Blockage vehicle data labeling system, method and device for automatic driving
By combining sensor data and obstacle recognition results from autonomous vehicles with pre-trained models to generate scene labels, the problem of identifying congested vehicles has been solved, and the efficiency and accuracy of data labeling in autonomous driving systems have been improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-03
- Publication Date
- 2026-03-10
AI Technical Summary
In autonomous driving scenarios, it is difficult to efficiently label blocked vehicles using deep learning models, leading to inaccurate identification and affecting driving safety and traffic efficiency.
By acquiring sensor data and obstacle recognition results from autonomous vehicles, and utilizing a pre-trained traffic jam data annotation model, combined with prior scene identifiers and key obstacle features, scene labels are automatically generated, reducing reliance on manual annotation.
It improved the speed and accuracy of traffic jam data annotation, reduced reliance on manual annotation, enhanced model recognition capabilities, and reduced dependence on data volume and upstream recognition results.
Smart Images
Figure CN121637069A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of artificial intelligence, and more particularly to the field of autonomous driving, specifically to a system, method, and apparatus for labeling traffic jam data for autonomous driving. Background Technology
[0002] In autonomous driving scenarios, identifying congested vehicles is a very challenging problem. Poor identification of congested vehicles can lead to traffic jams or violations of traffic regulations, thereby affecting driving safety and traffic efficiency.
[0003] Simply put, a traffic jam can be defined as follows: if there is a car stopped in front of you, there are two mutually exclusive scenarios: 1. Detour scenario: There is an illegally parked / long-term stopped vehicle ahead, so a detour is necessary; 2. Queuing scenario: There is a queue of vehicles waiting to pass in front of you, so you need to follow behind them in the queue to pass.
[0004] Traffic jams are primarily identified using deep learning models. However, training deep learning models requires a large amount of data, and relying entirely on manual annotation is time-consuming, labor-intensive, and the quality cannot be guaranteed. Summary of the Invention
[0005] This disclosure provides a system, method, apparatus, device, storage medium, and computer program product for labeling congested vehicle data for autonomous driving.
[0006] According to a first aspect of this disclosure, a method for labeling congested vehicle data for autonomous driving is provided, comprising: acquiring sensor data collected by an autonomous vehicle under different prior scenarios and the identification results of obstacles determined based on the sensor data; identifying key obstacles based on the obstacle identification results; determining the actual scenario associated with the key obstacles by using a pre-trained congested vehicle data labeling model based on prior scenario identifiers, sensor data, and the identification results of key obstacles; and generating scenario labels for the key obstacles based on the actual scenario.
[0007] According to a second aspect of this disclosure, a method for training a traffic congestion data annotation model is provided, comprising: acquiring training samples under different scenarios, wherein the training samples include scenario labels, prior scenario identifiers, sample sensor data, and identification results of sample obstacles, wherein the prior scenarios include queuing scenarios and detour scenarios; determining the actual scenarios associated with the sample obstacles through an initial traffic congestion data annotation model based on the prior scenario identifiers, sample sensor data, and identification results of sample obstacles; and adjusting the parameters of the initial traffic congestion data annotation model based on the differences between the actual scenarios and the scenario labels.
[0008] According to a third aspect of this disclosure, a traffic jam data labeling system for autonomous driving is provided, comprising: an autonomous vehicle configured to, in response to detecting that queuing or detour is required, collect sensor data and determine an obstacle identification result based on the sensor data, and report the sensor data and the obstacle identification result to a server; and a server configured to perform the method of either the first or second aspect.
[0009] According to a fourth aspect of this disclosure, a traffic jam data annotation device for autonomous driving is provided, comprising: an acquisition unit configured to acquire sensor data collected by an autonomous vehicle under different prior scenarios and obstacle identification results determined based on the sensor data; a filtering unit configured to identify key obstacles based on the obstacle identification results; a determination unit configured to determine the actual scenario associated with the key obstacles based on prior scenario identifiers, sensor data, and the key obstacle identification results, through a pre-trained traffic jam data annotation model; and an annotation unit configured to generate scene labels for the key obstacles based on the actual scenario.
[0010] According to a fifth aspect of this disclosure, an apparatus for training a traffic jam data annotation model is provided, comprising: a sampling unit configured to acquire training samples under different scenarios, wherein the training samples include scenario labels, prior scenario identifiers, sample sensor data, and identification results of sample obstacles, wherein the prior scenarios include queuing scenarios and detour scenarios; a prediction unit configured to determine the actual scenario associated with the sample obstacles based on the prior scenario identifiers, sample sensor data, and identification results of sample obstacles, through an initial traffic jam data annotation model; and an adjustment unit configured to adjust the parameters of the initial traffic jam data annotation model based on the difference between the actual scenario and the scenario labels.
[0011] According to a sixth aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to said at least one processor; wherein the memory stores instructions executable by said at least one processor, said instructions being executed by said at least one processor to enable said at least one processor to perform the method described in any one of the first aspects.
[0012] According to a seventh aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described in any one of the first aspects.
[0013] According to the eighth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in any one of the first aspects.
[0014] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description
[0015] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 This is an exemplary system architecture diagram to which one embodiment of this disclosure can be applied; Figure 2 This is a flowchart of an embodiment of the traffic jam data annotation method according to the present disclosure; Figure 3a This is a schematic diagram illustrating an application scenario of the traffic congestion data annotation method disclosed herein; Figure 3b This is a schematic diagram illustrating an application scenario of the method for training a data annotation model for congested vehicles according to this disclosure; Figure 4 This is a flowchart of an embodiment of a method for training a data annotation model for congested vehicles according to the present disclosure; Figure 5 This is a schematic diagram of a structure of one embodiment of the traffic jam data labeling device according to the present disclosure; Figure 6 This is a schematic diagram of one embodiment of the apparatus for training a data annotation model for blocked vehicles according to the present disclosure; Figure 7 This is a schematic diagram of the structure of a computer system suitable for implementing embodiments of the present disclosure. Detailed Implementation
[0016] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.
[0017] Figure 1 An exemplary system architecture 100 for which the traffic jam data annotation system of this disclosure can be applied is shown.
[0018] like Figure 1 As shown, system architecture 100 may include autonomous vehicles and servers. Servers may include data centers and back-end servers.
[0019] Autonomous vehicles are also known as driverless cars. Autonomous vehicles can be equipped with driving control devices, networks, and sensors. The network serves as the medium for providing a communication link between the driving control devices and sensors. The network can include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.
[0020] The driving control equipment (also known as the vehicle's brain) is responsible for the intelligent control of autonomous vehicles. The driving control equipment can be a standalone controller, such as a programmable logic controller (PLC), a microcontroller, an industrial control computer, etc.; it can also be a device composed of other electronic components with input / output ports and computational control functions; or it can be a computer device equipped with vehicle driving control applications.
[0021] Sensors can be, for example, cameras, radar, gravity sensors, wheel speed sensors, etc. In some cases, autonomous vehicles may also be equipped with GNSS (Global Navigation Satellite System) equipment and SINS (Strap-down Inertial Navigation System), etc.
[0022] Autonomous vehicles can report traffic congestion data to a server. For example, they can collect sensor data (including images, point cloud data, etc.) and send it to a backend server. The backend server then identifies obstacles based on the sensor data, obtains the identification results (i.e., attribute information), and saves it to a data center. Alternatively, autonomous vehicles can also identify obstacles based on sensor data, obtain the identification results (i.e., attribute information), and then send them to a data center.
[0023] Data centers can store traffic congestion data reported by autonomous vehicles. This traffic congestion data can serve as a training sample set for a traffic congestion data annotation model. The training sample set contains a large number of training samples. These samples may include scene labels, prior scene identifiers, sensor data, and obstacle identification results. The prior scenes include queuing scenarios and detour scenarios.
[0024] The backend server can use the training samples in the training sample set of the data center to train the initial traffic jam data annotation model, and use the trained traffic jam data annotation model to annotate the traffic jam data stored in the data center, and give the traffic jam a detour label or a queue label.
[0025] It's important to note that a server can be either hardware or software. When a server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When a server is software, it can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services), or as a single software program or software module. No specific limitations are made here. A server can also be a server for a distributed system, or a server integrated with blockchain technology. A server can also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.
[0026] It should be noted that the methods for labeling congested vehicle data and training congested vehicle data labeling models provided in the embodiments of this disclosure are generally executed by a server. Accordingly, the devices for labeling congested vehicle data and training congested vehicle data labeling models are generally located in the server.
[0027] It should be understood that Figure 1 The number of autonomous vehicles and servers shown is merely illustrative. Depending on implementation needs, any number of autonomous vehicles and servers can be used.
[0028] Continue to refer to Figure 2 The diagram illustrates a flow 200 of an embodiment of a traffic congestion data annotation method according to the present disclosure. The traffic congestion data annotation method includes the following steps: Step 201: Obtain sensor data collected by the autonomous vehicle in different prior scenarios and the obstacle recognition results determined based on the sensor data; In this embodiment, the execution body of the traffic jam data annotation method (e.g., Figure 1 The server shown can receive sensor data (including images and point clouds) collected by the autonomous vehicle under different prior scenarios. These prior scenarios can include queuing scenarios and detour scenarios. The autonomous vehicle can identify the macroscopic prior scenario of congestion, but further analysis is needed to determine which vehicle to queue behind, or which vehicle or obstacle to detour around.
[0029] The autonomous vehicle can identify the attribute information of at least one obstacle based on sensor data, i.e., the obstacle identification result, and then send the identification result to the server. The obstacle identification result includes at least one of the following information about the obstacle: type, speed, position, headlight status information, and door status information. The server can also identify the obstacle identification result. For example, the type of obstacle can be identified based on the acquired image using an image recognition algorithm. The position of the obstacle can also be determined based on the image and / or point cloud data.
[0030] Step 202: Based on the obstacle identification results, identify the key obstacles; In this embodiment, the key obstacle can be identified from multiple obstacles based on the obstacle identification results. For example, the vehicle closest to the autonomous vehicle can be identified as the key obstacle based on the location and type in the identification results.
[0031] Step 203: Based on prior scene identification, sensor data and key obstacle identification results, determine the actual scene associated with the key obstacles through a pre-trained traffic jam data annotation model; In this embodiment, the traffic jam data annotation model is a neural network model, such as a classifier. The traffic jam data annotation model can extract the features of key obstacles based on prior scene labels, sensor data, and the identification results of key obstacles, and then classify the features of key obstacles to obtain the actual scene associated with the key obstacles, such as needing to go around the key obstacle or needing to queue behind the key obstacle to pass.
[0032] Step 204: Generate scene labels for key obstacles based on the actual scene.
[0033] In this embodiment, the identification results of key obstacles can be labeled with the actual scene to obtain labeled data, which can be used to train the vehicle-side traffic jam model.
[0034] The method provided in the above embodiments of this disclosure proposes a scheme for automatic annotation of congested vehicle data for autonomous driving, which can significantly improve the speed of data annotation and lay the foundation for improving the model's recognition ability. Compared with rule-based automatic annotation algorithms, it reduces the need for manual rule design and is a sustainable automatic annotation scheme. Compared with traditional model recognition automatic annotation algorithms, it reduces the dependence on the amount of data. At the same time, since sensor data is also input, it also reduces the dependence on upstream recognition results. For congested vehicle recognition algorithms in autonomous driving scenarios, it can efficiently provide data for the training of vehicle-side congested vehicle models, reducing the dependence on manual annotation.
[0035] In some optional implementations of this embodiment, determining key obstacles based on the obstacle identification results includes: identifying obstacles whose obstacle identification results meet predetermined conditions as key obstacles, wherein the predetermined conditions include at least one of the following: the obstacle type is a vehicle, the obstacle speed is less than a predetermined speed threshold, the distance between the obstacle and the autonomous vehicle is less than a predetermined first distance threshold, and the distance between the obstacle and the navigation trajectory of the autonomous vehicle is less than a predetermined second distance threshold.
[0036] First, obtain the time point of the marked prior scene. At that time point, obtain the key obstacles using the following method: 1. Only retain obstacles that meet the following conditions: 1) The type of obstacle is a vehicle; 2) The speed is less than a predetermined speed threshold, for example, 0.3 m / s; 3) The distance between the obstacle and the autonomous vehicle is less than a predetermined first distance threshold, for example, within 15m to the left or right of the vehicle, within 80m in front of the vehicle, and within 10m behind the vehicle. 2. Calculate the distance between obstacles and the navigation trajectory of the autonomous vehicle: if it is greater than the second distance threshold (e.g., 2m), then filter it out.
[0037] It can quickly filter out non-critical obstacles and improve data processing speed.
[0038] In some optional implementations of this embodiment, the obstacle whose identification result meets the predetermined conditions is identified as a critical obstacle, including: in response to the existence of multiple obstacles that meet the predetermined conditions, the obstacle with the closest distance between the first projection point and the second projection point is identified as the critical obstacle, wherein the first projection point is the point on the navigation trajectory where the obstacle is projected, and the second projection point is the point on the navigation trajectory where the autonomous vehicle is projected.
[0039] Among all obstacles that meet the conditions, the obstacle closest to the vehicle's navigation trajectory projection point is identified as the critical obstacle. When multiple obstacles meet the predetermined conditions, if the vehicle is not traveling straight, the closest obstacle is not necessarily the critical obstacle; rather, the obstacle closest to the navigation trajectory projection point is the critical obstacle, thus improving the accuracy of identification.
[0040] In some optional implementations of this embodiment, the sensor data is collected by the autonomous vehicle when it detects a scenario where queuing or detour is required.
[0041] For driverless autonomous vehicles: If a queuing scenario is encountered, the vehicle's autonomous driving algorithm will mark it (e.g., "queueing required") and record the current time. If a detour scenario is encountered, the vehicle's autonomous driving algorithm will mark it (e.g., "detour required") and record the current time.
[0042] The obstacle recognition results collected at the moment when queuing or detour is detected have the most features, which can improve the accuracy of recognition.
[0043] In some optional implementations of this embodiment, the sensor data is collected when the autonomous vehicle detects a scenario where a detour is needed but manual intervention is required to queue, or when the autonomous vehicle detects a scenario where a detour is needed but manual intervention is required to queue.
[0044] For autonomous vehicles with a driver: In queuing scenarios, such as when there is a vehicle that should be waiting in line but the autonomous vehicle wants to go around it, the driver can take over and record the current time. In detour scenarios, such as when there is a vehicle that should be detoured but the autonomous vehicle wants to queue, the driver can take over and record the current time.
[0045] Human intervention, essentially labeling the data manually, is more accurate than the point-marking capabilities of autonomous driving algorithms. This improves the accuracy of recognition.
[0046] In some optional implementations of this embodiment, the data annotation model for autonomous vehicles is a Vision-Language Model (VLM).
[0047] Images from sensor data can be directly input into the VLM model. Obstacle recognition results can be encoded and then input into the VLM model.
[0048] Using a VLM model for fine-tuning has two advantages: 1. By leveraging the general knowledge of the VLM model, we can reduce our reliance on data and improve the model's generalization ability.
[0049] 2. Sensor data and obstacle recognition results can be used as input prompts without changing the model structure or retraining.
[0050] See also Figure 3a , Figure 3a This is a schematic diagram illustrating an application scenario of the traffic jam data annotation method according to this embodiment. Figure 3a In application scenarios, the following aspects are mainly involved: 1. Data Retrieval: First, retrieve the data from the data center and identify the data segments marked as needing to be queued (including obstacle identification results and other obstacle-related data). 2. Data Processing: Analyze and extract the sensor data and the identification results of obstacles on the line; 3. Key obstacle extraction: First, obtain the marked time point, and at that time point, obtain the key obstacle.
[0051] 4. For the moment a key obstacle appears, the trained VLM model is used to infer whether the obstacle is one that needs to be queued. Considering that there may be some scenarios in the data source that do not require queuing, only the identification results of obstacles that are considered to require queuing are retained after model identification as the automatic labeling results.
[0052] Further reference Figure 4 This illustrates a flow 400 of an embodiment of a method for training a traffic jam data annotation model. The flow 400 of the method for training a traffic jam data annotation model includes the following steps: Step 401: Obtain training samples for different scenarios.
[0053] In this embodiment, training samples can be obtained from a data center. The training samples include scene labels, prior scene identifiers, sample sensor data, and obstacle recognition results. The prior scenes include queuing scenarios and detour scenarios. The sample sensor data is collected by the autonomous vehicle when it detects a scenario requiring queuing or detouring. Alternatively, the sample sensor data can be collected when the autonomous vehicle detects a scenario requiring detouring but is manually intervened to queue, or when the autonomous vehicle detects a scenario requiring queuing but is manually intervened to detour.
[0054] Step 402: Based on the prior scene identification of the samples, the sensor data of the samples, and the identification results of the obstacles in the samples, the actual scene associated with the obstacles in the samples is determined through the initial blocked vehicle data annotation model; In this embodiment, compared to conventional model training methods, since the characteristics of the scene where the data is located are already known when the data is acquired (whether it requires queuing or detours), this information is used as the model input when creating the training data. This method can significantly reduce the difficulty of model recognition and improve its accuracy.
[0055] Step 403: Adjust the parameters of the initial traffic jam data annotation model based on the differences between the actual scene and the scene label.
[0056] In this embodiment, a loss value is calculated based on the difference between the actual scene and the scene label. The parameters of the initial traffic jam data labeling model are adjusted according to the loss value so that the loss value converges to a predetermined value or the iteration reaches a predetermined number of times. The loss function used to calculate the loss value can be a cross-entropy loss function, a logarithmic loss function, etc., and is not limited here.
[0057] The training process of the data annotation model for congested vehicles is as follows: Figure 3b As shown, the model's inputs include sensor data, prior scene information (queueing / detour required), and upstream structured information (obstacle identification results), and the output is the actual scene.
[0058] Further reference Figure 5 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a traffic jam data annotation device for autonomous driving, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.
[0059] like Figure 5 As shown, the traffic jam data annotation device 500 of this embodiment includes: an acquisition unit 501, a filtering unit 502, a determination unit 503, and an annotation unit 504. The acquisition unit 501 is configured to acquire sensor data collected by the autonomous vehicle under different prior scenarios and the obstacle identification results determined based on the sensor data; the filtering unit 502 is configured to determine key obstacles based on the obstacle identification results; the determination unit 503 is configured to determine the actual scenario associated with the key obstacles based on prior scenario identifiers, sensor data, and the key obstacle identification results, using a pre-trained traffic jam data annotation model; and the annotation unit 504 is configured to generate scene labels for the key obstacles based on the actual scenario.
[0060] In this embodiment, the specific processing of the acquisition unit 501, filtering unit 502, determination unit 503, and annotation unit 504 of the traffic jam data annotation device 500 can be referred to Figure 2 The corresponding steps are 201, 202, 203 and 204 in the embodiment.
[0061] In some optional implementations of this embodiment, the prior scenarios include queuing scenarios and detour scenarios; and the obstacle recognition results include at least one of the following obstacle information: type, speed, position, headlight status information, and door status information.
[0062] In some optional implementations of this embodiment, the filtering unit is further configured to: identify obstacles whose identification results meet predetermined conditions as key obstacles, wherein the predetermined conditions include at least one of the following: the type of obstacle is a vehicle, the speed of the obstacle is less than a predetermined speed threshold, the distance between the obstacle and the autonomous vehicle is less than a predetermined first distance threshold, and the distance between the obstacle and the navigation trajectory of the autonomous vehicle is less than a predetermined second distance threshold.
[0063] In some optional implementations of this embodiment, the filtering unit is further configured to: in response to the presence of multiple obstacles that meet predetermined conditions, identify the obstacle with the closest distance between the first projection point and the second projection point as the critical obstacle, wherein the first projection point is the point on which the obstacle is projected onto the navigation trajectory, and the second projection point is the point on which the autonomous vehicle is projected onto the navigation trajectory.
[0064] In some optional implementations of this embodiment, the sensor data is collected by the autonomous vehicle when it detects a scenario where queuing or detour is required.
[0065] In some optional implementations of this embodiment, the sensor data is collected when the autonomous vehicle detects a scenario where a detour is needed but manual intervention is required to queue, or when the autonomous vehicle detects a scenario where a detour is needed but manual intervention is required to queue.
[0066] In some optional implementations of this embodiment, the data annotation model for autonomous vehicles is a visual language model.
[0067] Further reference Figure 6 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of an apparatus for training a data annotation model for congested vehicles. This apparatus embodiment is similar to... Figure 4 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.
[0068] like Figure 6 As shown, the apparatus 600 for training a traffic jam data annotation model in this embodiment includes a sampling unit 601, a prediction unit 602, and an adjustment unit 603. The sampling unit 601 is configured to acquire training samples under different scenarios, wherein the training samples include scene labels, sample prior scene identifiers, sample sensor data, and sample obstacle recognition results, wherein the sample prior scenes include queuing scenarios and detour scenarios; the prediction unit 602 is configured to determine the actual scene associated with the sample obstacle based on the sample prior scene identifiers, the sample sensor data, and the sample obstacle recognition results, using an initial traffic jam data annotation model; the adjustment unit 603 is configured to adjust the parameters of the initial traffic jam data annotation model based on the difference between the actual scene and the scene label.
[0069] In this embodiment, the specific processing of the sampling unit 601, prediction unit 602, and adjustment unit 603 of the apparatus 600 for training the blocked vehicle data annotation model can be referred to... Figure 4 Steps 401, 402, and 403 in the corresponding embodiment.
[0070] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0071] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.
[0072] An electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method described in process 200 or 400.
[0073] A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described in process 200 or 400.
[0074] A computer program product includes a computer program that, when executed by a processor, implements the method described in process 200 or 400.
[0075] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0076] like Figure 7 As shown, device 700 includes a computing unit 701, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 702 or a computer program loaded into random access memory (RAM) 703 from storage unit 708. The RAM 703 may also store various programs and data required for the operation of device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via bus 704. Input / output (I / O) interface 705 is also connected to bus 704.
[0077] Multiple components in device 700 are connected to I / O interface 705, including: input unit 706, such as keyboard, mouse, etc.; output unit 707, such as various types of monitors, speakers, etc.; storage unit 708, such as disk, optical disk, etc.; and communication unit 709, such as network card, modem, wireless transceiver, etc. Communication unit 709 allows device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0078] The computing unit 701 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as the traffic jam data annotation method. For example, in some embodiments, the traffic jam data annotation method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the traffic jam data annotation method described above can be performed. Alternatively, in other embodiments, the computing unit 701 can be configured to perform the traffic jam data annotation method by any other suitable means (e.g., by means of firmware).
[0079] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0080] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0081] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0082] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0083] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0084] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.
[0085] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.
[0086] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-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 disclosure should be included within the scope of protection of this disclosure.
Claims
1. A method for automatic driving, comprising: obtaining sensor data collected by an automatic driving vehicle in different prior scenes and identification results of obstacles determined based on the sensor data; determining key obstacles based on the identification results of the obstacles; determining an actual scene associated with the key obstacles based on a prior scene identification, the sensor data and the identification results of the key obstacles by using a pre-trained blocked vehicle data labeling model; generating a scene label of the key obstacles based on the actual scene.
2. The method of claim 1, wherein, The prior scenes include a queuing scene and a detour scene; and The identification results of the obstacles include at least one of the following information of the obstacles: type, speed, position, light status information and door status information.
3. The method of claim 2, wherein, The determination of the key obstacles based on the identification results of the obstacles comprises: determining an obstacle as a key obstacle if the identification result of the obstacle meets a predetermined condition, wherein the predetermined condition includes at least one of the following: the type of the obstacle is a vehicle, the speed of the obstacle is less than a predetermined speed threshold, the distance between the obstacle and the automatic driving vehicle is less than a predetermined first distance threshold, and the distance between the obstacle and a navigation track of the automatic driving vehicle is less than a predetermined second distance threshold.
4. The method of claim 3, wherein, The determination of the key obstacles based on the identification results of the obstacles comprises: determining an obstacle as a key obstacle if the identification result of the obstacle meets a predetermined condition, wherein the predetermined condition includes at least one of the following:
5. The method of claim 1, wherein, the type of the obstacle is a vehicle, the speed of the obstacle is less than a predetermined speed threshold, the distance between the obstacle and the automatic driving vehicle is less than a predetermined first distance threshold, and the distance between the obstacle and a navigation track of the automatic driving vehicle is less than a predetermined second distance threshold.
6. The method of claim 1, wherein, The sensor data is collected by the automatic driving vehicle in a scene where the automatic driving vehicle detects a need for queuing or a need for detour.
7. The method of any one of claims 1-6, wherein, The sensor data is collected by the automatic driving vehicle in a scene where the automatic driving vehicle detects a need for detour but a human takes over for queuing, or a scene where the automatic driving vehicle detects a need for queuing but a human takes over for detour. The automatic driving vehicle data labeling model is a visual language model. 8.A method for training a blocked vehicle data labeling model, comprising: obtaining training samples in different scenes, wherein a training sample includes a scene label, a sample prior scene identification, sample sensor data and identification results of sample obstacles, wherein the sample prior scene includes a queuing scene and a detour scene; determining an actual scene associated with the sample obstacles based on the sample prior scene identification, the sample sensor data and the identification results of the sample obstacles by using an initial blocked vehicle data labeling model; and adjusting parameters of the initial blocked vehicle data labeling model based on a difference between the actual scene and the scene label. 9.A system for automatic driving, comprising: an automatic driving vehicle configured to collect sensor data and determine identification results of obstacles based on the sensor data in response to detecting a need for queuing or a need for detour, and report the sensor data and the identification results of the obstacles to a server. A server configured to perform the method of any one of claims 1-8.
10. The system of claim 9, wherein, The autonomous vehicle is further configured to: In response to detecting that the bypass is required but the manual takeover is queuing, or the queuing is required but the manual takeover is bypassing, collect sensor data and determine an identification result of an obstacle based on the sensor data, and report the sensor data and the identification result of the obstacle to a server.
11. A blocked vehicle data labeling apparatus for autonomous driving, comprising: an acquisition unit configured to acquire sensor data collected by an autonomous vehicle in different prior scenarios and an identification result of an obstacle determined based on the sensor data; a filtering unit configured to determine a key obstacle based on the identification result of the obstacle; a determination unit configured to determine an actual scenario associated with the key obstacle based on a prior scenario identification, the sensor data, and the identification result of the key obstacle by using a pre-trained blocked vehicle data labeling model; a labeling unit configured to generate a scenario label of the key obstacle based on the actual scenario.
12. An apparatus for training a blocked vehicle data labeling model, comprising: a sampling unit configured to acquire training samples in different scenarios, wherein a training sample comprises a scenario label, a sample prior scenario identification, sample sensor data, and a sample identification result of an obstacle, and wherein the sample prior scenario comprises a queuing scenario and a bypass scenario; a prediction unit configured to determine an actual scenario associated with the sample obstacle based on the sample prior scenario identification, the sample sensor data, and the sample identification result of the obstacle by using an initial blocked vehicle data labeling model; an adjustment unit configured to adjust parameters of the initial blocked vehicle data labeling model based on a difference between the actual scenario and the scenario label.
13. An electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.
14. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to cause the computer to perform the method of any one of claims 1-8.
15. A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-8.