Modeling method, system and application of automatic driving road model

Abstract

The invention belongs to the technical field of vehicle automatic driving, and provides an automatic driving road model modeling method and system and application, and the method comprises the steps: building a road model equation; building a recursive iteration framework of Kalman filtering, embedding the road model equation into the recursive iteration framework, and building a road model state equation; differentiated verification and sampling are carried out on the multi-source sensing information, so that all the information is uniformly converted to transverse and longitudinal position sampling points under a self-vehicle coordinate system; the method comprises the following steps: initializing a road model state quantity estimator, predicting the state and error covariance at the current moment based on a road model state equation, iteratively updating a plurality of sampling points based on a sequential Kalman filtering algorithm to obtain the optimal estimation of the road model state quantity estimator, and substituting the optimal estimation into the road model state equation to establish a road model. Through data format unification, dynamic verification screening and point-by-point iteration updating, the problem of multi-source data fusion is solved, and the accuracy and stability of a road model are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle automatic driving, in particular to an automatic driving road model modeling method and system and application. BACKGROUND

[0002] Among the many key technologies of automatic driving, the road model plays an indispensable role. Through the road model, the automatic driving vehicle can obtain the road conditions for driving, and can better make related decisions such as path planning, speed control, and collision detection. The accuracy and stability of the road model directly affect the quality of automatic driving decisions. Based on an accurate road model, the automatic driving vehicle can obtain the curvature of the front curve, slow down in advance and make corresponding lateral control, so as to safely and smoothly pass the curve. At the same time, in the scene of multiple vehicle interactions, the road model also plays a crucial role. Through the calculation of the positional relationship between surrounding vehicles and the road model, the automatic driving vehicle can predict the driving trajectory of the surrounding vehicles and decide whether to change lanes to overtake or slow down to follow.

[0003] With the advancement of technology, the data sources for constructing road models have become increasingly diverse, but how to effectively fuse these information has become a new challenge. The lane line information output by the camera perception contains type and geometric shape information, which was once the only source of road model information. With the development of automatic driving technology, high-precision maps containing rich road information provide more accurate and comprehensive reference for road models; the related technology of millimeter wave radar road edge detection provides edge information of the road, enhancing the edge information of the road model; and even the driving trajectory of surrounding vehicles can also become an important information source after reasonable screening. However, if these data are not reasonably utilized, it may lead to insufficient stability of the road model. For example, relying solely on visual lane lines may lead to misjudgment in poor lighting or lane line wear conditions; using high-precision map information alone may lead to dangerous situations where the data do not match the actual road in the case of inaccurate self-positioning. Reasonable use of various data sources can fully utilize their advantages and improve the accuracy and robustness of the road model. Through multi-source perception data, information complementation can be achieved to construct a more comprehensive and accurate road model. However, these data sources each have different characteristics and data formats, and the input types differ greatly, leading to difficulties in information fusion and complex modeling. If the data of these perception sources are not reasonably utilized, it may lead to inaccurate or incomplete road models, and even cause vehicle safety accidents.

[0004] To solve the problem of multi-source data fusion, relevant technical personnel have developed various fusion schemes. For example, the patent with publication number CN117649583B discloses an automatic driving vehicle driving real-time road model fusion method, which includes the following steps: obtaining visual perception parameters Is through a visual perception system of a vehicle, obtaining millimeter wave radar perception parameters Ih through a millimeter wave radar of the vehicle, and obtaining vehicle driving parameters Ix through a vehicle information system; obtaining visual information road model Ms based on the visual perception parameters Is, obtaining radar information road model Ms based on the millimeter wave radar perception parameters Ih, and obtaining driving information road model Mx based on the vehicle driving parameters Ix; fusing the visual information road model Ms, the radar information road model Ms, and the driving information road model Mx to obtain a fused comprehensive information road model. However, this method needs to model the three types of data respectively to obtain their respective road models before fusion, and the modeling method is complex and has poor universality.

[0005] For example, the patent with publication number CN114194219B discloses an automatic driving vehicle driving road model prediction method, which includes the following steps: initializing perception information, updating the vehicle driving road model based on road information, a high-precision map and a navigation map, and vehicle perception system perception information; updating the current time road model parameter information based on the above road information, high-precision map and navigation map, and the perception input of the vehicle perception system, and determining the type and confidence value of the current time road model based on the input perception type, obtaining the calculation result as the initial value for the next time vehicle driving road model update, and obtaining the prior predictive vehicle driving road model combined with the vehicle motion state, and combining the sensor input of the next time, repeating the above steps of road information, high-precision map and navigation map, and vehicle perception system perception update to obtain the final vehicle driving road model. However, this method has high requirements for the computing power and real-time performance of the vehicle-mounted computing platform, and has high complexity and computing burden, so the algorithm speed is limited in actual application. SUMMARY

[0006] In view of the above-mentioned shortcomings of the prior art, the present application provides an automatic driving road model modeling method, system and application, which solves the problem of multi-source data fusion by unifying the data format, dynamically checking and screening, and updating point by point, and improves the accuracy and stability of the road model.

[0007] To achieve the above and related purposes, the present application adopts the following technical solutions:

[0008] The first aspect of the present application provides an automatic driving road model modeling method, which includes the following steps:

[0009] Step S100: Based on the lateral offset, heading angle deviation, curvature and rate of curvature change of the road, establish the road model equations regarding the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system.

[0010] Step S200: Based on the sequential Kalman filter algorithm, a recursive iterative framework for Kalman filtering is built, and the road model equation is embedded in the recursive iterative framework to establish a road model state equation that can handle time-varying sequential data.

[0011] Step S300: Based on the input type, perform differential verification and sampling of multi-source perception information to uniformly convert all information to lateral and longitudinal position sampling points in the vehicle coordinate system;

[0012] Step S400: Initialize the road model state quantity estimator, predict the current state and error covariance based on the road model state equation, iteratively update multiple sampling points based on the sequential Kalman filter algorithm, obtain the optimal estimate of the road model state quantity estimator, and substitute it into the road model state equation to establish the road model.

[0013] Furthermore, in step S200, the road model state equations include state transition equations and observation equations:

[0014] (Formula 1),

[0015] In Formula 1, θ k A represents the road model state vector at time k; k-1 Represents the state transition matrix; θ k-1 B represents the road model state vector at time k-1; k-1 Represents the control input matrix; u k-1 Represents the control vector; w k-1 This represents process Gaussian white noise with zero mean and variance Q; v k Z represents observed Gaussian white noise with zero mean and variance R; k Represents the observation vector; C k This represents the observation matrix.

[0016] Furthermore, the state transition matrix includes:

[0017] (Formula 2),

[0018] In Formula 2, d represents the distance traveled by the vehicle within the update cycle;

[0019] Control vectors include:

[0020] (Formula 3),

[0021] In formula 3, This indicates the change in the vehicle's heading angle deviation within the update cycle;

[0022] The observation matrix includes:

[0023] (Formula 4),

[0024] In Formula 4, x represents the longitudinal distance of the observation point in the vehicle coordinate system.

[0025] Furthermore, in step S300, the multi-source perception information includes visual lane lines, high-precision map information of the vehicle, target vehicle, radar waypoints, and stable driving trajectory points.

[0026] Furthermore, in step S300, the differential verification and sampling include: after verifying the confidence, parallelism and curvature of the visual lane lines, performing sparse sampling at equal intervals on the lane lines that meet the quality requirements.

[0027] After verifying the high-precision map information of the vehicle, sampling is performed at fixed longitudinal distance intervals;

[0028] The stability and lane-changing behavior of the target vehicle are verified, the trajectory that meets the trajectory conditions is selected for curve fitting, and sparse sampling is performed on the fitted curve.

[0029] After verifying the signal-to-noise ratio and continuity of the radar curb points, the qualified curb points are fitted into a curve and sampled at intervals.

[0030] Further, step S400, initialization is performed at time k=0, including:

[0031] (Formula 5),

[0032] In formula 5, This indicates that the initialization is performed using the first frame of input data sensed by the camera. This indicates the initialization error set based on the camera's sensing accuracy.

[0033] Further, in step S400, the iterative update is performed at times k=1,2,......,N, where N is a positive integer greater than 1, including:

[0034] The current state and error covariance predicted based on the road model state equation include:

[0035] (Formula 6),

[0036] In formula 6, , These are the prior estimates of the state variables and the posterior update values ​​from the previous time step, respectively. This represents the prior estimation error covariance; Let represent the posterior estimation error covariance at time k-1; Represents the process noise covariance matrix;

[0037] Iterative updates of multiple sampling points based on the sequential Kalman filter algorithm include:

[0038] (Formula 7);

[0039] (Formula 8);

[0040] (Formula 9);

[0041] In formula 7, Indicates Kalman gain; This represents the prior error covariance before processing the i-th point; Represents the observation noise covariance matrix; Represents the observation matrix;

[0042] In formula 8, This represents the posterior state estimate after processing the i-th point; This represents the prior state estimate before processing the i-th point; This represents the actual measured value at the i-th observation point;

[0043] In formula 9, This represents the posterior error covariance after processing the i-th point; Represents the identity matrix.

[0044] A second aspect of the present invention provides an autonomous driving road modeling system, comprising:

[0045] The road model equation establishment module is used to establish road model equations about the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system based on the road's lateral offset, heading angle deviation, curvature, and rate of curvature change.

[0046] The road model state equation establishment module is used to build a recursive iterative framework for Kalman filtering based on the sequential Kalman filtering algorithm. The road model equation is embedded in the recursive iterative framework to establish a road model state equation that can handle time-varying sequential data.

[0047] The verification and sampling module is used to perform differentiated verification and sampling of multi-source sensing information based on the input type, so as to uniformly convert all information to the lateral and longitudinal position sampling points in the vehicle coordinate system.

[0048] The road model building module is used to initialize the road model state variable estimator, predict the current state and error covariance based on the road model state equation, iterate and update multiple sampling points based on the sequential Kalman filter algorithm to obtain the optimal estimate of the road model state variable estimator, and substitute it into the road model state equation to build the road model.

[0049] A third aspect of the present invention provides a computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by a computer's processor, cause the computer to perform the above-described autonomous driving road model modeling method.

[0050] A fourth aspect of the present invention provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described autonomous driving road model modeling method.

[0051] The beneficial technical effects of this invention are as follows:

[0052] This invention uses the lateral and longitudinal positions in the vehicle coordinate system as an intermediate carrier to convert multi-source data such as camera vision, high-precision maps, and millimeter-wave radar into a unified format, thus overcoming the format barriers in multi-source information fusion. It performs differentiated verification based on the characteristics of different data sources and reduces redundant data through sparse sampling, improving real-time performance while ensuring model accuracy and resolving fusion bias issues caused by differences in the reliability of multi-source data. Furthermore, the model update employs a sequential Kalman filter algorithm to perform "point-by-point updates" of observation points, replacing the "batch updates" of traditional Kalman filtering. This not only adapts to situations where the number of observations varies across different data sources but also reduces the computational load of large matrix inversion operations, enabling efficient operation even on embedded platforms with limited computing resources or low-cost hardware devices.

[0053] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0054] The accompanying drawings, incorporated in and forming part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without inventive effort. In the drawings:

[0055] Figure 1 This is a flowchart of the autonomous driving road modeling method of this application;

[0056] Figure 2 This is a flowchart of the road model state variable estimation process in this application;

[0057] Figure 3 This is a schematic diagram of multi-source sensing input sampling in this application;

[0058] Figure 4 This is a framework diagram of the autonomous driving road modeling system for this application;

[0059] Figure 5 A schematic diagram of the structure of a computer system suitable for an embodiment of this application is shown. Detailed Implementation

[0060] Unless otherwise defined, all technical and / or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should be understood that certain features of the invention (described in the context of separate embodiments for clarity) may also be provided in a single embodiment. Conversely, multiple features of the invention (described in the context of a single embodiment for brevity) may also be provided separately or in any suitable combination or, where appropriate, in any other described embodiment of the invention. Certain features described in the context of various embodiments will not be considered essential features of those embodiments unless the embodiment is inoperable without those elements. The invention is further illustrated below by specific examples; however, it should be noted that the specific process conditions and results described in the embodiments of the invention are merely illustrative and should not be construed as limiting the scope of protection of the invention. All equivalent changes or modifications made in accordance with the spirit and essence of the invention should be covered within the scope of protection of the invention.

[0061] Please see Figure 1 The flowchart of the autonomous driving road modeling method of this application is as follows:

[0062] Step S100: Based on the lateral offset, heading angle deviation, curvature, and rate of curvature change of the road, establish the road model equations regarding the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system.

[0063] Specifically, the expression for the road model equation in this application is as follows:

[0064] (Formula 10);

[0065] In Formula 10, offset represents lateral offset; heading represents heading angle deviation; c0 represents curvature; c1 represents the rate of change of curvature; x is the longitudinal position in the vehicle coordinate system; and y is the lateral position in the vehicle coordinate system.

[0066] More specifically, in this application, the lateral offset is the lateral distance of the road centerline relative to the origin of the vehicle's coordinate system from the vehicle's current position (x=0). The heading angle deviation is the angle between the tangent direction of the road centerline at the vehicle's current position (x=0) and the vehicle's longitudinal axis (i.e., the X-axis of the vehicle's coordinate system). Curvature and the rate of change of curvature directly describe the degree of curvature of the road centerline itself and how fast it changes; they are geometric characteristics of the road.

[0067] Step S200: Based on the sequential Kalman filter algorithm, a recursive iterative framework for Kalman filtering is built. The road model equation is embedded in the recursive iterative framework to establish a road model state equation that can handle time-varying sequential data.

[0068] Specifically, the state equations of the road model in this application include state transition equations and observation equations:

[0069] (Formula 1),

[0070] In Formula 1, θ k A represents the road model state vector at time k; k-1 Represents the state transition matrix; θ k-1 B represents the road model state vector at time k-1; k-1 Represents the control input matrix; u k-1 Represents the control vector; w k-1 This represents process Gaussian white noise with zero mean and variance Q; v k Z represents observed Gaussian white noise with zero mean and variance R; k Represents the observation vector; C k This represents the observation matrix.

[0071] Furthermore, the state transition matrix includes:

[0072] (Formula 2),

[0073] In Formula 2, d represents the distance traveled by the vehicle within the update cycle;

[0074] Control vectors include:

[0075] (Formula 3),

[0076] In formula 3, This indicates the change in the vehicle's heading angle deviation within the update cycle;

[0077] The observation matrix includes:

[0078] (Formula 4),

[0079] In Formula 4, x represents the longitudinal distance of the observation point in the vehicle coordinate system.

[0080] Furthermore, the road model state vector at time k includes:

[0081] (Formula 11);

[0082] The control input matrix includes:

[0083] (Formula 12);

[0084] In formula 12, Represents the identity matrix.

[0085] More specifically, this application establishes a road model state equation based on the static road model equation established in step S100, using a sequential Kalman filter algorithm. The state transition equation describes the automatic evolution of the road model equation parameters over time (vehicle motion). The elements in the state transition matrix are derived from the vehicle's travel distance d, thus reflecting the intrinsic dynamic relationship between vehicle motion and road parameter changes. An observation equation is used to establish a mathematical connection between road parameters and external sensor observations, where the observation matrix is ​​used to convert state variables into theoretically observable values. This application utilizes the road model state equation to improve the robustness of the system model, enabling the algorithm to tolerate occasional false detections, missed detections, or data fluctuations from sensors, resulting in more stable and reliable output results. Furthermore, by using the state transition matrix (based on vehicle motion d) for prediction and then using the observation matrix to correct the prediction using measured data, this "prediction-correction" closed-loop mechanism effectively reduces estimation errors and continuously improves the accuracy of road parameter estimation.

[0086] Step S300: Based on the input type, perform differential verification and sampling of multi-source perception information to uniformly convert all information to lateral and longitudinal position sampling points in the vehicle coordinate system.

[0087] Specifically, the multi-source perception information in this application includes visual lane lines, high-precision map information of the vehicle, target vehicles, radar waypoints, and stable driving trajectory points. Differential verification and sampling include: after verifying the confidence, parallelism, and curvature of the visual lane lines, sparsely sampling the qualified lane lines at equal intervals; after verifying the high-precision map information of the vehicle, sampling at fixed longitudinal distance intervals; verifying the stability and lane-changing behavior of the target vehicle, selecting trajectories that meet the trajectory conditions, performing curve fitting, and sparsely sampling on the fitted curve; after verifying the signal-to-noise ratio and continuity of the radar roadside points, fitting qualified roadside points into a curve and sampling at intervals.

[0088] More specifically, in combination Figure 3The sampling of visual lane line input in this application includes the following steps: First, the visual lane line is verified, including confidence, parallelism, curvature, etc., to obtain lane lines with good quality that conform to the actual road; then, sparse sampling is performed on the selected lane lines, and the horizontal and vertical positions of 8 sampling points are obtained at equal intervals for each lane line to obtain lane line sampling points.

[0089] More specifically, the sampling of high-precision map information input for vehicles in this application includes the following steps: First, the position of the vehicle on the high-precision map is obtained, and the high-precision map information of the front and rear 200m of the vehicle is extracted; then, after verifying the data of the high-precision map, a point is obtained every 10m along the longitudinal distance and converted to a sampling point under the coordinates of the vehicle.

[0090] More specifically, the sampling of the target vehicle input in this application includes the following steps: First, the motion trajectory of the vehicle in front after fusion perception tracking is recorded. By cross-validating and filtering the trajectories of multiple target vehicles in front, the driving trajectory with good stability, no lane change within 3 seconds, and similar to the road conditions in front is selected. Then, the trajectory is subjected to cubic curve fitting, and sparse sampling is performed at equal intervals on the fitted curve to extract 4 sampling points as auxiliary sampling points for road model updating.

[0091] More specifically, the sampling of radar curb point input in this application includes the following steps: first, extract curb points with high signal-to-noise ratio and continuous flow (lateral distance deviation less than 0.3m) from the radar point cloud; then, fit the curb points into a cubic curve and sample 4 points at equal intervals on it to obtain curb sampling points.

[0092] More specifically, for autonomous vehicles without rearward perception, sampling points at the location where the longitudinal distance of the visual lane line is 0 can be stored in a sliding window container with a window size of 8. As the vehicle moves, the positions of the points in the container are updated every 5 meters and then input into the rearward road model estimator for further updates, thereby compensating for the problem of incomplete road models caused by the lack of rearward perception.

[0093] More specifically, this application can also sample the trajectory points of the vehicle when it is driving stably, that is, without changing lanes or turning, and input them into the backward road model for updating.

[0094] More specifically, this application supports flexible access to new data sources, such as LiDAR point clouds, requiring only the addition of a corresponding verification and sampling module without modifying the core logic.

[0095] Step S400: Initialize the road model state quantity estimator, predict the current state and error covariance based on the road model state equation, iteratively update multiple sampling points based on the sequential Kalman filter algorithm, obtain the optimal estimate of the road model state quantity estimator, and substitute it into the road model state equation to establish the road model.

[0096] Specifically, the road model updated in this application is the lane centerline of the lane where the vehicle is located and the adjacent lane.

[0097] Specifically, in combination Figure 2 Initialization is performed at time k=0, including:

[0098] (Formula 5),

[0099] In formula 5, This indicates that the initialization is performed using the first frame of input data sensed by the camera. This indicates the initialization error set based on the camera's sensing accuracy.

[0100] Further, in step S400, the iterative update is performed at times k=1,2,......,N, where N is a positive integer greater than 1, including:

[0101] The current state and error covariance predicted based on the road model state equation include:

[0102] (Formula 6),

[0103] In formula 6, , These are the prior estimates of the state variables and the posterior update values ​​from the previous time step, respectively. This represents the prior estimation error covariance; Let represent the posterior estimation error covariance at time k-1; Represents the process noise covariance matrix;

[0104] Iterative updates of multiple sampling points based on the sequential Kalman filter algorithm include:

[0105] (Formula 7);

[0106] (Formula 8);

[0107] (Formula 9);

[0108] In Formula 7 (Kalman Gain Update), Indicates Kalman gain; This represents the prior error covariance before processing the i-th point; Represents the observation noise covariance matrix; Represents the observation matrix;

[0109] In Formula 8 (Posterior State Estimation Update), This represents the posterior state estimate after processing the i-th point; This represents the prior state estimate before processing the i-th point; This represents the actual measured value at the i-th observation point;

[0110] In Formula 9 (Update of Covariance of Posterior State Quantity Estimation Error), This represents the posterior error covariance after processing the i-th point; Represents the identity matrix.

[0111] More specifically, this application receives sample point data converted to a unified format, and optimally fuses the current prediction and the latest observation through the prediction-update mechanism of Kalman filtering, gradually converging to the most accurate estimate of the true state of the road. Sequential processing is used for computational efficiency and can asynchronously process observation data arriving from different sources and at different times. Therefore, this application uses a sequential Kalman filtering algorithm to iteratively fuse multi-source sensing data, calculate the optimal estimates of the state parameters describing the road geometry, and ultimately establish a lane-level road model that can be used for vehicle planning and control.

[0112] More specifically, this application employs a unified mapping mechanism for multi-source data, transforming various data sources with significantly different formats into standardized sampling points, thus resolving the fusion difficulties caused by data source differences; it also utilizes a dynamic adaptability design based on sequential Kalman filtering. Point-by-point updates replace traditional batch updates, supporting dynamically varying amounts of data input, avoiding large matrix inversion operations, reducing computational load, and improving adaptability to multi-source data input; finally, it employs a supplementary mechanism for the backward road model. For vehicles lacking backward perception, a sliding window is used to store the vehicle's historical trajectory and coordinates, dynamically updating them to address the incomplete road model caused by the lack of backward perception.

[0113] Please see Figure 4 The diagram shows the framework of the autonomous driving road modeling system 400 of this application, including:

[0114] The road model equation establishment module 410 is used to establish road model equations about the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system based on the lateral offset, heading angle deviation, curvature and rate of curvature change of the road.

[0115] The road model state equation establishment module 420 is used to build a recursive iterative framework for Kalman filtering based on the sequential Kalman filtering algorithm, embed the road model equation into the recursive iterative framework, and establish a road model state equation that can handle time-varying sequential data.

[0116] The verification and sampling module 430 is used to perform differentiated verification and sampling of multi-source perception information based on the input type, so as to uniformly convert all information to the lateral and longitudinal position sampling points in the vehicle coordinate system.

[0117] The road model building module 440 is used to initialize the road model state quantity estimator, predict the current state and error covariance based on the road model state equation, iteratively update multiple sampling points based on the sequential Kalman filter algorithm, obtain the optimal estimate of the road model state quantity estimator, and substitute it into the road model state equation to build the road model.

[0118] It should be noted that the autonomous driving road modeling system and the autonomous driving road modeling method provided in the above embodiments belong to the same concept. The specific methods of operation of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the autonomous driving road modeling system provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0119] Embodiments of this application also provide a computer device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the computer device to implement the autonomous driving road model modeling method provided in the above embodiments.

[0120] Figure 5 A schematic diagram of the structure of a computer system suitable for an embodiment of this application is shown. It should be noted that... Figure 5 The computer system 500 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0121] like Figure 5As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage section 508 into a random access memory (RAM) 503, such as performing the methods described in the above embodiments. Various programs and data required for system operation are also stored in the RAM 503. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504. The following components are connected to the I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN (local area network) card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A driver 510 is also connected to the I / O interface 505 as needed. Removable media 511, such as disks, optical discs, magneto-optical discs, semiconductor memories, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0122] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer tool programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs various functions defined in the system of this application.

[0123] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, flash memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. Computer programs contained on computer-readable media can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0124] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0125] The units described in the embodiments of this application can be implemented by tools or by hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the unit itself.

[0126] Another aspect of this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a computer's processor, causes the computer to perform the autonomous driving road model modeling method as described above. This computer-readable storage medium may be included in the computer device described in the above embodiments, or it may exist independently and not incorporated into the computer device.

[0127] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the autonomous driving road modeling method provided in the various embodiments described above.

[0128] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for modeling road models for autonomous driving, characterized in that, Includes the following steps: Step S100: Based on the lateral offset, heading angle deviation, curvature and rate of curvature change of the road, establish the road model equations regarding the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system. Step S200: Based on the sequential Kalman filter algorithm, a recursive iterative framework for Kalman filtering is built, and the road model equation is embedded in the recursive iterative framework to establish a road model state equation that can handle time-varying sequential data. Step S300: Based on the input type, perform differential verification and sampling of multi-source perception information to uniformly convert all information to lateral and longitudinal position sampling points in the vehicle coordinate system; Step S400: Initialize the road model state quantity estimator, predict the current state and error covariance based on the road model state equation, iteratively update multiple sampling points based on the sequential Kalman filter algorithm, obtain the optimal estimate of the road model state quantity estimator, and substitute it into the road model state equation to establish the road model.

2. The modeling method according to claim 1, characterized in that, In step S200, the road model state equation includes a state transition equation and an observation equation: (Formula 1), In Formula 1, θ k A represents the road model state vector at time k; k-1 Represents the state transition matrix; θ k-1 B represents the road model state vector at time k-1; k-1 Represents the control input matrix; u k-1 w represents the control vector. k-1 This represents process Gaussian white noise with zero mean and variance Q; v k Z represents observed Gaussian white noise with zero mean and variance R; k Represents the observation vector; C k This represents the observation matrix.

3. The modeling method according to claim 2, characterized in that, The state transition matrix includes: (Formula 2) In Formula 2, d represents the distance traveled by the vehicle within the update cycle; The control vector includes: (Formula 3) In formula 3, This indicates the change in the vehicle's heading angle deviation within the update cycle; The observation matrix includes: (Formula 4) In Formula 4, x represents the longitudinal distance of the observation point in the vehicle coordinate system.

4. The modeling method according to claim 3, characterized in that, In step S300, the multi-source perception information includes visual lane lines, high-precision map information of the vehicle, target vehicle, radar path points, and stable driving trajectory points.

5. The modeling method according to claim 4, characterized in that, In step S300, the differential verification and sampling include: after verifying the confidence, parallelism and curvature of the visual lane lines, performing sparse sampling at equal intervals on the lane lines that meet the quality requirements. After verifying the high-precision map information of the vehicle, sampling is performed at fixed longitudinal distance intervals; The stability and lane-changing behavior of the target vehicle are verified, the trajectory that meets the trajectory conditions is selected for curve fitting, and sparse sampling is performed on the fitted curve. After verifying the signal-to-noise ratio and continuity of the radar roadside points, the qualified roadside points are fitted into a curve and sampled at intervals.

6. The modeling method according to claim 5, characterized in that, Step S400, the initialization performed at time k=0, includes: (Formula 5) In formula 5, This indicates that the initialization is performed using the first frame of input data sensed by the camera. This indicates the initialization error set based on the camera's sensing accuracy.

7. The modeling method according to claim 6, characterized in that, In step S400, the iterative update is performed at times k=1,2,......,N, where N is a positive integer greater than 1, including: The prediction of the current state and error covariance based on the road model state equation includes: (Formula 6) In formula 6, , These are the prior estimates of the state variables and the posterior update values ​​from the previous time step, respectively. This represents the prior estimation error covariance; Let represent the posterior estimation error covariance at time k-1; Represents the process noise covariance matrix; Iterative updates of multiple sampling points based on the sequential Kalman filter algorithm include: (Formula 7); (Formula 8); (Formula 9); In formula 7, Indicates Kalman gain; This represents the prior error covariance before processing the i-th point; Represents the observation noise covariance matrix; Represents the observation matrix; In formula 8, This represents the posterior state estimate after processing the i-th point; This represents the prior state estimate before processing the i-th point; This represents the actual measured value at the i-th observation point; In formula 9, This represents the posterior error covariance after processing the i-th point; Represents the identity matrix.

8. An autonomous driving road modeling system, characterized in that, include: The road model equation establishment module is used to establish road model equations about the relationship between the lateral and longitudinal positions of the road in the vehicle coordinate system based on the road's lateral offset, heading angle deviation, curvature, and rate of curvature change. The road model state equation establishment module is used to build a recursive iterative framework for Kalman filtering based on the sequential Kalman filtering algorithm, embed the road model equation into the recursive iterative framework, and establish a road model state equation that can handle time-varying sequential data. The verification and sampling module is used to perform differentiated verification and sampling of multi-source sensing information based on the input type, so as to uniformly convert all information to the lateral and longitudinal position sampling points in the vehicle coordinate system. The road model building module is used to initialize the road model state quantity estimator, predict the current state and error covariance based on the road model state equation, iteratively update multiple sampling points based on the sequential Kalman filter algorithm, obtain the optimal estimate of the road model state quantity estimator, and substitute it into the road model state equation to build the road model.

9. A computer-readable storage medium, characterized in that, It stores computer-readable instructions, which, when executed by the computer's processor, cause the computer to perform the autonomous driving road model modeling method as described in any one of claims 1 to 7.

10. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the autonomous driving road model modeling method according to any one of claims 1 to 7.

Images