A tracking control method, device and equipment of adaptive hybrid path

Abstract

The application provides a tracking control method, device and equipment of an adaptive hybrid path, the tracking control method inputs system state information into an unmanned vehicle plane kinematics model, obtains deviation information of the unmanned vehicle relative to a reference path and curvature information of a reference path point; for a lateral control layer, inputs a plurality of candidate front wheel steering angles and the deviation information into a steering angle optimization model, obtains a candidate front wheel steering angle with minimum cost, and inputs the candidate front wheel steering angle into a front wheel steering angle correction model, outputs a steering angle residual correction amount, and superimposes the candidate front wheel steering angle and the residual correction amount to generate a target front steering angle; for a longitudinal control layer, inputs the curvature information and the target front steering angle into a target vehicle speed planning model, obtains a target vehicle speed; inputs a current vehicle speed and the target vehicle speed into an adaptive speed regulation relationship model, obtains a longitudinal acceleration, and the safety of path tracking control of the unmanned vehicle is improved in the aspects of tracking accuracy and steering smoothness under complex curvature road conditions.

Description

Technical Field

[0001] This application relates to the field of intelligent vehicle control technology, and more specifically, to an adaptive hybrid path tracking control method, apparatus, and device. Background Technology

[0002] As the automotive industry transforms towards intelligence, autonomous driving technology has become a research hotspot and development direction in the global automotive field. Path tracking control, as the core component of autonomous driving systems, directly determines whether driverless vehicles can drive safely, stably, and accurately along preset reference paths, and is a key support for ensuring reliable operation in complex scenarios.

[0003] Existing path tracking control technologies suffer from several shortcomings: Traditional reinforcement learning methods exhibit problems such as blind exploration, low sample utilization, slow policy convergence, and even divergence in the early stages of training. Purely data-driven control strategies lack effective prior guidance, are prone to getting trapped in local optima, or produce unstable control actions, making it difficult to meet the safety and interpretability requirements of actual vehicle control. Furthermore, existing research often decouples lateral path tracking from longitudinal speed control, only roughly considering speed constraints in upper-level planning, without fully taking into account the coupling relationship between road curvature, steering state, and longitudinal vehicle speed. This results in the longitudinal speed failing to adaptively adjust to road geometry and steering actions under conditions of high curvature curves, high-speed lane changes, or low adhesion, making the vehicle prone to excessive lateral deviation, tires approaching their limit of adhesion, or even instability, seriously affecting the driving safety of autonomous vehicles. Summary of the Invention

[0004] This application provides a projectile mount and a projectile to improve the safety of path tracking control, thereby enhancing the tracking accuracy and steering smoothness of unmanned vehicles under complex curvature road conditions.

[0005] In a first aspect, embodiments of this application provide an adaptive hybrid path tracking control method, which is applied to the vehicle control system of an unmanned vehicle. The tracking control method includes: Obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, obtain in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment. For the lateral control layer, the front wheel angle at the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle at the previous control moment, and the obtained deviation information are input to characterize the angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel angle within a preset finite prediction time domain. This model obtains the candidate front wheel angle at the next control moment that minimizes the cost of predicting that the vehicle will travel smoothly in the next control moment. The candidate front wheel steering angle and the system state information obtained at the current control moment are input into the front wheel steering angle correction model. The angle residual correction amount of the candidate front wheel steering angle is output through the reinforcement learning network. The candidate front wheel steering angle and the residual correction amount are superimposed to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path. For the longitudinal control layer, the curvature information of the current control cycle and the target forward steering angle at the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed at the next control moment. The current vehicle speed, current front wheel angle, and target vehicle speed are input into an adaptive speed adjustment relationship model that enables coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action. This model yields longitudinal acceleration, which is then used to control the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration. This enables the unmanned vehicle to perform adaptive hybrid path tracking in complex curvature road scenarios.

[0006] Secondly, this application also provides an adaptive hybrid path tracking control device, which is applied to the vehicle control system of an unmanned vehicle, and the tracking control device includes: The deviation curvature information determination unit is used to obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, it obtains the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment in real time. The front wheel steering angle determination unit is used for the lateral control layer to input the front wheel steering angle of the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood based on the front wheel steering angle of the previous control moment, and the obtained deviation information to characterize the candidate front wheel steering angle of the next control moment that minimizes the cost of predicting that the vehicle will travel in a smooth state in the next control moment in the steering angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel steering angle within a preset finite prediction time domain. The target forward steering angle determination unit is used to input the candidate front wheel steering angle and the system state information obtained at the current control moment into the front wheel steering angle correction model, output the steering angle residual correction amount of the candidate front wheel steering angle through the reinforcement learning network, and superimpose the candidate front wheel steering angle and the residual correction amount to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path. The target vehicle speed determination unit is used to input the curvature information of the current control cycle and the target forward steering angle at the next control moment into the target vehicle speed planning model for the longitudinal control layer, so as to obtain the target vehicle speed at the next control moment. The longitudinal acceleration determination unit is used to input the current vehicle speed, the current front wheel steering angle, and the target vehicle speed into an adaptive speed adjustment relationship model that has been constructed to achieve coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, so as to obtain the longitudinal acceleration. This allows the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration, thereby realizing adaptive hybrid path tracking of the unmanned vehicle in complex curvature road scenarios.

[0007] Thirdly, embodiments of this application also provide an electronic device, the electronic device including a readable storage medium and a processor; wherein, the readable storage medium is used to store machine-executable instructions; the processor is used to read the machine-executable instructions on the readable storage medium and execute the instructions to implement the steps of the tracking control method described in any embodiment of the first aspect.

[0008] The technical solution provided in this application can achieve at least the following beneficial effects: This application provides an adaptive hybrid path tracking control method, apparatus, and device. The tracking control method includes: inputting the obtained system state information into a constructed planar kinematic model of an unmanned vehicle; based on a set reference path, obtaining in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment; for the lateral control layer, inputting the front wheel angle of the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle of the previous control moment, and the obtained deviation information into a steering angle optimization model to obtain the candidate front wheel angle of the next control moment that minimizes the cost of predicting the vehicle's trajectory in the next control moment; and inputting the candidate front wheel angle and the obtained system state information of the current control moment into the front wheel steering... In the angle correction model, the reinforcement learning network outputs the angle residual correction amount for the candidate front wheel steering angle, and the candidate front wheel steering angle and the residual correction amount are superimposed to generate the target forward steering angle for the next control moment to drive the autonomous vehicle's lateral path tracking. For the longitudinal control layer, the curvature information of the current control cycle and the target forward steering angle for the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed for the next control moment. The current vehicle speed, the current front wheel steering angle and the target vehicle speed are input into the constructed adaptive speed adjustment relationship model to obtain the longitudinal acceleration. This enables the steering mechanism and drive actuator of the autonomous vehicle to form a lateral and longitudinal integrated closed-loop path tracking control based on the target forward steering angle and longitudinal acceleration, thereby realizing the autonomous vehicle's adaptive hybrid path tracking in complex curvature road scenarios. As can be seen, this embodiment comprehensively considers the coupling relationship between road curvature, steering state and longitudinal vehicle speed, so that the longitudinal speed can be adaptively adjusted according to the road geometry and steering action under conditions of large curvature curves, high-speed lane changes or low adhesion, thereby improving the safety of path tracking control with improved tracking accuracy and steering smoothness of unmanned vehicles under complex curvature road conditions. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating an exemplary embodiment of the adaptive hybrid path tracking control method of this application; Figure 2(a) is a schematic diagram of the trajectory of an unmanned vehicle based on lateral deviation, as illustrated in an exemplary embodiment of this application; Figure 2(b) is a schematic diagram of the trajectory of an unmanned vehicle based on heading deviation, as illustrated in an exemplary embodiment of this application; Figure 2(c) is a schematic diagram of the trajectory of an unmanned vehicle based on prediction deviation, as illustrated in an exemplary embodiment of this application; Figure 3 This is a state structure diagram of unmanned vehicle tracking shown in an exemplary embodiment of this application; Figure 4 This is a schematic diagram illustrating the structure of an adaptive hybrid path tracking control device according to an exemplary embodiment of this application; Figure 5 This is a schematic diagram illustrating the structure of an electronic device according to an exemplary embodiment of this application. Detailed Implementation

[0010] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0011] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to limit the application. Unless otherwise defined, the technical or scientific terms used in this application should be understood in their ordinary sense by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in this application specification and claims do not indicate any order, quantity, or importance, but are only used to distinguish different components. Similarly, the terms "a" or "one," etc., do not indicate a quantity limitation, but rather indicate the presence of at least one, which will be separately stated if only "a" is referred to. "A plurality" or "several" means two or more. Unless otherwise indicated, the terms "front," "rear," "lower," and / or "upper," "top," "bottom," etc., are for ease of description only and are not limited to a location or spatial orientation. The terms "comprising" or "including," etc., mean that the elements or objects preceding "comprising" or "including" encompass the elements or objects listed following "comprising" or "including" and their equivalents, and do not exclude other elements or objects. The word “connection” or “link” is not limited to physical or mechanical connections, but can also include electrical connections, whether direct or indirect.

[0012] Please see Figure 1 , Figure 1 The diagram shown is a flowchart illustrating an adaptive hybrid path tracking control method according to an exemplary embodiment of this application. This tracking control method is applied to the vehicle control system of an unmanned vehicle and includes the following steps: Step 101: Obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, obtain in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment.

[0013] In this embodiment, the system state information may include: the position of the vehicle's center of mass in the ground coordinate system, heading angle, longitudinal velocity, front wheel steering angle, and longitudinal acceleration. As an example, the unmanned vehicle's planar kinematic model is a planar kinematic model based on the assumption of a single track for the center of mass, and the specific construction process is as follows: the position of the vehicle's center of mass in the ground coordinate system, heading angle, and longitudinal velocity are selected as system state variables, and the front wheel steering angle and longitudinal acceleration are selected as control input variables. At the same time, based on a preset reference path, the lateral deviation, heading deviation, prediction deviation, and curvature information of the reference path points of the unmanned vehicle relative to the reference path are calculated in real time, and the system state information and deviation information are jointly constructed into the observation state vector of the vehicle control system.

[0014] Step 102: For the lateral control layer, the front wheel angle at the previous control time, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle at the previous control time, and the obtained deviation information are input to characterize the angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel angle within a preset finite prediction time domain, and obtains the candidate front wheel angle at the next control time that minimizes the cost of predicting that the vehicle will travel in a smooth state in the next control time.

[0015] In this embodiment, the finite prediction time domain can be understood as the cornering optimization model making a short-term prediction of the possible future motion path of the vehicle under the action of each candidate front wheel cornering angle within a pre-set finite prediction time range.

[0016] Furthermore, the minimum cost of smoothing can be understood as the optimal comprehensive evaluation result when the lateral deviation, heading deviation, and angle change smoothness of the predicted trajectory are all considered. The candidate front wheel angle for the next control moment with the minimum cost of smoothing is also the candidate front wheel angle for the next control moment that minimizes the comprehensive cost function for the smoothness of lateral deviation, heading deviation, and angle change.

[0017] The next control moment can be understood as the planned duration of a "short segment of the future". Based on this, this embodiment predicts the vehicle's driving trajectory within a set time range in the future for each candidate front wheel steering angle according to the autonomous vehicle's planar kinematics model, and calculates the deviation information of the driving trajectory relative to the reference path.

[0018] In this embodiment, the optimization objective is to minimize the cost incurred in predicting the vehicle's trajectory during the next control time step. The candidate front wheel angles for the next control time step are used as the optimization object, with constraints including lateral deviation, heading deviation, and control smoothness as boundary conditions. An angle optimization model is used as the optimization model. Based on this, in practical applications, the optimization objective is to minimize the cost incurred in predicting the vehicle's trajectory during the next control time step. Within the boundary conditions, the angle optimization model is optimized to obtain the candidate front wheel angle for the next control time step with the minimum cost. This embodiment evaluates each candidate path and selects the candidate front wheel angle with the minimum cost as the baseline angle, providing high-confidence prior guidance for strategy search.

[0019] Step 103: Input the candidate front wheel steering angle and the system state information obtained at the current control moment into the front wheel steering angle correction model, output the steering angle residual correction amount of the candidate front wheel steering angle through the reinforcement learning network, and superimpose the candidate front wheel steering angle and the residual correction amount to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path.

[0020] In this step, the reference steering angle and vehicle state information are input into the front wheel steering angle correction model. The reinforcement learning network outputs the steering angle residual correction amount for the reference steering angle. The reference steering angle and the residual correction amount are superimposed to generate the target forward steering angle at the next control moment to drive the autonomous vehicle to track the lateral path.

[0021] Step 104: For the longitudinal control layer, the curvature information of the current control cycle and the target forward turning angle at the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed at the next control moment.

[0022] In the longitudinal control layer, a target vehicle speed planning model is constructed based on the reference path curvature obtained in step 101 and the target forward turning angle obtained in step 103, so as to obtain the target vehicle speed at the next control moment.

[0023] The target speed is calculated using a target speed planning model based on the current path curvature and the target forward turning angle.

[0024] This step targets the longitudinal control layer, inputting the reference path curvature information corresponding to the current control cycle and the target front wheel steering angle for the next control moment output by the lateral control layer into the target vehicle speed planning model to obtain the target vehicle speed for the next control moment used for longitudinal speed adjustment.

[0025] Step 105: Input the current vehicle speed, current front wheel angle, and target vehicle speed into the constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, and obtain longitudinal acceleration. This allows the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration, thereby realizing adaptive hybrid path tracking of the unmanned vehicle in complex curvature road scenarios.

[0026] In this embodiment, the current vehicle speed, target vehicle speed, and current front wheel angle are input into the constructed adaptive speed adjustment relationship model. Based on the speed deviation between the current vehicle speed and the target vehicle speed, and combined with the current front wheel angle, the longitudinal acceleration is adaptively adjusted to obtain the longitudinal acceleration. Then, the autonomous vehicle's steering mechanism is controlled according to the target front wheel angle, and the autonomous vehicle's drive actuator is controlled according to the longitudinal acceleration, thereby forming a lateral and longitudinal integrated closed-loop path tracking control, realizing adaptive hybrid path tracking of the autonomous vehicle in complex curvature road scenarios.

[0027] An adaptive speed adjustment model is designed based on the deviation between the current vehicle speed and the target vehicle speed, and the longitudinal acceleration is output to achieve coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action.

[0028] Therefore, in the technical solution provided in this application, the tracking control method inputs the obtained system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, it obtains in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment. For the lateral control layer, the front wheel angle at the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle at the previous control moment, and the obtained deviation information are input into the angle optimization model to obtain the candidate front wheel angle for the next control moment with the minimum cost. The candidate front wheel angle and the obtained system state information at the current control moment are input into the front wheel angle correction model, and the model outputs the correction information through a reinforcement learning network. The candidate front wheel steering angle is corrected by the residual value, and the candidate front wheel steering angle is superimposed with the residual value to generate the target forward steering angle for the next control moment in driving the autonomous vehicle's lateral path tracking. For the longitudinal control layer, the curvature information of the current control cycle and the target forward steering angle for the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed for the next control moment. The current vehicle speed, current front wheel steering angle, and target vehicle speed are input into the constructed adaptive speed adjustment relationship model to obtain the longitudinal acceleration. This allows the steering mechanism and drive actuator of the autonomous vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and longitudinal acceleration, realizing adaptive hybrid path tracking of the autonomous vehicle in complex curvature road scenarios. It can be seen that this embodiment comprehensively considers the coupling relationship between road curvature, steering state, and longitudinal vehicle speed, so that the longitudinal speed can be adaptively adjusted according to the road geometry and steering action in high curvature curves, high-speed lane changes, or low adhesion conditions, thereby improving the safety of path tracking control for autonomous vehicles in complex curvature road conditions in terms of tracking accuracy and steering smoothness.

[0029] In one embodiment of this application, the implementation of step 101, which involves inputting the system state information into the constructed planar kinematic model of the unmanned vehicle, and obtaining the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment based on a set reference path, may include the following steps: Step A1: The position of the vehicle's center of mass in the ground coordinate system, heading angle, longitudinal velocity, front wheel steering angle, and longitudinal acceleration are determined by the front wheel steering angle and longitudinal acceleration as control inputs. These are then input into the following first and second expressions to obtain the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment. The first expression is:

[0030] in, and These are the position coordinates of the vehicle's center of gravity in the ground coordinate system at the current control time and the next control time, respectively. and These are the position coordinates of the vehicle's center of gravity in the ground coordinate system at the current control time and the next control time, respectively. and These are the vehicle's heading angles at the current control time and the next control time, respectively; and These are the longitudinal speeds of the vehicle at the current control moment and the next control moment, respectively; The front wheel steering angle at the current control moment; The longitudinal acceleration at the current control moment; The sideslip angle of the vehicle's center of gravity at the current control moment; To control the sampling period; The wheelbase of the vehicle, and ; This is the distance from the front axle to the vehicle's center of gravity. This is the distance from the rear axle to the vehicle's center of gravity. The current control moment is determined by the front wheel steering angle. A defined vehicle curvature; Step A2: Based on the set reference path and the curvature information of the reference path points, the lateral deviation, heading deviation, and prediction deviation of the unmanned vehicle relative to the reference path are obtained after solving the second expression. The second expression is: , ,

[0031] in, This is the lateral deviation. For heading deviation, To predict bias, This is the position for the next control time point. This is the sequence number of the next control time point in the discrete reference path; Sn For the current vehicle location point Reference path nearest point and adjacent path points The semi-perimeter of the triangle , Vehicle location point N To the nearest path point used as a reference path distance, , Vehicle location point N To the nearest path point Adjacent next path point distance, , The chord length between adjacent reference path points. , The next control time point Reference path curvature at that point Current vehicle location N To the next control time point P distance, The closest point on the reference path K To the next control time point P Path direction distance, Let yaw rate be the vehicle's angular velocity. For longitudinal velocity, The current vehicle heading angle, n The closest point on the reference path The serial number, and These are the current vehicle locations. The horizontal and vertical position coordinates, and The vehicle position points at the next control time point are respectively P The horizontal and vertical position coordinates.

[0032] In this embodiment, the lateral deviation, heading deviation, prediction deviation, and curvature information of the reference path points of the unmanned vehicle are calculated based on the reference path. The above state variables constitute the vehicle tracking state vector used by the unmanned vehicle path tracking control.

[0033] On the discrete reference path, the nearest path point corresponding to the current vehicle position. Starting from the reference path tangent direction, search forward for the distance to the next control time step. To obtain the next control time point Specifically, let the discrete reference path point sequence be... The approximate straight-line distance between the arc segments of adjacent path points is And define the cumulative path distance , To accumulate from the starting point or the shortest path point to the th The cumulative path distance of each path point, where , , Let be the total number of discrete reference path points. Find the one that satisfies... index , This is the index of the previous path point in the path interval where the next control time point is located. For the first Path points To the Path points The arc segment between them is approximately a straight line distance, which is the next control time point. Located at the waypoint and Between. Let The coordinates of the next control time point are then determined by the following formula: ; In the formula, The linear interpolation ratio is used to characterize the next control time point. At adjacent path points and The positions between.

[0034] In one embodiment of this application, step 102 involves inputting the front wheel angle from the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle from the previous control moment, and the obtained deviation information into a steering angle optimization model that characterizes the optimization evaluation of each candidate motion path corresponding to each candidate front wheel angle within a preset finite prediction time domain. The result is the candidate front wheel angle for the next control moment that minimizes the cost of predicting a smooth trajectory for the vehicle in the next control moment. This includes: For the front wheel steering angle at the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood are determined; The front wheel angle of the previous control moment, multiple candidate front wheel angles, and the obtained deviation information are input into the third expression. Based on the autonomous vehicle's planar kinematics model, the trajectory of the vehicle within a preset finite prediction time domain is predicted, and the candidate front wheel angle for the next control moment with the minimum cost is output. The third expression is: , ; in, For the lateral deviation weighting coefficient, For heading deviation weighting coefficient, For the prediction bias weighting coefficient, This is the corner smoothness weighting coefficient. The front wheel steering angle at the current control moment. The front wheel steering angle corresponding to the previous control moment. For the first k One candidate corner, k Candidate corner number, The candidate front wheel steering angle for the next control time is the one that minimizes the cost of predicting the vehicle's trajectory during the next control time. For the first The comprehensive cost function corresponding to each candidate front wheel steering angle Current control time The generated set of candidate front wheel steering angles.

[0035] In this embodiment, the tracking status of the unmanned vehicle is as follows: Figure 3As shown, the steering angle optimization model is used to generate a locally optimal reference front wheel steering angle. Specifically, it is the actual front wheel steering angle at the previous control moment. Centered on the preset corner neighborhood Multiple candidate front wheel steering angles are generated uniformly and discretely within the inner circle, forming a set of candidate front wheel steering angles:

[0036] in, The set of candidate front wheel steering angles generated at the current control moment; For the first One candidate front wheel angle; The actual front wheel steering angle at the previous control moment; For the candidate front wheel steering angle, the distance of the walk, ; The half-width of the preset corner search neighborhood; The discrete offset index of the candidate front wheel steering angle relative to the center steering angle; is a positive integer used to determine the discrete number of candidate front wheel steering angles.

[0037] For each candidate front wheel steering angle Set the current front wheel angle in the first expression to The system then uses a planar kinematics model of the unmanned vehicle to perform short-time prediction, obtaining the lateral coordinates of the predicted position of the vehicle at the next control moment under the action of the candidate front wheel steering angle. and vertical coordinate values and predicted heading angle superscript This indicates that the predicted quantity corresponds to the first... One candidate front wheel corner.

[0038] Furthermore, the distance to the next control moment is determined based on the current longitudinal speed of the vehicle: ; in, This represents the distance to the next control time from the current control time. The minimum distance to the next control time; This is the velocity gain coefficient; Let be the vehicle's longitudinal speed at the current control moment. The time of the next control moment can be expressed as: ; in, This is the time scale corresponding to the next control moment as the vehicle moves forward along the reference path.

[0039] In the local coordinate system, the lateral motion of the vehicle relative to the reference path can be expressed as: ; in, Here is the lateral coordinate of the vehicle's current position. The lateral speed of the vehicle at the current control moment. The lateral acceleration of the vehicle at the current moment. This represents the lateral deviation of the vehicle relative to the reference path at the current control moment; This represents the heading deviation between the vehicle's heading angle and the reference path direction at the current control moment. For the first The vehicle curvature corresponding to each candidate front wheel steering angle is expressed as follows: ; in, This refers to the vehicle's wheelbase.

[0040] Based on the vehicle's current lateral motion state, a second-order Taylor expansion is used to analyze the vehicle's aiming time. Predict the lateral position afterward: ; in, For the first The predicted lateral position corresponding to each candidate front wheel steering angle; , , These are the vehicle's lateral position, lateral velocity, and lateral acceleration at the current moment; Pre-aiming time; This is the current control moment.

[0041] For each candidate front wheel steering angle Based on the deviation between the predicted lateral position and the next control time point of the reference path, the prediction deviation is calculated. Based on this, a comprehensive cost function is constructed: ; in, For the first The comprehensive cost function corresponding to each candidate front wheel steering angle; This is the weighting coefficient for the lateral deviation; This is the heading deviation weighting coefficient; The prediction bias weighting coefficient; This represents the weighting coefficient for steering angle smoothness. The comprehensive cost function simultaneously evaluates the vehicle's current lateral deviation, heading deviation, prediction deviation at the next control moment, and the smoothness of the front wheel steering angle change.

[0042] Finally, in the candidate front wheel steering angle set The candidate front wheel angle that minimizes the comprehensive cost function is selected as the local optimal reference front wheel angle at the current control moment. ; in, This represents the locally optimal reference front wheel steering angle output by the steering angle optimization model at the current control moment; superscript This indicates that the variable is the optimal value obtained from the optimization solution. This locally optimal reference front wheel steering angle is used as input to the subsequent front wheel steering angle correction model, so as to further generate the target front wheel steering angle at the next control moment through steering angle residual correction.

[0043] In this step, the process is as follows: Figure 3 As shown, its reward function includes: (1) Obstacle Avoidance Reward: If a collision occurs during trajectory tracking, the reward function will be set to an extremely low value, where This represents the penalty value of the reward function when an autonomous vehicle collides with another vehicle in the environment.

[0044] (2) Guided Tracking Reward: In the control command information, a guided tracking reward function is constructed by utilizing the difference between the locally optimal reference front wheel angle output by the angle optimization model and the corrected front wheel angle output by the front wheel angle correction model. This function guides the front wheel angle correction model to perform residual correction within the neighborhood of the reference front wheel angle, avoiding excessive deviation in steering control actions and improving the stability and smoothness of steering commands. Specifically, the angle optimization model generates multiple candidate front wheel angles in a local neighborhood centered on the front wheel angle at the previous control moment. Based on the autonomous vehicle's planar kinematics model, it performs short-term prediction of the candidate motion paths corresponding to each candidate front wheel angle, and then selects the locally optimal reference front wheel angle according to the comprehensive cost function.

[0045] (3) Trajectory tracking reward: Represents the distance between the vehicle's desired position and its actual position. This distance is calculated at each time step to provide a dense reward, thereby guiding the vehicle to follow the reference path. This reward design effectively avoids the slow convergence speed or non-convergence problem caused by sparse rewards. The final constructed trajectory following reward function .

[0046] (4) Speed ​​tracking reward: used to measure the deviation between the vehicle’s current speed and the desired speed, thereby prompting the strategy to optimize the lateral trajectory tracking performance while achieving coordinated tracking of the longitudinal speed target during the learning process.

[0047] (5) Weighted reward function design: Based on the tasks to be completed during vehicle trajectory tracking, this application uses a weight vector to design the overall reward function. .

[0048] In one embodiment of this application, the implementation of step 103, which involves inputting the candidate front wheel steering angle and the obtained system state information at the current control moment into the front wheel steering angle correction model, and outputting the steering angle residual correction amount of the candidate front wheel steering angle through a reinforcement learning network, includes the following steps: Step B1: Input the candidate front wheel steering angle and the system state information obtained at the current control moment into the fourth expression to obtain the steering angle residual correction amount for the candidate front wheel steering angle; The fourth expression is: ; in, This is a model for correcting the front wheel steering angle. For state space, , This refers to the reinforcement learning state information corresponding to the current control moment; For the action space, , Current control time The control action vector, , For the maximum longitudinal acceleration, For maximum braking deceleration, This is the maximum permissible steering angle for the front wheels. For longitudinal acceleration control, This is the state transition function. , The current status of the vehicle. The next state for the vehicle. To assign values ​​to actions in the action space. For the overall reward function, , This is a reward item for lateral deviation. , Current control time t Lateral deviation, For heading deviation bonus items, , Current control time t The course deviation, Prediction bias reward items , Current control time t Prediction bias, For obstacle avoidance reward information, , The weighting coefficient for obstacle avoidance rewards. To guide the tracking of reward information, , To guide the weighting coefficient of the tracking reward, The front wheel angle residual correction is the output of the front wheel angle correction model determined by the SAC (Soft Actor-Critic) algorithm at the current control moment. The vehicle control system superimposes the candidate front wheel angle with the front wheel angle residual correction to generate the target front wheel angle for the next control moment. For speed tracking reward information, , The weighting coefficient of the velocity tracking excitation term. As a discount factor, Indicates the current control time. The actual longitudinal speed of the vehicle, Indicates the current control time. The corresponding target speed.

[0049] The front wheel steering angle correction model employs a maximum entropy reinforcement learning framework. Its state space includes at least the lateral bias solution: when the vehicle experiences a deviation, heading deviation, path curvature, vehicle speed, and the reference steering angle. The action space is the steering angle residual correction amount. In reinforcement learning modeling, for the vehicle path tracking task, the current state... After input, this control system will process the action set. Choose one action The vehicle transitions from its current state to the next state under this action and receives a reward based on tracking performance, safety, and guidance consistency. Figure 2(a) illustrates this lateral deviation. The calculation relationship. The current vehicle location is denoted as... Distance from vehicle location on the reference path The nearest path point is denoted as ,and The adjacent next reference path point is denoted as .in, Indicates the vehicle's location point to the nearest point on the reference path distance, Indicates the vehicle's location point To adjacent reference path points distance, Indicates reference path point and The chord length between. Based on the... , and The formed local triangle relationship allows for the calculation of the lateral distance deviation of the vehicle's current position relative to a local line segment of the reference path. Figure 2(b) illustrates the heading deviation. The calculation relationship. The current vehicle location is... The next control time point is The vehicle's current heading angle is Current location of the vehicle Point to the next control time point The direction is used to characterize the local guidance direction of the reference path, and the angle between the vehicle's current heading direction and this guidance direction is the heading deviation. This heading deviation is used to evaluate the degree of deviation of the autonomous vehicle's current attitude from the reference path direction. Figure 2(c) illustrates the prediction deviation. The calculation relationship. The vehicle control system is based on the current vehicle position. The current front wheel steering angle and candidate front wheel steering angles are used to predict the vehicle's position at the next control moment or within a preset finite prediction time domain, thus obtaining the predicted position point. The next control time point on the reference path is denoted as... Predict location points With the next control time point The deviation relationship between them is used to characterize the prediction bias. The gray area in the figure represents the local search neighborhood corresponding to the candidate front wheel turning angle. The closer the candidate turning angle is to the direction that makes the predicted trajectory closer to the reference path, the smaller the corresponding prediction deviation. When a collision is possible, the reward value is taken as follows: This is used to penalize collision behavior; when the vehicle does not collide, the reward value is 0. Used to assess whether a vehicle has been involved in a collision or entered a dangerous state.

[0050] In the current state The controller then executes the action. Then, the vehicle transitions from the current state to the next state. The state transition relationship. The current status of the vehicle. This will be the next state for the vehicle. To assign values ​​to actions in the action space. This is the single-step action vector executed at the current control moment.

[0051] The trajectory tracking reward information is used to evaluate the tracking error between the current trajectory of the autonomous vehicle and the reference path, and it is rewarded by lateral deviation. Heading deviation reward and prediction deviation reward These components together constitute the following: Lateral deviation reward characterizes the lateral distance error of the vehicle's current position relative to the reference path; heading deviation reward characterizes the angular error between the vehicle's current heading angle and the tangent direction of the reference path; and prediction deviation reward characterizes the error of the vehicle relative to the reference path at the next control time point or predicted position. The corresponding formula can be written as: , Used to evaluate the lateral deviation of the current vehicle position relative to the reference path. Used to evaluate the heading deviation of the current vehicle heading angle relative to the reference path direction. This trajectory tracking reward information is used to evaluate the vehicle's prediction deviation relative to the reference path at the next control time point or predicted position. It guides the control system to reduce path tracking error and improve the stability and accuracy of the vehicle's travel along the reference path.

[0052] The guiding reward item is used to evaluate the consistency between the SAC output rotation angle and the POP reference rotation angle. The speed tracking bonus is used to evaluate the deviation between the current vehicle speed and the target vehicle speed. This is a distance normalization scale used to normalize lateral or prediction biases, avoiding the influence of different dimensions on reward calculations.

[0053] In one embodiment of this application, the implementation of step 105, which involves inputting the current vehicle speed and the target vehicle speed into a pre-constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering, to obtain longitudinal acceleration, may include the following steps: Input the current vehicle speed and the target vehicle speed into the fourth expression to obtain the longitudinal acceleration; The fifth expression is: ,

[0054] in, For curvature feedback gain, For corner feedback gain, As the reference curvature of the current road, This is a function that cuts off the upper and lower limits. The target vehicle speed lower limit, To achieve the target vehicle speed limit, The upper limit of safe speed is obtained based on the front wheel steering angle constraint. , The target front wheel steering angle is output by the lateral control layer. The maximum permissible lateral acceleration for the vehicle. This refers to the vehicle's wheelbase. The upper limit of safe speed is obtained based on road curvature constraints. , To prevent the denominator from becoming singular when the curvature approaches zero, To prevent the denominator from being a singular positive variable when the front wheel steering angle approaches zero, The target vehicle speed is determined by geometric feedback. This is the maximum permissible steering angle for the front wheels. The target vehicle speed is output by the longitudinal control layer.

[0055] In this embodiment, The formula indicates that the greater the road curvature or the greater the target front wheel steering angle, the lower the geometric feedback target vehicle speed.

[0056] In this embodiment, a target vehicle speed based on curvature and steering feedback is constructed by coordinating longitudinal velocity with road geometry characteristics. This target vehicle speed is generated under the premise of satisfying dynamic constraints. ; The boundary constraints are set as follows: 1) During vehicle steering, curvature and velocity jointly determine the lateral acceleration. To satisfy the above kinematic constraints. 2) To ensure the lateral stability of the vehicle, the above lateral acceleration limit constraint should be satisfied, wherein the upper speed limit automatically decreases as the curvature increases, thereby preventing the vehicle from becoming unstable during sharp turns. 3) This upper speed limit automatically decreases when the curvature changes significantly or the coefficient of adhesion is low, thereby effectively preventing the vehicle from sideslipping or becoming unstable during sharp turns. Simultaneously, considering the physical constraint of the front wheel steering angle, another speed limit based on the steering angle is obtained. In summary, the comprehensive target speed is obtained.

[0057] In one embodiment of this application, the step of inputting the current vehicle speed, current forward steering angle, and target vehicle speed into a pre-constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, to obtain longitudinal acceleration, includes: Input the current vehicle speed, current forward turning angle, and target vehicle speed into the sixth expression to obtain the longitudinal acceleration; The sixth expression is: , .

[0058] in, This is the normalized longitudinal control adjustment coefficient. It is an adjustable scaling factor. The maximum speed limit allowed for vehicles. The current front wheel steering angle, The maximum front wheel steering angle, The maximum longitudinal acceleration of the vehicle. This is the longitudinal acceleration signal output to the drive / brake actuator at the current control moment. The target speed.

[0059] In this embodiment, when the target vehicle speed is close to the speed limit or the current front wheel steering angle is large, the longitudinal acceleration adjustment coefficient is adjusted. This will decrease, thereby reducing the longitudinal acceleration output.

[0060] In one embodiment of this application, the implementation of step 105, which involves controlling the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward turning angle and the longitudinal acceleration, may include the following steps: Step C1: Based on the target front wheel angle, which serves as a lateral control signal, control the steering actuator of the unmanned vehicle to drive the front wheels to rotate, thereby correcting lateral and heading deviations.

[0061] In this embodiment, the target front wheel steering angle is a lateral control signal used to control the target front wheel steering angle.

[0062] Step C2 converts the longitudinal acceleration signal, which serves as the lateral control signal, into the throttle opening of the unmanned vehicle. Based on the throttle opening, the drive and braking actuators of the unmanned vehicle are controlled to control the longitudinal movement of the vehicle, so as to follow the target vehicle speed and drive the unmanned vehicle to travel along the reference path.

[0063] The throttle opening determines the vehicle speed; a larger throttle opening results in greater acceleration for the autonomous vehicle, while a smaller throttle opening results in less acceleration. In this embodiment, the vehicle control system can drive the autonomous vehicle along a reference path based on a target vehicle speed.

[0064] Step C3 involves feeding back the updated vehicle status information in real time to form a closed-loop path tracking control system that integrates horizontal and vertical directions.

[0065] In this step, the vehicle status information is updated and fed back in real time, so that step 101 can obtain the latest updated vehicle status information in real time, thereby forming a horizontal and vertical integrated closed-loop path tracking control.

[0066] Secondly, such as Figure 4 As shown, Figure 4 This application also provides a schematic diagram of the structure of an adaptive hybrid path tracking control device 200, which is applied to the vehicle control system of an unmanned vehicle. The tracking control device includes: The deviation curvature information determination unit 201 is used to obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, it obtains the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment in real time. The front wheel steering angle determination unit 202 is used to input the front wheel steering angle of the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood based on the front wheel steering angle of the previous control moment, and the obtained deviation information into the steering angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel steering angle within a preset finite prediction time domain, and obtains the candidate front wheel steering angle of the next control moment that minimizes the cost of predicting that the vehicle will travel in a smooth state when driving its trajectory in the next control moment. The target forward steering angle determination unit 203 is used to input the candidate front wheel steering angle and the system state information obtained at the current control moment into the front wheel steering angle correction model, output the steering angle residual correction amount of the candidate front wheel steering angle through the reinforcement learning network, and superimpose the candidate front wheel steering angle and the residual correction amount to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path. The target vehicle speed determination unit 204 is used to input the curvature information of the current control cycle and the target forward turning angle at the next control moment into the target vehicle speed planning model for the longitudinal control layer, so as to obtain the target vehicle speed at the next control moment. The longitudinal acceleration determination unit 205 is used to input the current vehicle speed, the current front wheel steering angle, and the target vehicle speed into the constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, so as to obtain the longitudinal acceleration. This allows the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration, thereby realizing adaptive hybrid path tracking of the unmanned vehicle in complex curvature road scenarios.

[0067] Therefore, the tracking and control device includes: inputting the obtained system state information into the constructed planar kinematics model of the unmanned vehicle; obtaining in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment based on the set reference path; for the lateral control layer, inputting the front wheel angle of the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle of the previous control moment, and the obtained deviation information into the angle optimization model to obtain the candidate front wheel angle of the next control moment that minimizes the cost of predicting that the vehicle's trajectory will be smooth in the next control moment; and inputting the candidate front wheel angle and the obtained system state information of the current control moment into the front wheel angle correction model. The system outputs a residual correction value for the candidate front wheel steering angle through a reinforcement learning network. The candidate front wheel steering angle and the residual correction value are then superimposed to generate the target forward steering angle for the next control moment in driving the autonomous vehicle's lateral path tracking. For the longitudinal control layer, the curvature information of the current control cycle and the target forward steering angle for the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed for the next control moment. The current vehicle speed and the target vehicle speed are input into the constructed adaptive speed adjustment relationship model to obtain the longitudinal acceleration. This allows the steering mechanism and drive actuator of the autonomous vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and longitudinal acceleration, achieving adaptive hybrid path tracking of the autonomous vehicle in complex curvature road scenarios. It is evident that this embodiment comprehensively considers the coupling relationship between road curvature, steering state, and longitudinal vehicle speed, enabling the longitudinal speed to adaptively adjust according to road geometry and steering actions in high-curvature curves, high-speed lane changes, or low-adhesion conditions. This improves the safety of path tracking control, enhancing the tracking accuracy and steering smoothness of the autonomous vehicle under complex curvature road conditions.

[0068] Thirdly, this application also provides an electronic device. From a hardware perspective, a hardware architecture diagram can be found in [reference needed]. Figure 5 As shown, it includes a machine-readable storage medium and a processor, wherein: the machine-readable storage medium stores machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the tracking control operation disclosed in the above example.

[0069] The machine-readable storage medium provided in this application embodiment stores machine-executable instructions. When the machine-executable instructions are invoked and executed by a processor, the machine-executable instructions cause the processor to perform the tracking control operation disclosed in the above example.

[0070] Here, a machine-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, a machine-readable storage medium can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0071] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera, telephone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0072] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0073] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] Furthermore, these computer program instructions can also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An adaptive hybrid path tracking control method, characterized in that, The tracking control method is applied to the vehicle control system of an unmanned vehicle, and the tracking control method includes: Obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, obtain in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment. For the lateral control layer, the front wheel angle at the previous control moment, multiple candidate front wheel angles generated in the neighborhood based on the front wheel angle at the previous control moment, and the obtained deviation information are input to characterize the angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel angle within a preset finite prediction time domain. This model obtains the candidate front wheel angle at the next control moment that minimizes the cost of predicting that the vehicle will travel smoothly in the next control moment. The candidate front wheel steering angle and the system state information obtained at the current control moment are input into the front wheel steering angle correction model. The angle residual correction amount of the candidate front wheel steering angle is output through the reinforcement learning network. The candidate front wheel steering angle and the residual correction amount are superimposed to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path. For the longitudinal control layer, the curvature information of the current control cycle and the target forward steering angle at the next control moment are input into the target vehicle speed planning model to obtain the target vehicle speed at the next control moment. The current vehicle speed, current front wheel angle, and target vehicle speed are input into an adaptive speed adjustment relationship model that enables coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action. This model yields longitudinal acceleration, which is then used to control the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration. This enables the unmanned vehicle to perform adaptive hybrid path tracking in complex curvature road scenarios.

2. The tracking control method according to claim 1, characterized in that, The system status information includes: the position of the vehicle's center of gravity in the ground coordinate system, heading angle, longitudinal velocity, front wheel steering angle, and longitudinal acceleration, with the front wheel steering angle and longitudinal acceleration selected as the control inputs.

3. The tracking control method according to claim 2, characterized in that, The process of inputting the system state information into the constructed planar kinematic model of the unmanned vehicle, and obtaining in real time the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment based on the set reference path, includes: The vehicle's center of mass position in the ground coordinate system, heading angle, longitudinal velocity, front wheel steering angle, and longitudinal acceleration are determined by the front wheel steering angle and longitudinal acceleration as control inputs. These inputs are then fed into the first and second expressions below to obtain the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment. The first expression is: ; in, and These represent the current control time of the vehicle's center of gravity in the ground coordinate system. t and the next control moment t +1 position coordinate; and These represent the current control time of the vehicle's center of gravity in the ground coordinate system. t and the next control moment t +1 position coordinate; and These refer to the vehicle at the current control time and the next control time, respectively. t +1 heading angle; and Each represents the vehicle at the current moment of control. t And the longitudinal velocity at the next control moment; The front wheel steering angle at the current control moment; Current control time t longitudinal acceleration; Current control time t The vehicle's center of gravity sideslip angle; To control the sampling period; The wheelbase of the vehicle, and ; This is the distance from the front axle to the vehicle's center of gravity. This is the distance from the rear axle to the vehicle's center of gravity. The current control moment is determined by the front wheel steering angle. A defined vehicle curvature; Based on the set reference path and the curvature information of the reference path points, the lateral deviation, heading deviation, and prediction deviation of the unmanned vehicle relative to the reference path are obtained after solving the second expression. The second expression is: , , ; in, This is the lateral deviation. For heading deviation, To predict bias, This is the position for the next control time point. This is the sequence number of the next control time point in the discrete reference path; For the current vehicle location point Reference path nearest point and adjacent path points The semi-perimeter of the triangle , The nearest path point from vehicle location N to the reference path. distance, , For vehicle location point N, find the nearest path point. Adjacent next path point distance, , Indicates reference path point and The chord length between , The next control time point Reference path curvature at that point Current vehicle location N To the next control time point P distance, The closest point on the reference path K To the next control time point P Path direction distance, Let yaw rate be the vehicle's angular velocity. For longitudinal velocity, The current vehicle heading angle, n The closest point on the reference path The serial number, and These are the current vehicle locations. The horizontal and vertical position coordinates, and The vehicle position points at the next control time point are respectively P The horizontal and vertical position coordinates.

4. The tracking control method according to claim 3, characterized in that, The method of inputting the front wheel steering angle of the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood based on the front wheel steering angle of the previous control moment, and the obtained deviation information into the steering angle optimization model used to characterize the optimization evaluation of each candidate motion path corresponding to each candidate front wheel steering angle within a preset finite prediction time domain, to obtain the candidate front wheel steering angle of the next control moment that minimizes the cost of predicting that the vehicle's trajectory will be in a smooth state in the next control moment, includes: For the front wheel steering angle at the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood are determined; The front wheel angle of the previous control moment, multiple candidate front wheel angles, and the obtained deviation information are input into the third expression. Based on the planar kinematics model of the unmanned vehicle, the trajectory of the vehicle in the preset finite prediction time domain is predicted, and the candidate front wheel angle of the next control moment that minimizes the cost of the vehicle's trajectory driving in the next control moment is output. The third expression is: , ; in, For the lateral deviation weighting coefficient, For heading deviation weighting coefficient, For the prediction bias weighting coefficient, This is the corner smoothness weighting coefficient. The front wheel steering angle at the current control moment. For the previous control moment t -1 corresponds to the front wheel steering angle. For the first k One candidate corner, k Candidate corner number, The candidate front wheel steering angle for the next control time that minimizes the cost of predicting the vehicle's trajectory and ensuring a smooth driving state in the next control time. For the first The comprehensive cost function corresponding to each candidate front wheel steering angle Current control time The generated set of candidate front wheel steering angles.

5. The tracking control method according to claim 4, characterized in that, The step of inputting the candidate front wheel steering angle and the obtained system state information at the current control moment into the front wheel steering angle correction model, and outputting the steering angle residual correction amount for the candidate front wheel steering angle through a reinforcement learning network, includes: The candidate front wheel steering angle and the system state information obtained at the current control moment are input into the fourth expression to obtain the steering angle residual correction amount for the candidate front wheel steering angle; The fourth expression is: ; in, This is a model for correcting the front wheel steering angle. For state space, , This refers to the reinforcement learning state information corresponding to the current control moment; For the action space, , Current control time The control action vector, , For the maximum longitudinal acceleration, For maximum braking deceleration, This is the maximum permissible steering angle for the front wheels. For longitudinal acceleration control, This is the state transition function. , The current status of the vehicle. The next state for the vehicle. To assign values ​​to actions in the action space. For the overall reward function, , This is a reward item for lateral deviation. , Current control time t Lateral deviation, For heading deviation bonus items, , Current control time t The course deviation, Prediction bias reward items , Current control time t Prediction bias, For obstacle avoidance reward information, , The weighting coefficient for obstacle avoidance rewards. To guide the tracking of reward information, , To guide the weighting coefficient of the tracking reward, This is the front wheel angle residual correction amount output by the front wheel angle correction model determined using the SAC algorithm at the current control moment; For speed tracking reward information, , The weighting coefficient of the velocity tracking excitation term. As a discount factor, Indicates the current control time. The actual longitudinal speed of the vehicle, Indicates the current control time. The corresponding target speed.

6. The tracking control method according to claim 5, characterized in that, The current vehicle speed, current front wheel steering angle, and target vehicle speed are input into an already constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, to obtain longitudinal acceleration, including: Input the current vehicle speed and the target vehicle speed into the fifth expression to obtain the longitudinal acceleration; The fifth expression is: , ； in, For curvature feedback gain, For corner feedback gain, As the reference curvature of the current road, This is a function that cuts off the upper and lower limits. The target vehicle speed lower limit, To achieve the target vehicle speed limit, The upper limit of safe speed is obtained based on the front wheel steering angle constraint. , The target front wheel steering angle is output by the lateral control layer. The maximum permissible lateral acceleration for the vehicle. This refers to the vehicle's wheelbase. The upper limit of safe speed is obtained based on road curvature constraints. , To prevent the denominator from becoming singular when the curvature approaches zero, To prevent the denominator from being a singular positive variable when the front wheel steering angle approaches zero, The target vehicle speed is determined by geometric feedback. This is the maximum permissible steering angle for the front wheels. The target vehicle speed is output by the longitudinal control layer.

7. The tracking control method according to claim 4, characterized in that, The current vehicle speed, current front wheel steering angle, and target vehicle speed are input into an already constructed adaptive speed adjustment relationship model for achieving coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, to obtain longitudinal acceleration, including: Input the current vehicle speed, the current front wheel angle, and the target vehicle speed into the sixth expression to obtain the longitudinal acceleration; The sixth expression is: ; in, This is the normalized longitudinal control adjustment coefficient. It is an adjustable scaling factor. The maximum speed limit allowed for vehicles. The current front wheel steering angle, The maximum front wheel steering angle, The maximum longitudinal acceleration of the vehicle. This is the longitudinal acceleration signal output to the drive / brake actuator at the current control moment. The target speed.

8. The tracking control method according to claim 4, characterized in that, The step of controlling the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward turning angle and the longitudinal acceleration includes: The steering actuator of the unmanned vehicle is controlled by the target front wheel angle, which serves as a lateral control signal, to drive the front wheels to rotate in order to correct lateral and heading deviations. The longitudinal acceleration signal, which serves as the lateral control signal, is converted into the throttle opening of the unmanned vehicle. Based on the throttle opening, the drive and braking actuators of the unmanned vehicle are controlled to control the longitudinal movement of the vehicle, so as to follow the target vehicle speed and drive the unmanned vehicle to travel along the reference path. The updated vehicle status information is fed back in real time to form a closed-loop path tracking control that integrates horizontal and vertical directions.

9. An adaptive hybrid path tracking control device, characterized in that, The tracking control device is applied to the vehicle control system of the unmanned vehicle, and the tracking control device includes: The deviation curvature information determination unit is used to obtain the system state information set for the unmanned vehicle at the current control moment, and input the system state information into the constructed planar kinematic model of the unmanned vehicle. Based on the set reference path, it obtains the deviation information of the unmanned vehicle relative to the reference path and the curvature information of the reference path points at the current control moment in real time. The front wheel steering angle determination unit is used for the lateral control layer to input the front wheel steering angle of the previous control moment, multiple candidate front wheel steering angles generated in the neighborhood based on the front wheel steering angle of the previous control moment, and the obtained deviation information to characterize the candidate front wheel steering angle of the next control moment that minimizes the cost of predicting that the vehicle will travel in a smooth state in the next control moment in the steering angle optimization model that optimizes and evaluates each candidate motion path corresponding to each candidate front wheel steering angle within a preset finite prediction time domain. The target forward steering angle determination unit is used to input the candidate front wheel steering angle and the system state information obtained at the current control moment into the front wheel steering angle correction model, output the steering angle residual correction amount of the candidate front wheel steering angle through the reinforcement learning network, and superimpose the candidate front wheel steering angle and the residual correction amount to generate the target forward steering angle of the next control moment for driving the unmanned vehicle to track the lateral path. The target vehicle speed determination unit is used to input the curvature information of the current control cycle and the target forward steering angle at the next control moment into the target vehicle speed planning model for the longitudinal control layer, so as to obtain the target vehicle speed at the next control moment. The longitudinal acceleration determination unit is used to input the current vehicle speed, the current front wheel steering angle, and the target vehicle speed into an adaptive speed adjustment relationship model that has been constructed to achieve coordinated adaptive adjustment of vehicle speed to road geometry and lateral steering action, so as to obtain the longitudinal acceleration. This allows the steering mechanism and drive actuator of the unmanned vehicle to form an integrated lateral and longitudinal closed-loop path tracking control based on the target forward steering angle and the longitudinal acceleration, thereby realizing adaptive hybrid path tracking of the unmanned vehicle in complex curvature road scenarios.

10. An electronic device, characterized in that, The electronic device includes a readable storage medium and a processor; wherein the readable storage medium is used to store machine-executable instructions; and the processor is used to read the machine-executable instructions on the readable storage medium and execute the instructions to implement the steps of the tracking control method according to any one of claims 1-8.

Images