Deviation correction control method and system for automatic driving

Abstract

The invention discloses an automatic driving deviation correction control method and system, and relates to the technical field of automatic driving, and the method comprises the steps: carrying out the comparative analysis of a dynamically monitored real-time automatic driving path of a target vehicle and a predetermined automatic driving path, and obtaining a target deviation coefficient; reading a predetermined road condition feature, and extracting a first road condition feature; obtaining a first real-time characteristic parameter; extracting a first historical driving record in the historical driving vehicle log; obtaining a first prediction feature parameter; taking a mean value and recording the mean value as a first target parameter; and activating the deviation rectification control decision model to obtain a deviation rectification control decision of the target vehicle on the target road section. The technical problems that in the prior art, an automatic driving vehicle is difficult to deal with driving path deviation caused by dynamic road condition changes, then the driving track cannot be corrected in real time, and the automatic driving path control precision and stability are insufficient are solved, and the technical effect of improving the path control precision and stability of the automatic driving vehicle is achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to autonomous driving, and specifically to an error correction control method and control system for autonomous driving. Background Art

[0002] With the rapid development of autonomous driving technology, autonomous driving systems have shown great potential in improving driving safety and traffic efficiency. However, under actual road conditions, autonomous driving vehicles still face complex problems such as road deviation, road condition changes, and environmental interference. Especially in a changing traffic environment, how to achieve precise control of the autonomous driving path and prevent the vehicle from deviating from the predetermined trajectory is a core challenge. Autonomous driving systems usually include three main links: perception, decision-making, and control. The perception system uses on-board sensors (such as radar, lidar, cameras, etc.) to obtain real-time information about the surrounding environment and build a map of surrounding roads, vehicles, and obstacles; the decision-making system calculates the best driving route and behavior instructions based on the perception data and the predetermined path planning model; and the control system finally converts the decision into the actual action of the vehicle. However, factors such as uneven roads, emergencies or dynamic obstacles may still cause the vehicle to deviate from the planned path. Most of the current correction control methods are based on sensor data to correct the vehicle's driving trajectory in real time. The correction amount is usually calculated and the vehicle's control parameters are adjusted through dynamic information detected by on-board sensors, such as the vehicle's offset. Especially when facing complex changes in road conditions, historical driving data, changes in real-time road conditions, and the vehicle's ability to adapt to road changes during driving are ignored. Correction control is difficult to dynamically adjust according to different road conditions.

[0003] Therefore, in the current autonomous driving related technologies, there is a technical problem that autonomous driving vehicles find it difficult to cope with driving path deviations caused by dynamic road condition changes, and the driving trajectory cannot be corrected in real time, resulting in insufficient accuracy and stability of autonomous driving path control. Summary of the invention

[0004] The present application provides an automatic driving correction control method and control system to solve the technical problem in the prior art that automatic driving vehicles have difficulty coping with driving path deviations caused by dynamic road conditions changes, and the driving trajectory cannot be corrected in real time, resulting in insufficient automatic driving path control accuracy and stability, thereby achieving the technical effect of improving the path control accuracy and stability of automatic driving vehicles.

[0005] The present application provides a deviation correction control method for automatic driving, the method comprising: comparing and analyzing a real-time automatic driving path of a target vehicle under dynamic monitoring with a predetermined automatic driving path to obtain a target deviation coefficient of the target vehicle; reading a predetermined road condition feature and extracting a first road condition feature from the predetermined road condition feature; traversing and matching in real-time road information based on the first road condition feature to obtain a first real-time feature parameter, wherein the real-time road information refers to road information of a target section dynamically acquired by a vehicle-mounted sensor mounted on the target vehicle; retrieving a historical vehicle driving log and extracting a first historical driving record of a first vehicle in the historical vehicle driving log on the target section; activating a road condition integration fusion model to analyze the first historical driving feature information obtained by analyzing the first historical driving record to obtain a first predicted feature parameter of the first road condition feature; taking the average of the first real-time feature parameter and the first predicted feature parameter as the first target parameter of the first road condition feature; activating a deviation correction control decision model to analyze a road condition feature parameter set formed based on the first target parameter and the target deviation coefficient to obtain a deviation correction control decision of the target vehicle on the target section.

[0006] In a possible implementation, the correction control method for autonomous driving also performs the following processing: introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined autonomous driving path; obtaining a first real-time driving position under a first tracking according to the two-dimensional geometric constraint; and forming the real-time autonomous driving path based on the first real-time driving position; wherein, introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined autonomous driving path includes: obtaining a predetermined tracking aiming distance; establishing a target vehicle model according to a target size structure of the target vehicle; obtaining a rear axle center point of the target vehicle model, and drawing a first tracking circle with the rear axle center point as the center and the predetermined tracking aiming distance as the radius; recording the intersection of the first tracking circle and the predetermined autonomous driving path as a first tracking aiming point; and according to the predetermined pure tracking strategy, the first tracking aiming point, the rear axle center point, the predetermined tracking aiming distance and the predetermined autonomous driving path constitute the two-dimensional geometric constraint.

[0007] In a possible implementation, the deviation correction control method for autonomous driving further performs the following processing: the predetermined road condition characteristics include curvature, slope, slipperiness and bumpiness.

[0008] In a possible implementation, the error correction control method for autonomous driving further performs the following processing: reading predetermined driving characteristics, and traversing the first historical driving records with the predetermined driving characteristics to obtain the first historical driving characteristics information, wherein the predetermined driving characteristics include at least speed, acceleration, steering wheel angle, and steering angle.

[0009] In a possible implementation, the correction control method for autonomous driving also performs the following processing: obtaining a test record of a test vehicle test of the target vehicle; extracting a first test record from the test record, wherein the first test record corresponds to a first preset road condition; analyzing the first test record to obtain first test driving feature information of the test vehicle; matching a first preset parameter corresponding to the first road condition feature in the first preset road condition feature information of the first preset road condition; performing machine learning on a first data group formed based on the first test driving feature information and the first preset parameter to obtain a first prediction model; and fusing the first prediction model to obtain the road condition integrated fusion model.

[0010] In a possible implementation, the automatic driving correction control method further performs the following processing: the experimental test record refers to the curve automatic driving test record performed by the test vehicle at a predetermined speed threshold.

[0011] In a possible implementation, the autonomous driving correction control method further performs the following processing: determining whether a first correction test result in the first correction test information meets a predetermined correction constraint; if so, obtaining a first correction control decision in the first correction test information; and performing machine learning on a second data group formed based on the first preset road condition feature information and the first correction control decision to obtain the correction control decision model.

[0012] The present application also provides an automatic driving correction control system, including: a path comparison and analysis module, used to compare and analyze the real-time automatic driving path of the target vehicle dynamically monitored with the predetermined automatic driving path, and obtain the target deviation coefficient of the target vehicle; a predetermined road condition feature reading module, used to read the predetermined road condition feature, and extract the first road condition feature in the predetermined road condition feature; a first real-time feature parameter acquisition module, used to traverse and match in the real-time road information based on the first road condition feature, and obtain the first real-time feature parameter, wherein the real-time road information refers to the road information of the target road section dynamically acquired by the vehicle-mounted sensor installed on the target vehicle; a first historical driving record extraction module, used to retrieve the historical driving vehicle log, And extract the first historical driving record of the first vehicle in the historical driving vehicle log on the target section; a first prediction feature parameter acquisition module is used to activate the road condition integration fusion model to analyze the first historical driving feature information obtained by analyzing the first historical driving record, and obtain the first prediction feature parameter of the first road condition feature; a first target parameter acquisition module is used to take the average of the first real-time feature parameter and the first prediction feature parameter as the first target parameter of the first road condition feature; a correction control decision acquisition module is used to activate the correction control decision model to analyze the road condition feature parameter set based on the first target parameter and the target offset coefficient to obtain the correction control decision of the target vehicle on the target section.

[0013] The present application proposes a deviation correction control method and control system for automatic driving, which compares and analyzes the real-time automatic driving path of the target vehicle under dynamic monitoring with the predetermined automatic driving path to obtain the target deviation coefficient; reads the predetermined road condition characteristics and extracts the first road condition characteristics; obtains the first real-time characteristic parameter; extracts the first historical driving record in the historical driving vehicle log; obtains the first predicted characteristic parameter; takes the average as the first target parameter; activates the deviation correction control decision model to obtain the deviation correction control decision of the target vehicle on the target road section. The technical problem that the automatic driving vehicle is difficult to cope with the driving path deviation caused by the dynamic road condition changes, and the driving trajectory cannot be corrected in real time, resulting in insufficient automatic driving path control accuracy and stability, is solved in the prior art, and the technical effect of improving the path control accuracy and stability of the automatic driving vehicle is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A schematic flow chart of a correction control method for autonomous driving provided in an embodiment of the present application.

[0016] Figure 2 A schematic diagram of the structure of an automatic driving correction control system provided in an embodiment of the present application.

[0017] Explanation of the accompanying drawings: path comparison and analysis module 10, predetermined road condition feature reading module 20, first real-time feature parameter acquisition module 30, first historical driving record extraction module 40, first predicted feature parameter acquisition module 50, first target parameter acquisition module 60, deviation correction control decision acquisition module 70. DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0021] The present application provides a method for controlling an error correction in an automatic driving vehicle. Figure 1 As shown, the method includes: Step S100, comparing and analyzing the real-time automatic driving path of the target vehicle under dynamic monitoring with the predetermined automatic driving path to obtain a target deviation coefficient of the target vehicle.

[0022] Preferably, by comparing the current actual driving path of the target vehicle (i.e., the real-time autonomous driving path) with the predetermined ideal driving path (i.e., the predetermined autonomous driving path), the offset of the vehicle on the path is calculated as the target offset coefficient of the target vehicle. Specifically, the real-time autonomous driving path refers to the current driving trajectory of the vehicle monitored and mapped in real time by the sensors carried by the vehicle (such as lidar, camera, GPS, etc.). This path is dynamically updated and will be continuously adjusted as the position of the vehicle changes during driving. The predetermined autonomous driving path refers to the ideal driving route pre-set according to factors such as road planning and road condition information, and does not change with the real-time situation. By comparing the deviation (such as lateral deviation or longitudinal deviation) between the real-time path and the predetermined path, a value is obtained, namely the target offset coefficient, which represents the degree to which the vehicle deviates from the predetermined path on the current road section. The size and direction of the target offset coefficient usually reflect the direction and degree of the current deviation of the vehicle. For example, if the target vehicle deviates from the left side of the predetermined path, the target offset coefficient may be a negative value, indicating that the vehicle deviates to the left. If the vehicle deviates from the right side, the target offset coefficient may be a positive value.

[0023] Furthermore, step S100 also includes step S110, introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined autonomous driving path; step S120, obtaining a first real-time driving position under a first tracking according to the two-dimensional geometric constraint; and step S130, establishing the real-time autonomous driving path based on the first real-time driving position.

[0024] Preferably, the automatic driving system uses geometric constraints and tracking strategies to achieve precise control of the vehicle position to ensure that the vehicle travels along a predetermined trajectory. Specifically, a pure tracking strategy is a control method applied to path tracking to keep the target vehicle traveling along a predetermined path. The strategy calculates the deviation between the vehicle and the predetermined path and controls the vehicle's steering, speed and other parameters to maintain a match with the predetermined path. The pure tracking strategy is often based on the path error (the deviation between the vehicle's current position and the predetermined path) and combined with two-dimensional geometric constraints to determine how to adjust the vehicle's driving trajectory. Among them, the two-dimensional geometric constraints refer to the spatial constraints of the target vehicle relative to the predetermined automatic driving path. For example, the target vehicle is traveling on a curved road, and the error between the vehicle and the path (such as distance error, angle error) constitutes this geometric constraint.

[0025] Preferably, the position of the vehicle at the current moment (or a certain moment) is calculated according to the two-dimensional geometric constraints, and the deviation between the target vehicle and the predetermined path is minimized through a pure tracking strategy, and the first real-time driving position of the vehicle is calculated, which represents the precise position of the vehicle relative to the predetermined path at a specific point in time, which is not only the coordinate position of the vehicle on the road, but also includes the angle, direction and other information of the vehicle relative to the predetermined path; then a real-time automatic driving path is established based on the first real-time driving position, that is, a driving trajectory formed by dynamic adjustment based on the current position of the vehicle and the driving direction, speed and other information of the vehicle. This path is the target route that is gradually adjusted during the driving process of the vehicle, ensuring that the vehicle can always remain on the predetermined trajectory under actual road conditions, ensuring that the automatic driving system can flexibly respond to different real-time situations (such as obstacles ahead, temporary traffic signs, movement in front of the vehicle, etc.), while ensuring that the vehicle always maintains a safe and stable driving state, and continuously optimizing the path to ensure that it reaches the destination.

[0026] Furthermore, step S110 also includes step S111, obtaining a predetermined tracking aiming distance; step S112, establishing a target vehicle model according to the target size structure of the target vehicle; step S113, obtaining the center point of the rear axle of the target vehicle model, and drawing a first tracking circle with the center point of the rear axle as the center and the predetermined tracking aiming distance as the radius; step S114, recording the intersection of the first tracking circle and the predetermined automatic driving path as the first tracking aiming point; step S115, according to the predetermined pure tracking strategy, the two-dimensional geometric constraint is formed by the first tracking aiming point, the center point of the rear axle, the predetermined tracking aiming distance and the predetermined automatic driving path.

[0027] Preferably, the predetermined tracking distance (i.e., target tracking distance) refers to the ideal distance between the vehicle and the target vehicle during driving. The predetermined tracking distance is obtained to ensure that the target vehicle maintains an appropriate safety distance from the vehicle in front, and it is also convenient for the vehicle to track, overtake or adjust its position on the road. The target vehicle model is established through the actual size and geometric structure of the vehicle (such as vehicle length, width, wheelbase, etc.), which helps to better judge the vehicle's motion state and predict its trajectory. The rear axle center point of the target vehicle model, that is, the center position of the rear axle of the target vehicle, is obtained, and then the first tracking circle is drawn with the rear axle center point as the center and the predetermined tracking distance as the radius, representing the ideal distance between the vehicle and the target vehicle that is expected to be maintained during driving. The intersection of the first tracking circle and the predetermined autonomous driving path is recorded as the first tracking aiming point, which indicates the position that the target vehicle should focus on. The system will adjust the vehicle's control strategy according to this aiming point to ensure that it travels along the predetermined path and maintains a reasonable tracking distance with the vehicle in front. Finally, according to the predetermined pure tracking strategy, the first tracking aiming point, the center point of the rear axle, the predetermined tracking aiming distance and the predetermined autonomous driving path constitute a two-dimensional geometric constraint, that is, the spatial constraint of the target vehicle on the predetermined path. Through the two-dimensional geometric constraint, the autonomous driving system can judge the current driving state of the vehicle and calculate the direction, speed and other parameters that need to be adjusted, so as to ensure that the vehicle always stays on the path and maintains a safe distance from the vehicle in front.

[0028] Step S200, reading predetermined road condition features, and extracting a first road condition feature from the predetermined road condition features.

[0029] Step S200 further includes that the predetermined road condition characteristics include curvature, slope, slipperiness and bumpiness.

[0030] Preferably, in the automatic driving control system, predetermined and known road condition information (i.e., predetermined road condition features) is first accessed and obtained, and then a first road condition feature related to the current driving condition is selected or extracted from the predetermined information. Specifically, the predetermined road condition feature refers to road information that has been collected and organized in advance, which may come from map data, road planning, historical traffic flow data, weather conditions, etc., and describes specific conditions on the road, such as curvature, slope, slipperiness, and bumpiness. These predetermined road condition features are usually static and refer to fixed attributes of the road. For example, the predetermined road condition features may include road width, road surface type (such as asphalt, concrete, etc.), whether there are sharp bends, slopes, etc.; according to the needs of the current automatic driving task, one is selected from the predetermined road condition features as the first road condition feature. For example, the target vehicle is about to travel to a complex road section, which may contain sharp bends, intersections, or ups and downs. The automatic driving system will extract relevant features (such as curve radius, slope changes, etc.) from the predetermined road condition features and analyze them as the first road condition feature.

[0031] Step S300, traversing and matching in real-time road information based on the first road condition feature to obtain a first real-time feature parameter, wherein the real-time road information refers to the road information of the target road section dynamically acquired by the on-board sensor installed on the target vehicle.

[0032] Preferably, the extracted predetermined road condition features (such as road curves, slopes, number of lanes, etc.) are compared and matched with the road section information obtained in real time by the on-board sensors carried by the target vehicle, so as to obtain real-time feature parameters related to the current road section. Specifically, the first road condition feature describes the expected state of the road section that the vehicle is about to travel. Real-time road information refers to dynamic information about the current driving section collected in real time by various on-board sensors on the vehicle (such as cameras, lidars, radar sensors, GPS, inertial measurement units, etc.). Unlike the predetermined static road condition features, real-time road information changes dynamically and can reflect the immediate situation of the actual road conditions. For example, real-time road information includes the road conditions monitored in real time (such as whether there is water accumulation or ice and snow), whether there are new traffic signs, whether there are obstacles in front of the vehicle, the real-time speed limit of the road section, the number and width of lanes, changes in the slope of the road, etc.; automatic The driving system compares and matches the first road condition feature with the road information collected in real time by the sensor, and finds the dynamic road condition information that best matches the predetermined road condition feature according to the real-time situation of the current driving section. For example, if the first road condition feature includes a sharp bend or a ramp, the automatic driving system will use the on-board sensor to find the corresponding feature (such as the actual position of the bend, slope change, etc.) in the real-time road information, and perform traversal matching to obtain the first real-time feature parameter, which can be a specific state related to the first road condition feature under the real-time road conditions (such as the radius of the bend, the slope angle, the lane width, etc.). For example, if the first road condition feature is a sharp bend, the real-time road information can be measured by the on-board sensor to obtain the first real-time feature parameter, such as the radius of the bend or the curvature change rate, to help the automatic driving system make more accurate decisions, such as adjusting the vehicle speed, controlling the steering, etc., to ensure the safety and stability of the vehicle.

[0033] Step S400, retrieve a historical vehicle driving log, and extract a first historical driving record of a first vehicle driving on the target road section in the historical vehicle driving log.

[0034] Preferably, historical driving data related to the target road section is retrieved from the system, especially the records of vehicles that have traveled on the same road section before, so as to understand the historical traffic conditions, driving behaviors and road condition characteristics of the road section. Specifically, the historical driving vehicle log refers to the relevant data accumulated by the system during the vehicle's past driving process, which may include the vehicle's driving trajectory, speed, road condition information, driving behavior (such as acceleration, deceleration, steering, etc.), and road surface conditions collected by sensors, etc., reflecting the actual situation of a certain road section or road type. For example, the historical driving vehicle log may include vehicle driving records in the past few months or years, such as the speed change of a certain road section, the road condition of a specific road section, etc. conditions (for example, whether there is construction, whether there are common accident-prone spots on the road section, etc.); and randomly extract the first historical driving record of the first vehicle on the target road section, for example, the driving speed, braking conditions, whether there is deviation when a vehicle enters the road section, and other information. Through this historical record, it is possible to understand the historical driving behavior of the vehicle on the road section, and compare the driving conditions of the current road section with the performance of other vehicles in history to obtain reference information about the actual driving conditions of the road section. For example, if a road section has special traffic patterns (such as traffic congestion, common braking points, etc.), the historical driving records can reflect the frequency of these phenomena and help predict the traffic conditions that the target vehicle may encounter.

[0035] Furthermore, step S400 also includes reading a predetermined driving feature, and traversing the first historical driving record with the predetermined driving feature to obtain the first historical driving feature information, wherein the predetermined driving feature at least includes speed, acceleration, steering wheel angle, and steering angle.

[0036] Preferably, the predetermined driving characteristics include at least speed (vehicle speed during driving, such as the target speed of cruise control), acceleration (vehicle acceleration or deceleration rate, used to describe the speed of vehicle speed change), steering wheel angle (angle of rotation when the driver controls the steering wheel, reflecting the driver's control over the vehicle's driving direction), steering angle (vehicle wheel deflection angle, directly affecting the vehicle's driving trajectory), and the predetermined driving characteristics are used to perform traversal and comparison in the first historical driving record to obtain the first historical driving characteristic information, which reflects the data of the vehicle's actual driving behavior, namely the speed information, acceleration information, steering wheel angle information and steering angle information during the historical driving process, to provide basic data support for subsequent driving decisions, path planning and optimization of vehicle control strategies.

[0037] Step S500: activating the road condition integrated fusion model to analyze the first historical driving feature information obtained by analyzing the first historical driving record to obtain a first predicted feature parameter of the first road condition feature.

[0038] Preferably, the feature data extracted from the historical driving records are deeply analyzed through the road condition integration fusion model to predict the road condition characteristics that may be encountered on the same or similar road sections currently or in the future, and to derive corresponding prediction feature parameters. Specifically, the road condition integration fusion model refers to an integrated comprehensive model for fusing and processing data from various sources. It can combine information from multiple dimensions such as historical driving records, real-time road condition information, sensor data, etc. to predict an accurate analysis result. For example, the road condition integration fusion model may combine the driving characteristics in the historical data with the real-time sensor data (such as cameras, radars, lidars, etc.) to analyze the specific road condition characteristics that may affect vehicle driving, such as curves, slopes, traffic signals, obstacles, etc.; using road condition integration fusion model The road condition integration model analyzes the first historical driving feature information extracted from the first historical driving record, and predicts the first predicted feature parameter of the first road condition feature, that is, the prediction result of the future road condition or the current road condition, which is usually a parameter closely related to the vehicle behavior (such as vehicle speed, steering angle, etc.) and road condition characteristics (such as road surface condition, traffic flow, etc.). For example, the historical driving record shows that there are usually sharp turns on a certain road section. The road condition fusion integration model may predict that there will still be sharp turns on this road section in the future. The predicted feature parameters reflect the possible changes or performances of these road condition characteristics in the future, which may include the radius of the sharp turn, the distance at which the turn appears, etc., which helps the autonomous driving system to make more accurate decisions and optimize the driving strategy, thereby improving driving safety and comfort.

[0039] Furthermore, step S500 also includes step S510, obtaining the test record of the test vehicle test of the target vehicle; step S520, extracting the first test record in the test record, wherein the first test record corresponds to a first preset road condition; step S530, analyzing the first test record to obtain the first test driving characteristic information of the test vehicle; step S540, matching the first preset parameter corresponding to the first road condition characteristic in the first preset road condition characteristic information of the first preset road condition; step S550, performing machine learning on the first data group composed of the first test driving characteristic information and the first preset parameter to obtain a first prediction model; step S560, fusing the first prediction model to obtain the road condition integrated fusion model.

[0040] Preferably, the test record of the test vehicle of the target vehicle is obtained, that is, the test record of the test vehicle used by the autonomous driving system (that is, the vehicle used to collect driving data) under certain conditions, which usually includes the driving data of the vehicle under different road conditions, speeds and driving modes, wherein the test record may include the actual driving parameters of the vehicle, such as speed, acceleration, steering angle, accelerator pedal pressure, etc., and the driving status of the vehicle, such as whether an offset occurs, whether the vehicle deviates from a predetermined path, etc., and data records corresponding to specific preset road conditions are extracted from the test record to help understand the vehicle driving characteristics under different road conditions; the first test driving feature information of the test vehicle is obtained according to the analysis of the first test record, that is, the specific driving characteristics of the test vehicle, such as vehicle speed, acceleration, steering angle and vehicle stability, etc.

[0041] Preferably, the first road condition feature extracted from the first test record (such as the speed, acceleration, steering angle, etc. that appear during the test) is matched with the preset road condition feature to find the first preset parameter corresponding to the matched first road condition feature, and a model is constructed based on a machine learning algorithm (such as regression analysis, neural network, decision tree, etc.). The model is trained according to the extracted first test driving feature information and the data group constructed by the matched first preset parameter to learn the relationship between the historical test data and the preset road condition, and obtain a first prediction model, which can predict the performance of the vehicle under similar road conditions according to the input road condition information, especially how the vehicle adjusts the speed, acceleration, steering angle, etc. Finally, multiple prediction models for different road conditions are fused to obtain a road condition integrated fusion model, which combines the driving prediction results under different road conditions, so that the autonomous driving system can flexibly adjust the vehicle's driving strategy and optimize the driving performance when facing diverse road conditions (such as cities, highways, mountainous areas, etc.), ensuring that the autonomous driving vehicle can make appropriate decisions in a changing traffic environment and achieve a safe and efficient driving experience.

[0042] Furthermore, step S510 also includes that the test record refers to a record of the automatic driving test on a curve performed by the test vehicle at a predetermined speed threshold.

[0043] Preferably, the predetermined speed threshold refers to the target speed range that the vehicle needs to reach under specific test conditions. The test vehicle generally travels at a higher speed, that is, the behavior of the vehicle under high-speed driving conditions is mainly evaluated, such as the vehicle's control response, road adaptability, lane keeping, steering stability, etc. Curves represent complex road conditions, and the performance of the vehicle when driving at high speed in curves directly affects driving safety. The autonomous driving test record refers to the real-time data collected by sensors and recorders during the vehicle test, including vehicle driving data (such as speed, acceleration, steering wheel angle, steering angle, brake pedal position, etc.), such as the vehicle's steering response in curves (steering wheel angle, vehicle roll, turning radius, etc.), stability control, speed control, etc.

[0044] Step S600: taking the average of the first real-time feature parameter and the first predicted feature parameter as the first target parameter of the first road condition feature.

[0045] Preferably, the first real-time feature parameter obtained through real-time road information and the first predicted feature parameter obtained through historical prediction analysis are averaged to obtain a first target parameter of the first road condition feature, which represents the current road condition feature. Specifically, the real-time feature parameter and the predicted feature parameter are averaged, and the real-time situation and the predicted situation are comprehensively considered so that the final target parameter can reflect the current actual state of the road condition, while also taking into account the predictability of the prediction model. For example, the first real-time feature parameter is that the radius of the current curve is 50 meters, and the first predicted feature parameter is that the radius of the curve is predicted to be 55 meters. Then the first target parameter is 52.5 meters, and it is used as the target parameter of the target road condition feature. The first target parameter represents a comprehensive description of a certain road condition feature (such as curves, slopes, traffic flow, etc.), taking into account the influence of real-time and predicted information. The automatic driving system can make more accurate control decisions based on this target parameter, such as vehicle speed adjustment, steering control, etc.

[0046] Step S700, activating the deviation correction control decision model to analyze the road condition characteristic parameter set constructed based on the first target parameter and the target deviation coefficient, and obtaining the deviation correction control decision of the target vehicle on the target road section.

[0047] Preferably, based on the comprehensive road condition characteristics (i.e., the first target parameter), combined with the current vehicle deviation (target deviation coefficient), the deviation correction control decision model is used to make judgments and adjustments, thereby generating a control strategy to correct the vehicle deviation and ensure that the vehicle travels safely along the predetermined path. Specifically, the road condition characteristic parameter set is a parameter set composed of target parameters based on multiple relevant road condition characteristics (such as road form, lane width, curve, traffic signal, road surface condition, etc.), representing the comprehensive characteristics of the target road section, and used to provide comprehensive information support for control decisions. For example, if the target road section has curves, uphill and large traffic flow, the road condition characteristic parameter set may include multiple parameters such as curve radius, slope angle, and vehicle flow. The deviation correction control decision model is started for real-time analysis, and a deviation correction strategy is formulated based on the relationship between the target deviation coefficient and the road condition characteristics to ensure that the target vehicle can return to the predetermined path, wherein the deviation correction control decision model determines whether the vehicle's driving trajectory needs to be adjusted based on the real-time vehicle status, road condition information and deviation coefficient.

[0048] Preferably, the deviation correction control decision model analyzes all data related to the current road conditions, including the real-time characteristics of the target road section, historical predictions, current deviation conditions, etc., to determine whether correction measures need to be taken to correct the deviation. For example, the target deviation coefficient indicates that the target vehicle is deviating from the predetermined lane. The deviation correction control decision model determines whether it is necessary to guide the vehicle back to the center lane by adjusting the steering angle or vehicle speed. The deviation correction control decision model makes corresponding deviation correction control decisions based on the relationship between the target deviation coefficient (i.e., the deviation of the target vehicle) and the road condition characteristic parameter set (the specific conditions of the target road section), that is, how to adjust the vehicle's driving trajectory. For example, if the current road section is a curve and the vehicle deviates from the center of the lane, the system may decide to turn the steering wheel slightly or fine-tune the acceleration to correct the vehicle's deviation, ensuring that the vehicle can safely travel on the predetermined track and avoid collision or loss of control, thereby enhancing the intelligence and adaptability of the autonomous driving system and improving driving safety and driving experience.

[0049] Furthermore, step S700 also includes step S710, determining whether the first correction test result in the first correction test information meets the predetermined correction constraint; step S720, if it does, obtaining the first correction control decision in the first correction test information; and step S730, performing machine learning on the second data group based on the first preset road condition feature information and the first correction control decision to obtain the correction control decision model.

[0050] Preferably, the first deviation correction test information refers to a test performed by a vehicle control system of an autonomous driving vehicle in certain test scenarios to correct the vehicle's deviation from the path (for example, lane deviation, route deviation, etc.). The first deviation correction test result refers to the result of the vehicle's deviation correction based on certain dynamic feedback (such as sensor input, position monitoring, path tracking, etc.) during the test. Specifically, it is determined whether the deviation correction test result meets the predetermined deviation correction constraint. The predetermined deviation correction constraint refers to the performance standard set when performing the deviation correction test. For example, the driving path after correction is consistent with the preset path, and the deviation is less than a predetermined threshold. If the first deviation correction test result meets the predetermined deviation correction constraint, it means that the corrected path is consistent with the preset path. Consistent, obtain the first deviation correction control decision, including how to guide the vehicle back to the correct lane by turning the steering wheel, assist in deviation correction by adjusting the throttle and brake, etc., and then build a decision model based on machine learning (such as regression analysis, neural network, decision tree, etc.), and use the first preset road condition feature information (standardized preset road condition feature data, such as road type, lane width, traffic conditions, road conditions, etc.) and the second data group formed by the first deviation correction control decision (specific operations taken during the deviation correction test, such as steering adjustment, acceleration and deceleration, etc.) to train the model. The model learns how to make correct decisions based on historical deviation correction control decisions under specific road conditions, thereby improving the stability and safety of the autonomous driving system.

[0051] In the above, refer to Figure 1 A correction control method for an automatic driving according to an embodiment of the present invention is described in detail. Figure 2 An automatic driving error correction control system according to an embodiment of the present invention is described.

[0052] An automatic driving correction control system according to an embodiment of the present invention is used to solve the technical problem in the prior art that the automatic driving vehicle is difficult to cope with the driving path deviation caused by dynamic road condition changes, and the driving trajectory cannot be corrected in real time, resulting in insufficient automatic driving path control accuracy and stability, thereby achieving the technical effect of improving the path control accuracy and stability of the automatic driving vehicle. An automatic driving correction control system includes: a path comparison and analysis module 10, a predetermined road condition feature reading module 20, a first real-time feature parameter acquisition module 30, a first historical driving record extraction module 40, a first predicted feature parameter acquisition module 50, a first target parameter acquisition module 60, and a correction control decision acquisition module 70.

[0053] A path comparison and analysis module 10 is used to compare and analyze the real-time automatic driving path of the target vehicle under dynamic monitoring with the predetermined automatic driving path to obtain the target deviation coefficient of the target vehicle; a predetermined road condition feature reading module 20 is used to read the predetermined road condition feature and extract the first road condition feature in the predetermined road condition feature; a first real-time feature parameter acquisition module 30 is used to traverse and match in the real-time road information based on the first road condition feature to obtain the first real-time feature parameter, wherein the real-time road information refers to the road information of the target road section dynamically obtained by the on-board sensor mounted on the target vehicle; a first historical driving record extraction module 40 is used to retrieve the historical driving vehicle log and extract the historical driving vehicle The first historical driving record of the first vehicle in the log on the target road section; the first predicted characteristic parameter acquisition module 50, used to activate the road condition integration fusion model to analyze the first historical driving characteristic information obtained by analyzing the first historical driving record, and obtain the first predicted characteristic parameter of the first road condition characteristic; the first target parameter acquisition module 60, used to take the average of the first real-time characteristic parameter and the first predicted characteristic parameter as the first target parameter of the first road condition characteristic; the correction control decision acquisition module 70, used to activate the correction control decision model to analyze the road condition characteristic parameter set based on the first target parameter and the target offset coefficient, and obtain the correction control decision of the target vehicle on the target road section.

[0054] The specific configuration of the path comparison and analysis module 10 will be described in detail below. The path comparison and analysis module 10 further includes: introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined automatic driving path; obtaining a first real-time driving position under the first tracking according to the two-dimensional geometric constraint; and forming the real-time automatic driving path based on the first real-time driving position; wherein, introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined automatic driving path includes: obtaining a predetermined tracking aiming distance; establishing a target vehicle model according to the target size structure of the target vehicle; obtaining the rear axle center point of the target vehicle model, and drawing a first tracking circle with the rear axle center point as the center and the predetermined tracking aiming distance as the radius; recording the intersection of the first tracking circle and the predetermined automatic driving path as the first tracking aiming point; according to the predetermined pure tracking strategy, the first tracking aiming point, the rear axle center point, the predetermined tracking aiming distance and the predetermined automatic driving path constitute the two-dimensional geometric constraint.

[0055] The following will describe in detail the specific configuration of the predetermined road condition feature reading module 20. The predetermined road condition feature reading module 20 may further include: the predetermined road condition feature includes curvature, slope, slipperiness and bumpiness.

[0056] The specific configuration of the first prediction feature parameter acquisition module 50 will be described in detail below. The first prediction feature parameter acquisition module 50 further includes: reading a predetermined driving feature, and traversing the first historical driving record with the predetermined driving feature to obtain the first historical driving feature information, wherein the predetermined driving feature at least includes speed, acceleration, steering wheel angle, and steering angle.

[0057] The specific configuration of the first prediction characteristic parameter acquisition module 50 will be described in detail below. The first prediction characteristic parameter acquisition module 50 further includes: obtaining the test record of the test vehicle test of the target vehicle; extracting the first test record in the test record, wherein the first test record corresponds to the first preset road condition; analyzing the first test record to obtain the first test driving characteristic information of the test vehicle; matching the first preset parameter corresponding to the first road condition characteristic in the first preset road condition characteristic information of the first preset road condition; performing machine learning on the first data group formed based on the first test driving characteristic information and the first preset parameter to obtain a first prediction model; and fusing the first prediction model to obtain the road condition integrated fusion model.

[0058] The specific configuration of the first prediction characteristic parameter obtaining module 50 will be described in detail below. The first prediction characteristic parameter obtaining module 50 further includes: the test record refers to the curve automatic driving test record performed by the test vehicle at a predetermined speed threshold.

[0059] The specific configuration of the correction control decision acquisition module 70 will be described in detail below. The correction control decision acquisition module 70 further includes: determining whether the first correction test result in the first correction test information meets the predetermined correction constraint; if so, obtaining the first correction control decision in the first correction test information; and performing machine learning on the second data group formed based on the first preset road condition feature information and the first correction control decision to obtain the correction control decision model.

[0060] An automatic driving correction control system provided by an embodiment of the present invention can execute an automatic driving correction control method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0061] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0062] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A deviation correction control method for automatic driving, characterized in that: The method comprises: Comparing and analyzing the real-time automatic driving path of the dynamically monitored target vehicle with the predetermined automatic driving path to obtain a target deviation coefficient of the target vehicle; Reading predetermined road condition features, and extracting a first road condition feature from the predetermined road condition features; Based on the first road condition feature, traversing and matching in real-time road information to obtain a first real-time feature parameter, wherein the real-time road information refers to road information of a target road section dynamically acquired by an on-board sensor mounted on the target vehicle; Retrieving a historical vehicle driving log, and extracting a first historical driving record of a first vehicle driving on the target road section in the historical vehicle driving log; activating the road condition integrated fusion model to analyze the first historical driving feature information obtained by analyzing the first historical driving record to obtain a first predicted feature parameter of the first road condition feature; Taking the average of the first real-time characteristic parameter and the first predicted characteristic parameter as the first target parameter of the first road condition characteristic; The deviation correction control decision model is activated to analyze the road condition characteristic parameter set formed based on the first target parameter and the target deviation coefficient to obtain a deviation correction control decision for the target vehicle on the target road section.

2. The error correction control method for automatic driving according to claim 1, characterized in that: include: Introducing a predetermined pure tracking strategy to construct a two-dimensional geometric constraint between the target vehicle and the predetermined automatic driving path; Obtaining a first real-time driving position under a first tracking according to the two-dimensional geometric constraint; assembling the real-time autonomous driving path based on the first real-time driving position; Wherein, a predetermined pure tracking strategy is introduced to construct a two-dimensional geometric constraint between the target vehicle and the predetermined automatic driving path, including: Get the preset tracking distance; Establishing a target vehicle model according to the target size structure of the target vehicle; Obtaining the center point of the rear axle of the target vehicle model, and drawing a first tracking circle with the center point of the rear axle as the center and the predetermined tracking aiming distance as the radius; Recording the intersection of the first tracking circle and the predetermined automatic driving path as a first tracking aiming point; According to the predetermined pure tracking strategy, the two-dimensional geometric constraint is formed by the first tracking aiming point, the rear axle center point, the predetermined tracking aiming distance and the predetermined automatic driving path.

3. The error correction control method for automatic driving according to claim 1, characterized in that: The predetermined road condition characteristics include curvature, slope, slipperiness and bumpiness.

4. The error correction control method for automatic driving according to claim 1, characterized in that: The predetermined driving feature is read, and the first historical driving record is traversed with the predetermined driving feature to obtain the first historical driving feature information, wherein the predetermined driving feature at least includes speed, acceleration, steering wheel angle, and steering angle.

5. The error correction control method for automatic driving according to claim 4, characterized in that: include: Obtaining a test record of a test vehicle of the target vehicle; Extracting a first test record from the experimental test records, wherein the first test record corresponds to a first preset road condition; Analyzing the first test record to obtain first test driving characteristic information of the test vehicle; Matching a first preset parameter corresponding to the first road condition feature in the first preset road condition feature information; Performing machine learning on a first data set formed based on the first test driving characteristic information and the first preset parameters to obtain a first prediction model; The first prediction model is integrated and constructed to obtain the road condition integrated fusion model.

6. The error correction control method for automatic driving according to claim 5, characterized in that: The experimental test record refers to the curve automatic driving test record performed by the test vehicle at a predetermined speed threshold.

7. The error correction control method for automatic driving according to claim 5, characterized in that: The first test record also includes first deviation correction test information, activating a deviation correction control decision model to analyze a road condition characteristic parameter set based on the first target parameter and the target deviation coefficient, and obtaining a deviation correction control decision for the target vehicle on the target road section, including: Determining whether the first deflection correction test result in the first deflection correction test information meets the predetermined deflection correction constraint; If it is in compliance, obtaining a first deviation correction control decision in the first deviation correction test information; The deviation correction control decision model is obtained by performing machine learning on a second data group formed based on the first preset road condition feature information and the first deviation correction control decision.

8. An automatic driving correction control system, characterized in that: The system is used to implement the automatic driving correction control method according to any one of claims 1 to 7, and the system comprises: A path comparison and analysis module is used to compare and analyze the real-time automatic driving path of the dynamically monitored target vehicle with the predetermined automatic driving path to obtain a target deviation coefficient of the target vehicle; A predetermined road condition feature reading module, used for reading the predetermined road condition features and extracting the first road condition feature from the predetermined road condition features; A first real-time feature parameter acquisition module, configured to traverse and match the real-time road information based on the first road condition feature to obtain a first real-time feature parameter, wherein the real-time road information refers to road information of a target road section dynamically acquired by an on-board sensor mounted on the target vehicle; A first historical driving record extraction module, used to retrieve a historical driving vehicle log, and extract a first historical driving record of a first vehicle in the historical driving vehicle log on the target road section; A first prediction feature parameter acquisition module is used to activate the road condition integrated fusion model to analyze the first historical driving feature information obtained by analyzing the first historical driving record to obtain a first prediction feature parameter of the first road condition feature; A first target parameter obtaining module, used for taking the average of the first real-time characteristic parameter and the first predicted characteristic parameter as the first target parameter of the first road condition characteristic; The deviation correction control decision acquisition module is used to activate the deviation correction control decision model to analyze the road condition characteristic parameter set and the target offset coefficient formed based on the first target parameter to obtain the deviation correction control decision of the target vehicle on the target road section.

Images