A vehicle longitudinal control method and device

Abstract

The application provides a vehicle longitudinal control method and device, which can be applied to the technical field of intelligent traffic and automatic driving. The method comprises the following steps: constructing a discrete augmented state space model representing a vehicle longitudinal power system; obtaining a vehicle state prediction matrix and a vehicle control response matrix by performing offline calculation and online reuse of an augmented matrix on the discrete augmented state space model; performing analysis on a target optimization function by using the vehicle state prediction matrix and the vehicle control response matrix, so as to obtain a vehicle power Hessian matrix and a vehicle speed gradient vector; performing matrix decomposition on the vehicle power Hessian matrix after regularization by using the vehicle speed gradient vector, so as to obtain a vehicle longitudinal control adjustment; performing iterative updating and projection constraint on an initial longitudinal control amount of a current control period by using the vehicle longitudinal control adjustment, so as to obtain a longitudinal control increment; and determining a target longitudinal control amount of the current control period by using the longitudinal control increment and a target longitudinal control amount of a previous control period.

Description

Technical Field

[0001] This application relates to the fields of intelligent transportation and autonomous driving, and specifically to a method and apparatus for longitudinal control of a vehicle. Background Technology

[0002] In the field of vehicle longitudinal control, Model Predictive Control (MPC) is widely used to achieve closed-loop regulation of vehicle speed and acceleration. Its core relies on a Quadratic Programming (QP) solver to calculate the optimal control quantity. The main problems with this technology are as follows: general-purpose QP solvers depend on third-party libraries, have large code sizes, and high computational redundancy, making it difficult to meet the real-time operation requirements of 50Hz~100Hz on in-vehicle embedded platforms; traditional gradient descent methods have slow convergence speeds and require many iterations, resulting in significant control delays under varying load, gradient, and electric braking capabilities; regularized Newton-type QP solvers and code-generating solvers are generally designed for general QP problems, focusing on double-precision floating-point environments and optimality / convergence, without considering the fixed model structure of vehicle longitudinal control, the automotive-grade single-precision floating-point computing environment, and the functional safety requirements of "bounded single-frame computation time, non-divergent control, and control quantity not exceeding limits," making it difficult to directly meet the real-time and functional safety constraints of in-vehicle longitudinal control. Summary of the Invention

[0003] In view of the above problems, embodiments of this application provide a vehicle longitudinal control method and apparatus.

[0004] According to a first aspect of this application, a vehicle longitudinal control method is provided, comprising: constructing a discrete augmented state-space model representing the vehicle's longitudinal dynamic system based on a vehicle control mode; obtaining a vehicle state prediction matrix and a vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix on the discrete augmented state-space model; analyzing the objective optimization function used for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain a vehicle dynamic Hessian matrix and a vehicle speed gradient vector; regularizing the vehicle dynamic Hessian matrix to obtain a regularized vehicle dynamic Hessian matrix, and performing matrix decomposition on the regularized vehicle dynamic Hessian matrix using the vehicle speed gradient vector to obtain a vehicle longitudinal control adjustment amount; iteratively updating the initial longitudinal control amount of the current control cycle using the vehicle longitudinal control adjustment amount, and applying projection constraints to the iterative processing results of the initial longitudinal control amount during the iterative update process, obtaining the longitudinal control increment of the current control cycle under the condition of satisfying a preset convergence condition; and determining the target longitudinal control amount of the current control cycle using the longitudinal control increment of the current control cycle and the target longitudinal control amount of the previous control cycle.

[0005] According to an embodiment of this application, the above-mentioned construction of a discrete augmented state space model representing the longitudinal dynamic system of a vehicle based on a vehicle control mode includes: approximating the longitudinal dynamic response of the vehicle's longitudinal dynamic system as a first-order inertial model based on the vehicle control mode, wherein the vehicle control mode includes a vehicle speed control mode or a vehicle acceleration control mode; constructing a discrete augmented state of the vehicle using vehicle speed error, vehicle acceleration error, and vehicle longitudinal control increment error; and performing matrix augmentation expansion on the vehicle discrete state matrix, vehicle discrete state input matrix, and vehicle predicted state output matrix in the first-order inertial model using the vehicle's discrete augmented state to obtain a discrete augmented state space model.

[0006] According to an embodiment of this application, the above-mentioned method of obtaining the vehicle state prediction matrix and vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix in the discrete augmented state space model includes: performing offline power pre-calculation of the augmented matrix in the discrete augmented state space model to obtain the augmented matrix power calculation results corresponding to different control cycles; and performing online reuse of the augmented matrix power calculation results to obtain the vehicle state prediction matrix and vehicle control response matrix.

[0007] According to an embodiment of this application, the above-mentioned analysis of the objective optimization function for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamics Hessian matrix and vehicle speed gradient vector includes: constructing the objective optimization function for vehicle longitudinal control prediction using the vehicle tracking error weight, the vehicle longitudinal control increment weight, the Hessian matrix, the gradient vector, and the vehicle longitudinal control increment sequence corresponding to the control cycle; and analyzing the Hessian matrix and gradient vector in the objective optimization function using the vehicle state prediction matrix, the vehicle control response matrix, the vehicle tracking error weight, and the vehicle longitudinal control increment weight to obtain the vehicle dynamics Hessian matrix and vehicle speed gradient vector.

[0008] According to an embodiment of this application, the above-mentioned regularization of the vehicle dynamic Hessian matrix to obtain a regularized vehicle dynamic Hessian matrix includes: adding a Hernoff regularization term to the vehicle dynamic Hessian matrix using a preset regularization coefficient to obtain a regularized vehicle dynamic Hessian matrix.

[0009] According to an embodiment of this application, the above-mentioned matrix decomposition of the regularized vehicle dynamic Hessian matrix using the vehicle speed gradient vector to obtain the vehicle longitudinal control adjustment amount includes: decomposing the regularized vehicle dynamic Hessian matrix into a product of a unit lower triangular matrix and a diagonal matrix, and solving for the vehicle longitudinal control adjustment amount representing the Newton direction using the vehicle speed gradient vector during the decomposition of the regularized vehicle dynamic Hessian matrix to obtain the vehicle longitudinal control adjustment amount.

[0010] According to an embodiment of this application, the above-mentioned method of iteratively updating the initial longitudinal control quantity of the current control cycle using the vehicle longitudinal control adjustment amount, and subjecting the iterative update result of the initial longitudinal control quantity to projection constraints during the iterative update process, and obtaining the longitudinal control increment of the current control cycle under the condition of satisfying the preset convergence condition, includes: based on the hot start initialization strategy, taking the target longitudinal control quantity of the previous control cycle as the initial longitudinal control quantity of the current control cycle; during the iterative update process of the initial longitudinal control quantity, controlling the iteration direction of the initial longitudinal control quantity using the vehicle longitudinal control adjustment amount, subjecting the longitudinal control increment obtained in each iteration to projection constraints, and performing convergence judgment on the iterative update process of the initial longitudinal control quantity, and obtaining the longitudinal control increment of the current control cycle under the condition of satisfying the preset convergence condition.

[0011] According to an embodiment of this application, the above-mentioned control of the iterative direction of the initial longitudinal control quantity using the vehicle longitudinal control adjustment amount includes: when the vehicle longitudinal control adjustment amount is characterized as a downward direction, determining the initial longitudinal control quantity of the subsequent control cycle using the vehicle longitudinal control adjustment amount and the initial longitudinal control quantity of the current control cycle; when the vehicle longitudinal control adjustment amount is characterized as a non-downward direction, reverting the vehicle longitudinal control adjustment amount to the fastest downward direction, and then iteratively updating the initial longitudinal control quantity again.

[0012] According to an embodiment of this application, the above-mentioned projection constraint on the longitudinal control increment obtained in each iteration includes: comparing a preset minimum longitudinal control increment with the longitudinal control increment to obtain a first comparison result; comparing the first comparison result with a preset maximum longitudinal control increment to obtain a second comparison result; and using the second comparison result to apply upper and lower limit projection constraints on the longitudinal control increment to obtain the longitudinal control increment after projection constraint.

[0013] According to a second aspect of this application, a vehicle longitudinal control device is provided, comprising: an augmented model construction module for constructing a discrete augmented state-space model representing the vehicle's longitudinal dynamic system based on a vehicle control mode; an augmented matrix processing module for obtaining a vehicle state prediction matrix and a vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix on the discrete augmented state-space model; an objective optimization function analysis module for analyzing an objective optimization function used for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain a vehicle dynamic Hessian matrix and a vehicle speed gradient vector; and a longitudinal control adjustment acquisition module for processing the vehicle dynamic Hessian matrix. The system performs regularization to obtain the regularized vehicle dynamics Hessian matrix, and then uses the vehicle speed gradient vector to perform matrix decomposition on the regularized vehicle dynamics Hessian matrix to obtain the vehicle longitudinal control adjustment. The longitudinal control increment acquisition module is used to iteratively update the initial longitudinal control quantity of the current control cycle using the vehicle longitudinal control adjustment, and during the iterative update process, the iterative processing result of the initial longitudinal control quantity is subjected to projection constraints. Under the condition of satisfying the preset convergence condition, the longitudinal control increment of the current control cycle is obtained. The longitudinal control quantity acquisition module is used to determine the target longitudinal control quantity of the current control cycle using the longitudinal control increment of the current control cycle and the target longitudinal control quantity of the previous control cycle.

[0014] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0015] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0016] According to a fifth aspect of this application, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0017] The vehicle longitudinal control method provided in this application is a control method that does not rely on third-party libraries and is designed for automotive-grade embedded platforms. It employs techniques such as constructing a discrete augmented state-space model of the vehicle's longitudinal dynamics, offline pre-computing the power of the augmented matrix and reusing it online to obtain state prediction and control response matrices, analytically optimizing functions to obtain the dynamic Hessian matrix and vehicle speed gradient vector, regularizing the Hessian matrix and then decomposing it to solve for the control adjustment, and simultaneously applying projection constraints and combining convergence conditions to output the control increment during iterative updates of the control quantity. Therefore, it at least partially overcomes the technical problems of general-purpose QP solvers, such as reliance on third-party libraries, large code size, high computational redundancy, slow gradient descent convergence and numerous iterations, and the difficulty of meeting the high-frequency real-time operation requirements of 50Hz~100Hz and the significant control delay under complex slope and heavy load conditions on automotive embedded platforms. This achieves the technical effects of significantly reducing online real-time computation, not relying on third-party libraries, reducing computation time, adapting to automotive-grade single-precision floating-point environments, conforming to the fixed model structure of vehicle longitudinal control, and having controllable and bounded single-frame computation time. It meets the high-frequency real-time, stable, and robust operation requirements of automotive embedded platforms, balancing real-time performance and overall vehicle functional safety. Attached Figure Description

[0018] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0019] Figure 1 This diagram schematically illustrates an application scenario of vehicle longitudinal control according to an embodiment of this application.

[0020] Figure 2 A flowchart illustrating a vehicle longitudinal control method according to an embodiment of this application is shown schematically.

[0021] Figure 3 A schematic diagram illustrating the structure of a vehicle longitudinal control device according to an embodiment of this application is shown; and

[0022] Figure 4 A block diagram schematically illustrates an electronic device suitable for implementing a vehicle longitudinal control method according to an embodiment of this application. Detailed Implementation

[0023] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0024] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0025] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0026] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0027] This application provides a vehicle longitudinal control method and apparatus to overcome the problems of low operating efficiency, high resource consumption, and dependence on third-party libraries of general-purpose QP solvers on vehicle-mounted embedded platforms; it solves the defects of traditional solvers that are prone to matrix singularity, numerical instability, and control divergence in single-precision floating-point environments; it improves the convergence speed of QP problem solving, reduces the number of iterations, and meets the real-time requirements of vehicle longitudinal control above 50Hz; and it provides a lightweight, stable, and robust QP solution for vehicle longitudinal MPC that is independent of third parties.

[0028] Figure 1 The diagram illustrates an application scenario of vehicle longitudinal control according to an embodiment of this application.

[0029] like Figure 1 As shown, application scenario 100 according to an embodiment of this application may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables. For example, a user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send information, etc.

[0030] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be electronic devices such as smartphones, wearable devices, personal computers, intelligent voice interaction devices, smart home appliances, intelligent vehicles, in-vehicle terminals, aircraft, unmanned vending terminals, and extended reality devices. Extended reality devices can include virtual reality devices, augmented reality devices, and mixed reality devices. A client application for the target application can be installed and run on the terminal devices. This target application can include, but is not limited to, financial transaction applications, payment applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, and social media platform software (these are just examples). Furthermore, this application embodiment does not limit the form of the target application, and it can include, but is not limited to, applications, mini-programs, etc., installed on the terminal devices, and can also be in the form of web pages.

[0031] Server 105 can be a server providing various services, such as a backend management server supporting websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services such as cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and basic cloud computing services such as big data. The server can be the backend server of the aforementioned target application, used to provide backend services to the clients of the target application.

[0032] It should be noted that the vehicle longitudinal control method provided in this application embodiment can generally be executed by server 105 and / or terminal devices 101-103. Accordingly, the vehicle longitudinal control device provided in this application embodiment can generally be set in server 105 and / or terminal devices 101-103.

[0033] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0034] Figure 2 A flowchart illustrating a vehicle longitudinal control method according to an embodiment of this application is shown schematically.

[0035] like Figure 2 As shown, the vehicle longitudinal control method 200 according to an embodiment of this application may include operations S210 to S260.

[0036] In operation S210, a discrete augmented state-space model representing the vehicle's longitudinal dynamic system is constructed based on the vehicle control mode.

[0037] In operation S220, the vehicle state prediction matrix and vehicle control response matrix are obtained by offline power calculation and online reuse of the augmented matrix of the discrete augmented state space model.

[0038] The longitudinal power system of a vehicle is the power transmission and control system for the acceleration, deceleration, and uniform motion of a vehicle in the forward or backward direction.

[0039] The discrete augmented state-space model representing the longitudinal dynamic system of a vehicle provided in this application extends the traditional discrete state-space model by using the matrix representing the vehicle's operating state (or driving state), the input information matrix (e.g., the parameters for controlling the vehicle) of the current control cycle (e.g., with an image frame as one control cycle), and the output matrix as discretized augmented states, thus obtaining a discrete augmented state-space model representing the longitudinal dynamic system of a vehicle.

[0040] In vehicle longitudinal control, the vehicle state prediction matrix uses the vehicle's current actual state (such as current speed and acceleration) and the vehicle's motion patterns to predict the natural evolution of the vehicle's speed and acceleration over a future period of time without any additional intervention.

[0041] The vehicle control response matrix is ​​essentially an "effect amplifier" for control commands. It describes how, over a future period of time, if the vehicle's longitudinal powertrain applies a series of specific adjustment commands to the accelerator or brake, these commands will gradually change the vehicle's future speed and acceleration. It establishes a direct link between "control actions" and "future state changes."

[0042] In operation S230, the objective optimization function used for vehicle longitudinal control prediction is analyzed using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamic Hessian matrix and vehicle speed gradient vector.

[0043] The prediction of vehicle longitudinal control is transformed into a quadratic programming problem, and an objective optimization function for quadratic programming is constructed.

[0044] In longitudinal control, the vehicle's longitudinal dynamics system needs to find the smoothest and most energy-efficient acceleration and deceleration scheme. The vehicle dynamics Hessian matrix describes the curvature (i.e., the second-order rate of change) of the "objective function" that evaluates the effectiveness of this scheme in various directions. It tells the algorithm whether adjusting the throttle or brake will have an increasingly significant or gradual effect, thus helping the algorithm find the optimal control strategy more accurately and quickly.

[0045] In longitudinal control, the vehicle speed gradient vector reflects how far the current control scheme is from the "optimal scheme," and what the throttle or brake direction and force should be adjusted first in order to achieve the optimal scheme.

[0046] In operation S240, the vehicle dynamic Hessian matrix is ​​regularized to obtain the regularized vehicle dynamic Hessian matrix. Then, the vehicle speed gradient vector is used to perform matrix decomposition on the regularized vehicle dynamic Hessian matrix to obtain the vehicle longitudinal control adjustment amount.

[0047] The vehicle longitudinal control adjustment, i.e., the Newton direction, is a second-order optimal search path calculated to find the "optimal control increment." It is efficiently solved using the regularized vehicle dynamics Hessian matrix and vehicle speed gradient vector through LDLT decomposition, which is highly suitable for automotive chips. This ensures the absolute numerical stability of the control algorithm under extreme conditions while significantly reducing the computational load to meet the real-time requirements of 50Hz~100Hz.

[0048] In operation S250, the initial longitudinal control quantity of the current control cycle is iteratively updated using the vehicle longitudinal control adjustment quantity. During the iterative update process, the iterative processing result of the initial longitudinal control quantity is projected and constrained. Under the condition of satisfying the preset convergence condition, the longitudinal control increment of the current control cycle is obtained.

[0049] The hot-start strategy uses the optimal longitudinal control value of the previous control cycle as the initial longitudinal control value of the current control cycle. The control cycle, or one frame, is the time node in which the vehicle's longitudinal power system executes the vehicle's longitudinal control cyclically at fixed time intervals (e.g., 50Hz to 100Hz, i.e., every 10~20 milliseconds).

[0050] In operation S260, the target longitudinal control quantity of the current control cycle is determined by using the longitudinal control increment of the current control cycle and the target longitudinal control quantity of the previous control cycle.

[0051] The vehicle longitudinal control method provided in this application is a control method that does not rely on third-party libraries and is designed for automotive-grade embedded platforms. It employs techniques such as constructing a discrete augmented state-space model of the vehicle's longitudinal dynamics, offline pre-computing the power of the augmented matrix and reusing it online to obtain state prediction and control response matrices, analytically optimizing functions to obtain the dynamic Hessian matrix and vehicle speed gradient vector, regularizing the Hessian matrix and then decomposing it to solve for the control adjustment, and simultaneously applying projection constraints and combining convergence conditions to output the control increment during iterative updates of the control quantity. Therefore, it at least partially overcomes the technical problems of general-purpose QP solvers, such as reliance on third-party libraries, large code size, high computational redundancy, slow gradient descent convergence and numerous iterations, and the difficulty of meeting the high-frequency real-time operation requirements of 50Hz~100Hz and the significant control delay under complex slope and heavy load conditions on automotive embedded platforms. This achieves the technical effects of significantly reducing online real-time computation, not relying on third-party libraries, reducing computation time, adapting to automotive-grade single-precision floating-point environments, conforming to the fixed model structure of vehicle longitudinal control, and having controllable and bounded single-frame computation time. It meets the high-frequency real-time, stable, and robust operation requirements of automotive embedded platforms, balancing real-time performance and overall vehicle functional safety.

[0052] According to an embodiment of this application, the above-mentioned construction of a discrete augmented state space model representing the longitudinal dynamic system of a vehicle based on a vehicle control mode includes: approximating the longitudinal dynamic response of the vehicle's longitudinal dynamic system as a first-order inertial model based on the vehicle control mode, wherein the vehicle control mode includes a vehicle speed control mode or a vehicle acceleration control mode; constructing a discrete augmented state of the vehicle using vehicle speed error, vehicle acceleration error, and vehicle longitudinal control increment error; and performing matrix augmentation expansion on the vehicle discrete state matrix, vehicle discrete state input matrix, and vehicle predicted state output matrix in the first-order inertial model using the vehicle's discrete augmented state to obtain a discrete augmented state space model.

[0053] The following detailed description of the construction process of the discrete augmented state space model provided in this application will be further explained through specific implementation methods.

[0054] First, based on the two modes of vehicle speed control or acceleration control, the discrete state-space equations of the vehicle's longitudinal dynamic system are established, as shown in formulas (1) and (2):

[0055] (1),

[0056] (2).

[0057] The dynamic response is approximated as a first-order inertial element, meaning that the "vehicle state at the next moment" equals the "inertial continuation of the current state" plus the "change brought about by this control command"; the continuous model time constant is... Gain is According to the control cycle Discretization yields the vehicle's discrete state matrix. Vehicle discrete state input matrix With vehicle predicted state output matrix Further construct the augmented state matrix. , , Combine vehicle speed error, acceleration error, and control increment. By uniformly incorporating the augmented state, it becomes easier to simultaneously constrain tracking error and control smoothness within the objective function. Unlike general solvers that require an external input of a generalized matrix, this operation directly generates an augmented model with a fixed structure for the longitudinal model, eliminating redundant dimensions.

[0058] Formula (1) represents the first-order inertial characteristics of the vehicle power system (i.e., the vehicle response is delayed and does not change speed instantaneously), while the input matrix quantifies the specific impact of the control increment on the vehicle state. By discretizing according to the control cycle, this equation allows the controller to accurately predict the vehicle speed and acceleration trends at each future moment with a fixed time step.

[0059] Formula (2) acts like a "filter" or "sensor model" to accurately extract key feedback information such as vehicle speed error and acceleration error from complex internal states, which can then be used by subsequent optimization algorithms to calculate tracking deviation and control smoothness.

[0060] The specific implementation described above directly targets the augmented model with a fixed longitudinal dynamic generation structure, eliminating all redundant dimensions unrelated to longitudinal control; by directly using error and control increment as state variables, the subsequent construction of the optimization objective function can naturally take into account both "tracking accuracy" and "acceleration and deceleration smoothness" without the need for additional complex transformations, thus significantly reducing the computational overhead of single-precision chips.

[0061] According to an embodiment of this application, the above-mentioned method of obtaining the vehicle state prediction matrix and vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix in the discrete augmented state space model includes: performing offline power pre-calculation of the augmented matrix in the discrete augmented state space model to obtain the augmented matrix power calculation results corresponding to different control cycles; and performing online reuse of the augmented matrix power calculation results to obtain the vehicle state prediction matrix and vehicle control response matrix.

[0062] The offline power pre-calculation and online reuse process provided in this application will be further explained in detail below through specific implementation methods.

[0063] Pre-calculated vehicle state prediction matrix and vehicle control response matrix (Reduce online computation): Pre-compute the powers of the augmented state matrix during the initialization phase. , To predict the time domain length, and based on this, quickly construct the prediction domain α matrix (i.e., the vehicle state prediction matrix) and the control domain β matrix (i.e., the vehicle control response matrix), so that the prediction equation is as shown in formula (3):

[0064] (3).

[0065] in, To predict the output sequence in the time domain, To control the control increment sequence in the time domain, the time complexity of constructing the prediction matrix for each frame is reduced by pre-computing the exponentiation and reusing it. Down to ( For time domain length, (by reducing the state dimension), significantly reducing the amount of online computation on the vehicle, which is one of the keys to meeting the real-time requirements of 50Hz~100Hz.

[0066] According to an embodiment of this application, the above-mentioned analysis of the objective optimization function for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamics Hessian matrix and vehicle speed gradient vector includes: constructing the objective optimization function for vehicle longitudinal control prediction using the vehicle tracking error weight, the vehicle longitudinal control increment weight, the Hessian matrix, the gradient vector, and the vehicle longitudinal control increment sequence corresponding to the control cycle; and analyzing the Hessian matrix and gradient vector in the objective optimization function using the vehicle state prediction matrix, the vehicle control response matrix, the vehicle tracking error weight, and the vehicle longitudinal control increment weight to obtain the vehicle dynamics Hessian matrix and vehicle speed gradient vector.

[0067] The process of obtaining the Hessian matrix and gradient vector provided in this application will be further explained in detail below through specific implementation methods.

[0068] Construct the QP optimization problem and the MPC quadratic programming objective function, as shown in formula (4):

[0069] (4).

[0070] in, To track error weights, To control incremental weights; For Hessian matrix, This is the gradient vector, obtained from the aforementioned operations. , Matrix and weights , Analytically generate and This avoids the additional overhead caused by the generalized assembly of the general solver.

[0071] The pain point of related general solvers lies in the fact that traditional quadratic programming solvers (such as the general modes of OSQP and qpOASES) typically treat the objective function as a black box. In each control cycle, the Hessian matrix and gradient vector need to be recalculated using numerical methods or general matrix multiplication. For longitudinal control problems with fixed structures, this "generalized assembly" involves a large amount of redundant zero-element operations and memory addressing, which is a significant waste on automotive chips with limited computing power.

[0072] The advantage of this application is that: due to the state-space model of longitudinal dynamics ( , Matrix) and control objectives ( , The weights have a fixed structure and even constant values ​​during runtime, and the Hessian matrix and gradient vector can be derived in advance with explicit algebraic expressions. This means that at the code level, complex matrix operations are simplified to a few addition, subtraction, multiplication, division, and table lookup operations, compressing the original millisecond-level matrix assembly time to the microsecond level. This embodies the idea of ​​"compiler-level optimization for specific physical models".

[0073] According to an embodiment of this application, the above-mentioned regularization of the vehicle dynamic Hessian matrix to obtain a regularized vehicle dynamic Hessian matrix includes: adding a Hernoff regularization term to the vehicle dynamic Hessian matrix using a preset regularization coefficient to obtain a regularized vehicle dynamic Hessian matrix.

[0074] The regularization process of the Hessian matrix in this application will be further explained in detail below through specific implementation methods.

[0075] Hessian matrix preprocessing (one of the dual numerical stabilization mechanisms in this application): A Tikhonov regularization term is added to the Hessian matrix to obtain a regularized Hessian matrix. ,in For unit array, is the regularization coefficient (taken as 1e-6).

[0076] This regularization guarantees Strict positive definiteness, fundamentally avoiding the problems caused by single-precision floating-point operations in vehicles. Approaching singularity, leading to decomposition failure or control divergence—this is one of the key mechanisms that distinguishes this application from traditional single-precision solvers, which are prone to singularity and divergence.

[0077] According to an embodiment of this application, the above-mentioned matrix decomposition of the regularized vehicle dynamic Hessian matrix using the vehicle speed gradient vector to obtain the vehicle longitudinal control adjustment amount includes: decomposing the regularized vehicle dynamic Hessian matrix into a product of a unit lower triangular matrix and a diagonal matrix, and solving for the vehicle longitudinal control adjustment amount representing the Newton direction using the vehicle speed gradient vector during the decomposition of the regularized vehicle dynamic Hessian matrix to obtain the vehicle longitudinal control adjustment amount.

[0078] The matrix decomposition process provided in this application will be further described in detail below through specific implementation methods.

[0079] LDLT decomposition (a matrix decomposition method for symmetric matrices with non-zero principal minors of any order k) solves for the Newton direction (which is the second of the dual numerical stabilization mechanisms in this application): LDLT decomposition is performed on the positive definite regularized Hessian matrix, as shown in formula (5):

[0080] (5).

[0081] in For unit triangular array, It is a diagonal matrix.

[0082] Based on this decomposition, the Newtonian direction (i.e., the vehicle longitudinal control adjustment amount Δ) is solved: That is, first substitute back to solve. Then solve LDLT decomposition is more numerically stable, less computationally intensive, and does not require square root extraction compared to direct inversion, making it particularly suitable for single-precision embedded environments and further ensuring solution speed and numerical stability.

[0083] The Newton direction combines the Hessian matrix and the gradient vector. It doesn't blindly descend along the steepest direction, but rather takes into account the "terrain curvature" of the objective function. Therefore, compared to ordinary gradient descent, following the Newton direction requires fewer iterations.

[0084] According to an embodiment of this application, the above-mentioned method of iteratively updating the initial longitudinal control quantity of the current control cycle using the vehicle longitudinal control adjustment amount, and subjecting the iterative update result of the initial longitudinal control quantity to projection constraints during the iterative update process, and obtaining the longitudinal control increment of the current control cycle under the condition of satisfying the preset convergence condition, includes: based on the hot start initialization strategy, taking the target longitudinal control quantity of the previous control cycle as the initial longitudinal control quantity of the current control cycle; during the iterative update process of the initial longitudinal control quantity, controlling the iteration direction of the initial longitudinal control quantity using the vehicle longitudinal control adjustment amount, subjecting the longitudinal control increment obtained in each iteration to projection constraints, and performing convergence judgment on the iterative update process of the initial longitudinal control quantity, and obtaining the longitudinal control increment of the current control cycle under the condition of satisfying the preset convergence condition.

[0085] According to an embodiment of this application, the above-mentioned control of the iterative direction of the initial longitudinal control quantity using the vehicle longitudinal control adjustment amount includes: when the vehicle longitudinal control adjustment amount is characterized as a downward direction, determining the initial longitudinal control quantity of the subsequent control cycle using the vehicle longitudinal control adjustment amount and the initial longitudinal control quantity of the current control cycle; when the vehicle longitudinal control adjustment amount is characterized as a non-downward direction, reverting the vehicle longitudinal control adjustment amount to the fastest downward direction, and then iteratively updating the initial longitudinal control quantity again.

[0086] The process of obtaining the longitudinal control increment of the current control cycle provided in this application will be further described in detail below through specific implementation methods.

[0087] First, hot start initialization (reducing the number of iterations): the optimal longitudinal control value obtained in the previous frame (previous control cycle) is used as the initial value for the current frame (current control cycle). Since the operating conditions of adjacent control cycles are continuous and the optimal solution changes gradually, hot start allows the iteration to begin from near the optimal point, significantly reducing the number of iterations. This is the key to achieving convergence of 3 to 15 iterations per frame and a significant speedup compared to cold start.

[0088] Secondly, Newton's method iterative solution (including a robust mechanism for direction failure backoff): Execute Newton iterations, each step according to... Update control quantity ( (Step size). Each step first checks the Newtonian direction. Is the step size in the descent direction (i.e., If, due to numerical reasons, the Newtonian direction is not downward, it will automatically revert to the direction of steepest descent. Continue iterating. This backoff mechanism, which is based on Newton's method and uses the steepest descent as a fallback, ensures that the iteration always converges and the control remains stable and does not go out of control under extreme conditions such as large variations in load, gradient, and electric braking capacity. It is a robust design that is different from the single Newton method or the single gradient method.

[0089] According to an embodiment of this application, the above-mentioned projection constraint on the longitudinal control increment obtained in each iteration includes: comparing a preset minimum longitudinal control increment with the longitudinal control increment to obtain a first comparison result; comparing the first comparison result with a preset maximum longitudinal control increment to obtain a second comparison result; and using the second comparison result to apply upper and lower limit projection constraints on the longitudinal control increment to obtain the longitudinal control increment after projection constraint.

[0090] The constraint projection processing (i.e., hard constraint protection) provided in this application will be further described in detail below through specific implementation methods.

[0091] After each iteration, the control increment is forcibly projected element by element into the safe interval: .in, Projection lower limit, This is the upper limit of the projection.

[0092] By using projection, hard constraints are implemented to protect the control increment, ensuring that the output always falls within the physical and safety boundaries of the actuator, thus avoiding the impact caused by constraint violations.

[0093] In the technical solution provided in this application, multiple redundant convergence checks are performed during each iteration. The iteration process stops as long as any of the following conditions are met:

[0094] (1) Iterate at least 3 times, and the gradient norm and the change in control quantity are both less than their respective convergence thresholds (accuracy priority); (2) Reach the maximum safe number of iterations (real-time backup to prevent excessive iteration under extreme conditions); (3) Reach the configured maximum number of iterations.

[0095] The redundant triple criterion of "minimum number of iterations + gradient convergence + control variable change convergence" ensures the accuracy of the solution while strictly limiting the worst calculation time of a single frame, which is different from the traditional solver that stops based on a single threshold and has uncontrollable real-time performance.

[0096] Finally, output the optimal longitudinal control quantity for the current control cycle: the first control increment in the control time domain after iterative convergence. The control quantity is superimposed on the control quantity of the previous frame and used as the optimal longitudinal control quantity output for the current frame (i.e., the current control cycle) for vehicle longitudinal power / braking control; at the same time, the optimal solution of this frame is saved for use in the next frame's hot start, forming a closed loop.

[0097] This application addresses model predictive control (MPC) for vehicle longitudinal control, proposing a lightweight Newton-based QP solver and a vehicle longitudinal control method that is completely independent of third-party libraries and designed for automotive-grade embedded platforms. Unlike existing general / code-generated QP solvers such as qpOASES, OSQP, and FORCES, and regularized Newton-type solvers for general QP and double-precision environments, this application does not pursue a universally optimal solution. Instead, it addresses the characteristics of the vehicle longitudinal MPC problem—namely, a fixed model structure, automotive-grade single-precision floating-point operations, real-time operation at 50Hz or higher, and the need to meet functional safety constraints such as bounded computation time, non-divergent control, and non-outbounded output—by integrating the QP solver with the longitudinal control model and functional safety requirements. This is achieved through a collaborative mechanism of analytical pre-computation, hot-start closed-loop control, LDLT decomposition + Tikhonov regularization, redundant triple convergence criteria, and Newton direction failure backoff, resulting in a lightweight, stable, and robust real-time solution on an automotive embedded platform.

[0098] This application has the following advantages over related technologies:

[0099] (1) Deep coupling between the solver and the vehicle longitudinal control scenario: Based on the dual modes of vehicle speed control / acceleration control, an augmented state space model with a fixed structure is constructed. , , ), and by exponentiation of the augmented state matrix Offline pre-computation and online reuse, parsing to construct the prediction domain Matrix and Control Domain The time complexity of assembling the prediction matrix for each frame is reduced from... Down to Unlike general-purpose QP solvers such as qpOASES and OSQP, which require online generalization and assembly of dense matrices, this solver is designed for customized longitudinal control models, has a fixed structure, and requires less computation.

[0100] (2) Numerical stability combination for automotive-grade single-precision floating point: Tikhonov regularization (1e-6) is used to ensure that the Hessian matrix is ​​strictly positive definite, and the Newton direction is solved by superimposing the square root-free LDLT decomposition. The combination of the two is specifically designed to solve the decomposition failure and control divergence problem caused by the near singularity of Hessian in the automotive single-precision floating point environment. Unlike the regularized Newton-type solvers for general QP and double-precision floating point environments, this combination is customized for automotive-grade single-precision scenarios.

[0101] (3) Triple solution guarantee for functional safety: Through the triple mechanism of “maximum number of safe iterations (bounded computation time, no timeout) + automatic back-off from Newton direction failure with the fastest descent (no divergence in control) + control increment hard constraint projection (no output overshoot)”, the solver meets the automotive-grade functional safety requirements of “bounded single-frame computation time, no divergence in control, and no output violation of constraints” under any load, slope, and electric braking capacity variation conditions. Unlike the general QP solver that takes optimality / convergence as the only objective and has uncontrollable single-frame computation time, this application integrates the solver with functional safety requirements in its design.

[0102] (4) Hot start closed loop and dependency-free deployment: The optimal solution of the previous frame is used as the initial value of the current frame iteration and stored back. By utilizing the continuity of the optimal solutions of adjacent control cycles, the single frame iteration is compressed to 3~15 times. The whole process is implemented with pure Eigen (a high-level C open source library that focuses on linear algebra, matrix operations and numerical algorithms) without any third-party QP library. It can be directly run stably in real time at 50Hz~100Hz on automotive microcontroller units (MCUs) / vehicle controllers.

[0103] Based on the above-described vehicle longitudinal control method, embodiments of this application also provide a vehicle longitudinal control device. The following will be combined with... Figure 3 The device is described in detail.

[0104] Figure 3 A schematic block diagram of a vehicle longitudinal control device according to an embodiment of this application is shown.

[0105] like Figure 3 As shown, the vehicle longitudinal control device 300 in this embodiment includes an augmented model construction module 310, an augmented matrix processing module 320, an objective optimization function parsing module 330, a longitudinal control adjustment acquisition module 340, a longitudinal control increment acquisition module 350, and a longitudinal control quantity acquisition module 360.

[0106] The augmented model construction module 310 is used to construct a discrete augmented state space model representing the longitudinal dynamic system of the vehicle based on the vehicle control mode. In one embodiment, the augmented model construction module 310 can be used to perform the step S210 described above, which will not be repeated here.

[0107] The augmented matrix processing module 320 is used to obtain the vehicle state prediction matrix and the vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix on the discrete augmented state space model. In one embodiment, the augmented matrix processing module 320 can be used to execute the step S220 described above, which will not be repeated here.

[0108] The objective optimization function parsing module 330 is used to parse the objective optimization function for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamic Hessian matrix and the vehicle speed gradient vector. In one embodiment, the objective optimization function parsing module 330 can be used to execute the step S230 described above, which will not be repeated here.

[0109] The longitudinal control adjustment acquisition module 340 is used to regularize the vehicle dynamic Hessian matrix to obtain a regularized vehicle dynamic Hessian matrix, and to perform matrix decomposition on the regularized vehicle dynamic Hessian matrix using the vehicle speed gradient vector to obtain the vehicle longitudinal control adjustment amount. In one embodiment, the longitudinal control adjustment acquisition module 340 can be used to execute the step S240 described above, which will not be repeated here.

[0110] The longitudinal control increment acquisition module 350 is used to iteratively update the initial longitudinal control quantity of the current control cycle using the vehicle longitudinal control adjustment quantity, and to project constraints on the iterative processing result of the initial longitudinal control quantity during the iterative update process. Under the condition of satisfying the preset convergence condition, the longitudinal control increment of the current control cycle is obtained. In one embodiment, the longitudinal control increment acquisition module 350 can be used to execute the step S250 described above, which will not be repeated here.

[0111] The longitudinal control quantity acquisition module 360 ​​is used to determine the target longitudinal control quantity for the current control cycle using the longitudinal control increment of the current control cycle and the target longitudinal control quantity of the previous control cycle. In one embodiment, the longitudinal control quantity acquisition module 360 ​​can be used to execute step S260 described above, which will not be repeated here.

[0112] According to embodiments of this application, any and multiple modules among the augmented model construction module 310, augmented matrix processing module 320, objective optimization function parsing module 330, longitudinal control adjustment acquisition module 340, longitudinal control increment acquisition module 350, and longitudinal control quantity acquisition module 360 ​​can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the augmented model construction module 310, augmented matrix processing module 320, objective optimization function analysis module 330, longitudinal control adjustment acquisition module 340, longitudinal control increment acquisition module 350, and longitudinal control quantity acquisition module 360 ​​can be at least partially implemented as hardware circuits, such as field-programmable gate arrays, programmable logic arrays, systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits, or any other reasonable means of integrating or packaging circuits, or implemented in hardware or firmware, or in any one of software, hardware, and firmware implementations, or in a suitable combination of any of these. Alternatively, at least one of the augmented model construction module 310, augmented matrix processing module 320, objective optimization function analysis module 330, longitudinal control adjustment acquisition module 340, longitudinal control increment acquisition module 350, and longitudinal control quantity acquisition module 360 ​​can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0113] Figure 4 A block diagram schematically illustrates an electronic device suitable for implementing a vehicle longitudinal control method according to an embodiment of this application.

[0114] like Figure 4 As shown, an electronic device 400 according to an embodiment of this application includes a processor 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory 402 or a program loaded from a storage portion 408 into a random access memory 403. The processor 401 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a dedicated microprocessor. The processor 401 may also include onboard memory for caching purposes. The processor 401 may include a single processing unit or multiple processing units for executing different steps of the method flow according to an embodiment of this application.

[0115] Random access memory 403 stores various programs and data required for the operation of electronic device 400. Processor 401, read-only memory 402, and random access memory 403 are interconnected via bus 404. Processor 401 executes various steps of the method flow according to embodiments of this application by executing programs stored in read-only memory 402 and / or random access memory 403. It should be noted that the programs may also be stored in one or more memories other than read-only memory 402 and random access memory 403. Processor 401 may also execute various steps of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0116] According to embodiments of this application, the electronic device 400 may further include an input / output interface 405, which is also connected to a bus 404. The electronic device 400 may also include one or more of the following components connected to the input / output interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube, liquid crystal display, etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card, such as a local area network card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 410 as needed so that computer programs read from it can be installed into the storage section 408 as needed.

[0117] Embodiments of this application also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0118] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include the read-only memory 402 described above, and / or random access memory 403, and / or one or more memories other than read-only memory 402 and random access memory 403.

[0119] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0120] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via communication section 409, and / or installed from removable medium 411. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0121] In embodiments of this application, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by processor 401, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0122] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0124] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A vehicle longitudinal control method, characterized in that, The method includes: Construct a discrete augmented state-space model representing the vehicle's longitudinal dynamic system based on the vehicle control mode; By performing offline exponentiation of the augmented matrix and online reuse on the discrete augmented state space model, the vehicle state prediction matrix and the vehicle control response matrix are obtained. The objective optimization function for vehicle longitudinal control prediction is analyzed using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamic Hessian matrix and vehicle speed gradient vector. The vehicle dynamic Hessian matrix is ​​regularized to obtain a regularized vehicle dynamic Hessian matrix, and the vehicle speed gradient vector is used to perform matrix decomposition on the regularized vehicle dynamic Hessian matrix to obtain the vehicle longitudinal control adjustment amount. The initial longitudinal control quantity of the current control cycle is iteratively updated using the vehicle longitudinal control adjustment quantity, and during the iterative update process, the iterative processing result of the initial longitudinal control quantity is projected and constrained. Under the condition of satisfying the preset convergence condition, the longitudinal control increment of the current control cycle is obtained. The target longitudinal control quantity of the current control cycle is determined by using the longitudinal control increment of the current control cycle and the target longitudinal control quantity of the previous control cycle.

2. The method according to claim 1, characterized in that, The discrete augmented state-space model representing the vehicle's longitudinal dynamic system, constructed based on the vehicle control mode, includes: According to the vehicle control mode, the longitudinal dynamic response of the vehicle longitudinal power system is approximated as a first-order inertial model, wherein the vehicle control mode includes a vehicle speed control mode or a vehicle acceleration control mode. The discrete augmented state of the vehicle is constructed using vehicle speed error, vehicle acceleration error, and vehicle longitudinal control incremental error. The discrete augmented state space model is obtained by using the discrete augmented state of the vehicle to perform matrix augmentation on the vehicle discrete state matrix, vehicle discrete state input matrix, and vehicle predicted state output matrix in the first-order inertial model.

3. The method according to claim 1, characterized in that, By performing offline power calculation and online reuse of the augmented matrix on the discrete augmented state-space model, the vehicle state prediction matrix and vehicle control response matrix are obtained, including: Offline power pre-calculation is performed on the augmented matrix in the discrete augmented state space model to obtain the power calculation results of the augmented matrix corresponding to different control periods; The results of the augmented matrix power calculation are reused online to obtain the vehicle state prediction matrix and the vehicle control response matrix.

4. The method according to claim 1, characterized in that, Using the vehicle state prediction matrix and the vehicle control response matrix, the objective optimization function for vehicle longitudinal control prediction is analyzed to obtain the vehicle dynamics Hessian matrix and vehicle speed gradient vector, including: The objective optimization function for predicting vehicle longitudinal control is constructed using vehicle tracking error weights, vehicle longitudinal control increment weights, Hessian matrix, gradient vector, and vehicle longitudinal control increment sequence corresponding to the control cycle. The Hessian matrix and gradient vector in the objective optimization function are analyzed using the vehicle state prediction matrix, the vehicle control response matrix, the vehicle tracking error weight, and the vehicle longitudinal control increment weight to obtain the vehicle dynamic Hessian matrix and the vehicle speed gradient vector.

5. The method according to claim 1, characterized in that, The vehicle dynamics Hessian matrix is ​​regularized to obtain the regularized vehicle dynamics Hessian matrix, which includes: By using a preset regularization coefficient, the Heinrich's regularization term is added to the vehicle dynamics Heinrich's matrix to obtain the regularized vehicle dynamics Heinrich's matrix.

6. The method according to claim 5, characterized in that, Using the vehicle speed gradient vector to perform matrix decomposition on the regularized vehicle dynamics Hessian matrix, the longitudinal control adjustment of the vehicle is obtained, including: The regularized vehicle dynamics Hessian matrix is ​​decomposed into a product of a unit lower triangular matrix and a diagonal matrix. During the decomposition of the regularized vehicle dynamics Hessian matrix, the vehicle speed gradient vector is used to solve for the vehicle longitudinal control adjustment amount representing the Newton direction, thereby obtaining the vehicle longitudinal control adjustment amount.

7. The method according to claim 1, characterized in that, The initial longitudinal control quantity of the current control cycle is iteratively updated using the vehicle longitudinal control adjustment amount. During the iterative update process, the iterative update result of the initial longitudinal control quantity is subjected to projection constraints. Under the condition of satisfying a preset convergence condition, the longitudinal control increment of the current control cycle is obtained as follows: Based on the hot start initialization strategy, the target longitudinal control quantity of the previous control cycle is used as the initial longitudinal control quantity of the current control cycle. During the iterative update of the initial longitudinal control quantity, the iterative direction of the initial longitudinal control quantity is controlled by the vehicle longitudinal control adjustment quantity. The longitudinal control increment obtained in each iteration is projected and constrained. The convergence of the iterative update process of the initial longitudinal control quantity is judged. Under the condition that the preset convergence condition is met, the longitudinal control increment of the current control cycle is obtained.

8. The method according to claim 7, characterized in that, Controlling the iterative direction of the initial longitudinal control quantity using the vehicle longitudinal control adjustment includes: When the vehicle longitudinal control adjustment is characterized as a downward direction, the initial longitudinal control amount for the subsequent control cycle is determined using the vehicle longitudinal control adjustment and the initial longitudinal control amount of the current control cycle. When the vehicle longitudinal control adjustment is represented as a non-descending direction, the vehicle longitudinal control adjustment is reverted to the fastest descending direction, and the initial longitudinal control is iteratively updated again.

9. The method according to claim 7, characterized in that, The projection constraints on the longitudinal control increments obtained in each iteration include: The preset minimum longitudinal control increment is compared with the longitudinal control increment to obtain a first comparison result; The first comparison result is compared with the preset maximum longitudinal control increment to obtain the second comparison result; The longitudinal control increment is subjected to upper and lower limit projection constraints using the second comparison result to obtain the longitudinal control increment after projection constraints.

10. A vehicle longitudinal control device, characterized in that, The device includes: The augmented model construction module is used to construct a discrete augmented state-space model representing the vehicle's longitudinal dynamic system based on the vehicle control mode. The augmented matrix processing module is used to obtain the vehicle state prediction matrix and the vehicle control response matrix by performing offline power calculation and online reuse of the augmented matrix on the discrete augmented state space model. The objective optimization function parsing module is used to parse the objective optimization function for vehicle longitudinal control prediction using the vehicle state prediction matrix and the vehicle control response matrix to obtain the vehicle dynamic Hessian matrix and vehicle speed gradient vector. The longitudinal control adjustment acquisition module is used to regularize the vehicle dynamic Hessian matrix to obtain a regularized vehicle dynamic Hessian matrix, and to perform matrix decomposition on the regularized vehicle dynamic Hessian matrix using the vehicle speed gradient vector to obtain the vehicle longitudinal control adjustment. The longitudinal control increment acquisition module is used to iteratively update the initial longitudinal control quantity of the current control cycle using the vehicle longitudinal control adjustment quantity, and to project constraints on the iterative processing result of the initial longitudinal control quantity during the iterative update process, so as to obtain the longitudinal control increment of the current control cycle when the preset convergence condition is met. The longitudinal control quantity acquisition module is used to determine the target longitudinal control quantity of the current control cycle by using the longitudinal control increment of the current control cycle and the target longitudinal control quantity of the previous control cycle.

Images