A driving parameter identification model training method, a driving control method and device

Abstract

The application discloses a training method of a driving parameter identification model, a driving control method and device, and the method comprises the following steps: performing driving control parameter identification based on current deviation data and an initial driving parameter identification model to obtain initial driving control parameters and predicted driving control parameters; determining current return data based on the current deviation data and current position data, determining updated deviation data, thereby determining a current confidence index and a predicted confidence index, calculating a target confidence index according to the current return data and the predicted confidence index, training the initial driving parameter identification model based on the current confidence index and the target confidence index to obtain a target driving parameter identification model. The embodiment of the application can realize the rapid identification and calibration of the driving control parameters, and ensure the safety and stability of the automatic driving control.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a training method for a driving parameter recognition model, a driving control method, and a device. Background Technology

[0002] Autonomous driving optimization control is one of the core functions of vehicle automatic control systems. In actual operation, due to various external factors, the vehicle's actual trajectory is difficult to perfectly match the desired ideal curve. Therefore, the main objective of the control algorithm in an autonomous driving system is to minimize the error. Thus, the control algorithm is also a key technology for ensuring safe, stable, and energy-efficient vehicle operation. The control parameters of the driving control algorithm are calibrated quantities. Real-vehicle debugging is required during the algorithm development phase to determine a set of optimal control parameters for subsequent real-vehicle operation.

[0003] In existing technologies, the calibration of driving control parameters mainly relies on manual, repeated adjustments based on human experience. This often requires a large amount of repetitive work and parameter trials, is time-consuming, and rarely achieves the desired results. The problem of how to quickly calibrate driving control parameters while ensuring the safety, stability, and comfort of autonomous driving control urgently needs to be solved. Summary of the Invention

[0004] To address the aforementioned problems in the prior art, this invention discloses a training method for a driving parameter recognition model, a driving control method, and a device, which can achieve rapid calibration of driving control parameters while ensuring the safety and stability of autonomous driving control. The technical solution disclosed in this invention is as follows:

[0005] According to one aspect of the embodiments disclosed in this invention, a method for training a driving parameter recognition model is provided, comprising:

[0006] Obtain the current location data and current deviation data of the target vehicle, wherein the current deviation data is the deviation between the current location data and the first preset location data;

[0007] Based on the current deviation data and the initial driving parameter identification model, driving control parameters are identified to obtain initial driving control parameters and predicted driving control parameters.

[0008] The current return data is determined based on the current deviation data and the current position data;

[0009] Determine the updated deviation data corresponding to the initial driving control parameters;

[0010] The initial driving control parameters, the current deviation data, the predicted driving control parameters, and the updated deviation data are input into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index.

[0011] The target confidence index is calculated based on the current return data and the predicted confidence index.

[0012] The initial driving parameter recognition model is trained based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model.

[0013] According to another aspect of the disclosed embodiments of the present invention, a driving control method is provided, comprising:

[0014] Obtain the target position deviation data of the target vehicle;

[0015] The target position deviation data is input into the driving parameter recognition model to identify driving control parameters, thereby obtaining the target driving control parameters. The driving parameter recognition model is the target driving parameter recognition model trained using the driving parameter recognition model training method described above.

[0016] The target vehicle is controlled based on the target driving control parameters.

[0017] According to another aspect of the embodiments disclosed in this invention, a training apparatus for a driving parameter recognition model is provided, comprising:

[0018] The first acquisition module is used to acquire the current position data and current deviation data of the target vehicle, wherein the current deviation data is the deviation between the current position data and the first preset position data;

[0019] The first driving control parameter identification module is used to identify driving control parameters based on the current deviation data and the initial driving parameter identification model, so as to obtain the initial driving control parameters and the predicted driving control parameters.

[0020] The report data determination module is used to determine the current report data based on the current deviation data and the current position data;

[0021] The update deviation data determination module is used to determine the update deviation data corresponding to the initial driving control parameters;

[0022] The first confidence index determination module is used to input the initial driving control parameters, the current deviation data, the predicted driving control parameters and the updated deviation data into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index.

[0023] The second confidence index determination module is used to calculate the target confidence index based on the current return data and the predicted confidence index;

[0024] The training module is used to train the initial driving parameter recognition model based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model.

[0025] According to another aspect of the disclosed embodiments of the present invention, a driving control device is provided, comprising:

[0026] The second acquisition module is used to acquire the target position deviation data of the target vehicle.

[0027] The second driving control parameter identification module is used to input the target position deviation data into the driving parameter identification model to identify the driving control parameters and obtain the target driving control parameters. The driving parameter identification model is the target driving parameter identification model trained using the driving parameter identification model training method described above.

[0028] A control module is used to control the target vehicle based on the target driving control parameters.

[0029] According to another aspect of the disclosed embodiments of the present invention, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the training method of the driving parameter recognition model or the driving control method as described above.

[0030] According to another aspect of the disclosed embodiments of the present invention, a computer-readable storage medium is provided, wherein when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the training method for the driving parameter recognition model or the driving control method as described above.

[0031] According to another aspect of the disclosed embodiments of the present invention, a computer program product containing instructions is provided, which, when run on a computer, causes the computer to execute the training method for the driving parameter recognition model or the driving control method as described above.

[0032] The technical solutions provided by the embodiments disclosed in this invention bring at least the following beneficial effects:

[0033] The driving parameter recognition model training method provided by this invention identifies driving control parameters based on current deviation data and an initial driving parameter recognition model, obtaining initial driving control parameters and predicted driving control parameters. Based on current deviation data and current position data, it determines current return data and update deviation data, thereby determining current confidence index and predicted confidence index. Based on current return data and predicted confidence index, it calculates target confidence index, and then trains the initial driving parameter recognition model to obtain target driving parameter recognition model, realizing rapid identification and calibration of driving control parameters while ensuring safety and stability in vehicle driving control.

[0034] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0035] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the disclosure of this invention and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit the scope of this disclosure.

[0036] Figure 1 This is a flowchart illustrating a training method for a driving parameter recognition model according to an exemplary embodiment;

[0037] Figure 2 This is a flowchart illustrating a method for training a driving parameter recognition model based on a confidence index, according to an exemplary embodiment.

[0038] Figure 3 This is a flowchart illustrating a driving control method according to an exemplary embodiment;

[0039] Figure 4 This is a flowchart illustrating a method for driving control based on driving control parameters according to an exemplary embodiment;

[0040] Figure 5 This is a block diagram of a training device for a driving parameter recognition model according to an exemplary embodiment;

[0041] Figure 6 This is a block diagram of a driving control device according to an exemplary embodiment;

[0042] Figure 7 This is a block diagram illustrating a terminal electronic device for training a driving parameter recognition model or for driving control, according to an exemplary embodiment.

[0043] Figure 8 This is a block diagram illustrating a server electronic device for training a driving parameter recognition model or for driving control, according to an exemplary embodiment. Detailed Implementation

[0044] To enable those skilled in the art to better understand the technical solutions disclosed in this invention, the technical solutions in the disclosed embodiments will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0045] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention disclosed herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0046] In practical applications, vehicles with autonomous driving capabilities include an autonomous driving control module, which contains a control algorithm. The control algorithm includes driving control parameters, which are calibrated values.

[0047] When calibrating the control parameters in the control algorithm used by the control module of an autonomous vehicle, it is first necessary to establish a vehicle dynamics error model. The vehicle dynamics error model can include a vehicle longitudinal dynamics error model, a vehicle lateral dynamics error model, and a vehicle lateral-longitudinal coupling dynamics error model. The following explanation uses the vehicle lateral-longitudinal coupling dynamics error model as an example.

[0048] Specifically, based on the vehicle dynamics model, the following lateral and longitudinal coupling dynamics error model for autonomous vehicles can be obtained:

[0049]

[0050]

[0051] Among them, e l This represents the lateral positional deviation, which is the lateral distance between the vehicle's center of gravity and the desired position of a point on the planned driving trajectory. This represents the lateral velocity deviation, which is the difference between the vehicle's lateral velocity and the expected lateral velocity of a point on the planned travel trajectory, e.h This represents the heading angle deviation, which is the difference between the vehicle's heading angle and the expected heading angle of a point on the planned driving trajectory. This represents the deviation in heading angular velocity, which is the difference between the vehicle's heading angular velocity and the expected heading angular velocity of a point on the planned travel trajectory, e. st This represents the longitudinal position deviation, specifically the longitudinal distance between the vehicle's center of gravity and the desired position of a point on the planned driving trajectory, e. sp The longitudinal velocity deviation is the difference between the vehicle's longitudinal velocity and the expected longitudinal velocity of a point on the planned trajectory. V represents the longitudinal velocity, and δ represents the longitudinal velocity. f This represents the steering wheel angle, and 'a' represents the longitudinal acceleration. C represents the heading angle. f and C r These represent the lateral stiffness of the front and rear wheels, respectively. f and l r I represents the front overhang length and the rear overhang length, respectively. z Let m be the moment of inertia of the vehicle about its center of mass, and m be the mass of the vehicle.

[0052] After obtaining the vehicle dynamics error model for the autonomous driving control algorithm, it is also necessary to design the autonomous vehicle control algorithm. The principle is to solve the driving control quantity by the matrix composed of the vehicle position deviation, and then apply the driving control quantity to the vehicle control system to form a closed-loop feedback control, gradually reducing the deviation between the actual trajectory and the planned trajectory of the vehicle.

[0053] Specifically, control algorithms can include PID (Proportion Integral Differential) algorithm, LQR (Linear Quadratic Regulator) algorithm, MPC (Model Predictive Control) algorithm, and SMC (Sliding Mode Control) algorithm. Since the control principles of each control algorithm are similar, the MPC algorithm will be used as an example for explanation below.

[0054] Based on the MPC algorithm, the system state equations, after discretization and simplification of the above horizontal and vertical coupled dynamic error model, yield the following equation:

[0055] x(k+1)=Ax(k)+Bu(k)+C

[0056] y(k)=Dx(k)

[0057] Where x(k+1) represents the deviation data at time k+1, x(k) represents the deviation data at time k, u(k) represents the control quantity at time k, y(k) represents the system output quantity at time k, A represents the system matrix, B represents the control matrix, C represents the disturbance matrix, and D represents the observation matrix.

[0058] The optimization problem of model predictive control can be transformed into a quadratic programming problem. We design an optimization objective function to minimize this objective function, which is expressed as follows:

[0059]

[0060] Where J represents the optimization objective value, Q represents the state weight matrix, R represents the control weight matrix, M represents the prediction step size, x represents all deviations within the prediction step size, and u represents all control variables within the prediction step size.

[0061] Specifically, the state weight matrix is ​​a 6×6 diagonal matrix with 6 parameters q1, q2, q3, q4, q5, and q6, where q1-q6 represent the deviation data e in the aforementioned horizontal and vertical coupled dynamic error model. l , e h , e st and e sp The corresponding weights, the control weight matrix is ​​a 2×2 diagonal matrix with two parameters r1 and r2, where r1 and r2 represent the control quantity δ in the above-mentioned horizontal and vertical coupled dynamic error model. f The weights corresponding to 'a'.

[0062] For the MPC algorithm, q1-q6 and r1, r2 are the driving control parameters that need to be calibrated.

[0063] Solving the quadratic programming problem yields a set of optimal control sequences. The first term in the control sequence is output as the driving control variable at the current moment and applied to the control system. At the next moment, the deviation matrix is ​​updated, and the process of solving the quadratic programming problem is repeated to continuously control the vehicle's lateral and longitudinal coupling control system.

[0064] This invention provides a training method for a driving parameter recognition model, which can be applied to the identification and calibration of control parameters in the control algorithm of an autonomous driving module in a vehicle.

[0065] Figure 1 This is a flowchart illustrating a training method for a driving parameter recognition model according to an exemplary embodiment, such as... Figure 1 As shown, the training method for this driving parameter recognition model includes the following steps.

[0066] S101: Obtain the current location data and current deviation data of the target vehicle.

[0067] In one specific embodiment, the target vehicle can be the vehicle to be identified and calibrated for driving control parameters. The current position data can be the position data of the target vehicle at the current moment. Specifically, the position data can include data such as longitudinal position, longitudinal velocity, longitudinal acceleration, heading angle, lateral position, lateral velocity, and heading angular velocity. The current deviation data can be the deviation between the current position data and the first preset position data. Correspondingly, the current deviation data can include longitudinal position deviation data, longitudinal velocity deviation data, longitudinal acceleration deviation data, heading angle deviation data, lateral position deviation data, lateral velocity deviation data, and heading angular velocity deviation data.

[0068] In one specific embodiment, the first preset position data can be the planned position data corresponding to the current position data in the planned driving trajectory data of the target vehicle. Specifically, the planned driving trajectory data can be generated in a preset simulation environment system based on preset start-point information, preset end-point information and surrounding obstacle information of the target vehicle. The planned driving trajectory data can include the planned position data of multiple trajectory points. Specifically, the preset start-point information and preset end-point information can be set according to the actual application situation. In the parking control process of the vehicle, the preset end-point information can be parking space information.

[0069] In one specific embodiment, the aforementioned current location data, parking space information, and surrounding obstacle information can be obtained by the vehicle through its pre-installed high-precision map, positioning sensors, surround-view cameras, ultrasonic radar, and millimeter-wave radar.

[0070] S103: Based on the current deviation data and the initial driving parameter identification model, driving control parameters are identified to obtain initial driving control parameters and predicted driving control parameters.

[0071] In one specific embodiment, the initial driving parameter recognition model can be established based on reinforcement learning algorithms, such as Q-learning, DQN (Deep Q-Network), PG (Policy Gradient) and DDPG (Deep Deterministic Policy Gradient) algorithms. The DDPG algorithm will be used as an example for explanation below.

[0072] In one specific embodiment, the initial driving parameter recognition model may include a first driving parameter recognition model, a second driving parameter recognition model, and an evaluation model. Specifically, the evaluation model may include a first evaluation model and a second evaluation model. The second driving parameter recognition model may be the target model corresponding to the first driving parameter recognition model, and the second evaluation model may be the target model corresponding to the first evaluation model. The initial driving parameter recognition model identifies driving control parameters through the first driving parameter recognition model. The first evaluation model can be used to evaluate the recognition performance of the first driving parameter recognition model. The second driving parameter recognition model is used to predict the driving control parameter recognition result at the current time and the second evaluation model can be used to evaluate the recognition performance of the second driving parameter recognition model.

[0073] Specifically, the first driving parameter recognition model and the second driving parameter recognition model can have the same structure, and the first evaluation model and the second evaluation model can have the same structure. The first driving parameter recognition model, the second driving parameter recognition model, the first evaluation model, and the second evaluation model can each include their corresponding parameters. In the initial state, the first driving parameter recognition model and the second driving parameter recognition model can have the same parameters, and the first evaluation model and the second evaluation model can have the same parameters. During the training process, the parameters corresponding to the above four models can be continuously updated.

[0074] In an optional embodiment, the step of identifying driving control parameters based on the current deviation data and the initial driving parameter identification model to obtain initial driving control parameters and predicted driving control parameters may include:

[0075] The current deviation data is input into the first driving parameter recognition model to obtain the initial driving control parameters;

[0076] The updated deviation data is input into the second driving parameter recognition model to obtain the predicted driving control parameters.

[0077] In one specific embodiment, the initial driving control parameters can be the driving control parameters corresponding to the current deviation data, and the predicted driving control parameters can be obtained by performing driving control parameter identification on the updated deviation data based on the parameters corresponding to the second driving parameter identification model at the current moment.

[0078] In one specific embodiment, the updated deviation data can be the deviation between the updated position data of the target vehicle and the second preset position data. The updated position data can be the position data of the target vehicle when the target vehicle is controlled based on the initial driving control parameters, that is, the position data of the target vehicle at the next moment. Specifically, the updated position data can be the position data of the target vehicle at the next moment obtained by the target vehicle driving based on the driving control quantity output by the autonomous driving module of the target vehicle according to the initial driving control parameters when the target vehicle is controlled based on the initial driving control parameters. The second preset position data can be the planned position data corresponding to the updated position data in the planned driving trajectory data of the target vehicle.

[0079] S105: Determine the current return data based on the current deviation data and the current position data.

[0080] In one specific embodiment, the current report data can be the real-time report data obtained when the initial driving control parameters are obtained based on the current deviation data.

[0081] In an optional embodiment, the current location data and the determination of the current return data based on the current location data may include:

[0082] The current report data is determined based on the current deviation data and the current longitudinal velocity, current longitudinal jerk, and current heading angular acceleration in the current position data.

[0083] In one specific embodiment, the current longitudinal jerk and the current heading angular acceleration may be negatively correlated with the current reported data, while the current longitudinal velocity may be positively correlated with the current reported data.

[0084] Specifically, determining the current report data based on the current deviation data and the current longitudinal velocity, current longitudinal jerk, and current heading angular acceleration in the current position data may include combining the following formula:

[0085]

[0086] Where, r t This represents the current return data, v t Represents the current longitudinal velocity, j t The current longitudinal jerk is represented by k. j α represents the first preset adjustment coefficient corresponding to the current longitudinal acceleration. t k represents the current heading angular acceleration. α The second preset adjustment coefficient, Δ, represents the current heading angular acceleration. x k represents the current deviation data. ΔThis represents the third preset adjustment coefficient corresponding to the current deviation data.

[0087] In one specific embodiment, the first preset adjustment coefficient can characterize the sensitivity of the current reported data to the current longitudinal jerk, the second preset adjustment coefficient can characterize the sensitivity of the current reported data to the current heading angular acceleration, and the third preset adjustment coefficient can characterize the sensitivity of the current reported data to the current deviation data. Specifically, the first preset adjustment coefficient, the second preset adjustment coefficient, and the third preset adjustment coefficient can be set according to the actual application.

[0088] S107: Determine the updated deviation data corresponding to the initial driving control parameters.

[0089] In one specific embodiment, the current deviation data, initial driving control parameters, current report data, and updated deviation data are treated as a single sample data set. This sample data is stored in a pre-established experience pool, which can store a number of sample data sets for training the initial driving parameter recognition model. When the capacity of the experience pool reaches a preset upper limit threshold, for each new sample data set added, a historical sample data set can be deleted; specifically, the oldest historical sample data set can be deleted.

[0090] During each training session, samples are randomly drawn from the experience pool for training. This eliminates the correlation between the selected samples, resulting in a more accurate model.

[0091] S109: Input the initial driving control parameters, the current deviation data, the predicted driving control parameters, and the updated deviation data into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index.

[0092] In one specific embodiment, the current confidence index can characterize the current recognition performance of the driving control parameters corresponding to the first driving parameter recognition model in the initial driving parameter recognition model, and the prediction confidence index can characterize the recognition performance of the driving control parameters corresponding to the second driving parameter recognition model in the initial driving parameter recognition model.

[0093] In an optional embodiment, inputting the initial driving control parameters, the current deviation data, the predicted driving control parameters, and the updated deviation data into the evaluation model of the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index may include:

[0094] The initial driving control parameters and the current deviation data are input into the first evaluation model in the initial driving parameter identification model to obtain the current confidence index.

[0095] The predicted driving control parameters and the updated deviation data are input into the second evaluation model in the initial driving parameter identification model to obtain the prediction confidence index.

[0096] S111: Calculate the target confidence index based on the current return data and the predicted confidence index.

[0097] In one specific embodiment, the target confidence index can characterize the expected recognition performance of the driving control parameters corresponding to the first driving parameter recognition model.

[0098] Specifically, the target confidence index can be determined using the following formula:

[0099] y t =r t +γQ′(s t+1 ,μ′(s t+1 |θ′)|ω′), γ∈(0,1)

[0100] Among them, y t r represents the target confidence index at the current moment. t Represents the current return data, Q′(s) t+1 ,μ′(s t+1 |θ′)|ω′) represents the prediction confidence index mentioned above, ω′ represents the second evaluation parameter of the second evaluation model, and μ′(s) t+1 |θ′) represents the above predictive control parameter, s t+1 This indicates the update of the deviation data, θ′ represents the second identification parameter of the second driving parameter identification model, and γ represents the discount factor.

[0101] S113: The initial driving parameter recognition model is trained based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model.

[0102] In one specific embodiment, the parameters of the first evaluation model in the initial driving parameter identification model are updated based on the error between the current confidence index and the target confidence index, and the parameters of the first driving parameter identification model in the initial driving parameter identification model are updated based on the current confidence index, thereby updating the parameters of the second driving parameter identification model and the second evaluation model, to obtain the target driving parameter identification model corresponding to the first driving parameter identification model.

[0103] In the above embodiments, the model training direction is to obtain the maximum reward data. By designing the reward function, the longitudinal acceleration and yaw acceleration terms of the vehicle are added, and the longitudinal acceleration and yaw acceleration are negatively correlated with the reward data, thereby improving the stability and comfort of vehicle driving control. In addition, the longitudinal speed of the vehicle is added to the reward function, and the longitudinal speed is positively correlated with the reward data, which can improve driving control efficiency.

[0104] In an optional embodiment, Figure 2 This is a flowchart illustrating a method for training a driving parameter recognition model based on a confidence index, according to an exemplary embodiment. Figure 2 As shown, training the initial driving parameter recognition model based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model may include:

[0105] S201: Determine the first loss information based on the target confidence index and the current confidence index;

[0106] In one specific embodiment, the first loss information can be determined based on the error between the current confidence index and the target confidence index. Specifically, the error can be the mean square error. The mean square error loss function between the current confidence index and the target confidence index is as follows:

[0107]

[0108] Where L(ω) represents the first loss information mentioned above, N represents the number of training samples, and y t Q(s) represents the target confidence index. t ,a t |ω) represents the current confidence index, where ω represents the first evaluation parameter of the first evaluation model, and s t This represents the current deviation data, a t This represents the initial driving control parameters, where a t =μ(s) t |θ), where θ represents the first recognition parameter of the first driving parameter recognition model.

[0109] S203: Update the first evaluation parameter of the first evaluation model based on the first loss information;

[0110] In one specific embodiment, the first evaluation parameters of the first evaluation model can be updated through backpropagation based on the first loss information.

[0111] S205: Determine the second loss information based on the current confidence index.

[0112] In one specific embodiment, the second loss information can be determined according to the following formula:

[0113]

[0114] Where J(θ) represents the aforementioned second loss information.

[0115] S207: Update the first recognition parameter of the first driving parameter recognition model based on the second loss information.

[0116] In one specific embodiment, the first identification parameters of the first driving parameter identification model can be updated through backpropagation based on the second loss information.

[0117] S209: Based on the first evaluation parameters, update the second evaluation parameters of the second evaluation model.

[0118] In one specific embodiment, the target confidence index can be determined through current return data and a predicted confidence index. The predicted confidence index can be obtained through a second evaluation model and a second driving parameter identification model. The parameters of the second evaluation model and the second driving parameter identification model need to be updated to ensure the correctness of the target confidence index. Specifically, the second evaluation parameters can be updated using the following formula:

[0119] ω ′ =ω+(1-)ω ′

[0120] Where τ represents the update coefficient.

[0121] S211: Based on the first identification parameters, update the second identification parameters of the second driving parameter identification model.

[0122] In one specific embodiment, the second identification parameter can be updated using the following formula:

[0123] θ ′ =θ+(1-)θ ′

[0124] In the above embodiments, the second identification parameter and the second evaluation parameter are slowly slid through the above parameter update method, which can reduce the fluctuation of the target confidence index and enhance the stability of the training process.

[0125] S213: Use the updated location data as the current location data, and use the updated deviation data as the current deviation data.

[0126] S215: Based on the updated first driving parameter recognition model, second driving parameter recognition model, first evaluation model, second evaluation model, current position data, and current deviation data, repeat the step of identifying driving control parameters based on the current deviation data and the initial driving parameter recognition model to obtain initial driving control parameters and predicted driving control parameters, until the step of using the updated position data corresponding to the initial driving control parameters as the current position data and the updated deviation data as the current deviation data is met, until the preset training convergence condition is satisfied.

[0127] In one specific embodiment, the preset training convergence condition can be that the cumulative deviation data during the training process is less than a preset deviation threshold. Specifically, the cumulative deviation data can be the sum of deviation data at multiple time points, and the preset deviation threshold can be set according to the actual application.

[0128] S217: The first driving parameter recognition model that satisfies the preset training convergence condition is used as the target driving parameter recognition model.

[0129] In one specific embodiment, the driving parameter recognition model can be continuously updated as the current location data and current deviation data are updated and trained. The control parameters of the lateral and longitudinal coupling control algorithm for autonomous driving are continuously updated until the training convergence condition is met, achieving the optimal lateral and longitudinal coupling control effect.

[0130] As can be seen from the technical solutions provided in the embodiments of this specification above, this specification identifies driving control parameters based on current deviation data and an initial driving parameter identification model to obtain initial driving control parameters and predicted driving control parameters. It determines current reward data based on the current longitudinal velocity, current longitudinal jerk, and current yaw angular acceleration from the current deviation data and current position data, and determines updated deviation data, thereby determining the current confidence index and predicted confidence index. The target confidence index is calculated based on the current reward data and predicted confidence index, and then the initial driving parameter identification model is trained to obtain the target driving parameter identification model. This achieves rapid identification and calibration of driving control parameters while ensuring the safety and stability of autonomous driving control. Furthermore, the model training direction is to obtain the maximum reward data. By designing the reward function, the longitudinal jerk and yaw angular acceleration terms of the vehicle are added, and the longitudinal jerk and yaw angular acceleration are negatively correlated with the reward data, improving the comfort of vehicle driving control. In addition, adding the vehicle's longitudinal velocity to the reward function, and the longitudinal velocity being positively correlated with the reward data, can improve driving control efficiency. Furthermore, the vehicle's autonomous driving control module adopts a horizontal and vertical coupling control method, which simultaneously obtains optimized driving control parameters in both directions, thereby improving efficiency.

[0131] Based on the above-described training method for the driving parameter recognition model, the following describes an embodiment of a driving control method disclosed herein. Figure 3 This is a flowchart illustrating a driving control method according to an exemplary embodiment, such as... Figure 3 As shown, the driving control method includes the following steps.

[0132] S301: Obtain the target position deviation data of the target vehicle.

[0133] S303: Input the target position deviation data into the driving parameter recognition model to identify driving control parameters and obtain the target driving control parameters.

[0134] In one specific embodiment, the driving parameter recognition model can be the target driving parameter recognition model trained using the driving parameter recognition model training method described above.

[0135] S305: Control the target vehicle based on the target driving control parameters.

[0136] In an optional embodiment, Figure 4 This is a flowchart illustrating a method for driving control based on driving control parameters according to an exemplary embodiment, such as... Figure 4 As shown, controlling the target vehicle based on the target driving control parameters may include:

[0137] S401: Based on the target driving control parameters, construct an optimized objective function corresponding to the target driving control quantity.

[0138] S403: Based on the target constraints, the optimization objective function is solved to obtain the target driving control quantity.

[0139] In one specific embodiment, the driving control quantity may include longitudinal acceleration and steering wheel angle, and the constraint condition may be a preset value range corresponding to the driving control quantity. Specifically, the preset value range can be set according to the actual application situation.

[0140] S405: Control the target vehicle based on the target driving control quantity.

[0141] In an optional embodiment, obtaining the target position deviation data of the target vehicle may include:

[0142] The current location data and target driving trajectory data of the target vehicle are obtained, wherein the target driving trajectory data includes target location data of multiple trajectory points;

[0143] The deviation between the current location data and the target location data corresponding to the current location data is used as the target location deviation data.

[0144] In one specific embodiment, the target position data for each trajectory point may include data such as the target longitudinal position, target longitudinal velocity, target longitudinal acceleration, target heading angle, target lateral position, target lateral velocity, and target heading angular velocity for each trajectory point.

[0145] In the above embodiments, the target driving parameter recognition model can realize the rapid identification and calibration of autonomous driving control parameters, and based on the design of the reward function during model training, it can improve the stability, comfort and driving control efficiency of vehicle driving control.

[0146] Figure 5 This is a block diagram illustrating a calibration apparatus according to an exemplary embodiment. (Refer to...) Figure 5 The device includes:

[0147] The first acquisition module 510 is used to acquire the current position data and current deviation data of the target vehicle, wherein the current deviation data is the deviation between the current position data and the first preset position data;

[0148] The first driving control parameter identification module 520 is used to identify driving control parameters based on the current deviation data and the initial driving parameter identification model, and to obtain the initial driving control parameters and the predicted driving control parameters.

[0149] The report data determination module 530 is used to determine the current report data based on the current deviation data and the current position data;

[0150] The update deviation data determination module 540 is used to determine the update deviation data corresponding to the initial driving control parameters;

[0151] The first confidence index determination module 550 is used to input the initial driving control parameters, the current deviation data, the predicted driving control parameters and the updated deviation data into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index.

[0152] The second confidence index determination module 560 is used to calculate the target confidence index based on the current return data and the predicted confidence index;

[0153] The training module 570 is used to train the initial driving parameter recognition model based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model.

[0154] Optionally, the return data determination module 530 may include:

[0155] The report data determination unit is used to determine the current report data based on the current longitudinal velocity, current longitudinal jerk, and current heading angular acceleration in the current deviation data and the current position data. The current longitudinal jerk and the current heading angular acceleration are negatively correlated with the current report data, and the current longitudinal velocity is positively correlated with the current report data.

[0156] Optionally, the return data determination unit may include:

[0157] The return data determines the sub-unit, which is used in conjunction with the following formula:

[0158]

[0159] Where, r t This represents the current return data, v t Represents the current longitudinal velocity, j t The current longitudinal jerk is represented by k. j α represents the first preset adjustment coefficient corresponding to the current longitudinal acceleration. t k represents the current heading angular acceleration. α The second preset adjustment coefficient, Δ, represents the current heading angular acceleration. x k represents the current deviation data. Δ This represents the third preset adjustment coefficient corresponding to the current deviation data.

[0160] Optionally, the first driving control parameter recognition module 520 may include:

[0161] An initial driving control parameter determination unit is used to input the current deviation data into the first driving parameter recognition model to obtain the initial driving control parameters.

[0162] The predictive driving control parameter determination unit is used to input the updated deviation data into the second driving parameter recognition model to obtain the predictive driving control parameters.

[0163] Optionally, the evaluation model includes a first evaluation model and a second evaluation model, and the training module 570 may include:

[0164] The first loss information determination unit is used to determine first loss information based on the target confidence index and the current confidence index;

[0165] The first evaluation parameter update unit is used to update the first evaluation parameter of the first evaluation model based on the first loss information.

[0166] The second loss information determination unit is used to determine second loss information based on the current confidence index;

[0167] The first identification parameter update unit is used to update the first identification parameter of the first driving parameter identification model based on the second loss information.

[0168] The second evaluation parameter update unit is used to update the second evaluation parameter of the second evaluation model based on the first evaluation parameter.

[0169] The second identification parameter update unit is used to update the second identification parameter of the second driving parameter identification model based on the first identification parameter;

[0170] A location data update unit is used to use updated location data as the current location data and updated deviation data as the current deviation data. The updated location data is the location data of the target vehicle when the target vehicle is controlled based on the initial driving control parameters.

[0171] The training unit is used to repeat the steps of identifying driving control parameters based on the current deviation data and the initial driving parameter identification model, obtaining initial driving control parameters and predicted driving control parameters, based on the updated first driving parameter identification model, second driving parameter identification model, first evaluation model, second evaluation model, current position data, and current deviation data, until the steps of using the updated position data corresponding to the initial driving control parameters as the current position data and the updated deviation data as the current deviation data are met, until the preset training convergence condition is satisfied.

[0172] The target driving parameter recognition model determination unit is used to select the first driving parameter recognition model that meets the preset training convergence condition as the target driving parameter recognition model.

[0173] Figure 6 This is a block diagram illustrating a driving control device according to an exemplary embodiment. (Refer to...) Figure 6 The device includes:

[0174] The second acquisition module 610 is used to acquire the target position deviation data of the target vehicle.

[0175] The second driving control parameter identification module 620 is used to input the target position deviation data into the driving parameter identification model to identify the driving control parameters and obtain the target driving control parameters. The driving parameter identification model is the target driving parameter identification model trained using the training method of the driving parameter identification model in the embodiment of the present invention.

[0176] The control module 630 is used to control the target vehicle based on the target driving control parameters.

[0177] Optionally, the control module 630 may include:

[0178] An optimization objective function construction unit is used to construct an optimization objective function corresponding to the target driving control quantity based on the target driving control parameters.

[0179] The target driving control quantity determination unit is used to solve the optimization objective function based on the target constraints to obtain the target driving control quantity;

[0180] A control unit for controlling the target vehicle based on the target driving control quantity.

[0181] Optionally, the second acquisition module 610 may include:

[0182] A location data acquisition unit is used to acquire the current location data and target driving trajectory data of the target vehicle, wherein the target driving trajectory data includes target location data of multiple trajectory points;

[0183] The target position deviation data determination unit is used to take the deviation between the current position data and the target position data corresponding to the current position data as the target position deviation data.

[0184] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0185] Figure 7 This is a block diagram illustrating an electronic device for training a driving parameter recognition model or for driving control, according to an exemplary embodiment. The electronic device may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown, the electronic device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a training method for a driving parameter recognition model or a driving control method. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0186] Figure 8This is a block diagram illustrating an electronic device for training a driving parameter recognition model or for driving control, according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, the electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for training a driving parameter recognition model or a driving control method.

[0187] Those skilled in the art will understand that Figure 7 or Figure 8 The structures shown are merely block diagrams of some structures related to the disclosed solutions of this invention, and do not constitute a limitation on the electronic devices to which the disclosed solutions of this invention are applied. Specific electronic devices may include more or fewer components than those shown in the figures, or combine certain components, or have different component arrangements.

[0188] In an exemplary embodiment, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement a training method for a driving parameter recognition model or a driving control method as disclosed in the embodiments of the present invention.

[0189] In an exemplary embodiment, a computer-readable storage medium is also provided, wherein when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the training method for the driving parameter recognition model or the driving control method of the disclosed embodiments of the present invention.

[0190] In an exemplary embodiment, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the training method or driving control method of the driving parameter recognition model in the embodiments disclosed in this invention.

[0191] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), and double data rate RAM.

[0192] SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), Rambus Direct RAM (RDRAM), Direct Memory Bus Dynamic RAM (DRDRAM), and Memory Bus Dynamic RAM (RDRAM), etc.

[0193] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles disclosed herein and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0194] It should be understood that the present invention is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A training method for a driving parameter recognition model, characterized in that, include: Obtain the current location data and current deviation data of the target vehicle, wherein the current deviation data is the deviation between the current location data and the first preset location data; Based on the current deviation data and the initial driving parameter identification model, driving control parameters are identified to obtain initial driving control parameters and predicted driving control parameters. Determining current report data based on the current deviation data and the current position data includes: determining current report data based on the current longitudinal velocity, current longitudinal jerk, and current angular acceleration in the current deviation data and the current position data, wherein the current longitudinal jerk and the current angular acceleration are negatively correlated with the current report data, and the current longitudinal velocity is positively correlated with the current report data; Determine the updated deviation data corresponding to the initial driving control parameters; The initial driving control parameters, the current deviation data, the predicted driving control parameters, and the updated deviation data are input into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index. The target confidence index is calculated based on the current return data and the predicted confidence index. The initial driving parameter recognition model is trained based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model in the initial driving parameter recognition model.

2. The training method for a driving parameter recognition model according to claim 1, characterized in that, The determination of the current report data based on the current deviation data and the current longitudinal velocity, current longitudinal jerk, and current angular acceleration in the current position data includes combining the following formula: in, This indicates the current return data. This indicates the current longitudinal velocity. This indicates the current longitudinal jerk. This represents the first preset adjustment coefficient corresponding to the current longitudinal jerk. This indicates the current heading angular acceleration. This represents the second preset adjustment coefficient corresponding to the current heading angular acceleration. This indicates the current deviation data. This represents the third preset adjustment coefficient corresponding to the current deviation data.

3. The training method for a driving parameter recognition model according to claim 1, characterized in that, The process of identifying driving control parameters based on the current deviation data and the initial driving parameter identification model to obtain initial driving control parameters and predicted driving control parameters includes: The current deviation data is input into the first driving parameter recognition model to obtain the initial driving control parameters; The updated deviation data is input into the second driving parameter recognition model in the initial driving parameter recognition model to obtain the predicted driving control parameters.

4. The training method for a driving parameter recognition model according to claim 1, characterized in that, The evaluation model includes a first evaluation model and a second evaluation model. The step of training the initial driving parameter recognition model based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model includes: The first loss information is determined based on the target confidence index and the current confidence index; The first evaluation parameter of the first evaluation model is updated based on the first loss information. The second loss information is determined based on the current confidence index; The first recognition parameter of the first driving parameter recognition model is updated based on the second loss information; Based on the first evaluation parameters, update the second evaluation parameters of the second evaluation model; Based on the first identification parameter, update the second identification parameter of the second driving parameter identification model in the initial driving parameter identification model; The updated position data is used as the current position data, and the updated deviation data is used as the current deviation data. The updated position data is the position data of the target vehicle when the target vehicle is controlled based on the initial driving control parameters. Based on the updated first driving parameter recognition model, second driving parameter recognition model, first evaluation model, second evaluation model, current position data, and current deviation data, the process of identifying driving control parameters based on the current deviation data and the initial driving parameter recognition model is repeated to obtain initial driving control parameters and predicted driving control parameters, until the steps of using the updated position data corresponding to the initial driving control parameters as the current position data and the updated deviation data as the current deviation data are met, until the preset training convergence condition is satisfied. The first driving parameter recognition model that meets the preset training convergence condition is used as the target driving parameter recognition model.

5. A driving control method, characterized in that, include: Obtain the target position deviation data of the target vehicle; The target position deviation data is input into the driving parameter recognition model to identify driving control parameters, thereby obtaining the target driving control parameters. The driving parameter recognition model is the target driving parameter recognition model trained using the training method described in any one of claims 1-4. The target vehicle is controlled based on the target driving control parameters.

6. The driving control method according to claim 5, characterized in that, The control of the target vehicle based on the target driving control parameters includes: Based on the target driving control parameters, construct the optimization objective function corresponding to the target driving control quantity; Based on the target constraints, the optimization objective function is solved to obtain the target driving control quantity; The target vehicle is controlled based on the target driving control quantity.

7. A driving control method according to claim 5, characterized in that, The process of obtaining the target position deviation data of the target vehicle includes: The current location data and target driving trajectory data of the target vehicle are obtained, wherein the target driving trajectory data includes target location data of multiple trajectory points; The deviation between the current location data and the target location data corresponding to the current location data is used as the target location deviation data.

8. A training device for a driving parameter recognition model, characterized in that, include: The first acquisition module is used to acquire the current position data and current deviation data of the target vehicle, wherein the current deviation data is the deviation between the current position data and the first preset position data; The first driving control parameter identification module is used to identify driving control parameters based on the current deviation data and the initial driving parameter identification model, so as to obtain the initial driving control parameters and the predicted driving control parameters. The report data determination module is used to determine the current report data based on the current deviation data and the current position data; The update deviation data determination module is used to determine the update deviation data corresponding to the initial driving control parameters; The first confidence index determination module is used to input the initial driving control parameters, the current deviation data, the predicted driving control parameters and the updated deviation data into the evaluation model in the initial driving parameter identification model to obtain the current confidence index and the predicted confidence index. The second confidence index determination module is used to calculate the target confidence index based on the current return data and the predicted confidence index; The training module is used to train the initial driving parameter recognition model based on the current confidence index and the target confidence index to obtain the target driving parameter recognition model corresponding to the first driving parameter recognition model in the initial driving parameter recognition model; The return data determination module includes: The report data determination unit is used to determine the current report data based on the current longitudinal velocity, current longitudinal jerk, and current heading angular acceleration in the current deviation data and the current position data. The current longitudinal jerk and the current heading angular acceleration are negatively correlated with the current report data, and the current longitudinal velocity is positively correlated with the current report data.

9. A driving control device, characterized in that, include: The second acquisition module is used to acquire the target position deviation data of the target vehicle. The second driving control parameter identification module is used to input the target position deviation data into the driving parameter identification model to identify the driving control parameters and obtain the target driving control parameters. The driving parameter identification model is the target driving parameter identification model trained by the training method described in any one of claims 1-4. A control module is used to control the target vehicle based on the target driving control parameters.

10. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the training method of the driving parameter recognition model as described in any one of claims 1 to 4, or to implement the driving control method as described in any one of claims 5 to 7.

11. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device is able to perform the training method of the driving parameter recognition model as described in any one of claims 1 to 4, or implement the driving control method as described in any one of claims 5 to 7.

Images