Track line processing method and device, equipment and storage medium

Abstract

The invention discloses a trajectory processing method and device, equipment and a storage medium, and relates to the technical field of intelligent driving. The method comprises the following steps: acquiring first intersection environment information; the first intersection environment information is used for indicating characteristics of a first intersection where a vehicle is about to pass; acquiring a first virtual speed based on the first intersection environment information; obtaining a first pre-decision trajectory based on the first intersection environment information and the first virtual speed; the first pre-decision trajectory is used for indicating a reference driving path of the vehicle passing through the first intersection; under the condition that the first pre-decision trajectory meets the precision requirement, determining the first pre-decision trajectory as a target trajectory of the first intersection; the precision requirement is used for indicating the curvature range of the target trajectory. According to the invention, the smoothness of the vehicle trajectory can be ensured.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent driving technology, and in particular to a trajectory processing method, device, equipment and storage medium. Background Art

[0002] With the improvement of computing power, the maturity of sensor technology and the development of artificial intelligence algorithms, intelligent driving is gradually moving from theory to market application.

[0003] Crossroads and other intersections are traffic fortresses where multiple roads converge. Since there are usually no left-turn or right-turn lanes at intersections, it is necessary to provide vehicles with trajectories through intersections during decision-making and planning. In related technologies, lane lines can be obtained through perception and map fusion, and then a Vernet coordinate system is established based on the lane lines to plan the vehicle's path, which is then converted into a Cartesian coordinate system for subsequent operations.

[0004] However, the error of the lane line will affect the stability of the Vernet coordinate system, and then affect the smoothness of the vehicle trajectory. Summary of the invention

[0005] The embodiments of the present application provide a trajectory processing method, device, equipment and storage medium, which can ensure the smoothness of the vehicle trajectory. The technical solution proposed in the present application is as follows:

[0006] According to one aspect of an embodiment of the present application, a trajectory processing method is provided, the method comprising:

[0007] Acquire first intersection environment information; the first intersection environment information is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the first intersection environment information includes at least one of the width of a first lane, the width of a second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane that the vehicle exits through the first intersection, and the second lane is the lane that the vehicle enters through the first intersection;

[0008] Based on the first intersection environment information, a first virtual speed is acquired; the first virtual speed is used to indicate the estimated driving speed of the vehicle passing through the first intersection;

[0009] Based on the first intersection environment information and the first virtual speed, a first pre-decision trajectory is acquired; the first pre-decision trajectory is used to indicate a reference driving path of the vehicle through the first intersection;

[0010] When the first pre-decision trajectory meets the accuracy requirement, the first pre-decision trajectory is determined as the target trajectory of the first intersection; the accuracy requirement is used to indicate the curvature range of the target trajectory, and the target trajectory is used to indicate the driving path of the vehicle through the first intersection.

[0011] According to one aspect of an embodiment of the present application, a trajectory processing device is provided, the device comprising:

[0012] An information acquisition module, used to acquire first intersection environment information; the first intersection environment information is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the first intersection environment information includes at least one of the width of a first lane, the width of a second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane that the vehicle exits through the first intersection, and the second lane is the lane that the vehicle enters through the first intersection;

[0013] A speed acquisition module, configured to acquire a first virtual speed based on the first intersection environment information; the first virtual speed is used to indicate an estimated driving speed of the vehicle passing through the first intersection;

[0014] A path acquisition module, configured to acquire a first pre-decision trajectory line based on the first intersection environment information and the first virtual speed; the first pre-decision trajectory line is used to indicate a reference driving path of the vehicle passing through the first intersection;

[0015] A path determination module is used to determine the first pre-decision trajectory line as the target trajectory line of the first intersection when the first pre-decision trajectory line meets the accuracy requirement; the accuracy requirement is used to indicate the curvature range of the target trajectory line, and the target trajectory line is used to indicate the driving path of the vehicle through the first intersection.

[0016] In some embodiments, the speed acquisition module is used to query the intersection environment information of the second intersection matching the first intersection environment information from the stored information based on the first intersection environment information; the stored information is used to indicate the corresponding relationship between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection;

[0017] The speed acquisition module is used to acquire the first virtual speed based on the virtual speed of the target trajectory line corresponding to the intersection environment information of the second intersection in the stored information.

[0018] In some embodiments, the speed acquisition module is used to input the first intersection environment information into a prediction model to obtain the first virtual speed;

[0019] The prediction model is a machine learning model trained based on sample intersection environment information and sample virtual speed.

[0020] In some embodiments, the trajectory processing device further includes: a training module, configured to input the sample intersection environment information into the prediction model to obtain a predicted virtual speed;

[0021] The training module is used to obtain a loss function value based on the difference between the predicted virtual speed and the sample virtual speed;

[0022] The training module is used to update the parameters of the prediction model based on the loss function value.

[0023] In some embodiments, the trajectory processing device further includes: an adjustment module, configured to obtain a second virtual speed based on the first pre-decision trajectory when the first pre-decision trajectory does not meet the accuracy requirement;

[0024] The adjustment module is used to obtain a second pre-decision trajectory based on the first intersection environment information and the second virtual speed;

[0025] The adjustment module is used to determine the second pre-decision trajectory line as the target trajectory line when it is determined that the second pre-decision trajectory line meets the accuracy requirement.

[0026] In some embodiments, the adjustment module is used to obtain a curvature integral of the first pre-decision trajectory line when the first pre-decision trajectory line does not meet the accuracy requirement;

[0027] The adjustment module is configured to determine the second virtual speed as a virtual speed greater than the first virtual speed when the curvature integral of the first pre-decision trajectory decreases;

[0028] The adjustment module is configured to determine the second virtual speed as a virtual speed less than the first virtual speed when the curvature integral of the first pre-decision trajectory increases;

[0029] The adjustment module is used to obtain the second virtual speed by an average sampling method when the curvature integral of the first pre-decision trajectory line decreases first and then increases.

[0030] According to one aspect of an embodiment of the present application, a computer device is provided, the computer device comprising a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the above-mentioned trajectory processing method.

[0031] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is loaded and executed by a processor to implement the above-mentioned trajectory processing method.

[0032] According to one aspect of an embodiment of the present application, a computer program product is provided. The computer program product includes a computer program. The computer program is loaded and executed by a processor to implement the above-mentioned trajectory processing method.

[0033] The technical solution provided in the embodiments of the present application can bring the following beneficial effects:

[0034] In a scenario where a vehicle is about to automatically pass through an intersection, the virtual speed of the vehicle passing through the first intersection can be predicted based on the obtained intersection environment information; thereafter, the pre-decision trajectory of the vehicle passing through the first intersection can be predicted based on the virtual speed and the intersection environment information; through the curvature range of the pre-stored target trajectory, it is determined whether the curvature of the predicted pre-decision trajectory meets the curvature range, and when the pre-decision trajectory meets the curvature range, the pre-decision trajectory is determined as the target trajectory; this scheme can determine the smoothness of the predicted pre-decision trajectory through the curvature range indicated by the accuracy requirement, thereby ensuring the smoothness of the target trajectory and the stability of the vehicle passing through the first intersection. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 is a schematic diagram of an implementation environment provided by an exemplary embodiment of the present application;

[0037] Figure 2 is a schematic diagram of a vehicle passing through an intersection provided by an exemplary embodiment of the present application;

[0038] Figure 3 is a flow chart of a trajectory processing method provided by an exemplary embodiment of the present application;

[0039] Figure 4 is a schematic diagram of a target trajectory provided by an exemplary embodiment of the present application;

[0040] Figure 5 is a curvature image of an arc trajectory line provided by an exemplary embodiment of the present application;

[0041] Figure 6 is a schematic diagram of a training method and an inference method of a prediction model provided by an exemplary embodiment of the present application;

[0042] Figure 7 is a flow chart of a trajectory processing method provided by an exemplary embodiment of the present application;

[0043] Figure 8 An exemplary embodiment of the present application provides a high-precision ground-based multi-decision maker and an evaluator. Figure 10 Flowchart of the adaptive trajectory generation method for intersections;

[0044] Fig. 9 An exemplary embodiment of the present application provides a high-precision ground-based multi-decision maker and an evaluator. Figure 10 The framework diagram of the adaptive trajectory generation method for intersections;

[0045] Fig.10 is a block diagram of a trajectory processing device provided by an exemplary embodiment of the present application;

[0046] Fig.11 It is a structural block diagram of a computer device provided by an embodiment of the present application.

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0049] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0050] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. The singular forms of "a", "said" and "the" used in this disclosure and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0051] In the embodiments of the present application, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0052] It should be understood that although the terms first, second, etc. may be used in the present disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present disclosure, the first virtual speed may also be referred to as the second virtual speed, and similarly, the second virtual speed may also be referred to as the first virtual speed. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0053] First, several terms involved in this application are introduced and explained:

[0054] 1) Intelligent driving technology: It combines advanced technologies such as AI algorithms, sensors, computer vision, radar, monitoring devices and global positioning systems to enable vehicles to complete driving tasks autonomously without driver intervention. The core process of autonomous driving technology can be divided into three main parts:

[0055] Environmental perception and positioning: refers to obtaining information about the vehicle's surroundings through a variety of sensors such as cameras, lidar, millimeter-wave radar, ultrasonic sensors, etc., and accurately positioning the vehicle in combination with high-precision maps and the Global Positioning System (GPS); these sensors can help the vehicle identify road signs, pedestrians, other vehicles, and other obstacles;

[0056] Decision-making and planning: Based on the collected information, the decision-maker needs to make reasonable judgments and plan the best driving route; the decision-maker also needs to consider factors such as safety, comfort and efficiency to ensure that each operation is the optimal solution;

[0057] Execution control: It converts decisions into actual actions, that is, controlling the vehicle's acceleration, deceleration, steering and other activities; this part of the work is completed by the wire control system, which uses electrical signals instead of traditional mechanical connections to achieve more precise and smooth operation.

[0058] 2) Artificial Intelligence (AI) Algorithms: refers to a class of technologies and methods that enable computers to simulate or realize human intelligence. These technologies enable machines to have the ability to learn, reason, perceive, understand language, solve problems and self-correct, so as to perform similar or better than humans in specific tasks. More specifically, artificial intelligence algorithms are programs that automatically perform operations to achieve specific goals through a series of pre-set rules and instructions. The core of this type of algorithm is that it can not only process information according to established rules, but also autonomously learn patterns from data and use these patterns to make predictions and decisions.

[0059] 3) Machine learning: It is a method that allows computers to autonomously learn patterns from data. It is the most important part of artificial intelligence algorithms. Through continuous training and optimization, machine learning algorithms can gradually master various skills and thus achieve the ability to complete tasks autonomously.

[0060] Please refer to Figure 1 , which shows a schematic diagram of an implementation environment provided by an exemplary embodiment of the present application. Figure 1 As shown, the implementation environment may include: a vehicle 110 and a server 120. The vehicle 110 and the server 120 communicate with each other via a network. Optionally, the vehicle 110 and the server 120 are directly or indirectly connected via wired or wireless communication, which is not limited in this application.

[0061] The vehicle 110 includes an ADCC domain controller 110a, a navigation-level map device 110b, a fused positioning device 110c, and a perception device 110d.

[0062] Among them, the Autonomous Driving Control Center (ADCC) domain controller 110a is a key component designed specifically for the vehicle's intelligent driving system. It is capable of receiving data from multiple sensors, such as cameras, millimeter-wave radars, lidars, etc., and processing these data to support autonomous driving functions. The ADCC domain controller is not only responsible for perceiving the environment, but also undertakes decision-making tasks such as path planning and generation of control instructions. In some implementations, the ADCC domain controller works with the cockpit domain controller (Integrated Cockpit Controller, ICC) to ensure the memory and display consistency of the setting items. In addition, as part of a centralized control system, ADCC can implement all necessary autonomous driving functions, from basic assisted driving to higher levels of fully autonomous operation.

[0063] Among them, the navigation-level map device 110b is a map product specially designed to meet the needs of autonomous driving, providing absolute and relative accuracy at the centimeter level. Exemplarily, the navigation-level map device 110b can be a high-definition map (HD Maps). The navigation-level map device 110b contains detailed geographic information, such as lane geometry, slope, curvature, heading, roll and other parameters, as well as the location of traffic signs and other road facilities. Navigation-level map devices 110b are crucial for intelligent driving because they allow vehicles to accurately locate their own positions and understand the static environment around them.

[0064] The fusion positioning device 110c refers to a device that combines the advantages of multiple positioning technologies and can obtain a more accurate and reliable position estimate than a single method. Exemplarily, the fusion positioning device 110c involves the comprehensive use of multiple technologies such as the Global Navigation Satellite System (GNSS), Bluetooth Low Energy (BLE), Ultra-Wide Band (UWB) wireless communication, and Inertial Measurement Unit (IMU).

[0065] Among them, the perception device 110d plays the role of "eyes" in intelligent driving, providing rich image information about the surrounding world. Exemplarily, the perception device 110d includes a visual sensor (such as a camera); optionally, the perception device 110d also includes other types of sensors, such as a lidar or a millimeter-wave radar. Among them, the camera can identify the color, texture and type of the object, which is very important for distinguishing different traffic participants (such as pedestrians, cyclists) and interpreting the status of traffic lights. Depending on the installation location, the camera can be divided into multiple types such as front view, side view, rear view, built-in and surround view, each serving specific functional requirements, such as Lane Keeping Assist (Lane Keeping Assist, LKA), Blind Spot Detection (Blind Spot Detection, BSD), Driver Monitoring System (Driver Monitoring System, DMS), etc.

[0066] In addition, with the advancement of computer vision algorithms, especially the application of deep learning, perception devices 110d have become an indispensable part of achieving complex perception tasks, including but not limited to target classification, distance estimation, and behavior prediction.

[0067] The server 120 is used to provide background services for the vehicle 110. Optionally, the server 120 is a server, or a server cluster or distributed system composed of multiple servers, or a cloud computing service center, or a cloud server, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, content delivery network (CDN) that provides cloud computing services. Optionally, the server 120 provides background services for multiple vehicles 110 at the same time.

[0068] like Figure 2 As shown, taking the example of vehicle 210a about to turn left at intersection 220, vehicle 210a can obtain intersection environment information by fusing positioning device 110c and perception device 110d or navigation-level map device 110b; the intersection environment information is used to indicate the characteristics of intersection 220 that vehicle 210a is about to pass, and the intersection environment information includes at least one of the width of the first lane 201, the width of the second lane 202, and the positional relationship between the first lane 201 and the second lane 202. Afterwards, vehicle 210a sends the first intersection environment information to server 120 through the network, and server 120 obtains a first virtual speed based on the first intersection environment information, and the first virtual speed is used to indicate the expected driving speed of vehicle 210a through intersection 220; afterward, server 120 obtains a first pre-decision trajectory based on the first intersection environment information and the first virtual speed, and the first pre-decision trajectory is used to indicate a reference driving path for vehicle 210a through intersection 220; when the first pre-decision trajectory meets the accuracy requirement, server 120 determines the first pre-decision trajectory as the target trajectory 203 of intersection 220, and sends the target trajectory 203 of intersection 220 to ADCC domain controller 110a of vehicle 210a; wherein, the accuracy requirement is used to indicate the curvature range of target trajectory 203, and target trajectory 203 is used to indicate the driving path of the vehicle turning left through intersection 220.

[0069] like Figure 2As shown, taking the example of vehicle 210b about to turn right at intersection 220, vehicle 210b can obtain intersection environment information by fusing positioning device 110c and perception device 110d or navigation-level map device 110b; the intersection environment information is used to indicate the characteristics of intersection 220 that vehicle 210b is about to pass, and the intersection environment information includes at least one of the width of the first lane 204, the width of the second lane 205, and the positional relationship between the first lane 204 and the second lane 205. Afterwards, vehicle 210b sends the first intersection environment information to server 120 through the network, and server 120 obtains a first virtual speed based on the first intersection environment information, and the first virtual speed is used to indicate the expected driving speed of vehicle 210b through intersection 220; afterward, server 120 obtains a first pre-decision trajectory based on the first intersection environment information and the first virtual speed, and the first pre-decision trajectory is used to indicate a reference driving path for vehicle 210b through intersection 220; when the first pre-decision trajectory meets the accuracy requirement, server 120 determines the first pre-decision trajectory as the target trajectory 206 of intersection 220, and sends the target trajectory 206 of intersection 220 to ADCC domain controller 110a of vehicle 210b; wherein, the accuracy requirement is used to indicate the curvature range of target trajectory 206, and target trajectory 206 is used to indicate the driving path of the vehicle turning right through intersection 220.

[0070] Please refer to Figure 3 , which shows a flow chart of a trajectory processing method provided by an exemplary embodiment of the present application. The method is executed by a computer device, and optionally, the computer device may be Figure 1 The vehicle 110 in the system shown; or, it can also be the server 120, or, it can also be the vehicle 110 and the server 120. Figure 3 As shown, the method may include steps 310 to 340.

[0071] Step 310: Obtain environmental information of the first intersection; the environmental information of the first intersection is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the environmental information of the first intersection includes at least one of the width of the first lane, the width of the second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane where the vehicle exits from the first intersection, and the second lane is the lane where the vehicle enters from the first intersection.

[0072] In the embodiment of the present application, the vehicle is about to turn left or right at the first intersection. The first intersection may be a crossroads, or other types of intersections such as a T-shaped intersection, a Y-shaped intersection, a roundabout, etc., and the present application does not limit this.

[0073] Optionally, the computer device can capture images and distance data of the surrounding environment through vehicle-mounted sensors (such as cameras, lidars), and thereby obtain the above-mentioned first intersection environmental information; or, the computer device can obtain the above-mentioned first intersection environmental information through static information provided by a high-precision map.

[0074] The positional relationship between the first lane and the second lane may indicate an angle between the first lane and the second lane. For example, taking a standard intersection as an example, the angle between the first lane and the second lane is 90°.

[0075] The first intersection environment information refers to characteristic information such as the geometric shape, size, coordinate position, etc. of the road related to the first intersection. For example, when there is no static obstacle at the first intersection, the computer device can obtain the length and width information of the first intersection and the rotation radius of the first intersection according to the width of the first lane, the width of the second lane, and the positional relationship between the first lane and the second lane.

[0076] Optionally, the first intersection environment information may also include static obstacle information, such as location information of traffic signs, signal lights, cameras, etc. Exemplarily, the computer device may obtain the length, width, and height information of the first intersection based on the width of the first lane, the width of the second lane, the positional relationship between the first lane and the second lane, and the static obstacle information.

[0077] Step 320: Based on the first intersection environment information, a first virtual speed is obtained; the first virtual speed is used to indicate the estimated driving speed of the vehicle passing through the first intersection.

[0078] Optionally, the computer device constructs a memory pool, which is used to output the corresponding virtual speed based on the intersection environment information. That is, the computer device can input the first intersection environment information into the memory pool and use the output result of the memory pool as the first virtual speed.

[0079] Exemplarily, the memory pool stores the correspondence between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection. Exemplarily, the computer device can predict the first virtual speed based on the intersection environment information of each intersection, the virtual speed of the target trajectory line of each intersection and the first intersection environment information.

[0080] The virtual speed ranges from 2 to 30 m / s. For example, the virtual speed ranges from 2 to 8 m / s; or, for another example, the virtual speed ranges from 2 to 8 m / s.

[0081] It should be noted that the above virtual speed is the driving speed of the vehicle passing through the first intersection at a constant speed under an ideal state.

[0082] Step 330: Based on the first intersection environment information and the first virtual speed, a first pre-decision trajectory is obtained; the first pre-decision trajectory is used to indicate a reference driving path for the vehicle to pass through the first intersection.

[0083] Based on the above Figure 2 ,like Figure 4 As shown, the target trajectory 203 of the vehicle passing through the intersection is composed of sub-trajectory 203a and sub-trajectory 203b, wherein sub-trajectory 203a is the straight path of the vehicle in the first lane, and sub-trajectory 203b is the curved path of the vehicle passing through the intersection. The starting point of sub-trajectory 203b is the first position, and the end point of sub-trajectory 203b is the second position. The core goal of this solution is to generate a curved path for the vehicle to pass through the intersection, i.e., sub-trajectory 203b.

[0084] That is to say, the first pre-decision trajectory is composed of the straight trajectory of the first lane and the arc trajectory between the first lane and the second lane. The first position is the starting point of the arc trajectory between the first lane and the second lane, and the second position is the end point of the arc trajectory between the first lane and the second lane. After that, the vehicle can drive straight along the straight trajectory of the second lane, or change lanes as needed.

[0085] The computer device may determine the first position and the second position according to the first intersection environment information. Exemplarily, the computer device establishes a Cartesian coordinate system according to the position of the vehicle, and then the computer device represents the first position and the second position through coordinates.

[0086] Optionally, the computer device may construct a multi-decision maker, which is used to output an arc trajectory line from the starting point to the end point according to the virtual speed based on the starting point, the end point and the virtual speed. That is, the computer device may input the first position, the second position and the first virtual speed into the multi-decision maker to obtain the arc trajectory line between the first lane and the second lane; then, the straight trajectory line of the first lane formed by the current position of the vehicle and the first position is spliced ​​with the arc trajectory line between the first lane and the second lane to obtain the first pre-decision trajectory line.

[0087] Optionally, the multi-decision maker can use a fifth-order polynomial to generate a trajectory, and cast a suitable trajectory according to the virtual speed to generate a reasonable arc trajectory line through the intersection. Exemplarily, the curve formula of the fifth-order polynomial in the Cartesian coordinate system is as follows:

[0088] x(t)=a 0 t 5 +a 1 t 4 +a 2 t 3 +a3 t 2 +a 4 t+a 5 ;

[0089] y(t)=b 0 t 5 +b 1 t 4 +b 2 t 3 +b 3 t 2 +b 4 t+b 5 ;

[0090] Among them, x(t) is the horizontal coordinate of the arc trajectory line, y(t) is the vertical coordinate of the arc trajectory line, and a 0 、a 1 、a 2 、a 3 、a 4 、a 5 , b 0 , b 1、 b 2 , b 3 , b 4 , b 5 , t is a parameter, and the range of t is [0, 1].

[0091] Among them, the first-order derivative of the arc trajectory is the above-mentioned first virtual speed, and the second-order derivative of the arc trajectory is 0. According to the first-order derivative and the second-order derivative of the curve formula, the above a can be solved 0 、a 1 、a 2 、a 3 、a 4 、a 5 , b 0 , b 1 , b 2 , b 3 , b 4 , b 5 .

[0092] Exemplarily, the computer device uses a normalization method to draw a curvature image of the arc trajectory line. Figure 5 As shown in part (a) of FIG. 1 , the curvature image of the arc trajectory corresponding to the first virtual speed of 15 m / s; Figure 5 As shown in part (b) of FIG. 1 , the curvature image of the arc trajectory corresponding to the first virtual speed of 20 m / s; Figure 5 As shown in part (c) of FIG. 1 , the curvature image of the arc trajectory corresponding to the first virtual speed of 25 m / s.

[0093] Step 340: When the first pre-decision trajectory meets the accuracy requirement, the first pre-decision trajectory is determined as the target trajectory of the first intersection; the accuracy requirement is used to indicate the curvature range of the target trajectory, and the target trajectory is used to indicate the driving path of the vehicle through the first intersection.

[0094] Since the smoothness of the arc-shaped trajectory between the first lane and the second lane is related to the smoothness of the vehicle passing through the first intersection, the embodiment of the present application constrains the smoothness of the first pre-decision trajectory through curvature.

[0095] Among them, the computer device can pre-store the accuracy requirements for evaluating the smoothness of the first pre-decision trajectory; when the first pre-decision trajectory meets the accuracy requirements, it means that the first pre-decision trajectory can ensure the safety and ride comfort of the vehicle. At this time, the computer device determines the first pre-decision trajectory as the final generated target trajectory.

[0096] Exemplarily, the computer device may obtain the curvature of the first pre-decision trajectory line; thereafter, compare the curvature of the first pre-decision trajectory line with the curvature range indicated by the accuracy requirement, and when the curvature of the first pre-decision trajectory line satisfies the curvature range indicated by the accuracy requirement, determine the first pre-decision trajectory line as the target trajectory line finally generated.

[0097] Optionally, the computer device establishes and stores a correspondence between the intersection environment information of the first intersection and the first virtual speed in a memory pool to update data in the memory pool.

[0098] For example, the above accuracy requirement may be a curvature range less than or equal to 0.15, or other curvature ranges, which are not limited in this application.

[0099] Optionally, the computer device may construct an evaluator that determines whether the first pre-decision trajectory meets the accuracy requirement. That is, the computer device may input the first pre-decision trajectory into the evaluator to obtain a result of whether the first pre-decision trajectory meets the accuracy requirement.

[0100] It should be noted that, when the vehicle actually passes through the first intersection along the driving path according to the target trajectory, the driving speed of the vehicle is not greater than the first virtual speed to ensure the smoothness of the vehicle passing through the first intersection.

[0101] In summary, the technical solution provided by the embodiment of the present application can predict the virtual speed of the vehicle passing through the first intersection based on the acquired intersection environment information when the vehicle is about to automatically pass through the intersection; thereafter, the pre-decision trajectory of the vehicle passing through the first intersection can be predicted based on the virtual speed and the intersection environment information; through the curvature range of the pre-stored target trajectory, it is determined whether the curvature of the predicted pre-decision trajectory meets the curvature range, and when the pre-decision trajectory meets the curvature range, the pre-decision trajectory is determined as the target trajectory; this solution can determine the smoothness of the predicted pre-decision trajectory through the curvature range indicated by the accuracy requirement, thereby ensuring the smoothness of the target trajectory and the smoothness of the vehicle passing through the first intersection.

[0102] Based on the above Figure 3 In the solution in the illustrated embodiment, in a possible implementation manner, the above step 320 can be implemented as step 320a and step 320b.

[0103] Step 320a: Based on the first intersection environment information, query the second intersection environment information matching the first intersection environment information from the stored information; the stored information is used to indicate the correspondence between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection.

[0104] Exemplarily, the computer device creates a database table or a file structure for storing the corresponding relationship between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection.

[0105] Optionally, the above-mentioned storage information is constructed by the computer device according to the intersection environment information of each intersection and the corresponding virtual speed of the target trajectory line that meets the accuracy requirements.

[0106] The above-mentioned intersection environment information of the second intersection that matches the first intersection environment information means that the similarity between the first intersection environment information and the intersection environment information of the second intersection reaches a specified threshold. Exemplarily, the specified threshold is preset.

[0107] The intersection environment information of the second intersection that matches the first intersection environment information may be one or more.

[0108] Step 320b: Obtain a first virtual speed based on the virtual speed of the target trajectory corresponding to the intersection environment information of the second intersection in the stored information.

[0109] If there is intersection environment information of a second intersection that matches the first intersection environment information, the computer device can determine the virtual speed of the target trajectory line corresponding to the intersection environment information of the second intersection as the first virtual speed; or, based on the difference between the first intersection environment information and the intersection environment information of the second intersection, adjust the virtual speed of the target trajectory line corresponding to the intersection environment information of the second intersection to obtain the first virtual speed.

[0110] If there are multiple second intersection environment information matching the first intersection environment information, the computer device can determine the first virtual speed according to the multiple virtual speeds by linear difference. For example, if the rotation radius of the first intersection is between the rotation radii of the two second intersections, the computer device can interpolate and calculate the appropriate speed according to the virtual speeds of the target trajectory corresponding to the intersection environment information of the two second intersections.

[0111] In an embodiment of the present application, by establishing and maintaining a correspondence between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection, the first intersection environment information can be matched with the intersection environment information of each intersection in the correspondence to obtain the intersection environment information of the second intersection that matches the first intersection environment information, and then the first virtual speed can be obtained based on the virtual speed of the target trajectory line corresponding to the intersection environment information of the second intersection, thereby improving the accuracy of the first virtual speed.

[0112] Based on the solutions in the above embodiments of the present application, in a possible implementation manner, the above step 320 can be implemented as follows:

[0113] Inputting the first intersection environment information into the prediction model to obtain a first virtual speed;

[0114] Among them, the prediction model is a machine learning model trained based on sample intersection environment information and sample virtual speed.

[0115] The sample intersection environment information and sample virtual speed are collected in advance. Exemplarily, the sample intersection environment information is used to indicate the characteristics of each intersection, and the sample intersection environment information includes at least one of the first intersection environment information including the width of the first lane, the width of the second lane, and the positional relationship between the first lane and the second lane. The sample virtual speed is the virtual speed of the target trajectory corresponding to the sample intersection environment information.

[0116] Exemplarily, the prediction model obtains the probability distribution of virtual speeds according to the first intersection environment information, and then the computer device determines a virtual speed with the highest probability as the first virtual speed.

[0117] In an embodiment of the present application, a prediction model is pre-trained through a machine learning algorithm, and the first intersection environment information is input into the prediction model. The prediction model can automatically calculate the corresponding virtual speed within milliseconds, which can improve the accuracy of determining the first virtual speed based on the first intersection environment information.

[0118] Based on the solutions in the above embodiments of the present application, in a possible implementation, the above trajectory processing method further includes:

[0119] Input the sample intersection environment information into the prediction model to obtain the predicted virtual speed;

[0120] Based on the difference between the predicted virtual speed and the sample virtual speed, a loss function value is obtained;

[0121] Based on the loss function value, the parameters of the prediction model are updated.

[0122] In the example of the present application, the computer device obtains N pre-collected sample intersection environment information and N sample virtual speeds; then, the N sample intersection environment information is input into the prediction model to obtain N predicted virtual speeds.

[0123] The above loss function value may be a mean square error (MSE). For example, the loss function value is:

[0124]

[0125] Here, N represents the number of samples, Vpred,i is the predicted virtual speed of the i-th sample, and Vtrue,i is the sample virtual speed of the i-th sample.

[0126] Afterwards, the computer device updates the parameters of the prediction model using the gradient descent method based on the loss function value.

[0127] The embodiment of the present application illustrates a supervised training process of a prediction model, which can improve the accuracy of the prediction model when performing prediction tasks, thereby improving the accuracy of determining the first virtual speed based on the first intersection environment information.

[0128] Please refer to Figure 6 , which shows a schematic diagram of a training method and an inference method of a prediction model provided by an exemplary embodiment of the present application. Figure 6 As shown in part (a) of the figure, the training method of the prediction model includes:

[0129] Step A1: Obtain sample intersection environment information 601 and sample virtual speed 603;

[0130] Step A2: input the sample intersection environment information 601 into the prediction model to obtain the predicted virtual speed 602;

[0131] Step A3: obtaining a loss function value 604 based on the difference between the predicted virtual speed 602 and the sample virtual speed 603;

[0132] Step A4: based on the loss function value 604, update the parameters of the prediction model;

[0133] Step A5: Repeat the above steps A1 to A4 until the convergence condition is reached.

[0134] like Figure 6 As shown in part (b) of the paper, the inference method of the prediction model includes:

[0135] Step B1: inputting the first intersection environment information 605 into the prediction model;

[0136] Step B2: Obtain the first virtual speed 606 output by the prediction model.

[0137] Based on the solutions in the above embodiments of this application, please refer to Figure 7 , which shows a flow chart of a trajectory processing method provided by an exemplary embodiment of the present application. Figure 7 As shown, the above-mentioned trajectory processing method also includes steps 350 to 370.

[0138] Step 350: When the first pre-decision trajectory does not meet the accuracy requirement, obtain a second virtual speed based on the first pre-decision trajectory.

[0139] When the first pre-decision trajectory does not meet the accuracy requirement, it means that the first pre-decision trajectory cannot ensure the safety and ride comfort of the vehicle. At this time, the computer device should adjust the above-mentioned first virtual speed, obtain the second virtual speed, and regenerate the second pre-decision trajectory.

[0140] In the embodiment of the present application, the first pre-decision trajectory line does not meet the accuracy requirement, indicating that the first pre-decision trajectory line has a problem of too large or too small curvature.

[0141] When the curvature of the first pre-decision trajectory is too large, the computer device can reduce the first virtual speed to obtain a second virtual speed that is smaller than the first virtual speed, thereby reducing the turning radius; when the curvature of the first pre-decision trajectory is too small, the computer device can increase the first virtual speed to obtain a second virtual speed that is larger than the first virtual speed, so as to better adapt to smooth road sections.

[0142] Step 360: Based on the first intersection environment information and the second virtual speed, obtain a second pre-decision trajectory.

[0143] Once the new virtual speed, ie, the second virtual speed, is determined, the next step is to perform trajectory planning again to obtain the second pre-decision trajectory line.

[0144] The execution method of step 360 is the same as the execution method of step 330. That is, the computer device can input the first position, the second position and the second virtual speed into the multi-decision maker to obtain the arc trajectory between the first lane and the second lane; then, the straight trajectory of the first lane formed by the current position of the vehicle and the first position is spliced ​​with the arc trajectory between the first lane and the second lane to obtain the second pre-decision trajectory.

[0145] Step 370: When it is determined that the second pre-decision trajectory meets the accuracy requirement, the second pre-decision trajectory is determined as the target trajectory.

[0146] The execution manner of step 370 is the same as the execution manner of step 340. That is, the computer device can obtain the curvature of the second pre-decision trajectory line; then, compare the curvature of the second pre-decision trajectory line with the curvature range indicated by the accuracy requirement, and when the curvature of the second pre-decision trajectory line meets the curvature range indicated by the accuracy requirement, determine the second pre-decision trajectory line as the target trajectory line finally generated.

[0147] Optionally, the computer device establishes and stores a correspondence between the intersection environment information of the first intersection and the second virtual speed in a memory pool to update data in the memory pool.

[0148] In a possible implementation, when it is determined that the second pre-decision trajectory does not meet the accuracy requirement, the above steps 350 to 370 are looped. That is, when it is determined that the second pre-decision trajectory does not meet the accuracy requirement, the third virtual speed is obtained based on the second pre-decision trajectory; the third pre-decision trajectory is obtained based on the first intersection environment information and the third virtual speed; when it is determined that the third pre-decision trajectory meets the accuracy requirement, the third pre-decision trajectory is determined as the target trajectory.

[0149] The embodiment of the present application shows a feasible solution for regenerating a pre-decision trajectory line that meets the accuracy requirements when the predicted pre-decision trajectory line does not meet the accuracy requirements, so as to ensure the smoothness of the target trajectory line, and further ensure the smoothness of the vehicle passing through the first intersection.

[0150] Based on the solutions in the above embodiments of the present application, in a possible implementation manner, the above step 350 can be implemented as follows:

[0151] Step 350a: When the first pre-decision trajectory does not meet the accuracy requirement, obtain the curvature integral of the first pre-decision trajectory.

[0152] The curvature integral reflects the cumulative effect of the curvature change in the entire path of the first pre-decision trajectory, which can help understand the overall smoothness and complexity of the first pre-decision trajectory.

[0153] Exemplarily, the computer device may obtain the curvature integral by integrating the curvature value at each point on the first pre-decision trajectory line.

[0154] Step 350b: When the curvature integral of the first pre-decision trajectory decreases, determine the second virtual speed as a virtual speed greater than the first virtual speed.

[0155] If the curvature integral of the first pre-decision trajectory shows a decreasing trend, it means that as the vehicle moves forward, the first pre-decision trajectory gradually becomes straight. In this case, in order to avoid unnecessary deceleration, the virtual speed can be appropriately increased, that is, the second virtual speed is set to be greater than the first virtual speed.

[0156] Optionally, the computer device may obtain the decreasing amount of the curvature integral of the first pre-decision trajectory; then, according to the decreasing amount of the curvature integral of the first pre-decision trajectory, determine the increase of the first virtual speed, and then determine the second virtual speed. For example, the corresponding relationship between the decreasing amount of the curvature integral and the increase of the first virtual speed may be preset in the computer device; accordingly, the computer device may query the increase of the first virtual speed from the above corresponding relationship according to the decreasing amount of the curvature integral of the first pre-decision trajectory.

[0157] Step 350c: When the curvature integral of the first pre-decision trajectory increases, determine the second virtual speed as a virtual speed that is smaller than the first virtual speed.

[0158] If the curvature integral of the first pre-decision trajectory is increasing, it means that as the vehicle moves forward, the first pre-decision trajectory becomes more and more curved. At this time, in order to ensure safety and maintain a good riding experience, the virtual speed should be reduced so that the second virtual speed is less than the first virtual speed.

[0159] Optionally, the computer device may obtain the incremental change of the curvature integral of the first pre-decision trajectory; then, according to the incremental change of the curvature integral of the first pre-decision trajectory, determine the decrease of the first virtual speed, and then determine the second virtual speed. For example, the corresponding relationship between the incremental change of the curvature integral and the decrease of the first virtual speed may be preset in the computer device; accordingly, the computer device may query the decrease of the first virtual speed from the above corresponding relationship according to the incremental change of the curvature integral of the first pre-decision trajectory.

[0160] Step 350d: When the curvature integral of the first pre-determination trajectory first decreases and then increases, the second virtual speed is obtained by an average sampling method.

[0161] If the curvature integral of the first pre-decision trajectory shows a trend of decreasing first and then increasing, it means that as the vehicle moves forward, the first pre-decision trajectory becomes straight first and then curved. At this time, directly selecting a fixed high speed or low speed is not the optimal solution.

[0162] Among them, by means of average sampling, a balance point can be found between maintaining a certain ratio of acceleration and deceleration. For example, the curvature integral of the first pre-decision trajectory line decreases first and then increases, indicating that the curvature integral just presents a good "convex characteristic", and the computer device can take two convex speeds for average speed sampling.

[0163] Optionally, the above-mentioned acquisition of the second virtual speed by the average sampling method can be implemented as follows:

[0164] Determine key points: identify the first key point and the second key point on the first pre-decision trajectory line that best represent the convex feature;

[0165] Calculate the velocity increment: according to the velocity value v corresponding to the first key point s The velocity value v corresponding to the second key point e , and the number of speed steps n, calculate the speed increment Δv in each small segment = (v e -v s ) / n;

[0166] Assign speed values: Increase speed step by step along the path until a final second virtual speed is reached.

[0167] The embodiment of the present application performs a detailed analysis of the curvature integral of the first pre-decision trajectory line, and adjusts the virtual speed according to the changing trend of the curvature integral, thereby improving the accuracy of the second virtual speed, so that the second pre-decision trajectory line generated based on the second virtual speed meets the accuracy requirements.

[0168] With the development of intelligent driving technology and artificial intelligence technology, the penetration rate of assisted driving is getting higher and higher. In the intelligent driving algorithm, decision planning is an important module for automobile decision-making, path planning, and trajectory generation. Traditional decision planning is based on high-precision maps, but due to the high production and maintenance costs of high-precision maps, it cannot meet the needs of rapid map updates, so a decision planning method without high-precision maps is needed to complete the tasks of path planning and trajectory generation. However, intersections are difficult problems in path planning and trajectory generation without high-precision maps.

[0169] First, the lane lines obtained by the non-HD map system with visual perception and map fusion can replace the lane lines on the HD map. Due to recognition errors, the lane lines are unstable, and the Vernet coordinate system established on this basis will be unstable. Secondly, a curve method is needed to generate a trajectory that meets the vehicle dynamics requirements at the intersection. Finally, due to the limitation of computing resources, a method is needed to optimize the search method, reduce the amount of calculation, and effectively search for a trajectory curve that meets the requirements.

[0170] Based on the solutions shown in the above embodiments, the present application proposes a method based on multiple decision makers and evaluators without high precision. Figure 10 The adaptive trajectory generation method for intersections aims to solve the problem of adaptive trajectory generation for intersections without high-precision map planning. The prerequisites for this method are: a small vehicle with autonomous driving function, a non-high-precision map system with visual perception and map fusion, an intersection with an appropriate rotation radius, and a system-on-chip (SoC) with moderate computing power.

[0171] Please refer to Figure 8 , which shows a high-precision ground-based multi-decision maker and evaluator provided by an exemplary embodiment of the present application Figure 10 Flowchart of the adaptive trajectory generation method for intersections. Figure 8 As shown in the above, the non-high-precision Figure 10 The method for generating adaptive trajectories at intersections comprises the following steps:

[0172] Step 801, read the experience pool:

[0173] At the beginning of the program, the multi-decision maker reads the upper and lower bound constraints and historical curve information in the experience pool (that is, reads the historical estimation information) to obtain the virtual speed.

[0174] Step 802, scatter multiple groups of curves:

[0175] The multi-decision maker casts out multiple groups of curves under virtual speed constraints as pre-decision trajectories (i.e., pre-decision trajectory lines) according to the fifth-order polynomial.

[0176] Step 803, evaluation curve:

[0177] The evaluator evaluates the effects of multiple groups of curves and writes the evaluation results into the experience pool, storing historical estimation information for use in subsequent steps.

[0178] If the curve is determined to meet the accuracy requirement or reach the limit number of times, step 804 is executed and the program ends; otherwise, step 805 is executed first and then step 802 is executed until the accuracy requirement is met.

[0179] Step 804, output trajectory:

[0180] When the curve meets the accuracy requirement or reaches the limit number of times, the curve is output as a trajectory and the optimal virtual speed in the environment is stored.

[0181] Step 805, generate search direction and range decision:

[0182] When the curve does not meet the accuracy requirements or does not reach the limit number of times, the historical curve evaluation information is read and combined with the existing evaluation information to generate a search direction and range decision.

[0183] Please refer to Fig. 9 , which shows a high-precision ground-based multi-decision maker and evaluator provided by an exemplary embodiment of the present application Figure 10 Framework diagram of the adaptive trajectory generation method for intersections.

[0184] like Fig. 9 As shown in the figure, the multi-decision maker uses the memory pool in the experience pool to read past experience for the first time, searches for the optimal virtual speed for the exit environment, and scatters multiple groups of pre-decision trajectory lines for the first round. The multi-decision maker scatters multiple groups of pre-decision trajectory lines based on the search direction and search range decisions of the evaluator, combined with the upper and lower bounds in the experience pool and local memory, and continues to perform speed sampling to scatter the line. Here, the curvature integral obtained by the multi-decision maker from multiple groups of sampling is passed to the evaluator to determine the gradient change method, so as to use iterative search with multiple optimized search directions to replace brute force search, thereby reducing the amount of calculation. The curvature integral here can also be replaced by the maximum curvature value.

[0185] The evaluator uses curvature to evaluate whether the line cast by the multi-decision maker meets the requirements (accuracy requirements or number requirements). If the requirements are met, the process ends and returns a curve that meets the requirements. If the requirements are not met, an evaluation is made based on the curvature integral of the pre-decision line cast in this round, and a decision on the search direction and range of the virtual speed is generated. If there is a decreasing trend, the virtual speed is increased to continue casting the line; if there is an increasing trend, the virtual speed is reduced to continue casting the line; if there is a convex shape, two convex points are taken for average speed sampling, and the line is cast again. The curvature integral here can also be replaced by the maximum curvature.

[0186] Among them, the curvature formula of the quintic polynomial in the Cartesian coordinate system is as follows:

[0187]

[0188] The formula for the curvature integral in Cartesian coordinates is as follows:

[0189]

[0190] The formula for the maximum curvature in Cartesian coordinates is as follows:

[0191] f(t)=max{k(1),k(2),…,k(t)}

[0192] Each time, the evaluator will mark the pre-decision trajectory line with a curvature integral as a penalty value, and store it in the local memory in the experience pool, which will be used by multiple decision makers to find past experience line spreading in the future. However, due to the differences in intersections, the experience of similar intersections may not be applicable to the optimal solution for each intersection. The curvature integral here can be replaced by the maximum curvature. The experience pool includes two modules, one is the local memory, which is used to store and search for the experience of this line spreading. The other is the memory pool, which is used to improve the initial hit rate based on historical data when the line is first spread. The structure of the memory pool can be based on rule matching, machine learning regression, or neural network generation strategy methods, etc. This depends on how the experience data is used and the computing power support of the chip.

[0193] In summary, path planning and trajectory generation in the Cartesian coordinate system can avoid the instability of the coordinate system caused by the fluctuation of lane lines after visual perception and map fusion, and can also maintain the stability of the starting point.

[0194] Path planning and trajectory generation in Cartesian coordinates can avoid the coordinate conversion cost of converting input from Cartesian coordinates to Vernet coordinates for path planning, and then converting back to Cartesian coordinates for subsequent planning tasks.

[0195] This method can provide a crossing trajectory for an intersection when there is no high-precision map planning task or when there is no map lane line at the intersection in a memory driving task, so that the vehicle can pass through the intersection.

[0196] This method provides a multi-decision maker-evaluator framework that can effectively reduce the amount of search, and when valid lane lines are searched, the amount of calculation also meets the limitations of computing resources.

[0197] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0198] Please refer to Fig.10 , which shows a block diagram of a trajectory processing device provided by an exemplary embodiment of the present application. The device has the functions of implementing the above examples, and the functions can be implemented by hardware or by hardware executing corresponding software. Fig.10 As shown, the device may include: an information acquisition module 1001 , a speed acquisition module 1002 , a path acquisition module 1003 and a path determination module 1004 .

[0199] The information acquisition module 1001 is used to acquire first intersection environment information; the first intersection environment information is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the first intersection environment information includes at least one of the width of the first lane, the width of the second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane where the vehicle exits from the first intersection, and the second lane is the lane where the vehicle enters from the first intersection;

[0200] The speed acquisition module 1002 is used to acquire a first virtual speed based on the first intersection environment information; the first virtual speed is used to indicate the estimated driving speed of the vehicle passing through the first intersection;

[0201] The path acquisition module 1003 is used to acquire a first pre-decision trajectory line based on the first intersection environment information and the first virtual speed; the first pre-decision trajectory line is used to indicate a reference driving path for the vehicle to pass through the first intersection;

[0202] The path determination module 1004 is used to determine the first pre-decision trajectory line as the target trajectory line of the first intersection when the first pre-decision trajectory line meets the accuracy requirement; the accuracy requirement is used to indicate the curvature range of the target trajectory line, and the target trajectory line is used to indicate the driving path of the vehicle through the first intersection.

[0203] In some embodiments, the speed acquisition module 1002 is used to query the intersection environment information of the second intersection matching the first intersection environment information from the stored information based on the first intersection environment information; the stored information is used to indicate the corresponding relationship between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection;

[0204] The speed acquisition module 1002 is used to acquire a first virtual speed based on the virtual speed of the target trajectory corresponding to the intersection environment information of the second intersection in the stored information.

[0205] In some embodiments, the speed acquisition module 1002 is used to input the first intersection environment information into the prediction model to obtain a first virtual speed;

[0206] Among them, the prediction model is a machine learning model trained based on sample intersection environment information and sample virtual speed.

[0207] In some embodiments, the trajectory processing device further includes: a training module, configured to input sample intersection environment information into a prediction model to obtain a predicted virtual speed;

[0208] A training module, for obtaining a loss function value based on a difference between a predicted virtual speed and a sample virtual speed;

[0209] The training module is used to update the parameters of the prediction model based on the loss function value.

[0210] In some embodiments, the trajectory processing device further includes: an adjustment module, configured to obtain a second virtual speed based on the first pre-decision trajectory when the first pre-decision trajectory does not meet the accuracy requirement;

[0211] An adjustment module, configured to obtain a second pre-decision trajectory based on the first intersection environment information and the second virtual speed;

[0212] The adjustment module is used to determine the second pre-decision trajectory line as the target trajectory line when it is determined that the second pre-decision trajectory line meets the accuracy requirement.

[0213] In some embodiments, the adjustment module is used to obtain the curvature integral of the first pre-decision trajectory line when the first pre-decision trajectory line does not meet the accuracy requirement;

[0214] An adjustment module, configured to determine the second virtual speed as a virtual speed greater than the first virtual speed when the curvature integral of the first pre-decision trajectory decreases;

[0215] An adjustment module, configured to determine the second virtual speed as a virtual speed less than the first virtual speed when the curvature integral of the first pre-decision trajectory increases;

[0216] The adjustment module is used to obtain the second virtual speed by an average sampling method when the curvature integral of the first pre-decision trajectory line decreases first and then increases.

[0217] It should be noted that the device provided in the above embodiment, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the content structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0218] Please refer to Fig.11 , which shows a block diagram of a computer device provided in one embodiment of the present application. The computer device 1100 may be any electronic device with data calculation, processing and storage capabilities. The computer device 1100 may be used to implement the trajectory processing method provided in the above embodiment.

[0219] Typically, the computer device 1100 includes a processor 1101 and a memory 1102 .

[0220] The processor 1101 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 1101 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1101 may also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 1101 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 1101 may also include an artificial intelligence (AI) processor, which is used to process computing operations related to machine learning.

[0221] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 1102 is used to store a computer program, which is configured to be executed by one or more processors to implement the above-mentioned trajectory processing method.

[0222] Those skilled in the art will understand that Fig.11 The structure shown in the figure does not constitute a limitation on the computer device 1100, and the computer device 1100 may include more or less components than shown in the figure, or combine some components, or adopt a different component arrangement.

[0223] In an exemplary embodiment, a computer-readable storage medium is also provided, in which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned trajectory processing method is implemented. Optionally, the computer-readable storage medium may include: ROM (Read-Only Memory), RAM (Random Access Memory), SSD (Solid State Drives) or an optical disk, etc. Among them, the random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0224] In an exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program, the computer program being stored in a computer-readable storage medium. A processor of a computer device reads the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device performs the above-mentioned trajectory processing method.

[0225] It should be understood that the "multiple" mentioned in this article refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. In addition, the step numbers described in this article only illustrate a possible execution sequence between the steps. In some other embodiments, the above steps may not be executed in the order of the numbers, such as two steps with different numbers are executed at the same time, or two steps with different numbers are executed in the opposite order to the diagram. The embodiments of the present application are not limited to this.

[0226] The above description is only an exemplary embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A trajectory processing method, characterized in that: The method comprises: Acquire first intersection environment information; the first intersection environment information is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the first intersection environment information includes at least one of the width of a first lane, the width of a second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane that the vehicle exits through the first intersection, and the second lane is the lane that the vehicle enters through the first intersection; Based on the first intersection environment information, a first virtual speed is acquired; the first virtual speed is used to indicate the estimated driving speed of the vehicle passing through the first intersection; Based on the first intersection environment information and the first virtual speed, a first pre-decision trajectory is acquired; the first pre-decision trajectory is used to indicate a reference driving path of the vehicle through the first intersection; When the first pre-decision trajectory meets the accuracy requirement, the first pre-decision trajectory is determined as the target trajectory of the first intersection; the accuracy requirement is used to indicate the curvature range of the target trajectory, and the target trajectory is used to indicate the driving path of the vehicle through the first intersection.

2. The method according to claim 1, characterized in that The acquiring a first virtual speed based on the first intersection environment information includes: Based on the first intersection environment information, querying the intersection environment information of the second intersection matching the first intersection environment information from the stored information; the stored information is used to indicate the corresponding relationship between the intersection environment information of each intersection and the virtual speed of the target trajectory line of each intersection; The first virtual speed is acquired based on the virtual speed of the target trajectory corresponding to the intersection environment information of the second intersection in the stored information.

3. The method according to claim 1, characterized in that The acquiring a first virtual speed based on the intersection environment information includes: Inputting the first intersection environment information into a prediction model to obtain the first virtual speed; The prediction model is a machine learning model trained based on sample intersection environment information and sample virtual speed.

4. The method according to claim 3, characterized in that The method further comprises: Inputting the sample intersection environment information into the prediction model to obtain a predicted virtual speed; Obtaining a loss function value based on a difference between the predicted virtual speed and the sample virtual speed; Based on the loss function value, the parameters of the prediction model are updated.

5. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: When the first pre-decision trajectory does not meet the accuracy requirement, obtaining a second virtual speed based on the first pre-decision trajectory; Based on the first intersection environment information and the second virtual speed, obtaining a second pre-decision trajectory; In the case where it is determined that the second pre-decision trajectory meets the accuracy requirement, the second pre-decision trajectory is determined as the target trajectory.

6. The method according to claim 5, characterized in that The acquiring a second virtual speed based on the first pre-decision trajectory line when the first pre-decision trajectory line does not meet the accuracy requirement includes: When the first pre-decision trajectory does not meet the accuracy requirement, obtaining a curvature integral of the first pre-decision trajectory; In a case where the curvature integral of the first pre-decision trajectory decreases, determining the second virtual speed as a virtual speed greater than the first virtual speed; In a case where the curvature integral of the first pre-decision trajectory increases, determining the second virtual speed as a virtual speed that is less than the first virtual speed; When the curvature integral of the first pre-decision trajectory first decreases and then increases, the second virtual speed is acquired by an average sampling method.

7. A trajectory processing device, characterized in that: The device comprises: An information acquisition module, used to acquire first intersection environment information; the first intersection environment information is used to indicate the characteristics of the first intersection that the vehicle is about to pass; the first intersection environment information includes at least one of the width of a first lane, the width of a second lane, and the positional relationship between the first lane and the second lane, the first lane is the lane that the vehicle exits through the first intersection, and the second lane is the lane that the vehicle enters through the first intersection; A speed acquisition module, configured to acquire a first virtual speed based on the first intersection environment information; the first virtual speed is used to indicate an estimated driving speed of the vehicle passing through the first intersection; A path acquisition module, configured to acquire a first pre-decision trajectory line based on the first intersection environment information and the first virtual speed; the first pre-decision trajectory line is used to indicate a reference driving path of the vehicle passing through the first intersection; A path determination module is used to determine the first pre-decision trajectory line as the target trajectory line of the first intersection when the first pre-decision trajectory line meets the accuracy requirement; the accuracy requirement is used to indicate the curvature range of the target trajectory line, and the target trajectory line is used to indicate the driving path of the vehicle through the first intersection.

8. A computer device, characterized in that: The computer device includes a processor and a memory, wherein a computer program is stored in the memory, and the computer program is loaded and executed by the processor to implement the trajectory processing method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is loaded and executed by a processor to implement the trajectory processing method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises a computer program, the computer program is stored in a computer-readable storage medium, and a processor reads and executes the computer program from the computer-readable storage medium to implement the trajectory processing method according to any one of claims 1 to 6.

Images