Predictive learning tracking control method and device for self-driving automobile

Abstract

The invention discloses a predictive learning tracking control method and device for an autonomous vehicle. The method comprises the steps that a monorail bicycle model of the vehicle is established according to the transverse degree of freedom of the vehicle; establishing a vehicle dynamics model according to the monorail bicycle model and the vehicle parameters; an automobile augmentation model is obtained according to the automobile dynamics model, and an automobile input and output relation model is obtained according to the automobile augmentation model; constructing an automobile prediction model according to the automobile augmentation model and a preset control step length, and determining a target function according to the automobile prediction model; constructing an event-triggered data transmission mechanism according to event triggering conditions, and constructing an event-triggered controller updating mechanism according to the automobile input and output relation model; and constructing a trajectory tracking controller according to an iterative learning model predictive control algorithm, the vehicle dynamics model, a data transmission mechanism, a controller updating mechanism and a target function to realize tracking control of a vehicle trajectory.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a predictive learning tracking control method and device for an autonomous driving vehicle. Background Art

[0002] Autonomous driving cars are an important development direction in the future automotive industry. Their main advantages are improving road traffic safety, reducing road traffic congestion, and reducing vehicle fuel consumption, thereby reducing environmental pollution. With the increasing global traffic pressure, the research and development of autonomous driving cars and intelligent transportation technologies is particularly important. Trajectory tracking control, as one of the key issues in the field of autonomous driving cars, is the core foundation for achieving autonomous driving.

[0003] The existing trajectory tracking control methods for autonomous vehicles generally control the vehicle trajectory by modeling and adjusting the parameters of the vehicle dynamics, but it usually requires a time-consuming and labor-intensive parameter adjustment process to improve the controller performance. At the same time, this type of method does not have learning capabilities and has poor control accuracy; or it uses ILC control algorithms and iterative learning to learn historical driving data to improve vehicle control performance, but this method has weak anti-real-time interference capabilities, making it difficult to ensure the closed-loop stability and robustness of the vehicle control system, and has high computational complexity. Summary of the invention

[0004] The present invention provides a predictive learning tracking control method and device for an autonomous driving vehicle, so as to solve the problem that the existing trajectory tracking control of an autonomous driving vehicle is difficult to reduce the communication and computing burden while ensuring the control accuracy.

[0005] In order to solve the above technical problems, the present invention provides a predictive learning tracking control method for an autonomous driving vehicle, comprising:

[0006] Establishing a single-track bicycle model of the vehicle according to the lateral degrees of freedom of the vehicle; and establishing a vehicle dynamics model according to the single-track bicycle model and vehicle parameters;

[0007] Acquire a vehicle augmented model according to the vehicle dynamics model, and acquire a vehicle input-output relationship model according to the vehicle augmented model;

[0008] Building a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model;

[0009] Constructing an event-triggered data transmission mechanism according to the event triggering conditions, and constructing an event-triggered controller update mechanism according to the automobile input-output relationship model;

[0010] A trajectory tracking controller is constructed according to an iterative learning model predictive control algorithm, the vehicle dynamics model, a data transmission mechanism, a controller update mechanism and an objective function to achieve tracking control of the vehicle trajectory.

[0011] Furthermore, the monorail bicycle model of the vehicle is established according to the lateral degree of freedom of the vehicle; and the vehicle dynamics model is established according to the monorail bicycle model and vehicle parameters, including:

[0012] Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system;

[0013] Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error;

[0014] The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

[0015] Further, the step of obtaining a vehicle augmented model according to the vehicle dynamics model, and obtaining a vehicle input-output relationship model according to the vehicle augmented model, includes:

[0016] Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model;

[0017] The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

[0018] Furthermore, constructing a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model, includes:

[0019] Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model;

[0020] An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

[0021] Furthermore, the event-triggered data transmission mechanism is constructed according to the event triggering condition, and the event-triggered controller update mechanism is constructed according to the automobile input-output relationship model, including:

[0022] Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment;

[0023] A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering.

[0024] In a second aspect, the present invention provides a predictive learning tracking control device for an autonomous driving vehicle, comprising: a dynamic model building module, a relational model building module, a vehicle prediction model building module, an event trigger mechanism building module, and a controller module;

[0025] The dynamic model building module is used to establish a single-track bicycle model of the vehicle according to the lateral degree of freedom of the vehicle; and to establish a vehicle dynamic model according to the single-track bicycle model and vehicle parameters;

[0026] The relationship model construction model is used to obtain a vehicle augmented model according to the vehicle dynamics model, and to obtain a vehicle input-output relationship model according to the vehicle augmented model;

[0027] The automobile prediction model building module is used to build an automobile prediction model according to the automobile augmented model and a preset control step length, and determine an objective function according to the automobile prediction model;

[0028] The event trigger mechanism construction module is used to construct an event-triggered data transmission mechanism according to the event trigger condition, and to construct an event-triggered controller update mechanism according to the automobile input-output relationship model;

[0029] The controller module is used to construct a trajectory tracking controller according to the iterative learning model predictive control algorithm, the vehicle dynamics model, the data transmission mechanism, the controller update mechanism and the objective function to achieve vehicle trajectory tracking control.

[0030] Furthermore, the kinetic model construction module includes:

[0031] Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system;

[0032] Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error;

[0033] The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

[0034] Furthermore, the relationship model building module includes:

[0035] Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model;

[0036] The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

[0037] Furthermore, the automobile prediction model building module includes:

[0038] Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model;

[0039] An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

[0040] Furthermore, the event triggering mechanism construction module includes:

[0041] Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment;

[0042] A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a flow chart of a predictive learning tracking control method for an autonomous driving vehicle provided in an embodiment of the present invention;

[0044] Figure 2A schematic diagram of the structure of a predictive learning tracking control device for an autonomous driving vehicle provided by an embodiment of the present invention;

[0045] Figure 3 Another schematic flow chart of the predictive learning tracking control method for an autonomous driving vehicle provided by an embodiment of the present invention;

[0046] Figure 4 A schematic diagram of vehicle state tracking performance under different iteration times provided by an embodiment of the present invention;

[0047] Figure 5 A schematic diagram of a vehicle control input curve under different iteration times provided by an embodiment of the present invention;

[0048] Figure 6 A schematic diagram of a maximum tracking error curve along an iteration axis provided by an embodiment of the present invention;

[0049] Figure 7 A schematic diagram of an initial tracking error curve along an iteration axis provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0051] The terms "first" and "second" and the like in the specification, claims and drawings of this application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0052] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0053] First, some terms in this application are explained to facilitate understanding by those skilled in the art.

[0054] (1) Linear Quadratic Regulator (LQR) is a strategy for optimal control. Its main purpose is to find a control strategy that minimizes a certain performance indicator (usually the weighted square error of the system and the weighted square of the control variable) under a given dynamic system model.

[0055] (2) Model Predictive Control (MPC) is an advanced control strategy that generates control instructions by optimizing a performance indicator within a prediction time range. Unlike traditional control methods, MPC uses a mathematical model of the system to predict future system behavior and solves an optimization problem in each control cycle to determine the optimal control input.

[0056] (3) Iterative Learning Control (ILC) is a control method for repetitive tasks. Its main feature is to improve the control performance of the current task by learning from the experience of previous tasks. ILC is particularly suitable for systems that need to repeatedly perform the same or similar operations, such as production lines, robot trajectory tracking, etc.

[0057] (4) Iterative Learning Model Predictive Control (ILMPC) combines the advantages of iterative learning control (ILC) and model predictive control (MPC), aiming to improve the control performance of the system in repetitive tasks.

[0058] (5) Event-Triggered Control (ETC) is a control strategy used to improve system efficiency and resource utilization, especially when communication bandwidth is limited or computing resources are constrained. Unlike traditional time-triggered control strategies, ETC performs control updates only when specific events occur, rather than at fixed time intervals.

[0059] Example 1

[0060] See also Figure 1 , Figure 1 The flowchart of the predictive learning tracking control method of the autonomous driving vehicle provided in the embodiment of the present invention is shown in FIG. The predictive learning tracking control method of the autonomous driving vehicle provided in the embodiment of the present invention comprises steps 101 to 105, which are as follows:

[0061] Step 101: establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle; and establishing a vehicle dynamics model according to the monorail bicycle model and vehicle parameters;

[0062] In this embodiment, the monorail bicycle model of the vehicle is established according to the lateral degree of freedom of the vehicle; and the vehicle dynamics model is established according to the monorail bicycle model and vehicle parameters, including:

[0063] Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system;

[0064] Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error;

[0065] The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

[0066]

[0067] Table 1

[0068] Please refer to Table 1, which is a vehicle parameter variable table provided in an embodiment of the present invention.

[0069] Please refer to Figure 2 , Figure 2 A schematic diagram of a monorail bicycle model provided in an embodiment of the present invention.

[0070] In this embodiment, a single-track bicycle model of the vehicle is established based on the two degrees of freedom of lateral translation and yaw, OXY is an inertial coordinate system, and oxy is a vehicle body coordinate system.

[0071] In this embodiment, the vehicle parameters are determined according to the inertial coordinate system and the body coordinate system of the monorail bicycle model, and the vehicle parameters include the yaw angle error. It represents the yaw angle error of the center of mass relative to the reference trajectory. Specifically:

[0072]

[0073] in, is the desired heading angle of the vehicle determined based on the reference trajectory, is the vehicle yaw angle.

[0074] In this embodiment, the counterclockwise direction of the coordinate system in the monorail bicycle model is set as the positive direction, and the road curvature c is determined R, then the desired vehicle yaw rate Specifically:

[0075]

[0076] Among them, x is the vehicle longitudinal velocity.

[0077] In this embodiment, according to the desired vehicle yaw rate and vehicle yaw angle Calculate vehicle yaw error And according to the vehicle longitudinal speed υ x and the lateral velocity of the vehicle's center of mass υ x and vehicle yaw angle error Calculate the vehicle lateral error e y .

[0078] In this embodiment, the vehicle yaw angle error and the vehicle lateral error e y The derivative of is specifically:

[0079]

[0080]

[0081]

[0082]

[0083] In this embodiment, the vehicle lateral acceleration a is determined by the force balance of the vehicle model along the y-axis. y , specifically:

[0084] ma y =F yf +F yr (7)

[0085] Among them, F yf and F yr Represent the steering forces of the front and rear wheels respectively. Under the assumption that the tire slip angle is small, the tire model is linearized. The specific linear tire model is:

[0086] F yf =C αf α f (8)

[0087] F yr =C αr α r (9)

[0088] Among them, C αf and Cαr Respectively represent the cornering stiffness of the front and rear wheels of the vehicle, α f and α r are the side slip angles of the front and rear wheels of the vehicle, respectively.

[0089] In this embodiment, α f and α r The side slip angles of the front and rear wheels of the vehicle are approximately:

[0090]

[0091]

[0092] Among them, l f and l r are the distances from the front and rear wheels to the center of mass of the vehicle, δ f is the front wheel steering angle, is the inverse of the vehicle's yaw angle.

[0093] In this embodiment, the vehicle lateral error e is calculated based on the vehicle lateral acceleration, the linearized tire model, the side slip angles of the front and rear wheels of the vehicle, and the vehicle yaw angle error. y .

[0094] In this embodiment, the second derivative of the vehicle lateral error is Specifically:

[0095]

[0096] Where m is the vehicle mass, υ x is the vehicle longitudinal velocity, C αf and C αr Denote the cornering stiffness of the front and rear wheels of the vehicle, δ f is the front wheel steering angle.

[0097] In this embodiment, the second derivative of the vehicle yaw angle is calculated based on the moment balance in the yaw direction. Specifically:

[0098]

[0099] Among them, M z is the external yaw moment, F yf and F yr are the front wheel and rear wheel steering forces in the vehicle coordinate system, I z is the vehicle yaw moment of inertia.

[0100] In this embodiment, the linearized tire model, the side slip angles of the front and rear wheels of the vehicle, and the second derivative of the vehicle yaw angle are Calculate vehicle yaw error

[0101] In this embodiment, the second derivative of the vehicle yaw angle error is Specifically:

[0102]

[0103] Among them, M z is the external yaw moment, I z is the vehicle yaw moment of inertia, l f and l r are the distances from the front and rear wheels to the center of mass of the vehicle, m is the mass of the vehicle, and υ x is the vehicle longitudinal velocity, C αf and C αr Denote the cornering stiffness of the front and rear wheels of the vehicle, δ f is the front wheel steering angle.

[0104] In this embodiment, according to the vehicle yaw angle error and the vehicle lateral error e y Construct state variables, specifically:

[0105]

[0106] In this embodiment, according to the front wheel steering angle δ f and the external yaw moment M z Build control input, specifically:

[0107]

[0108] In this embodiment, according to the vehicle yaw angle error and the vehicle lateral error e y Build output parameters, specifically:

[0109]

[0110] In this embodiment, the curvature of the road tracked by the vehicle is set to 0, and an initial vehicle dynamics model is constructed. Specifically:

[0111]

[0112] Among them, A0, B0 and C are the weight coefficients corresponding to the control input, state variable and output parameter respectively.

[0113] In this embodiment, the weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters.

[0114] In this embodiment, the first coefficient a1 is calculated according to the cornering stiffness of the front wheels and the rear wheels of the vehicle, the second coefficient a2 and the third coefficient a3 are calculated according to the cornering stiffness of the front wheels and the rear wheels of the vehicle and the distances from the front wheels and the rear wheels to the center of mass of the vehicle, and the weight coefficient A0 of the control input is calculated according to the first coefficient a1, the second coefficient a2, the third coefficient a3, the longitudinal speed of the vehicle, the yaw moment of inertia of the vehicle and the mass of the vehicle. Specifically:

[0115]

[0116] Among them, the first coefficient a1=C αf +C αr , the second coefficient a2 = -(l f C αf -l rα C) r , the third coefficient

[0117] In this embodiment, the weight coefficient B0 of the state variable is calculated according to the cornering stiffness of the front and rear wheels of the vehicle, the yaw moment of inertia of the vehicle and the vehicle mass, specifically:

[0118]

[0119] In this embodiment, the weight coefficient corresponding to the output parameter is specifically:

[0120]

[0121] Step 102: obtaining a vehicle augmented model according to the vehicle dynamics model, and obtaining a vehicle input-output relationship model according to the vehicle augmented model;

[0122] In this embodiment, the step of acquiring the vehicle augmented model according to the vehicle dynamics model, and acquiring the vehicle input-output relationship model according to the vehicle augmented model includes:

[0123] Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model;

[0124] The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

[0125] In this embodiment, the vehicle dynamics model is iteratively processed according to a preset sampling period to obtain a vehicle dynamics model with iterative characteristics, specifically:

[0126]

[0127] Where k is the number of iterations.

[0128] In this embodiment, the iterative vehicle dynamics model is discretized to obtain a discretized model, specifically:

[0129]

[0130] in, T is the system sampling period. Where k∈Z + and t∈{0,1,…,N-1} represent iteration and time index respectively, Z + is a set of positive integers, and N is the length of time for each iteration.

[0131] In this embodiment, u k (t) is the control input, which has an upper bound According to the upper limit of the control input, the front wheel steering angle δ can be obtained. f and the external yaw moment M z The upper bound of , specifically:

[0132] In this embodiment, the discretized model is transformed to obtain the vehicle augmented model, specifically:

[0133] Y k =GU k +G0x k (0) (24)

[0134] Among them, the control input sequence Output parameter sequence Control input sequence coefficients Initial state coefficient x k (0) represents the initial state of the discretized model.

[0135] In this embodiment, the automobile input-output relationship model is obtained according to the automobile augmented model, specifically:

[0136]

[0137] in, And g t and In fact, they are the tth row of the control input sequence coefficient G and the initial state coefficient G0 respectively.

[0138] Step 103: constructing a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model;

[0139] In this embodiment, the step of constructing a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model includes:

[0140] Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model;

[0141] An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

[0142] In this embodiment, the preset prediction step size and control step size are both p, and the initial vehicle prediction model is constructed according to the vehicle augmented model, the prediction step size and the control step size. Specifically:

[0143]

[0144] Among them, i∈[1,p]. Represents the output y at the kth iteration time t k The estimated value of (t+i), and represents the predicted input from time t to time t+i-1,

[0145] In this embodiment, considering that the duration of each batch is fixed, the length of the prediction range cannot exceed the remaining time. Therefore, the concept of fallback range is introduced, and a control step model is constructed according to the preset initial prediction range and iteration time length. Specifically:

[0146]

[0147] where p0 is a constant initial prediction horizon.

[0148] In this embodiment, the initial vehicle prediction model is optimized according to the step length model to obtain the final vehicle prediction model, specifically:

[0149]

[0150]

[0151]

[0152]

[0153]

[0154] in, It is the augmented vehicle dynamics prediction model of p-step prediction, that is, the final vehicle prediction model.

[0155] In this embodiment, the objective function is constructed according to the vehicle prediction model and the reference trajectory. is the reference trajectory, then the objective function stipulates: for an autonomous driving vehicle with initial value offset and input saturation, determine the control input u k (t),k∈Z + , making the tracking error It converges to zero when k→∞. It also appropriately reduces the communication and computational burden of the vehicle system while ensuring the tracking accuracy of the autonomous vehicle.

[0156] Step 104: constructing an event-triggered data transmission mechanism according to the event triggering condition, and constructing an event-triggered update mechanism according to the automobile input-output relationship model;

[0157] In this embodiment, in order to improve the efficiency of communication resources between the controlled system and the controller of the autonomous driving vehicle, an event triggering mechanism is constructed to reduce the data transmission of the vehicle lateral error and yaw angle error in the time direction.

[0158] In this embodiment, the event-triggered data transmission mechanism is constructed according to the event triggering condition, and the event-triggered update mechanism is constructed according to the automobile input-output relationship model, including:

[0159] Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment;

[0160] A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering.

[0161] In this embodiment, t is defined as l is the event triggering time, is the system output after the event trigger mechanism is introduced. For the event trigger mechanism, it is only necessary to collect and update the output information at the triggering moment. Then, when t = t l hour, If t≠t l , the system output remains the same as the previous moment, that is,

[0162] In this embodiment, according to the actual system output y k (t) and event trigger output The difference between them builds the event-triggered output error model, specifically:

[0163]

[0164] In this embodiment, when the event is triggered, will be reset to zero.

[0165] In this embodiment, a first event trigger condition model is constructed according to the event trigger output error, specifically:

[0166]

[0167] Among them, θ>0 is a constant.

[0168] In this embodiment, the output data is obtained according to the first event trigger condition model. At the triggering time t l Sent to the controller, and build a data transmission mechanism based on the event-triggered output error model and the first event trigger condition model. Specifically:

[0169]

[0170] It is assumed that the initial triggering time is t0=0.

[0171] In this embodiment, in order to save computing resources of the controlled system of the autonomous driving vehicle, an event-triggered update mechanism is constructed along the iteration axis, thereby reducing the update frequency of the vehicle's front wheel steering angle and external yaw moment.

[0172] In this embodiment, k is defined as s ,s=0,1,2,…as the iterative sequence triggered by the event, the controller inputs u k (t) will only be s Update every iteration. When k∈(k s-1 ,k s ), the actuator is not triggered, which means that u k (t) will keep the control input of the previous iteration unchanged, that is, δu k (t) = 0, δ represents the difference operator along the iteration direction, that is,

[0173] In this embodiment, the output relationship between two adjacent iterations is obtained according to the input-output relationship model, specifically:

[0174]

[0175] in,

[0176] In this embodiment, the relationship between the tracking error and the output parameter is obtained according to the output relationship between two adjacent iterations: δek (t+1)=-δy k (t+1), and when k∈(k s-1 ,k s ), the control input δu k (t)=0, then the output parameter Therefore, according to the relationship between the tracking error and the output parameter, the relationship between the tracking error and the control input sequence can be obtained, which is:

[0177]

[0178] In this embodiment, the second event trigger condition model is constructed based on the relationship between the Lyapunov function and the tracking error and the control input sequence. Specifically: define the Lyapunov function as V k (t+1)=||e k (t+1)||, when k∈(k s-1 ,k s ), the relationship between the tracking error and the control input sequence is:

[0179]

[0180] When δV k (t+1)<0, the second event trigger condition model along the iteration axis is specifically:

[0181]

[0182] Where ε>0 is a positive scalar. If the tracking error and the control input sequence meet the second event trigger condition model, the vehicle will be updated along the iteration axis if the second event trigger condition model is followed by the vehicle.

[0183] In this embodiment, the convergence speed is reduced by constructing a tracking error threshold condition to avoid wasting resources, wherein the tracking error threshold condition || e k-1 (t+1)||≥ε. Construct an event-triggered iteration sequence based on the tracking error, control input sequence, and tracking error threshold conditions. Specifically:

[0184]

[0185] Step 105: construct a trajectory tracking controller according to the iterative learning model predictive control algorithm, the vehicle dynamics model, the data transmission mechanism, the update mechanism and the objective function to achieve tracking control of the vehicle trajectory.

[0186] In this embodiment, when k=k s When , the control input is:

[0187]

[0191] in, is an unconstrained control input.

[0192] In this embodiment, the increment It consists of two parts: ILC part and MPC section ILC Section Generated by D-type Iterative Learning Control (D-type ILC), MPC part It is obtained by constructing the first optimization function.

[0193] In this embodiment, the ILC part Specifically:

[0194]

[0195] When k∈(k s-1 ,k s ), the actuator is not triggered, that is, the control input is Therefore, according to the initial offset and input saturation of the autonomous vehicle, an initial trajectory tracking controller equipped with an initial state learning scheme is constructed, specifically:

[0196] u k (t) = sat(v k (t),u * ) (46)

[0197]

[0198] x k (0) = x k-1 (0)+Le k-1 (0) (48)

[0199] Where L and K are the preset control gains.

[0200] In this embodiment, the predicted tracking error vector of the kth iteration from time t+1 to t+p is calculated according to the tracking error, and a first optimization function is constructed according to the predicted tracking error vector and the control input sequence, and the MPC part is determined according to the first optimization function. The first optimization function is specifically:

[0201]

[0202] Among them, Q and W are positive definite weight matrices, represents the predicted tracking error vector from time t+1 to t+p at the kth iteration, which is specifically:

[0203]

[0204] in,

[0205] In this embodiment, the equation is solved according to the tracking error vector and the first optimization function The optimal solution can be obtained as

[0206]

[0207] in, Only the optimal control (optimal solution) of the first row will be used. Therefore, the MPC part It can be expressed as follows

[0208]

[0209] in,

[0210] In this embodiment, a trajectory tracking controller is constructed according to the HTIDET ILMPC scheme, wherein the trajectory tracker includes a vehicle dynamics model, a data transmission mechanism, an update mechanism and an objective function, and increments and initial states of LLC and MPC.

[0211] In this embodiment, the main results about the monotonic convergence of the tracking error can be obtained according to the trajectory tracking controller:

[0212] For the vehicle dynamics model, if is non-singular, ||I2-CL||<1 and the first inequality holds, then the trajectory tracking controller based on the HTIDET ILMPC scheme can ensure that the tracking error e k (t) converges to zero when k→∞, and the first inequality is:

[0213]

[0214] Here, these symbols are matrices related to system parameters A, B, C and controller parameters K, Q, W, θ. The first inequality represents a condition that the controller parameters K, Q, W, θ need to satisfy when the system converges.

[0215] Therefore, for the autonomous vehicle system, the event triggering time in the trajectory tracker, the iterative sequence based on the event triggering, the initial trajectory tracking controller, the MPC It can ensure that the vehicle's tracking error e k (t) converges to zero when k→∞.

[0216] When k = k s When , the difference in tracking errors between adjacent iterations is:

[0217]

[0218] The error improvement vector is:

[0219]

[0220]

[0221]

[0222] Then, the input signal generated by the MPC part can be expressed as:

[0223]

[0224] in, Since the matrix G is nonsingular, the following inequality holds:

[0225]

[0226] The reference trajectory is expressed as Similarly, for the initial state at the kth iteration, there is the following relationship:

[0227]

[0228] Next, substituting equations (59) and (60) into the input signal generated by the MPC part, that is, equation (58), we can obtain:

[0229]

[0230] when When , the control input of the MPC part can be rewritten as:

[0231]

[0232] in,

[0233] By substituting (62) into the tracking error equation (58), we can obtain the following equation:

[0234]

[0235] in,

[0236] Defining the Matrix

[0237] and Then E k It can be expressed as:

[0238]

[0239] in Taking the Euclidean norm on both sides of equation (64), we have:

[0240]

[0241] and and δU k The following inequality is satisfied:

[0242]

[0243]

[0244] Among them, the matrix By substituting inequalities (66) and (67) into (64), we can obtain:

[0245]

[0246] in:

[0247]

[0248] but:

[0249]

[0250] In addition, according to the initial value theorem, that is, formula (48) x k (0) = x k-1 (0)+Le k-1 (0), we can get

[0251] e k+1 (0)-e k (0)

[0252] =r(0)-y k+1 (0)-[r(0)-y k (0)]

[0253] =-CLe k (0) (71)

[0254] It can be expressed as:

[0255] e k+1 (0) = (I2-CL)e k(0) (72)

[0256] Taking the norms of both sides of equation (72), we have:

[0257] ||e k+1 (0)||=||I2-CL||||e k (0)|| (73)

[0258] According to the condition ||I2-CL||<1, we have Therefore, using Stolz's theorem, we can get therefore, That is, the tracking error converges to zero along the triggered iterative sequence.

[0259] For the event-triggered iteration sequence interval, that is, k∈(k s-1 ,k s ), since the current control input remains unchanged from the input of the previous iteration, according to the second event trigger condition model, that is, formula (39), the tracking error satisfies ||e k (t+1)||<||e k-1 (t+1)||.

[0260] Therefore, when k→∞, the tracking error e k Converges to zero.

[0261] Please refer to Figure 2 , Figure 2 A schematic diagram of the structure of a predictive learning tracking control device for an autonomous driving vehicle provided by an embodiment of the present invention, comprising: a dynamic model building module 201, a relationship model building module 202, a vehicle prediction model building module 203, an event trigger mechanism building module 204, and a controller module 205;

[0262] The dynamic model building module 201 is used to build a single-track bicycle model of the vehicle according to the lateral degree of freedom of the vehicle; and to build a vehicle dynamic model according to the single-track bicycle model and vehicle parameters;

[0263] The relationship model construction model 202 is used to obtain a vehicle augmented model according to the vehicle dynamics model, and to obtain a vehicle input-output relationship model according to the vehicle augmented model;

[0264] The vehicle prediction model building module 203 is used to build a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determine an objective function according to the vehicle prediction model;

[0265] The event trigger mechanism construction module 204 is used to construct an event-triggered data transmission mechanism according to the event trigger condition, and to construct an event-triggered update mechanism according to the automobile input-output relationship model;

[0266] The controller module 205 is used to construct a trajectory tracking controller according to the iterative learning model predictive control algorithm, the vehicle dynamics model, the data transmission mechanism, the update mechanism and the objective function to achieve vehicle trajectory tracking control.

[0267] In this embodiment, the kinetic model building module includes:

[0268] Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system;

[0269] Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error;

[0270] The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

[0271] In this embodiment, the relationship model building module includes:

[0272] Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model;

[0273] The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

[0274] In this embodiment, the automobile prediction model building module includes:

[0275] Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model;

[0276] An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

[0277] In this embodiment, the event trigger mechanism construction module includes:

[0278] Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment;

[0279] A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering.

[0280] Please refer to Figure 3 , Figure 3 Another flowchart of the predictive learning tracking control method for an autonomous driving vehicle provided in an embodiment of the present invention.

[0281] In this embodiment, a dynamic model of an autonomous driving vehicle is established to predict the Moses of the autonomous driving vehicle, so as to design an event triggering mechanism of mixed time and iteration direction. The event triggering mechanism includes a data transmission mechanism based on event triggering and a controller update mechanism based on event triggering. An event-triggered iterative learning model predictive control algorithm of mixed time and iteration direction is designed according to the event triggering mechanism of mixed time and iteration direction, and the control method is applied to the autonomous driving vehicle to realize the tracking control of the vehicle trajectory.

[0282] In this embodiment, please refer to Table 2, which is a vehicle setting parameter provided by an embodiment of the present invention.

[0283]

[0284] Table 2

[0285] Vehicle longitudinal speed x =15m / s, the initial state is Control input upper bound The system sampling period is T = 0.05s, and the duration of each batch is 2.5s.

[0286] Set the controller parameters according to the vehicle parameters. The specific controller parameters can be set as:

[0287] The initial prediction range p0=3, the positive definite weight matrix Q=10I, the positive definite weight matrix W=0.001I, the difference operator θ=1, and the tracking error threshold condition ε=0.5.

[0288] Please refer to Figure 4 , Figure 4 A schematic diagram of the tracking performance of a vehicle state under different iteration numbers provided by an embodiment of the present invention, wherein 4(a) is the lateral error and 4(b) is the yaw angle error.

[0289] In this embodiment, the higher the number of iterations, the smaller the lateral error and the yaw angle error.

[0290] Please refer to Figure 5 , Figure 5 A schematic diagram of a vehicle control input curve under different iteration numbers provided in an embodiment of the present invention, wherein 5(a) is the front wheel steering angle, and 5(b) is the external yaw moment.

[0291] In this embodiment, when the initial state of the autonomous driving vehicle is offset, the reference trajectory can still be accurately tracked, and the input saturation problem of the vehicle system can be effectively solved.

[0292] Please refer to Figure 6 and Figure 7 , Figure 6 A schematic diagram of a maximum tracking error curve along an iteration axis provided by an embodiment of the present invention, Figure 7 A schematic diagram of an initial tracking error curve along an iteration axis provided by an embodiment of the present invention.

[0293] In this embodiment, the initial tracking error includes the initial vehicle lateral error and the initial vehicle yaw angle error. The repeated operation characteristics of the vehicle system are used for learning and prediction capabilities to improve the vehicle's trajectory tracking control accuracy. When the number of iterations is larger, the initial tracking error trend and the maximum tracking error trend are smaller.

[0294] In this embodiment, a dynamic model of an autonomous driving vehicle is established to predict the Moses of the autonomous driving vehicle, so as to design an event triggering mechanism of mixed time and iteration direction. The event triggering mechanism includes a data transmission mechanism based on event triggering and a controller update mechanism based on event triggering. An event-triggered iterative learning model predictive control algorithm of mixed time and iteration direction is designed according to the event triggering mechanism of mixed time and iteration direction, and the control method is applied to the autonomous driving vehicle to realize the tracking control of the vehicle trajectory.

[0295] In this embodiment, by constructing a vehicle dynamics model and designing a data update based on event triggering and a data transmission mechanism based on event triggering, an event-triggered iterative learning model is designed to predict trajectory tracking control, making full use of the repeated operation characteristics of the vehicle system for learning and prediction capabilities, and improving the vehicle's trajectory tracking control accuracy; by introducing an event triggering strategy along the time axis and iteration axis, the data transmission of the vehicle's controlled system and the actuator update frequency can be greatly reduced; in the case of an initial state deviation of the autonomous driving vehicle, the reference trajectory can still be accurately tracked. And it can effectively solve the input saturation problem of the vehicle system.

[0296] In an embodiment of the present invention, a multi-device access platform processing device is also provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the above-mentioned multi-device access platform processing method is implemented.

[0297] In an embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned multi-device access platform processing method.

[0298] Exemplarily, the computer program may be divided into one or more modules, one or more modules are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the multi-device access platform processing device.

[0299] The multi-device access platform processing device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The multi-device access platform processing device may include, but is not limited to, a processor, a memory, and a display. Those skilled in the art will appreciate that the above components are merely examples of the multi-device access platform processing device and do not constitute a limitation on the multi-device access platform processing device. It may include more or fewer components than the components, or a combination of certain components, or different components. For example, the multi-device access platform processing device may also include input and output devices, network access devices, buses, and the like.

[0300] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the multi-device access platform processing device, and uses various interfaces and lines to connect various parts of the entire multi-device access platform processing device.

[0301] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the multi-device access platform processing device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, a text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0302] Among them, if the module based on the processing of the multi-device access platform is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0303] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A predictive learning tracking control method for an autonomous driving vehicle, characterized in that: include: A single-track bicycle model of the vehicle is established based on the vehicle's lateral degrees of freedom; and establishing a vehicle dynamics model based on the single-track bicycle model and vehicle parameters; Acquire a vehicle augmented model according to the vehicle dynamics model, and acquire a vehicle input-output relationship model according to the vehicle augmented model; Building a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model; Constructing an event-triggered data transmission mechanism according to the event triggering conditions, and constructing an event-triggered controller update mechanism according to the automobile input-output relationship model; A trajectory tracking controller is constructed according to an iterative learning model predictive control algorithm, the vehicle dynamics model, a data transmission mechanism, a controller update mechanism and an objective function to achieve tracking control of the vehicle trajectory.

2. The predictive learning tracking control method for an autonomous driving vehicle according to claim 1, wherein: The monorail bicycle model of the vehicle is established according to the lateral degree of freedom of the vehicle; A vehicle dynamics model is established according to the single-track bicycle model and vehicle parameters, including: Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system; Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error; The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

3. The predictive learning tracking control method for an autonomous driving vehicle according to claim 2, wherein: The step of obtaining an augmented model of the vehicle according to the vehicle dynamics model, and obtaining an input-output relationship model of the vehicle according to the augmented model of the vehicle, comprises: Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model; The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

4. The predictive learning tracking control method for an autonomous driving vehicle according to claim 3, wherein: The step of constructing a vehicle prediction model according to the vehicle augmented model and a preset control step length, and determining an objective function according to the vehicle prediction model includes: Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model; An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

5. The predictive learning tracking control method for an automatic driving vehicle according to claim 4, characterized in that: The method of constructing an event-triggered data transmission mechanism according to the event triggering condition and constructing an event-triggered controller update mechanism according to the automobile input-output relationship model includes: Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment; A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering.

6. A predictive learning tracking control device for an autonomous driving vehicle, characterized in that: include: Dynamic model building module, relational model building module, automobile prediction model building module, event trigger mechanism building module, controller module; The dynamic model building module is used to establish a single-track bicycle model of the vehicle according to the lateral degree of freedom of the vehicle; and to establish a vehicle dynamic model according to the single-track bicycle model and vehicle parameters; The relationship model construction model is used to obtain a vehicle augmented model according to the vehicle dynamics model, and to obtain a vehicle input-output relationship model according to the vehicle augmented model; The automobile prediction model building module is used to build an automobile prediction model according to the automobile augmented model and a preset control step length, and determine an objective function according to the automobile prediction model; The event trigger mechanism construction module is used to construct an event-triggered data transmission mechanism according to the event trigger condition, and to construct an event-triggered controller update mechanism according to the automobile input-output relationship model; The controller module is used to construct a trajectory tracking controller according to the iterative learning model predictive control algorithm, the vehicle dynamics model, the data transmission mechanism, the controller update mechanism and the objective function to achieve vehicle trajectory tracking control.

7. The predictive learning tracking control device for an automatic driving vehicle according to claim 6, characterized in that: The kinetic model building module comprises: Establishing a monorail bicycle model of the vehicle according to the lateral degree of freedom of the vehicle, the monorail bicycle model includes an inertial coordinate system and a vehicle body coordinate system, and determining vehicle parameters according to the inertial coordinate system and the vehicle body coordinate system; Determine state variables, control inputs and output parameters according to the vehicle parameters, the state variables include vehicle yaw angle error and vehicle lateral error, the control inputs include front wheel steering angle and external yaw moment, and the output parameters include vehicle yaw angle error and vehicle lateral error; The weight coefficients corresponding to the state variables, control input and output parameters are calculated respectively according to the vehicle parameters, and a vehicle dynamics model is constructed according to the state variables, control input and output parameters and their corresponding weight coefficients.

8. The predictive learning tracking control device for an automatic driving vehicle according to claim 7, characterized in that: The relationship model building module includes: Discretize the vehicle dynamics model according to a preset system sampling period to obtain a discretized model; The discretized model is transformed to obtain a vehicle augmented model, and a vehicle input-output relationship model is obtained according to the vehicle augmented model.

9. The predictive learning tracking control device for an automatic driving vehicle according to claim 8, characterized in that: The automobile prediction model building module includes: Constructing a control step length model according to a preset initial prediction range and an iteration time length, and constructing a vehicle prediction model according to the vehicle augmented model and the step length model; An objective function is constructed based on the vehicle prediction model and the reference trajectory, wherein the objective function includes control input and tracking error.

10. The predictive learning tracking control device for an automatic driving vehicle according to claim 9, characterized in that: The event trigger mechanism building module includes: Constructing an event-triggered output error model and a first event-triggered condition model, and constructing a data transmission mechanism according to the event-triggered output error model and the first event-triggered condition model, wherein the data transmission mechanism includes data output at the current moment and the event-triggered moment; A second event trigger condition model is constructed according to the input-output relationship model and the Lyapunov function, and an event-triggered controller update mechanism is constructed according to the tracking error condition and the second event trigger condition model. The event-triggered update mechanism includes an iterative sequence based on event triggering.

Images