. It comprehensively considers the mutual relationship and importance of various vehicle state parameters in vehicle danger level identification. By constructing a hierarchical structure model and determining the weight vector, it systematically analyzes the vehicle state parameters, thereby more scientifically and reasonably dividing the overall danger level of vehicles, enabling better adaptive control of vehicle stability and improving the real-time performance of vehicle trajectory tracking.

[0017] In addition, based on the real-time lateral error and the overall danger level of the vehicle, this solution introduces the ant colony algorithm to dynamically adjust the prediction horizon of the MPC and the state error weight coefficient of the objective function. Compared with traditional MPC, it significantly enhances the adaptability to complex environments and better balances the stability, accuracy, and real-time performance of vehicle trajectory tracking.

[0018] Details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more concise and understandable. Description of the Drawings

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0020] Figure 1 is a flowchart of an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels according to an embodiment of the present application.

[0021] Figure 2 is a diagram of a dynamic model of the inertial coordinate system and the forces acting on the vehicle constructed.

[0022] Figure 3 Flowchart of identifying the overall danger level of the vehicle by the analytic hierarchy process.

[0023] Figure 4 Flowchart of optimizing the prediction horizon and state error weight coefficients by the ant colony algorithm.

[0024] Figure 5 is a schematic diagram of the framework of an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels.

[0025] Figure 6 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. Detailed Embodiments

[0026] Here, exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. On the contrary, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0027] It should be noted that: In other embodiments, the steps of the corresponding methods are not necessarily executed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may also be combined into a single step for description in other embodiments.

[0028] Embodiment 1

[0029] like Figure 1 As shown, this scheme provides an adaptive MPC vehicle trajectory tracking control method for online identification of danger levels, including the following steps:

[0030] Step S1: obtaining a vehicle reference trajectory of the controlled vehicle;

[0031] Step S2: establishing a three-degree-of-freedom vehicle dynamics model;

[0032] Step S3: obtaining vehicle status information of the controlled vehicle;

[0033] Step S4: Identify the overall danger level of the currently controlled vehicle based on the vehicle status information;

[0034] Step S5: using the overall danger level and lateral error of the controlled vehicle as input variables of the ant algorithm to adaptively adjust the state error weight coefficient of the prediction time domain and the objective function;

[0035] Step S6: An MPC trajectory tracking controller is established according to the three-degree-of-freedom vehicle dynamics model and the reference trajectory, and the prediction time domain and state error weight coefficients are assigned to the MPC trajectory tracking controller. The MPC trajectory tracking controller performs rolling optimization on the vehicle state information of the controlled vehicle with the vehicle reference trajectory as reference information to obtain an adaptive MPC vehicle trajectory.

[0036] In step S1, the vehicle reference trajectory output by the planning layer is obtained as reference information for the MPC trajectory planner, wherein the vehicle reference trajectory includes the ordinate and abscissa of the reference trajectory. It should be noted that the vehicle reference trajectory is generated by the planning layer by integrating multiple information, which is not an improvement point of this solution, so it will not be described in detail here.

[0037] In step S2, an inertial coordinate system is established and a mechanical equilibrium equation is established according to the forces on the vehicle in the x-axis, y-axis and z-axis. The longitudinal stiffness and lateral stiffness of the front and rear wheels of the vehicle are obtained based on the magic formula tire model. The longitudinal stiffness and lateral stiffness of the front and rear wheels of the vehicle are substituted into the mechanical equilibrium equation to establish a three-degree-of-freedom vehicle dynamics model.

[0038] Specifically, Figure 2 As shown, Figure 2 It is a schematic diagram of the constructed inertial coordinate system. The inertial coordinate system XOY is established with the center of mass of the vehicle as the origin O, the longitudinal axis of the vehicle as the X-axis, and the transverse axis of the vehicle as the Y-axis. The forces on the vehicle in the x-axis, y-axis, and z-axis are defined, including the lateral force on the front tires of the vehicle. F cf , the lateral force on the rear tires of the vehicle F cr , the longitudinal force on the front tireF lf , the longitudinal force received by the rear tire F lr , the force in the X direction received by the front tire F xf , the force in the X direction received by the rear tire F xr , the force in the Y direction received by the front tire F yf , the force in the Y direction received by the rear tire F yr .

[0039] Based on the forces on the vehicle in the x-axis, y-axis, and z-axis, the following mechanical equilibrium equations are established:

[0040] ;

[0041] where is the curb weight of the vehicle, and a and b are the distances between the vehicle's center of mass position and the centers of the front and rear axles respectively; is the front wheel steering angle; is the longitudinal velocity; is the longitudinal acceleration; is the lateral velocity; is the lateral acceleration; is the yaw angular velocity; is the yaw angular acceleration; F cf is the lateral force received by the vehicle's front tire; F cr is the lateral force received by the vehicle's rear tire; F lf is the longitudinal force received by the front tire; F lr is the longitudinal force received by the rear tire, F xf is the force in the x direction received by the front tire; F xr is the force in the x direction received by the rear tire; F yf is the force in the y direction received by the front tire; F yr is the force in the y direction received by the rear tire , I z is the moment of inertia of the vehicle about the z-axis.

[0042] In addition, when conducting vehicle dynamics simulation research, the handling stability of the vehicle is closely related to the mechanical properties of the tires. Therefore, this solution obtains the longitudinal stiffness and cornering stiffness of the vehicle's front and rear wheels based on the Magic Formula tire model, as follows:

[0043] Based on the magic formula tire model, the longitudinal force is simplified as the product of the longitudinal stiffness and the slip ratio within a specific longitudinal force range, which is expressed as follows:

[0044] ;

[0045] ;

[0046] where is the longitudinal stiffness of the vehicle tire, and its value is determined by the vehicle model and tire type; s is the slip ratio, v represents the vehicle's center-of-mass velocity, r represents the wheel radius, represents the wheel angular velocity; is the longitudinal force.

[0047] Based on the magic formula tire model, the lateral force is simplified as the product of the cornering stiffness and the sideslip angle within a specific sideslip angle range, which is expressed as follows:

[0048] ;

[0049] where is the cornering stiffness, and its value is determined by the vehicle model and tire type, α is the sideslip angle, is the lateral force.

[0050] It should be noted that the specific longitudinal force range refers to [-0.05 * 10 4 kN, 0.05 * 10 4 kN], and the specific sideslip angle range refers to [-5°, 5°].

[0051] Due to the existence of various non-linear relationships in the model, it will lead to complex calculations and affect the real-time performance of the controller during the design process of the controller. Therefore, in this solution, the longitudinal stiffness and cornering stiffness of the front and rear wheels of the vehicle are substituted into the mechanical equilibrium equation to establish a three-degree-of-freedom vehicle dynamics model under the small-angle assumption:

[0052] ;

[0053] where is the cornering stiffness of the front-wheel tire; is the cornering stiffness of the rear-wheel tire; is the longitudinal stiffness of the front-wheel tire; is the longitudinal stiffness of the rear-wheel tire; is the longitudinal velocity in the inertial coordinate system; is the lateral velocity in the inertial coordinate system; is the curb weight of the vehicle, and a and b are the distances between the vehicle's center of mass and the centers of the front and rear axles respectively; is the front-wheel steering angle; is the longitudinal velocity; is the longitudinal acceleration; is the lateral velocity; is the lateral acceleration; is the yaw rate; is the yaw acceleration.

[0054] It should be noted that the cornering stiffness and longitudinal stiffness of the front and rear tires are obtained according to the above formulas for calculating cornering stiffness and longitudinal stiffness.

[0055] Furthermore, the three-degree-of-freedom vehicle dynamics model is expressed as a non-linear dynamic state equation using the state space equation, and is transformed into a linear state equation through Taylor expansion based on the non-linear dynamic state equation, and then discretized using the forward Euler method to obtain the discretized linear state equation. The advantage of this is that the non-linear dynamic state equation can intuitively describe the complex relationships between the state variables of the vehicle system, comprehensively present the vehicle dynamics characteristics, and help researchers grasp the essence of vehicle motion as a whole. Further transformation into a linear state equation greatly reduces the analysis difficulty, enabling researchers to more clearly understand the system behavior and performance.

[0056] In addition, the linear state equation is also more suitable for model predictive control of the MPC trajectory tracking controller. Taking MPC as an example, its receding horizon optimization process depends on the prediction of the future state of the system. The discretized linear state equation can be conveniently used for prediction calculations to determine the optimal control input sequence, achieve precise tracking control of the vehicle trajectory, and improve the control effect and system performance.

[0057] The three-degree-of-freedom vehicle dynamics model is expressed as a non-linear dynamic state equation using the state space equation, and is transformed into a linear state equation through Taylor expansion based on the non-linear dynamic state equation, and then discretized using the forward Euler method to obtain the discretized linear state equation, where the state space equation is expressed as:

[0058] ;

[0059] ;

[0060] ;

[0061] ;

[0062] ;

[0063] ;

[0064] where is the vehicle state parameter at time k; is the vehicle state parameter at time (k + 1); is the state transition matrix; is the input coefficient matrix; is the output coefficient matrix, represents the output at time k, represents t the control applied at the 0 sampling time, represents at the state trajectory under the control, T is the sampling time, and I represents the identity matrix, is the vehicle control at time k, is the Jacobian matrix of the system state transition function f with respect to ; is the Jacobian matrix of the system state transition function f with respect to u, , is a certain sampling time.

[0065] In step S3, vehicle state information of the controlled vehicle is obtained, where the vehicle state information is from the built-in sensors of the controlled vehicle.

[0066] In some embodiments, considering the vehicle state information related to vehicle stability includes the following vehicle state parameters: longitudinal speed , longitudinal acceleration , longitudinal speed , lateral acceleration A y , yaw rate , roll angle Roll , roll angular acceleration , sideslip angle at the center of mass , sideslip angular velocity at the center of mass , where and cannot be directly measured by sensors and need to be estimated. The vehicle state information is organized into a state matrix S as follows:

[0067] .

[0068] In step S4, based on the dataset of vehicle state parameters of other vehicles collected by Carsim software, an online recognition system for danger levels is constructed using the k-means clustering algorithm and the analytic hierarchy process. The vehicle state information obtained in real time of the current controlled vehicle is input into the online recognition system for danger levels to obtain the overall vehicle level of the vehicle.

[0069] Further, based on the dataset, the k-means clustering algorithm is used to determine the risk level of each vehicle state parameter in the vehicle state information. The analytic hierarchy process is used to determine the criterion layer weight vector of the influence of each vehicle state parameter on the overall vehicle risk level and the scheme layer weight vector of each vehicle state parameter under each risk level, so as to construct an online risk level identification system.

[0070] It should be noted that in this solution, the dataset used in constructing the online risk level identification system is the vehicle state parameters of other vehicles obtained by using simulation software.

[0071] The steps for determining the risk level of each vehicle state parameter in the vehicle state information based on the dataset using the k-means clustering algorithm are as follows:

[0072] Specifically, a dataset of vehicle state parameters of other vehicles is collected based on the Carsim software. And to avoid the influence of dimension on the clustering result, the data in the dataset is standardized to obtain a normalized dataset.

[0073] Among them, the formula for standardization is as follows: ;

[0074] where s is each vehicle state parameter in the vehicle state information, is the vehicle state parameter after standardization, min(s) represents the minimum value of the vehicle state parameter, and max(s) represents the maximum value of the vehicle state parameter.

[0075] Further, the k-means clustering algorithm is used for clustering based on the normalized dataset to obtain the risk level of each vehicle state parameter. Specifically, the number of clustering centers is preset. A data point is randomly selected as a clustering center in the normalized dataset corresponding to each type of vehicle state parameter, and other clustering centers are determined according to the criterion that the distance between the clustering centers is as far as possible. Based on the initially determined clustering centers, the normalized dataset of the current type of vehicle state parameter is clustered iteratively until the clustering centers no longer change or reach the preset number of iterations. The risk level of each vehicle state parameter is determined based on the clustering centers.

[0076] In each iteration of the clustering iteration process, the distance between each data point and the existing clustering centers is calculated, and the data point is assigned to the cluster where the nearest clustering center is located. The average value of the data points in each cluster is calculated as the new clustering center, where the distance is the Euclidean distance.

[0077] In some embodiments, according to the professional knowledge of vehicle driving safety and the actual application background, the number of clustering centers is set to 5, corresponding to risk levels 1 to 5 respectively.

[0078] Such asFigure 3 As shown in Figure 3 , the steps for determining the interrelationship and importance degree among each vehicle state information by using the analytic hierarchy process are as follows:

[0079] Determine the hierarchical structure, where the target layer in the hierarchical structure is the overall vehicle danger level, the criterion layer is each vehicle state parameter, and the scheme layer is the number of levels corresponding to different danger levels;

[0080] Use the scale method to make pairwise comparisons of each vehicle state parameter in the criterion layer, construct the criterion layer judgment matrix, and form a scheme layer judgment matrix for each vehicle state parameter:

[0081] Perform consistency verification on the criterion layer judgment matrix and the scheme layer judgment matrix. Calculate the criterion layer weight vector for the criterion layer judgment matrix that passes the consistency verification, and calculate the scheme layer weight vector for the scheme layer judgment matrix that passes the consistency verification.

[0082] In some embodiments, the scheme layer has 5 levels, where level 1 represents the lowest danger level and level 5 represents the highest danger level.

[0083] In some embodiments, pairwise comparisons are made of each vehicle state parameter in the criterion layer to determine the relative importance degree of the influence of each vehicle state parameter on the overall vehicle danger level. The constructed criterion layer judgment matrix A is expressed as:

[0084] ;

[0085] where represents the scale value of the comparison between state parameter and state parameter j , for example, a 38 represents the scale value of the 3rd state parameter and the 8th state parameter, and satisfies:

[0086] 。

[0087] where represents the scale value of the comparison between state parameter j and state parameter i,

[0088] The scheme layer judgment matrix formed for each vehicle state parameter is expressed as:

[0089] ;

[0090] where represents the standard value of the comparison between the th danger level and the th danger level of the current vehicle state parameter, for example, b 35Indicates the standard value for comparing the 3rd and 6th hazard levels of the current vehicle status parameters, which also satisfies:

[0091] ;

[0092] where represents the standard value for comparing the th hazard level and the i th hazard level of the current vehicle status parameters

[0093] Perform consistency verification on the criterion layer judgment matrix and the scheme layer judgment matrix as follows:

[0094] For the consistency verification of the criterion layer judgment matrix, first use Matlab software to calculate the maximum eigenvalue A of the criterion layer judgment matrix , calculate the consistency index based on the maximum eigenvalue, where n is the number of vehicle status parameters, and CI A represents the consistency index. Calculate the consistency ratio based on the maximum eigenvalue, and CR A represents the consistency ratio. According to relevant materials take 1.41. When CR A <0.1, the criterion layer judgment matrix A has satisfactory consistency. If CR A ≥0.1, then the element values in the matrix A need to be readjusted until the consistency verification is passed.

[0095] Perform consistency verification on each scheme layer judgment matrix. First, calculate the maximum eigenvalue ( i represents the scheme layer judgment matrix corresponding to the i th vehicle status parameter), then calculate the consistency index based on the maximum eigenvalue, where z is the number of hazard levels, and CI i represents the consistency index. Calculate the consistency ratio based on the maximum eigenvalue, and CR i represents the consistency ratio. According to relevant materials take 1.12. When <0.1, the scheme layer judgment matrix has satisfactory consistency; if ≥0.1, then the corresponding matrix needs to be readjusted until the consistency verification is passed.

[0096] The weight vector of the criterion layer calculated from the judgment matrix of the criterion layer that passes the consistency check is specifically as follows: Calculate the eigenvector of the judgment matrix of the criterion layer, and after normalizing the eigenvector, obtain the weight vector of the criterion layer for each vehicle state parameter affecting the overall vehicle danger level, which is expressed as follows:

[0097] ;

[0098] where is the weight vector of the criterion layer for the i-th vehicle state parameter affecting the overall vehicle danger level, n is the number of vehicle state parameters, 1 < i ≤ n, is the weight vector of the criterion layer for each vehicle state parameter affecting the overall vehicle danger level.

[0099] The weight vector of the scheme layer calculated from the judgment matrix of the scheme layer that passes the consistency check is specifically as follows: Calculate the eigenvector of the judgment matrix of the scheme layer, and after normalizing the eigenvector, obtain the weight vector of the scheme layer for each vehicle state parameter at different danger levels.

[0100] Taking the longitudinal speed v x as an example, the weight vector of the scheme layer is obtained as follows:

[0101] ;

[0102] where is the weight vector of the scheme layer for the longitudinal speed, is the weight vector of the scheme layer for the longitudinal speed with respect to the j-th danger level, m is the number of danger levels, 1 < j ≤ m.

[0103] Similarly, the weight vectors of the vehicle state parameters of longitudinal acceleration , longitudinal speed , lateral acceleration A y , yaw rate , roll angle Roll , roll angular acceleration , sideslip angle at the center of mass , sideslip angular velocity at the center of mass are respectively: 、 、 、 、 、 、 、 .

[0104] In the step of inputting the vehicle state information obtained in real time of the current controlled vehicle into the online hazard level recognition system to obtain the overall vehicle level of the vehicle, the hazard levels of the respective vehicle state parameters in the vehicle state information obtained in real time are determined according to the hazard levels of the vehicle state parameters, and then the overall vehicle level of the vehicle is calculated by combining the criterion layer weight vector and the scheme layer weight vector, as shown below: ;

[0105] where L represents the overall vehicle level of the vehicle, and i represents the vehicle state parameter represents the hazard level of the vehicle state parameter, is the criterion layer weight vector of the i-th vehicle state parameter, is the scheme layer weight vector of the vehicle state parameter i corresponding to this hazard level .

[0106] In some embodiments, the overall vehicle level of the vehicle is divided into 5 levels, level 1: L ∈[0, 0.2), level 2: L ∈[0.2, 0.4), level 3: L ∈[0.4, 0.6), level 4: L ∈[0.6, 0.8), level 5: L ∈[0.8, 1].

[0107] Since the main parameters affecting the performance of the MPC trajectory tracking controller include the prediction horizon, control horizon, state error weight function of the objective function, and control quantity error weight coefficient, and among them, the prediction horizon and the state error weight coefficient of the objective function are for the tracking accuracy and stability of the vehicle trajectory of the MPC trajectory tracking controller, so step S5 of this solution further adaptively adjusts the prediction horizon and the state error weight coefficient of the objective function.

[0108] Specifically, the present invention uses the ant colony algorithm for adaptive adjustment, which has the advantages of improving the search efficiency, enhancing the global search ability, adapting to different problem scales and characteristics, strong robustness, and reducing manual intervention, to adaptively adjust the prediction horizon of the MPC and the weight coefficients of the objective function.

[0109] Furthermore, as Figure 4As shown in the figure, a fitness function is constructed based on the initial prediction time domain of the MPC trajectory tracking controller, the overall danger level of the vehicle, and the lateral error. The value ranges of the number of ants, the prediction time domain, and the weight coefficients of the objective function in the ant colony algorithm are determined, and an initial pheromone matrix is created. For each ant, the ant searches for a suitable prediction time domain and weight coefficients of the objective function within the value ranges of the prediction time domain and the weight coefficients of the objective function according to the pheromone matrix and the transition probability formula, and calculates the value of the fitness function corresponding to each ant. After multiple iterative searches, the prediction time domain and the weight coefficients of the objective function corresponding to the ant with the largest value of the fitness function are used as the optimal solution.

[0110] The construction of the fitness function is as follows:

[0111] ;

[0112] where F is the fitness function, is the initial prediction time domain, k represents the vehicle driving time, is the overall danger level of the vehicle output at time k, is the lateral error at time k, that is, the difference between the reference trajectory coordinate and the actual lateral coordinate of the vehicle; is a very small parameter, w1 is the weight coefficient of the lateral error, and w2 is the weight coefficient of the overall danger level of the vehicle.

[0113] The meaning of this fitness function is to minimize the combined influence of the lateral error and the overall danger level of the vehicle under the consideration of the overall danger level of the vehicle. The larger the value, the better the current combination of the prediction time domain and the weight coefficients.

[0114] In the step of "determining the number of ants in the ant colony algorithm", the number of ants participating in the search is set, and the number of ants is 10 - 30.

[0115] Preferably, the number of ants is 30.

[0116] It should be noted that the number of ants determines the exploration intensity of the ant colony algorithm in the search space. If the number of ants is too small, the algorithm will converge to the local optimal solution prematurely, while if the number of ants is too large, the computational amount of operations such as calculating the ant path and fitness in each iteration will be very large.

[0117] In the step of "determining the value ranges of the prediction time domain and the weight coefficients of the objective function", reasonable value ranges are determined for the prediction time domain N p and the weight coefficients Q of the objective function, N p is set to N p min , N p max , Q is set toQ min , Q max 。

[0118] In the step of "creating the initial pheromone matrix", a pheromone matrix is created , which is used to record the pheromone concentration at each position in the search space by the ants. The dimension of the pheromone matrix is determined according to the number of decision variables and the number after discretization of the value range. Initially, the elements of the pheromone matrix are uniformly set to a relatively small constant 。

[0119] In the step of "for each ant, the ant searches for appropriate prediction horizons and weight coefficients of the objective function within the prediction horizon and the value range of the weight coefficients of the objective function according to the pheromone matrix and the transition probability formula and calculates the value of the fitness function corresponding to each ant", for each ant k b , k b denoted as 1, 2, ……, r , the ant selects the values of the prediction horizon and the weight coefficients of the objective function within the given prediction horizon and the value range of the weight coefficients of the objective function according to the pheromone matrix and the transition probability formula, applies the selected values of the prediction horizon and the weight coefficients of the objective function to the offline MPC control algorithm, and combines the current lateral error and the overall vehicle risk level to perform system control simulation to calculate the value of the fitness function corresponding to each ant.

[0120] The transition probability formula of this scheme can be expressed as:

[0121] ;

[0122] In the formula: is the probability that the ant transfers from the current parameter combination to another parameter combination , is the pheromone concentration at the corresponding position of the ant , is the heuristic information of the ant , and are the parameters that control the relative importance of the pheromone and the heuristic information, and r is the total number of ants.

[0123] Within the range of the search scope limit prediction time domain \(N_p\) of the ants and the value range of the weight coefficient \(Q\) of the objective function, the ants establish a close connection according to the transition probability formula through pheromone, heuristic information, and related parameters and parameter combinations within the corresponding value range, guiding the ants to continuously explore and adjust in the search space to obtain the values of the prediction time domain and the weight coefficient of the objective function, and calculating the fitness function value of each ant under the current parameter combination based on the selected values of the prediction time domain and the weight coefficient of the objective function.

[0124] It should be noted that during the iterative process of the ant search, each ant selects different search paths under its respective parameter combinations, and updates the pheromone matrix according to the pheromone update formula for the pheromone on the search paths. Specifically, different parameter combinations will cause the ants to select different search paths, thus leaving different concentrations of pheromone on different search paths. The pheromone matrix is updated according to the pheromone update formula. The pheromone will volatilize over time, while the pheromone on the excellent solution paths will be enhanced. Through the volatilization and increment mechanisms of the pheromone, the ant colony algorithm can find a balance between exploring new parameter combinations and utilizing historical experience, and gradually converge to a better parameter combination.

[0125] The update formula during the iterative process of the ant search is as follows:

[0126] ;

[0127] ;

[0128] Where represents the pheromone concentration on the search path from state \(i\) to state \(j\) within the parameter search space, is the pheromone evaporation sparsity, \(0 \lt \lt 1\), is the pheromone concentration at the corresponding position, is the pheromone increment, The value rule of k b is: if the ant has passed through the corresponding position during this search, is , is the value of the fitness function, otherwise

[0129] is 0, and \(r\) is the number of ants. N p best ,Q best ), the optimal solution, and the weight coefficients of the best prediction time domain and the objective function obtained at this time ( N p best , Q best ) are applied to the MPC trajectory tracking controller for actual control.

[0130] In step S6, an MPC trajectory tracking controller is built based on the reference trajectory and the three-degree-of-freedom vehicle dynamics model, and the prediction time domain is input into the MPC trajectory tracking controller to obtain the output quantity within the prediction time domain. A quadratic objective function of the MPC trajectory tracking controller is constructed based on the output quantity and the state error weight coefficient, and the control increment is obtained by optimizing and solving the quadratic objective function. Combining the control quantity and the control increment in the reference trajectory, an adaptive control quantity is output to the controlled vehicle to obtain an adaptive MPC vehicle trajectory.

[0131] Specifically, the state quantity at the current moment and the control quantity at the previous moment are combined into a new state quantity, and the three-degree-of-freedom vehicle dynamics model is expressed as a new space state space model, as follows:

[0132] ;

[0133] Among them ; ; ; .

[0134] Among them; is the state transition matrix; is the input coefficient matrix; is the output coefficient matrix; is the vehicle state parameter at time k; is the output quantity at time k; is the 0 matrix at time k; is the 0 matrix; I represents the identity matrix; represents the vehicle control quantity at time (k - 1); is the new combined vehicle state quantity at time k; is the new combined vehicle state quantity at time (k + 1); is the control increment at time k.

[0135] In the step of inputting the prediction time domain into the MPC trajectory tracking controller to obtain the output quantity within the prediction time domain, first, the prediction state variables are constructed based on the prediction time domain and the control time domain as follows:

[0136] ;

[0137] Among them ; ;

[0138] The output quantity in the new state - space model of space can be obtained based on the predicted state variable as:

[0139] ;

[0140] where ; ; ; ;

[0141] ;

[0142] where ; ; ; ;

[0143] where is the output quantity in the new state - space model of space at time k, k represents time, is the state increment matrix within the control time domain, is the vehicle control increment at the Nc - th control step at time k, is the state quantity coefficient matrix, is the control increment coefficient matrix, is the vehicle state parameter at the 0 - th prediction step at time k, is the state transition matrix; is the input coefficient matrix; is the output coefficient matrix; is the vehicle state parameter at time k; is the output quantity at time k; is the 0 - matrix at time k; is the 0 - matrix; I represents the identity matrix; represents the vehicle control quantity at time (k - 1); is the combined new vehicle state quantity at time k; is the combined new vehicle state quantity at time (k + 1); is the vehicle control increment at time k; Np is the prediction time domain, is the control time domain, is the increment of the control quantity at time (k + 1), is the increment of the control quantity at time k, is the increment of the control quantity at the ( N (c + 1)) - th control step at time k, refers to ( N (p - 2)) the state transition matrix in the prediction time domain, Refers to the state transition matrix at time step k for the N p-step prediction, Refers to the state transition matrix at time step k for the N (p - 1)-step prediction, Refers to the state transition matrix at time step k for the (Np - 2)-step prediction, Refers to the state transition matrix at time step k for the N (p - N c - 2)-step prediction.

[0144] In the quadratic objective function for constructing the MPC trajectory tracking controller based on the output and state error weight coefficients, the quadratic objective function is expressed as:

[0145] ;

[0146] ;

[0147] ;

[0148] where , are weight coefficient matrices, Q is the state error weight coefficient, R is the control quantity error weight coefficient, diag is the diagonal matrix function, is the relaxation factor, is the control quantity increment at time step k, is the relaxation factor, refers to the reference abscissa at time step k for the N p-step prediction, is the reference trajectory abscissa at time step k, is the actual trajectory abscissa of the vehicle, is the relaxation factor weight.

[0149] And to ensure the feasibility of the MPC trajectory tracking controller, constraints also need to be added to the control quantity and control quantity increment. Subsequently, the control increment is obtained through optimization based on the quadratic objective function using the QP function solver in Matlab, and the optimal front wheel steering angle control increment is obtained as: ;

[0150] where is the increment of the control quantity at the (k + N c + 1) - th time step;

[0151] Then, the control increment and the control quantity in the reference trajectory are combined and output as the control quantity to the controlled vehicle, thereby controlling the steering of the controlled vehicle to obtain an adaptive trajectory.

[0152] Figure 5It is a framework diagram of the adaptive MPC vehicle trajectory tracking control method for online identification of the danger level of this solution. It can be seen that the state matrix of the controlled vehicle is sent to the online danger level identification system to obtain the overall vehicle danger level L. The overall vehicle danger level and the lateral error are input into the ant colony algorithm optimizer to calculate the prediction horizon and the state error weight coefficient of the objective function. The prediction horizon and the state error weight coefficient of the objective function are input into the MPC trajectory tracking controller for adaptive adjustment to obtain the control quantity delivered to the controlled vehicle.

[0153] In summary, the present invention collects vehicle state information and combines the k-means clustering algorithm with the analytic hierarchy process to more scientifically and reasonably divide the danger levels of different state parameters, thereby better performing adaptive control on vehicle stability. Based on the real-time lateral error and the vehicle danger level, the ant colony algorithm is introduced to dynamically adjust the prediction horizon and the state error weight coefficient of the objective function of the MPC, better taking into account the stability, accuracy, and real-time performance of vehicle trajectory tracking.

[0154] Embodiment 2

[0155] This embodiment also provides an electronic device. Refer to Figure 6 , including a memory 404 and a processor 402. A computer program is stored in the memory 404, and the processor 402 is configured to run the computer program to execute the steps in any of the above embodiments of the adaptive MPC vehicle trajectory tracking control method for online identification of the danger level.

[0156] Specifically, the above processor 402 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits implementing the embodiments of the present application.

[0157] Among them, the memory 404 may include a mass memory 404 for data or instructions. By way of example and not limitation,

[0158] The processor 402 reads and executes the computer program instructions stored in the memory 404 to implement any of the above embodiments of the adaptive MPC vehicle trajectory tracking control method for online identification of the danger level.

[0159] Optionally, the above electronic device may further include a transmission device 406 and an input / output device 408. Among them, the transmission device 406 is connected to the above processor 402, and the input / output device 408 is connected to the above processor 402.

[0160] The transmission device 406 can be used to receive or send data via a network. Specific examples of the above-mentioned network can include wired or wireless networks provided by a communication provider of an electronic device. In one example, the transmission device includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 406 can be a Radio Frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0161] The input / output device 408 is used to input or output information. In this embodiment, the input information can be a reference trajectory, etc., and the output information can be an adaptive trajectory, etc.

[0162] Optionally, in this embodiment, the above-mentioned processor 402 can be set to execute the following steps through a computer program:

[0163] Step S1: Obtain the vehicle reference trajectory of the vehicle to be controlled;

[0164] Step S2: Establish a three-degree-of-freedom vehicle dynamics model;

[0165] Step S3: Obtain the vehicle state information of the vehicle to be controlled;

[0166] Step S4: Identify the overall risk level of the current vehicle to be controlled based on the vehicle state information;

[0167] Step S5: Use the overall risk level and lateral error of the vehicle to be controlled as input variables of the ant algorithm to adaptively adjust the prediction horizon and the state error weight coefficient of the objective function;

[0168] Step S6: Establish an MPC trajectory tracking controller according to the three-degree-of-freedom vehicle dynamics model and the reference trajectory, and assign the prediction horizon and the state error weight coefficient to the MPC trajectory tracking controller. The MPC trajectory tracking controller performs rolling optimization on the vehicle state information of the vehicle to be controlled with the vehicle reference trajectory as the reference information to obtain an adaptive MPC vehicle trajectory.

[0169] It should be noted that the specific examples in this embodiment can refer to the examples described in the above-mentioned embodiments and alternative embodiments, and will not be elaborated here.

[0170] In general, various embodiments can be implemented in hardware or special-purpose circuits, software, logic, or any combination thereof. Some aspects of the present invention can be implemented in hardware, while other aspects can be implemented by firmware or software executed by a controller, a microprocessor, or other computing devices, but the present invention is not limited thereto. Although various aspects of the present invention may be shown and described as block diagrams, flowcharts, or using some other graphical representation, it should be understood that, by way of non-limiting example, the blocks, devices, systems, techniques, or methods described herein can be implemented in hardware, software, firmware, special-purpose circuits or logic, general-purpose hardware or a controller or other computing devices, or some combination thereof.

[0171] Embodiments of the present invention can be implemented by computer software that is executable by a data processor of a mobile device, such as in a processor entity, or by hardware, or by a combination of software and hardware. A computer software or program (also referred to as a program product), including software routines, applets, and / or macros, can be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. The computer program product can include one or more computer-executable components that are configured to perform the embodiments when the program runs. One or more computer-executable components can be at least one software code or a part thereof. Additionally, in this regard, it should be noted that any box in the logical flow in the figure can represent a program step, or interconnected logical circuits, boxes, and functions, or a combination of program steps and logical circuits, boxes, and functions. The software can be stored on physical media such as memory chips or storage blocks implemented within a processor, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and their data variants, CDs. The physical media are non-transitory media.

[0172] Those skilled in the art should understand that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0173] The above embodiments only represent several implementation manners of the present application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An adaptive MPC vehicle trajectory tracking control method for online identification of danger level, characterized in that, It includes the following steps: Step S1: Obtain the vehicle reference trajectory of the vehicle to be controlled; Step S2: Establish a three-degree-of-freedom vehicle dynamics model; Step S3: Obtain the vehicle state information of the vehicle to be controlled; Step S4: Identify the overall risk level of the current vehicle to be controlled based on the vehicle state information; Step S5: Use the overall risk level and lateral error of the vehicle to be controlled as input variables of the ant algorithm to adaptively adjust the prediction horizon and the state error weight coefficient of the objective function. Among them, construct a fitness function with the initial prediction horizon of the MPC trajectory tracking controller, the overall risk level of the vehicle, and the lateral error, determine the value range of the number of ants, prediction horizon, and state error weight coefficient of the objective function in the ant colony algorithm and create an initial pheromone matrix. For each ant, the ant searches for appropriate prediction horizons and state error weight coefficients of the objective function within the value range of the pheromone matrix and the transition probability formula and calculates the value of the fitness function corresponding to each ant. After multiple iterative searches, use the prediction horizon and state error weight coefficient of the objective function corresponding to the ant with the largest fitness function value as the optimal solution; Step S6: Build an MPC trajectory tracking controller based on the reference trajectory and the three-degree-of-freedom vehicle dynamics model, input the prediction horizon into the MPC trajectory tracking controller to obtain the output quantity within the prediction horizon, construct the quadratic objective function of the MPC trajectory tracking controller based on the output quantity and the state error weight coefficient, optimize and solve based on the quadratic objective function to obtain the control increment, combine the control quantity and the control increment in the reference trajectory and output the adaptive control quantity to the vehicle to be controlled to obtain the adaptive MPC vehicle trajectory.

2. The adaptive MPC vehicle trajectory tracking control method for online identification of danger level according to claim 1, wherein In step S2, establish an inertial coordinate system and establish a mechanical equilibrium equation according to the forces on the vehicle in the x-axis, y-axis, and z-axis. Obtain the longitudinal stiffness and cornering stiffness of the front and rear wheels of the vehicle based on the magic formula tire model, and substitute the longitudinal stiffness and cornering stiffness of the front and rear wheels of the vehicle into the mechanical equilibrium equation to establish a three-degree-of-freedom vehicle dynamics model.

3. The adaptive MPC vehicle trajectory tracking control method for online identification of danger level according to claim 2, wherein , Represent the three-degree-of-freedom vehicle dynamics model as a nonlinear dynamics state equation using the state space equation, and transform it into a linear state equation through Taylor expansion based on the nonlinear dynamics state equation, and then discretize it using the forward Euler method to obtain the discretized linear state equation.

4. The adaptive MPC vehicle trajectory tracking control method for online identification of danger level according to claim 1, wherein , In step S4, obtain the vehicle state information of other vehicles as a data set, construct an online risk level identification system based on the data set using the k-means clustering algorithm and the analytic hierarchy process, and input the vehicle state information obtained in real time of the current vehicle to be controlled into the online risk level identification system to obtain the overall vehicle level of the vehicle.

5. The adaptive MPC vehicle trajectory tracking control method for online identification of danger level according to claim 4, characterized in that , Determine the risk level of each vehicle state parameter in the vehicle state information using the k-means clustering algorithm based on the data set, determine the criterion layer weight vector of the influence of each vehicle state parameter on the overall risk level of the vehicle using the analytic hierarchy process, and determine the scheme layer weight vector of each vehicle state parameter under each risk level to construct an online risk level identification system.

6. The online identification method for danger level-based adaptive MPC vehicle trajectory tracking control according to claim 5, wherein , Determine the hierarchy, where the target layer in the hierarchy is the overall vehicle danger level, the criterion layer is each vehicle state parameter, and the scheme layer is the number of levels corresponding to different danger levels; , Use the scaling method to make pairwise comparisons of each vehicle state parameter within the criterion layer, construct the criterion layer judgment matrix, and form a scheme layer judgment matrix for each vehicle state parameter: perform consistency verification on the criterion layer judgment matrix and the scheme layer judgment matrix, calculate the criterion layer weight vector for the criterion layer judgment matrix that passes the consistency verification, and calculate the scheme layer weight vector for the scheme layer judgment matrix that passes the consistency verification.

7. The adaptive MPC vehicle trajectory tracking control method for online identification of danger level according to claim 5, characterized in that , Preset the number of cluster centers, randomly select a data point from the normalized dataset corresponding to each type of vehicle state parameter as the cluster center, and determine other cluster centers according to the criterion that the distances between the cluster centers are as far as possible. Based on the initially determined cluster centers, perform clustering iteration on the normalized dataset of the current type of vehicle state parameter until the cluster centers no longer change or reach the preset number of iterations, and determine the danger levels of each vehicle state parameter based on the cluster centers.

8. An electronic device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to execute the adaptive MPC vehicle trajectory tracking control method for online identification of the danger level according to any one of claims 1 to 7.

Images