Automatic driving method and system, terminal and storage medium

Abstract

The invention discloses an automatic driving method and system, a terminal and a storage medium, and relates to the technical field of vehicle motion control, and the method comprises the steps: obtaining environment perception data and motion state data, the environment perception data comprises an RGB image and corresponding depth information, and the motion state data comprises the angular velocity, acceleration and wheel rotation information of a vehicle; target detection is carried out on the RGB image, navigation elements are determined, and the navigation elements comprise guide lines and / or navigation target points; projecting key points of the navigation elements according to the RGB image and the depth information to obtain two-dimensional horizontal coordinates of the key points in a pre-constructed world coordinate system; determining vehicle state information corresponding to the vehicle according to the two-dimensional horizontal coordinates and the motion state data; according to the navigation elements and the vehicle state information, a vehicle motion control instruction is generated through a PID controller; and performing motion control on the vehicle according to the vehicle motion control instruction. Therefore, the automatic driving effect of the vehicle is improved.

Description

Technical Field

[0001] This application relates to the field of vehicle motion control technology, and in particular to an autonomous driving method, system, terminal and storage medium. Background Technology

[0002] With the development of science and technology, various intelligent control schemes are being used more and more widely, such as autonomous driving schemes.

[0003] In existing technologies, autonomous driving technology is typically applied to structured scenarios. Therefore, existing autonomous driving solutions usually predefine the vehicle's driving route, controlling the vehicle to travel along the pre-defined route in simple structured scenarios. The problem with existing technologies is that they typically only control the vehicle to travel along the pre-defined route, making it difficult to perform real-time autonomous driving control of the vehicle in complex scenarios and specific real-time environments, which is not conducive to improving the autonomous driving effect.

[0004] Therefore, the relevant technologies still need to be improved and developed. Summary of the Invention

[0005] The main purpose of this application is to provide an autonomous driving method, system, terminal, and storage medium, which aims to solve the technical problem that in related technologies, vehicles are usually only controlled to drive along a pre-defined route, which makes it difficult to perform real-time autonomous driving control of vehicles in specific real-time environments under complex scenarios, thus hindering the improvement of the autonomous driving effect of vehicles.

[0006] To achieve the above objectives, a first aspect of this application provides an autonomous driving method, wherein the method includes: The vehicle acquires environmental perception data and motion state data, wherein the environmental perception data includes an RGB image and depth information corresponding to the RGB image, and the motion state data includes the vehicle's angular velocity, acceleration, and wheel rotation information. Target detection is performed on the above RGB image to determine navigation elements in the above RGB image, wherein the navigation elements include guide lines and / or navigation target points; Based on the above RGB image and the above depth information, the key points of the above navigation elements are projected to obtain the two-dimensional horizontal coordinates of the above key points in the pre-constructed world coordinate system. Based on the above two-dimensional horizontal coordinates and the above motion state data, the vehicle state information corresponding to the above vehicle is determined, wherein the above vehicle state information includes the vehicle's position information, speed information and attitude information. Based on the above navigation elements and vehicle status information, vehicle motion control commands are generated through a preset PID controller; The vehicle is controlled to move according to the aforementioned vehicle motion control commands.

[0007] Optionally, the above-mentioned object detection of the RGB image to determine navigation elements in the RGB image includes: The above RGB image is subjected to target detection by a preset deep learning-based target detection model to determine the navigation elements in the above RGB image, wherein the navigation elements include guide lines, navigation target points, parking signs and dynamic obstacles.

[0008] Optionally, the above-mentioned projection of key points of the navigation elements based on the RGB image and the depth information to obtain the two-dimensional horizontal coordinates of the key points in a pre-constructed world coordinate system includes: Obtain the pixel coordinates of the aforementioned key points in the aforementioned RGB image; Based on the pixel coordinates and depth information mentioned above, the two-dimensional horizontal coordinates of the key points in the world coordinate system are calculated using a preset perspective projection model.

[0009] Optionally, determining the vehicle status information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion state data includes: Based on the aforementioned two-dimensional horizontal coordinates and motion state data, the vehicle state information is calculated using the Kalman filter algorithm.

[0010] Optionally, the aforementioned preset PID controller includes a parallel heading PID controller and a longitudinal speed PID controller; The aforementioned vehicle motion control commands include heading control commands and longitudinal speed control commands; Based on the aforementioned navigation elements and vehicle status information, the vehicle motion control commands are generated through a preset PID controller, including: Based on the aforementioned navigation elements and vehicle status information, the aforementioned heading PID controller generates the aforementioned heading control command, and the aforementioned longitudinal speed PID controller generates the aforementioned longitudinal speed control command.

[0011] Optionally, the parameter of any target item in the above-mentioned preset PID controller is determined according to the preset reference parameter and dynamic adjustment parameter corresponding to the target item; The aforementioned dynamic adjustment parameters are determined in real time based on the aforementioned navigation elements and vehicle status information. The aforementioned objective terms include proportional terms, integral terms, and differential terms.

[0012] Optionally, the dynamic adjustment parameters corresponding to any target item are determined in real time based on the following steps: The path curvature is determined based on the geometry of the guide lines in the RGB image. The vehicle pitch angle is determined based on the attitude information in the vehicle status information above, and the road slope is determined based on the vehicle pitch angle above. Based on the aforementioned path curvature, road surface slope, and speed information in the aforementioned vehicle status information, the dynamic adjustment parameters corresponding to the aforementioned target item are determined by searching the preset parameter lookup table corresponding to the aforementioned target item.

[0013] A second aspect of this application provides an autonomous driving system, wherein the system includes: The data acquisition module is used to acquire environmental perception data and motion state data corresponding to the vehicle. The environmental perception data includes RGB images and depth information corresponding to the RGB images. The motion state data includes the angular velocity, acceleration and wheel rotation information of the vehicle. The target detection module is used to perform target detection on the above RGB image to determine the navigation elements in the above RGB image, wherein the navigation elements include guide lines and / or navigation target points; The key point projection module is used to project the key points of the navigation elements based on the RGB image and the depth information to obtain the two-dimensional horizontal coordinates of the key points in a pre-built world coordinate system. The vehicle status determination module is used to determine the vehicle status information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion status data, wherein the vehicle status information includes the vehicle's position information, speed information and attitude information. The control command generation module is used to generate vehicle motion control commands based on the above navigation elements and vehicle status information through a preset PID controller. The control module is used to control the motion of the vehicle according to the aforementioned vehicle motion control commands.

[0014] A third aspect of this application provides a terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements any of the steps of the aforementioned autonomous driving method.

[0015] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described autonomous driving methods.

[0016] As can be seen from the above, the present application provides an autonomous driving method, specifically, acquiring environmental perception data and motion state data corresponding to the vehicle. The environmental perception data includes an RGB image and corresponding depth information, while the motion state data includes the vehicle's angular velocity, acceleration, and wheel rotation information. Target detection is performed on the RGB image to determine navigation elements, including guide lines and / or navigation target points. Based on the RGB image and the depth information, key points of the navigation elements are projected to obtain their two-dimensional horizontal coordinates in a pre-constructed world coordinate system. Based on the two-dimensional horizontal coordinates and the motion state data, vehicle state information is determined, including the vehicle's position, speed, and attitude information. Based on the navigation elements and the vehicle state information, a vehicle motion control command is generated using a preset PID controller. The vehicle's motion is then controlled according to the vehicle motion control command.

[0017] Thus, in the process of autonomous driving, the vehicle is not directly controlled to follow a pre-defined route. Instead, environmental perception data and motion state data are acquired in real time. Object detection is performed on the RGB images in the environmental perception data to identify navigation elements in the environment. Combined with depth information from the environmental perception data, key points of the navigation elements are projected onto the ground in the world coordinate system to obtain two-dimensional horizontal coordinates. Then, combining the two-dimensional horizontal coordinates with the motion state data, vehicle state information is determined. Based on the navigation elements and vehicle state information, PID control is applied to the vehicle to achieve autonomous driving control. In this way, even in complex scenarios, real-time autonomous driving control can be performed on the vehicle according to the specific real-time environment, which is beneficial to improving the effectiveness of autonomous driving. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an autonomous driving method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the constituent modules of an autonomous driving system provided in an embodiment of this application; Figure 3 This is a block diagram illustrating the internal structure of a terminal provided in an embodiment of this application. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0027] Currently, autonomous driving solutions are being applied more and more widely. Existing technologies typically consider applying autonomous driving technology to structured scenarios. Therefore, existing autonomous driving solutions usually predefine the vehicle's route, controlling the vehicle to travel along the pre-defined route in simple, structured scenarios. Thus, simply controlling the vehicle to travel along a pre-defined route is usually insufficient for real-time autonomous driving control in complex scenarios and specific real-time environments, which is detrimental to improving the effectiveness of autonomous driving.

[0028] Autonomous driving technology faces unique technical bottlenecks in unstructured, enclosed environments such as mining areas, logistics parks, and industrial plants. Unlike structured urban roads, unstructured, enclosed environments typically feature fixed work processes but complex scene elements, such as irregular terrain, blurred visual guide lines, visibility fluctuations caused by dust, and dynamic obstacles arising from mixed human-vehicle operations. These characteristics place higher demands on the perception reliability, real-time decision-making, and task adaptability of autonomous driving systems. Furthermore, traditional autonomous driving systems often employ a modular architecture with sequential execution, relying on relatively simple decision rules. In such environments, problems frequently arise, including the accumulation of perception errors along the task chain, rigid decision logic responses to unexpected situations, and difficulty adapting to the differentiated needs of different operational stages within the same industrial park.

[0029] In complete, streamlined operational scenarios, existing solutions either focus on understanding dynamic interactions while lacking the ability to model fixed task sequences, or are limited to static optimization of perception resources and fail to adapt to dynamically changing task stages, or rely on cloud commands at the expense of the vehicle's local autonomous response speed. The core problem is that existing architectures generally fail to use the "context of the task flow" as the core thread driving system operation, resulting in a relatively fragmented state between perception, decision-making, and control modules. This makes it difficult to support the autonomous, smooth, and reliable execution of the entire operational process, from "global path navigation" to "local guide line tracking" and then to "precise destination positioning."

[0030] It is evident that existing technologies still have the following shortcomings: the system control logic lacks understanding and modeling of the overall task flow context, making it difficult to adapt to the control requirements of multi-stage and differentiated tasks; the operation mode switching mechanism is rigid and passive, relying on a single or a few triggering conditions, and lacks intelligent decision-making capabilities based on multi-source information fusion; the system architecture design often suffers from an imbalance between excessive centralized control and local functional optimization, making it difficult to balance overall system efficiency, local autonomy, and real-time robustness.

[0031] To address at least one of the aforementioned technical problems, this application proposes an autonomous driving method. Specifically, the method involves acquiring environmental perception data and motion state data corresponding to a vehicle. The environmental perception data includes an RGB image and corresponding depth information. The motion state data includes the vehicle's angular velocity, acceleration, and wheel rotation information. Target detection is performed on the RGB image to determine navigation elements, including guide lines and / or navigation target points. Based on the RGB image and the depth information, key points of the navigation elements are projected to obtain their two-dimensional horizontal coordinates in a pre-constructed world coordinate system. Based on the two-dimensional horizontal coordinates and the motion state data, vehicle state information is determined, including the vehicle's position, speed, and attitude information. Based on the navigation elements and the vehicle state information, a vehicle motion control command is generated using a preset PID controller. Finally, the vehicle's motion is controlled according to the vehicle motion control command.

[0032] Thus, in the process of autonomous driving, the vehicle is not directly controlled to follow a pre-defined route. Instead, environmental perception data and motion state data are acquired in real time. Object detection is performed on the RGB images in the environmental perception data to identify navigation elements in the environment. Combined with depth information from the environmental perception data, key points of the navigation elements are projected onto the ground in the world coordinate system to obtain two-dimensional horizontal coordinates. Then, combining the two-dimensional horizontal coordinates with the motion state data, vehicle state information is determined. Based on the navigation elements and vehicle state information, PID control is applied to the vehicle to achieve autonomous driving control. In this way, even in complex scenarios, real-time autonomous driving control can be performed on the vehicle according to the specific real-time environment, which is beneficial to improving the effectiveness of autonomous driving.

[0033] Specifically, the proposed solution can be applied to autonomous driving control systems and methods in unstructured closed environments (such as mining areas, logistics parks, and industrial plants). These environments are characterized by fixed work processes but complex environmental factors. Therefore, this solution utilizes a multimodal perception and hierarchical decision-making architecture, combined with a task state machine and confidence assessment mechanism, to achieve reliable switching and autonomous control of autonomous vehicles across multiple operating modes. This solution aligns with the fixed work process characteristics of unstructured closed environments, employing multimodal perception fusion and a progressive hierarchical decision-making architecture to construct a quantifiable system state assessment method oriented towards work processes. This method comprehensively considers task context, real-time environmental characteristics, and system performance to generate dynamic, tiered confidence assessments, thereby driving the autonomous driving system to achieve safe, reasonable, and timely switching between multiple operating modes, significantly improving overall robustness, safety, and operational efficiency in specific working environments.

[0034] like Figure 1 As shown in the figure, this application provides an autonomous driving method, which specifically includes the following steps: Step S100: Obtain environmental perception data and motion state data corresponding to the vehicle. The environmental perception data includes an RGB image and depth information corresponding to the RGB image. The motion state data includes the angular velocity, acceleration and wheel rotation information of the vehicle. Step S200: Perform target detection on the above RGB image to determine navigation elements in the above RGB image, wherein the navigation elements include guide lines and / or navigation target points; Step S300: Based on the above RGB image and the above depth information, project the key points of the above navigation elements to obtain the two-dimensional horizontal coordinates of the above key points in the pre-constructed world coordinate system. Step S400: Based on the above two-dimensional horizontal coordinates and the above motion state data, determine the vehicle state information corresponding to the above vehicle, wherein the above vehicle state information includes the vehicle's position information, speed information and attitude information; Step S500: Based on the above navigation elements and the above vehicle status information, generate vehicle motion control commands through a preset PID controller. Step S600: Perform motion control on the vehicle according to the aforementioned vehicle motion control command.

[0035] It should be noted that the above-mentioned autonomous driving method can be applied to an autonomous driving system, and the autonomous driving system can execute the above method, but no specific limitation is made here.

[0036] The aforementioned environmental perception data and motion state data can be acquired by data acquisition sensors installed on the vehicle, or by additional data acquisition sensors, without specific limitations.

[0037] In some application scenarios, a multi-source sensor acquisition system for autonomous driving in unstructured closed parks can be set up to achieve comprehensive, reliable, and real-time acquisition of environmental and vehicle status information, providing support for subsequent perception, localization, and decision-making control.

[0038] Specifically, the above system collects data through the following sensors to obtain two types of data: vehicle motion state data and external environmental perception data: Visual sensor: An RGB-D camera is used as the core environmental perception unit, which can simultaneously output RGB images and corresponding depth information. The RGB images are used to identify the texture and semantic information of key navigation elements such as guide lines, parking signs, work targets, and dynamic obstacles; the depth information directly provides the three-dimensional geometric structure of the scene, providing key elevation data for subsequent processing of terrain undulations and achieving accurate spatial positioning.

[0039] Inertial Measurement Unit (IMU): Used to directly measure the vehicle's three-dimensional motion state. It acquires the vehicle's angular velocity around its three axes using gyroscopes and measures linear acceleration along those axes using accelerometers. IMU data is updated frequently, providing a foundation for vehicle attitude estimation, motion prediction, and multi-sensor time synchronization.

[0040] Wheel-mounted odometer: This type of odometer measures wheel rotation information via encoders mounted on the drive wheels. This rotation information includes the wheel rotation angle or the number of pulses. Based on this wheel rotation information and a vehicle model, the vehicle's longitudinal displacement and velocity can be calculated. On flat roads, this sensor is crucial for trajectory estimation.

[0041] It should be noted that the aforementioned sensors achieve time synchronization at the hardware level, ensuring that data from different sources are aligned in time. The acquired raw data then enters the preprocessing stage: visual images undergo distortion correction (depending on camera calibration parameters); IMU data undergoes zero-bias correction; and odometry data undergoes calibration and conversion. The resulting standardized data stream will be used in subsequent processing steps to further support the system's perception and decision-making functions.

[0042] The aforementioned target detection of the RGB image to determine navigation elements in the RGB image includes: The above RGB image is subjected to target detection by a preset deep learning-based target detection model to determine the navigation elements in the above RGB image, wherein the navigation elements include guide lines, navigation target points, parking signs and dynamic obstacles.

[0043] It should be noted that the aforementioned deep learning-based object detection model can be pre-set and adjusted according to actual needs, and no specific limitations are made here. Furthermore, the specific navigation elements to be detected can also be set and adjusted according to actual needs, and no specific limitations are made here either.

[0044] Furthermore, based on the aforementioned RGB image and depth information, the key points of the navigation elements are projected to obtain their two-dimensional horizontal coordinates in a pre-constructed world coordinate system, including: Obtain the pixel coordinates of the aforementioned key points in the aforementioned RGB image; Based on the pixel coordinates and depth information mentioned above, the two-dimensional horizontal coordinates of the key points in the world coordinate system are calculated using a preset perspective projection model.

[0045] In unstructured, closed-campus autonomous driving scenarios, vehicles need to convert the two-dimensional image information captured by their onboard RGB-D cameras into two-dimensional spatial coordinates in the real world, specifically two-dimensional horizontal coordinates in a pre-defined world coordinate system. This world coordinate system can be pre-constructed according to actual needs and is not specifically limited here.

[0046] First, the parameters of lens distortion involved in the camera are obtained through offline calibration, and the original image is corrected to eliminate radial and tangential distortion.

[0047] For ideal planar regions in an environment, traditional planar projection models based on homography matrices establish the pixel coordinates of the image. ground coordinates The mapping relationship between them is used to achieve rough positioning, and its mathematical expression is as follows: ; in, A 3×3 reference homography matrix is ​​obtained through calibration; is the scale factor. This model is based on a strict planar assumption, mapping image points directly to an assumed ground plane.

[0048] This model is based on a strict flat-ground assumption. However, in real-world scenarios, ramps, cargo piles, and uneven surfaces cause variations in ground elevation. Traditional visual localization methods assume a flat ground; therefore, directly mapping points from an image onto a map would be like incorrectly projecting objects from a ramp onto flat ground, resulting in significant localization errors.

[0049] To address this issue, this application utilizes depth information provided by an RGB-D camera, eliminating errors introduced by the planar assumption through coordinate system transformation. Specifically, instead of relying on a single planar mapping, it uses the camera's intrinsic and extrinsic parameter matrices, combined with depth information, to directly calculate the 3D coordinates of key points in the image in the world coordinate system. The mathematical expression of this model (i.e., the preset perspective projection model) is as follows: ; in, This is the depth value of the key point. This is the camera intrinsic parameter matrix (determined by the camera hardware's own properties). This is the camera extrinsic parameter matrix (the position and attitude of the camera relative to the world coordinate system are determined through offline calibration). This is the actual elevation value of the key point. Specifically, the depth value of the key point is determined based on the pixel coordinates corresponding to the key point, and the depth value corresponding to those pixel coordinates from the aforementioned depth information.

[0050] Specifically, in this application, the precise 3D coordinates of key points in the world coordinate system are calculated using a preset perspective projection model based on the RGB image and depth information perceived and constructed by an RGB-D camera. (The calculated actual elevation value of the key point is used as the vertical coordinate of the key point.) The complete steps are as follows: First, based on the correspondence between the RGB image and depth information, the depth value corresponding to the original pixel coordinates (i.e., the pixel coordinates of the key point in the image) in the depth information is used as the depth value of the key point. Substitute the pixel coordinates and corresponding depth values ​​into the mathematical expression of the preset perspective projection model to calculate the precise 3D coordinates of the key points in the world coordinate system; then extract the 2D horizontal coordinates of the key points in the world coordinate system from the calculated 3D coordinates. .

[0051] In some application scenarios, the output two-dimensional horizontal coordinates will be used as key information to generate reliable vehicle pose estimation together with data collected by sensors such as IMU and wheel odometer; at the same time, the coordinates are also used to calculate the tracking path and generate specific steering and speed control commands.

[0052] In this embodiment of the application, determining the vehicle state information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion state data includes: Based on the aforementioned two-dimensional horizontal coordinates and motion state data, the vehicle state information is calculated using the Kalman filter algorithm.

[0053] In real-world park operations, vehicles often face complex perception challenges: dust or nighttime conditions can cause camera malfunctions, slippery surfaces can lead to inaccurate wheel speed measurements, and IMUs themselves have drift errors. The failure of any of these sensors can result in positioning failures, causing operational interruptions or safety accidents.

[0054] To address these issues, this application employs a multi-sensor fusion algorithm based on Kalman filtering to achieve deep fusion of visual, IMU, and wheel odometer information.

[0055] Specifically, firstly, the displacement measurement of the wheel odometer and the inertial calculation of the IMU are combined, and then the position information is obtained by visual observation of environmental characteristics to form a precise positioning; then, based on the acceleration data of the IMU, supplemented by the longitudinal velocity provided by the wheel speedometer and the lateral velocity obtained by visual analysis, the actual speed of the vehicle is determined; finally, the attitude information is obtained by real-time calculation of the gyroscope and accelerometer data of the IMU.

[0056] Based on the above information, vehicle state information is obtained. This vehicle state information can be a system state vector representing the complete motion state of the vehicle, as shown below: ; in, Representative moment The system state vector, This represents the vehicle's x-coordinate in the world coordinate system. This represents the vehicle's ordinate in the world coordinate system. This represents the vehicle's actual elevation value (i.e., vertical coordinate) in the world coordinate system. , and These represent the vehicle's velocity along the corresponding coordinate axis in the world coordinate system; , and These represent the vehicle's roll angle, pitch angle, and yaw angle, respectively. This represents the transpose of the vector. As can be seen, the above system state vector integrates the vehicle's position, velocity, and attitude information.

[0057] In this application, based on the classic Kalman filter algorithm, the vehicle state is continuously estimated optimally through two iterative steps: "prediction" and "update". In the prediction stage, based on the vehicle motion model (e.g., using a simplified uniform speed model) and the state at the previous moment, combined with the current IMU and wheel speed sensor data, the "prior estimate" of the vehicle's current position, speed, and attitude is predicted. In the update stage, the actual observation data from sensors such as vision and wheel speed sensors are compared with the predicted values, and the corrected optimal state estimate is obtained by dynamically fusing the calculated "Kalman gain".

[0058] Through this iterative filtering process, the optimal fusion estimate of the vehicle state is finally obtained. It is used directly to generate precise motion control commands.

[0059] At the same time, by constructing a confidence assessment system, quantitative basis is provided for the upper-level decision-making system to support the intelligent switching of operating modes.

[0060] Specifically, visual detection confidence Used to quantify the output reliability of the visual perception subsystem in the current environment: ; Among them, the basic test score Image quality assessment: Derived from the confidence output of the deep learning object detection model for key navigation elements, it directly reflects the model's grasp of the recognition results. Image sharpness is assessed by calculating the Laplacian variance of the image and by combining it with information entropy to measure the information richness of the image. Then, a normalization function is used to fuse these two metrics into a scalar value in the interval [0, 1], thereby collaboratively determining the overall image quality of the current environment; Target stability index. : Calculate the rate of change of the center point position and area of ​​the target detection box in the current frame compared to the detection boxes in the previous 5 frames; temporal consistency A lightweight Kalman filter is used to make short-term predictions of the target's motion in the image. The intersection-over-union (IoU) ratio is calculated between the actual detection box and the predicted box in the current frame. , , and It is a pre-set ratio coefficient, which can be set and adjusted according to actual needs, and is not specifically limited here.

[0061] This evaluation mechanism enables the system to identify and reduce unreliable perception results caused by environmental interference or detection fluctuations, thereby improving the robustness of overall decision-making.

[0062] Positioning accuracy confidence Used to assess the health status and reliability of the output of a fusion positioning system: ; Among them, observation residuals This residual is the observation value within the Kalman filter framework. Compared with the predicted observed values The difference norm between them; sensor cross-validation differences The system periodically performs sensor cross-validation to calculate the distance between the positioning results of the two systems (visual positioning and inertial navigation positioning). If the two differ significantly, it indicates that one of them may be unreliable; proportionality coefficient and It is based on experience and historical data to optimize pre-set parameters, which are then fixed before system deployment. This mechanism enables the system to perceive the degradation of its own positioning accuracy and sensor anomalies in real time, even without an external absolute reference.

[0063] The system performs real-time multi-dimensional fusion calculations, weighting and fusing information from two key dimensions to achieve intelligent decision-making for state transitions. Ultimately, it outputs an overall confidence score that dynamically reflects the system's comprehensive reliability, environmental adaptability, and task completion assurance. : ; Among them, the weighting coefficient and These are fixed values ​​pre-set through a combination of experience and offline optimization, and .

[0064] However, in a state management system, in a dynamically changing unstructured campus environment, The absolute value of the threshold will drift due to factors such as lighting and weather. Using a fixed threshold would cause the system to be overly sensitive in harsh environments or overly insensitive in favorable environments. Therefore, this system employs a dynamic threshold mechanism based on the statistical characteristics of a sliding window. The algorithm implementation consists of the following three steps: Confidence sequence cache: The system maintains a cache of length [length missing]. sliding window Continuously record recent Historical confidence sequence for each calculation period: ; Statistical feature extraction: Real-time calculation of the mean of all confidence scores within the window. with standard deviation : ; ; Among them, the mean It reflects the system's recent average performance level, while the standard deviation This quantifies the range of confidence level fluctuations.

[0065] Dynamic threshold calculation: Based on the above statistics, dynamically set the high and low thresholds for state transitions. Upgrade threshold : ; Degradation threshold : ; In the formula, and This is a configurable sensitivity coefficient used to control the level of conservatism during state transitions.

[0066] Through the above calculations, the system obtains a set of thresholds that dynamically adjust with the environment: upgrade thresholds. With the downgrade threshold This threshold can be adaptively adjusted based on the average level and fluctuation range of the system's recent confidence level, thereby achieving the effect of automatically relaxing the switching conditions when the overall environment deteriorates and tightening the switching conditions when the environment is favorable.

[0067] In some application scenarios, multiple control levels can be set for vehicles. For example, the first level can be set to manual control, the second level to follow a guide line, and the third level to follow a navigation target point. If the vehicle is currently at the second level, it will upgrade to the third level if the calculated overall confidence score is higher than the upgrade threshold; it will downgrade to the first level if the calculated overall confidence score is lower than the downgrade threshold; and it will maintain the current control level if the calculated overall confidence score is between the downgrade and upgrade thresholds.

[0068] When controlling based on the above-mentioned automatic control levels of traveling along the guide line or traveling based on the navigation target point, the system generates specific path tracking points in real time according to navigation elements (as tracking points for the PID controller).

[0069] Specifically, tracking points are generated based on detected guide lines or directions pointing to the navigation target point. Simultaneously, the system detects dynamic obstacles in the environment in real time.

[0070] If a dynamic obstacle is detected, the pixel coordinates of the obstacle's key points are converted into two-dimensional coordinates in the world coordinate system and compared with the vehicle's position coordinates. If the comparison determines that the dynamic obstacle is located within a preset danger zone in front of the vehicle, an offset is applied to the lateral coordinates of the generated tracking point to maintain a set safe lateral distance between the tracking point and the dynamic obstacle. The direction of the offset is away from the center of the obstacle. The tracking point adjusted by the offset is used as the final path tracking point, enabling the vehicle to avoid dynamic obstacles in real time and locally while maintaining its overall tracking intention.

[0071] If no dynamic obstacles are detected, the tracking points generated above will be used as the final path tracking points.

[0072] To address the challenges posed by varying road conditions, slopes, curves, and load changes within the park, this application proposes a hierarchical control architecture consisting of a hierarchical PID controller and a parameter adaptive engine. Two parallel adaptive PID controllers are employed to handle heading control and longitudinal speed control, respectively.

[0073] Specifically, the aforementioned preset PID controller includes a parallel heading PID controller and a longitudinal speed PID controller; The aforementioned vehicle motion control commands include heading control commands and longitudinal speed control commands; Based on the aforementioned navigation elements and vehicle status information, the vehicle motion control commands are generated through a preset PID controller, including: Based on the aforementioned navigation elements and vehicle status information, the aforementioned heading PID controller generates the aforementioned heading control command, and the aforementioned longitudinal speed PID controller generates the aforementioned longitudinal speed control command.

[0074] Among them, the parameter of any target item in the above-mentioned preset PID controller is determined according to the preset reference parameter and dynamic adjustment parameter corresponding to the above-mentioned target item; The aforementioned dynamic adjustment parameters are determined in real time based on the aforementioned navigation elements and vehicle status information. The aforementioned objective terms include proportional terms, integral terms, and differential terms.

[0075] Furthermore, the dynamic adjustment parameters corresponding to any target item are determined in real time based on the following steps: The path curvature is determined based on the geometry of the guide lines in the RGB image. The vehicle pitch angle is determined based on the attitude information in the vehicle status information above, and the road slope is determined based on the vehicle pitch angle above. Based on the aforementioned path curvature, road surface slope, and speed information in the aforementioned vehicle status information, the dynamic adjustment parameters corresponding to the aforementioned target item are determined by searching the preset parameter lookup table corresponding to the aforementioned target item.

[0076] For longitudinal speed control, its basic form is: ; in, represent The longitudinal velocity control output at any given moment, represent The speed error at any given moment (the difference between the target speed and the actual speed). The gain coefficient representing the proportional term. The gain coefficient representing the integral term. The gain coefficient represents the differential term.

[0077] Unlike traditional PID controllers, the solution in this application, , and These are dynamic parameters that are adjusted in real time by the adaptive engine.

[0078] In this architecture, the parameter adaptive engine can dynamically and globally optimize all parameters of the PID controller based on real-time perceived vehicle status and environmental context. Its specific implementation is based on multi-dimensional lookup tables and linear interpolation methods, with the following steps: The parameter adaptive function is defined as: ; ; ; in, , and These are preset baseline parameters. , and Based on vehicle speed Path curvature and road slope The dynamic adjustment amount. Among them, the path curvature is obtained from the geometry of the guide line extracted by visual perception; the road surface slope is determined by the pitch angle determined by the IMU, and the pitch angle can be used as the road surface slope.

[0079] The dynamic adjustment amount is calculated by integrating the independent components of the three dimensions. Taking the proportional term as an example: ; The weighting coefficients satisfy Its final value is preset and fixed as a set of empirical parameters before the system is deployed. and The calculation follows the same pattern, so it will not be repeated here.

[0080] Implementation using a lookup table for component functions: Each one-dimensional component function , and All of these are achieved through predefined lookup tables. The system adopts a lookup and interpolation method based on experience data: First, based on historical operation data and experience, a lookup table is pre-established for each influencing factor, its effective range is divided into continuous intervals, and the optimal parameter adjustment amount is calibrated for each interval; during online operation, the system locates the corresponding interval in the table according to the real-time measurement value, and calculates the precise adjustment amount through linear interpolation, thereby ensuring that the parameters change continuously and smoothly with the operating conditions, effectively avoiding abrupt changes in control commands while maintaining the reliability of experience.

[0081] The PID controller mechanism for handling lateral heading control is similar to the method described above. Its error signals are lateral position deviation and heading angle deviation, and the output is a steering angle command. The specific calculation process can be referred to the calculation process described above, and will not be repeated here.

[0082] This layered design, which incorporates a parameter adaptive engine, not only reduces the design complexity of the controller through modular decomposition but also endows the system with the crucial ability to dynamically optimize performance.

[0083] It should be noted that when generating vehicle motion control commands based on the PID controller, the identified navigation target point can be used as the target, or, if no navigation target point is detected, movement along the guide line can be used as the control target. The specific settings and adjustments can be made based on the actual application scenario, and no specific limitations are made here.

[0084] In some application scenarios, remote monitoring and safety intervention mechanisms can also be set up for the autonomous driving control process. Specifically, a complete remote monitoring and safety intervention mechanism has been built to address the complexity and safety requirements of unstructured closed-loop operations. This mechanism achieves seamless and safe switching and collaborative operation between autonomous driving and remote control through multi-level state machine management and quantified safety verification algorithms.

[0085] Specifically, a hierarchical control state machine is constructed, and the permission transfer relationship between autonomous driving and remote control is clearly defined. The control state set is defined as follows: ; The transitions between states are driven by system events, including system anomalies, communication interruptions, path deviations, and requests for manual intervention. In "Auto Navigation" mode, the system operates completely autonomously. When the system detects a complex or uncertain local path, it can switch to "Cooperative Control" mode, in which the system retains basic tracking capabilities while receiving high-level instructions from the remote operator. In "Remote Control" mode, vehicle control is completely transferred to the remote terminal, where the operator directs the vehicle's actions in real time. Once a serious anomaly or safety rule is triggered, the system will immediately enter "Emergency Braking" mode.

[0086] Furthermore, a multi-level security verification mechanism is constructed, including communication link quality verification and operator identity and permission verification.

[0087] Communication link quality verification: The system maintains a connection with the remote monitoring center through continuous bidirectional heartbeat packets to ensure that the communication link meets minimum security standards before switching to remote control mode. This prevents control command loss, delay, or interruption due to network instability, which could lead to security incidents. Link quality Due to communication delay and packet loss rate Joint assessment: ; Among them, only when the link quality Switching from "Auto Navigation" to "Cooperative Control" or "Remote Control" is only permitted when the system is rated "Available" or "Excellent".

[0088] Operator Identity and Authorization Verification: Before the transfer of control, the system requires remote operators to authenticate their identities. This invention employs a two-factor authentication protocol based on digital certificates. The operator first logs into the monitoring system using a username / password. When requesting vehicle control, the operator must sign the request command using their unique private key digital certificate. The vehicle uses a pre-set public key to verify the signature's validity, ensuring the operation command is legitimate and has not been tampered with.

[0089] During the permission switching process, the system employs a smooth interpolation strategy for control commands to prevent vehicle oscillations or instability caused by sudden changes in control inputs. Taking steering control as an example, when switching from automatic control to remote control, the system will... (The sentence is incomplete and requires more context to translate accurately.) Within 200ms, the steering angle of automatic control and the turning angle of remote commands Perform weighted fusion: ; in, The steering angle after fusion; It is a mixed weighting factor used to ensure the smoothness of the vehicle control process. The system smoothly transitions from 0 to 1. When switching from automatic control to remote control, the system uses the algorithm described above to obtain a steering angle that continuously changes from the current value to the target value along a preset smooth curve. This ensures that the steering angle does not undergo abrupt changes following remote commands, effectively suppressing potential vehicle oscillations or instability. In some applications, the sigmoid function can be used as a weighting factor, but this is not a specific limitation.

[0090] This mechanism ensures the reliable operation of a single vehicle while also supporting centralized monitoring and scheduling in multi-vehicle collaborative operation scenarios, further expanding the system's adaptability and deployability in complex park applications.

[0091] By systematically integrating multi-dimensional confidence assessment, hierarchical control architecture, and parameter adaptive mechanisms, this application constructs a complete, intelligent, and adaptive autonomous driving decision-making and control system. This system can reliably achieve autonomous decision-making and smooth switching of operating states in complex, unstructured park environments, ensuring high-precision path tracking while possessing excellent motion smoothness, operational adaptability, and system robustness, thus meeting the safety and efficiency requirements of autonomous driving operations in closed scenarios.

[0092] Furthermore, the vehicle's motion is controlled according to the aforementioned vehicle motion control commands. Specifically, the high-level control intentions (steering angle commands, target speed commands) generated at the upper level are translated into precise physical actions at the vehicle's lower level, achieving stable closed-loop control. The system sends the calculated steering angle and speed commands to the steer-by-wire system and the steer-by-wire drive / braking system in real time through a standardized vehicle control interface.

[0093] During execution, the system continuously receives real-time vehicle status feedback (such as actual yaw rate and actual longitudinal speed) and PID control commands. By comparing the actual motion state with the expected commands, a closed-loop control loop is formed to ensure that the vehicle can accurately and smoothly track the planned path. In addition, actuator status monitoring and safety limiting logic can be set to perform final verification and smoothing of output commands, preventing vehicle instability caused by sudden command changes or actuator malfunctions. This physically ensures the safety, comfort, and final execution accuracy of autonomous driving operations.

[0094] This application provides an autonomous driving method, specifically, acquiring environmental perception data and motion state data corresponding to a vehicle. The environmental perception data includes an RGB image and corresponding depth information, while the motion state data includes the vehicle's angular velocity, acceleration, and wheel rotation information. Target detection is performed on the RGB image to determine navigation elements, including guide lines and / or navigation target points. Based on the RGB image and depth information, key points of the navigation elements are projected to obtain their two-dimensional horizontal coordinates in a pre-constructed world coordinate system. Based on the two-dimensional horizontal coordinates and the motion state data, vehicle state information is determined, including the vehicle's position, speed, and attitude information. Based on the navigation elements and vehicle state information, a vehicle motion control command is generated using a preset PID controller. The vehicle's motion is then controlled according to the vehicle motion control command.

[0095] Thus, in the process of autonomous driving, the vehicle is not directly controlled to follow a pre-defined route. Instead, environmental perception data and motion state data are acquired in real time. Object detection is performed on the RGB images in the environmental perception data to identify navigation elements in the environment. Combined with depth information from the environmental perception data, key points of the navigation elements are projected onto the ground in the world coordinate system to obtain two-dimensional horizontal coordinates. Then, combining the two-dimensional horizontal coordinates with the motion state data, vehicle state information is determined. Based on the navigation elements and vehicle state information, PID control is applied to the vehicle to achieve autonomous driving control. In this way, even in complex scenarios, real-time autonomous driving control can be performed on the vehicle according to the specific real-time environment, which is beneficial to improving the effectiveness of autonomous driving.

[0096] like Figure 2 As shown, corresponding to the above-described autonomous driving method, this application embodiment also provides an autonomous driving system, which includes: The data acquisition module 210 is used to acquire environmental perception data and motion state data corresponding to the vehicle. The environmental perception data includes an RGB image and depth information corresponding to the RGB image. The motion state data includes the angular velocity, acceleration and wheel rotation information of the vehicle. The target detection module 220 is used to perform target detection on the above RGB image to determine navigation elements in the above RGB image, wherein the navigation elements include guide lines and / or navigation target points; The key point projection module 230 is used to project the key points of the navigation elements according to the above RGB image and the above depth information to obtain the two-dimensional horizontal coordinates of the key points in the pre-constructed world coordinate system. The vehicle state determination module 240 is used to determine the vehicle state information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion state data, wherein the vehicle state information includes the vehicle's position information, speed information and attitude information. The control command generation module 250 is used to generate vehicle motion control commands through a preset PID controller based on the above navigation elements and the above vehicle status information. The control module 260 is used to control the motion of the vehicle according to the vehicle motion control command.

[0097] Thus, in the process of autonomous driving, the vehicle is not directly controlled to follow a pre-defined route. Instead, environmental perception data and motion state data are acquired in real time. Object detection is performed on the RGB images in the environmental perception data to identify navigation elements in the environment. Combined with depth information from the environmental perception data, key points of the navigation elements are projected onto the ground in the world coordinate system to obtain two-dimensional horizontal coordinates. Then, combining the two-dimensional horizontal coordinates with the motion state data, vehicle state information is determined. Based on the navigation elements and vehicle state information, PID control is applied to the vehicle to achieve autonomous driving control. In this way, even in complex scenarios, real-time autonomous driving control can be performed on the vehicle according to the specific real-time environment, which is beneficial to improving the effectiveness of autonomous driving.

[0098] It should be noted that the specific structure and implementation of the above-mentioned autonomous driving system and its various modules or units can be referred to the corresponding descriptions in the above method embodiments, and will not be repeated here.

[0099] It should be further noted that the division of the various modules of the above-mentioned autonomous driving system is not unique and is not intended as a specific limitation.

[0100] Based on the above embodiments, this application also provides a terminal, the principle block diagram of which can be as follows: Figure 3 As shown. The terminal includes a processor, memory, network interface, and display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media 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 media. 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 the steps of any of the above-described automatic driving methods. The display screen can be a liquid crystal display (LCD) or an e-ink display.

[0101] Those skilled in the art will understand that Figure 3 The block diagram shown is only a partial structural diagram related to the solution of this application and does not constitute a limitation on the terminal on which the solution of this application is applied. The specific terminal may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0102] In one embodiment, a terminal is provided, the terminal including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of any of the autonomous driving methods provided in the embodiments of this application.

[0103] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the autonomous driving methods provided in this application.

[0104] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0105] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0106] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0107] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0108] In the embodiments provided in this application, it should be understood that the disclosed systems / terminal devices and methods can be implemented in other ways. For example, the system / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units described above is merely a logical functional division, and in actual implementation, it can be divided in other ways. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0109] If the integrated modules / units described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, and software distribution media, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0110] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions are not in essence a departure from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An autonomous driving method, characterized in that, The method includes: The vehicle acquires environmental perception data and motion state data, wherein the environmental perception data includes an RGB image and depth information corresponding to the RGB image, and the motion state data includes the vehicle's angular velocity, acceleration, and wheel rotation information. Target detection is performed on the RGB image to determine navigation elements in the RGB image, wherein the navigation elements include guide lines and / or navigation target points; Based on the RGB image and the depth information, the key points of the navigation element are projected to obtain the two-dimensional horizontal coordinates of the key points in a pre-constructed world coordinate system. Based on the two-dimensional horizontal coordinates and the motion state data, the vehicle state information corresponding to the vehicle is determined, wherein the vehicle state information includes the vehicle's position information, speed information, and attitude information; Based on the navigation elements and the vehicle status information, vehicle motion control commands are generated through a preset PID controller; The vehicle is controlled to move according to the vehicle motion control command.

2. The autonomous driving method according to claim 1, characterized in that, The step of performing target detection on the RGB image to determine navigation elements in the RGB image includes: The RGB image is subjected to target detection using a pre-defined deep learning-based target detection model to determine navigation elements in the RGB image, wherein the navigation elements include guide lines, navigation target points, parking signs, and dynamic obstacles.

3. The autonomous driving method according to claim 1, characterized in that, The step of projecting key points of the navigation element based on the RGB image and the depth information to obtain the two-dimensional horizontal coordinates of the key points in a pre-constructed world coordinate system includes: Obtain the pixel coordinates of the key points in the RGB image; Based on the pixel coordinates and the depth information, the two-dimensional horizontal coordinates of the key point in the world coordinate system are calculated using a preset perspective projection model.

4. The autonomous driving method according to claim 1, characterized in that, The step of determining the vehicle status information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion state data includes: The vehicle state information is calculated using the Kalman filter algorithm based on the two-dimensional horizontal coordinates and the motion state data.

5. The autonomous driving method according to claim 1, characterized in that, The preset PID controller includes a parallel heading PID controller and a longitudinal speed PID controller. The vehicle motion control commands include heading control commands and longitudinal speed control commands; The step of generating vehicle motion control commands through a preset PID controller based on the navigation elements and the vehicle status information includes: Based on the navigation elements and the vehicle status information, the heading control command is generated by the heading PID controller, and the longitudinal speed control command is generated by the longitudinal speed PID controller.

6. The autonomous driving method according to any one of claims 1 to 5, characterized in that, The parameter of any target item in the preset PID controller is determined according to the preset baseline parameter and dynamic adjustment parameter corresponding to the target item; The dynamic adjustment parameters are determined in real time based on the navigation elements and the vehicle status information; The target term includes a proportional term, an integral term, and a differential term.

7. The autonomous driving method according to claim 6, characterized in that, The dynamic adjustment parameters for any target item are determined in real time based on the following steps: The path curvature is determined based on the geometry of the guide lines in the RGB image within the navigation elements; The vehicle pitch angle is determined based on the attitude information in the vehicle status information, and the road slope is determined based on the vehicle pitch angle; Based on the path curvature, the road surface slope, and the speed information in the vehicle status information, the dynamic adjustment parameters corresponding to the target item are determined by searching the preset parameter lookup table corresponding to the target item.

8. An autonomous driving system, characterized in that, The system includes: The data acquisition module is used to acquire environmental perception data and motion state data corresponding to the vehicle. The environmental perception data includes an RGB image and depth information corresponding to the RGB image. The motion state data includes the angular velocity, acceleration and wheel rotation information of the vehicle. The target detection module is used to perform target detection on the RGB image to determine navigation elements in the RGB image, wherein the navigation elements include guide lines and / or navigation target points; The key point projection module is used to project the key points of the navigation element according to the RGB image and the depth information to obtain the two-dimensional horizontal coordinates of the key points in a pre-constructed world coordinate system; The vehicle status determination module is used to determine the vehicle status information corresponding to the vehicle based on the two-dimensional horizontal coordinates and the motion status data, wherein the vehicle status information includes the vehicle's position information, speed information and attitude information; The control command generation module is used to generate vehicle motion control commands based on the navigation elements and the vehicle status information through a preset PID controller. The control module is used to control the motion of the vehicle according to the vehicle motion control command.

9. A terminal, characterized in that, The terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the autonomous driving method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the autonomous driving method as described in any one of claims 1 to 7.

Images