Automatic driving target scene identification method and device, terminal equipment and automobile
By acquiring the time-series information of the current vehicle and the target vehicle, as well as high-precision maps, the driving status of the target vehicle is identified, which solves the problem of low prediction accuracy in complex traffic scenarios of autonomous driving technology and realizes high-precision target selection and strategy optimization under complex working conditions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHONGQING CHANGAN TECH CO LTD
- Filing Date
- 2023-03-30
- Publication Date
- 2026-07-24
AI Technical Summary
Existing autonomous driving technologies have low prediction accuracy in complex traffic scenarios and are difficult to adapt to diverse operating conditions.
By acquiring the time-series information of the current vehicle and the target vehicle, as well as high-precision maps, the relative position and driving status of the vehicles are determined. The high-precision map information is then used to determine the scene of the target vehicle and identify its driving status.
It improves the accuracy of identifying the driving status of target vehicles, provides data samples for different scenarios, supports offline prediction training, and optimizes the performance of autonomous driving strategies under complex conditions.
Smart Images
Figure CN116198509B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of autonomous driving technology, specifically to a method for identifying autonomous driving target scenes, a device for identifying autonomous driving target scenes, a terminal device, and a vehicle. Background Technology
[0002] Driving in dynamically changing traffic scenarios is an extremely challenging task for autonomous vehicles, especially on urban roads. When vehicles face complex traffic situations, the ability to interpret the scenarios they are in and to predict potential traffic participants will play a crucial role.
[0003] Scene recognition for target vehicles comprises both static and dynamic components. Static recognition refers to the established scene constituted by a static environment, such as highways, ramps, and intersections formed by road types and structures. Dynamic recognition refers to the dynamic interactions between traffic participants and the vehicle itself during movement, such as adjacent vehicles cutting in or vehicles braking ahead. Since different target prediction technologies exhibit variations in computational efficiency and performance across different application scenarios, evaluating and optimizing different prediction algorithms through scene recognition, and selectively activating and training these algorithms, can improve the overall computational efficiency and performance of the prediction algorithm.
[0004] Existing natural driving scene recognition and extraction methods are usually based on the perspective of the vehicle to identify driving scenes, which directly affect the behavior decision-making and planning of the intelligent vehicle itself. However, scene extraction and prediction for target vehicles are mostly concentrated in a certain type of specific scene. They use machine learning or deep learning methods to directly predict behavior or trajectory, which has low prediction accuracy and is difficult to adapt to complex working conditions. Summary of the Invention
[0005] One objective of this invention is to provide a method for recognizing target scenes in autonomous driving, so as to solve the problems of low prediction accuracy and difficulty in adapting to complex working conditions in the prior art; a second objective is to provide a device for recognizing target scenes in autonomous driving; a third objective is to provide a terminal device; and a fourth objective is to provide a vehicle.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0007] A method for identifying target scenes in autonomous driving, comprising:
[0008] When a target vehicle is identified, the first time-series information of the current vehicle and the second time-series information of the target vehicle are obtained. The first time-series information includes the first vehicle status information of the current vehicle at each consecutive sampling moment within a preset sampling period. The second time-series information includes the second vehicle status information of the target vehicle at each consecutive sampling moment within the preset sampling period. The first vehicle status information includes the first location information of the current vehicle. The second vehicle status information includes the second location information of the target vehicle.
[0009] Acquire high-precision map information of the target area including the current vehicle and the target vehicle; determine the first lane information of the current vehicle based on the high-precision map information and the first location information; determine the second lane information of the target vehicle based on the high-precision map information and the second location information.
[0010] If, based on the first lane information and the second lane information, it is determined that the target vehicle is in a first target scenario, and if the second time-series information contains second vehicle status information that meets the first preset condition, the target vehicle is determined to be in a first driving state; and
[0011] If the target vehicle is determined to be in the second target scenario based on the first lane information and the second lane information, and if the second time sequence information contains second vehicle status information that meets the second preset conditions, the target vehicle is determined to be in the second driving state.
[0012] Furthermore, after obtaining the first time-series information of the current vehicle and the second time-series information of the target vehicle, the method further includes:
[0013] Determine the target sampling time;
[0014] The set of first vehicle state information with sampling times shorter than the target sampling time in the first time series information is defined as the first sub-time series information, and the set of first vehicle state information with sampling times longer than the target sampling time in the first time series information is defined as the second sub-time series information; and
[0015] The set of second vehicle state information whose sampling time is less than the target sampling time in the second time sequence information is determined as the third sub-time sequence information, and the set of second vehicle state information whose sampling time is greater than the target sampling time in the second time sequence information is determined as the fourth sub-time sequence information.
[0016] Furthermore, based on the high-precision map information and the first location information, the first lane information of the current vehicle is determined, including:
[0017] Based on the high-precision map information and the first location information corresponding to the target sampling time, the lane with the shortest vertical distance between the current vehicle and the center line of each lane in the target area is determined as the lane where the current vehicle is located, and the first lane information of the current vehicle is generated based on the lane where the current vehicle is located.
[0018] Determining the second lane information of the target vehicle based on the high-precision map information and the second location information includes:
[0019] Based on the second location information corresponding to the target sampling time in the high-precision map information, the vertical distance between the target vehicle and the center line of each lane in the target area is determined;
[0020] The vertical distance with the minimum value among all vertical distances corresponding to the target vehicle is determined as the target vertical distance.
[0021] If the vertical distance to the target is not greater than the first distance threshold, the lane corresponding to the vertical distance to the target is determined as the lane where the target vehicle is located, and the second lane information of the target vehicle is generated based on the lane where the target vehicle is located.
[0022] Furthermore, determining that the target vehicle is in the first target scenario based on the first lane information and the second lane information includes:
[0023] If the lane where the target vehicle is located is the same lane as the current vehicle, or if the lane where the target vehicle is located is an adjacent lane to the lane following the current vehicle, the target vehicle is determined to be in the first target scenario.
[0024] Determining that the target vehicle is in the second target scenario based on the first lane information and the second lane information includes:
[0025] If the lane following the target vehicle is the same as the lane currently occupied by the vehicle, or if the lane following the target vehicle is the same as the lane currently occupied by the vehicle, the target vehicle is determined to be in the second target scenario.
[0026] Furthermore, if the second time-series information contains second vehicle status information that meets the first preset condition, determining the target vehicle as the first driving state includes: if the fourth sub-time-series information contains second vehicle status information that meets the first preset condition, determining the target vehicle as the first driving state; the first preset condition includes:
[0027] Based on the second location information and the high-precision map information, it is determined that the target vehicle is in the lane where the current vehicle is located, and the vertical distance between the target vehicle and the lane centerline of the lane where the current vehicle is located is lower than the second distance threshold.
[0028] Furthermore, if the second time-series information contains second vehicle status information that meets the second preset condition, determining the target vehicle as being in the second driving state includes: if the fourth sub-time-series information contains second vehicle status information that meets the second preset condition, determining the target vehicle as being in the second driving state; the second preset condition includes:
[0029] Based on the second location information and the high-precision map information, it is determined that the target vehicle is in the lane following the current vehicle's lane, and that the target vehicle entered the lane following the current vehicle's lane before the current vehicle.
[0030] Furthermore, determine the successor lane to the lane where the target vehicle enters the current vehicle's lane before the current vehicle, including:
[0031] The sampling time in the fourth sub-time series information that represents the target vehicle entering the lane of the current vehicle is determined as the first merging time, and the sampling time in the second sub-time series information that represents the current vehicle entering the lane of the current vehicle is determined as the second merging time.
[0032] If the first merging time is less than the second merging time, it is determined that the target vehicle enters the next lane of the lane where the current vehicle is located before the current vehicle.
[0033] An autonomous driving target scene recognition device, comprising:
[0034] The data acquisition module is configured to acquire first time-series information of the current vehicle and second time-series information of the target vehicle when the target vehicle is identified. The first time-series information includes first vehicle status information of the current vehicle at each consecutive sampling moment within a preset sampling period. The second time-series information includes second vehicle status information of the target vehicle at each consecutive sampling moment within the preset sampling period. The first vehicle status information includes first location information of the current vehicle and the second vehicle status information includes second location information of the target vehicle.
[0035] The lane information determination module is configured to acquire high-precision map information of the target area including the current vehicle and the target vehicle, determine the first lane information of the current vehicle based on the high-precision map information and the first location information, and determine the second lane information of the target vehicle based on the high-precision map information and the second location information.
[0036] The target vehicle driving state determination module is configured to, when determining that the target vehicle is in a first target scenario based on the first lane information and the second lane information, determine that the target vehicle is in the first driving state if the second time-series information contains second vehicle state information that meets a first preset condition; and
[0037] If the target vehicle is determined to be in the second target scenario based on the first lane information and the second lane information, and if the second time sequence information contains second vehicle status information that meets the second preset conditions, the target vehicle is determined to be in the second driving state.
[0038] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for recognizing autonomous driving target scenes.
[0039] A vehicle includes: the aforementioned autonomous driving target scene recognition device.
[0040] The beneficial effects of this invention are:
[0041] Based on the location information of the target vehicle and the current vehicle, the relative position of the target vehicle and the current vehicle is determined. Based on the driving trajectory of the target vehicle within a preset sampling period and the relative position of the target vehicle and the current vehicle, the driving state of the target vehicle is determined. This can accurately identify the driving state of the target vehicle, thereby improving the accuracy of data labeling of the driving state of the target vehicle. This is beneficial for providing data samples for offline prediction training under different scenarios, thus meeting the requirements for target selection, strategy activation and performance optimization for autonomous driving under complex and diverse working conditions. Attached Figure Description
[0042] Figure 1 This is a flowchart of the autonomous driving target scene recognition method of the present invention;
[0043] Figure 2 This is a schematic diagram illustrating the target scene discrimination of the present invention;
[0044] Figure 3 This is a schematic diagram illustrating the interaction relationship determination between the target vehicle and the current vehicle according to the present invention;
[0045] Figure 4 This is a schematic block diagram of the autonomous driving target scene recognition device of the present invention;
[0046] Figure 5 This is a schematic diagram of a terminal device structure.
[0047] Among them, 10-terminal device, 100-processor, 101-memory, and 102-computer program. Detailed Implementation
[0048] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.
[0049] like Figure 1 As shown, this embodiment proposes a method for recognizing target scenes in autonomous driving, including:
[0050] S100. When a target vehicle is identified, the first time sequence information of the current vehicle and the second time sequence information of the target vehicle are obtained. The first time sequence information includes the first vehicle status information of the current vehicle at each consecutive sampling moment within a preset sampling period. The second time sequence information includes the second vehicle status information of the target vehicle at each consecutive sampling moment within a preset sampling period. The first vehicle status information includes the first position information of the current vehicle. The second vehicle status information includes the second position information of the target vehicle.
[0051] S200: Obtain high-precision map information of the target area including the current vehicle and the target vehicle; determine the first lane information of the current vehicle based on the high-precision map information and the first location information; and determine the second lane information of the target vehicle based on the high-precision map information and the second location information.
[0052] S300: If, based on the first lane information and the second lane information, it is determined that the target vehicle is in the first target scenario, and if the second time-series information contains second vehicle state information that meets the first preset condition, the target vehicle is determined to be in the first driving state; and
[0053] If the target vehicle is determined to be in the second target scenario based on the first lane information and the second lane information, and if the second time sequence information contains second vehicle status information that meets the second preset conditions, the target vehicle is determined to be in the second driving state.
[0054] Thus, this application determines the relative position of the target vehicle and the current vehicle based on the target vehicle's location information and the current vehicle's location information, and determines the target vehicle's driving state based on the target vehicle's driving trajectory within a preset sampling period and its relative position to the current vehicle. This enables accurate identification of the target vehicle's driving state, thereby improving the accuracy of data labeling of the target vehicle's driving state. It is beneficial for providing data samples for offline prediction training under different scenarios, thus meeting the requirements for target selection, strategy activation, and performance optimization in complex and diverse working conditions for autonomous driving.
[0055] In step S100, the preset sampling period can be set to 121ms. Therefore, the first and second time-series information include vehicle state information arranged sequentially from the 1st ms to the 121st ms. For example, the first time-series information can be represented as {Q1, Q2, ..., Qn}, where Q1 represents the first vehicle state information collected at the 1st ms, and Qn represents the first vehicle state information collected at the nth ms. The vehicle state information includes, but is not limited to, the vehicle's position, speed, and direction of travel. It is understood that the target vehicle's position information can be determined through information fusion based on the current vehicle's position, speed, and other information.
[0056] To further improve the accuracy of target scene recognition, after obtaining the first temporal information of the current vehicle and the second temporal information of the target vehicle, the method of this application further includes:
[0057] S101. Determine the target sampling time. In this application, the target sampling time is determined to be the 41st ms, and the target sampling time is used as the current time.
[0058] S102. The set of vehicle state information with sampling times less than the target sampling time in the first time-series information is defined as the first sub-time-series information, and the set of vehicle state information with sampling times greater than the target sampling time in the first time-series information is defined as the second sub-time-series information. Using 41ms as the boundary, the first time-series information is divided into first sub-time-series information such as {Q1, Q2, ... Q40} and second sub-time-series information such as {Q42, Q43, ... Q121}; and the set of vehicle state information with sampling times less than the target sampling time in the second time-series information is defined as the third sub-time-series information, and the set of vehicle state information with sampling times greater than the target sampling time in the second time-series information is defined as the fourth sub-time-series information. Similarly, the second time-series information is divided into third sub-time-series information such as {Q1, Q2, ... Q40} and fourth sub-time-series information such as {Q42, Q43, ... Q121}. In this application, the first sub-time series information is used as the historical time series information of the current vehicle, and the second sub-time series information is used as the future time series information of the current vehicle. For example, the preset sampling period of 121ms includes a historical observation period of 40ms and a future observation period of 80ms. Similarly, the third sub-time series information is used as the historical time series information of the target vehicle, and the fourth sub-time series information is used as the future time series information of the target vehicle.
[0059] In step S200, the first lane information of the current vehicle is determined based on the high-precision map information and the first position information. This includes: determining the lane with the shortest perpendicular distance between the current vehicle and the center lines of all lanes in the target area based on the high-precision map information and the first position information corresponding to the target sampling time; and generating the first lane information of the current vehicle based on the lane it is in. Specifically, all lane center lines in the target area are traversed using the high-precision map information, and the coordinates of the beginning and end points of each lane segment are extracted. The vector formed by the lane start point A, the lane end point B, and the target position point (i.e., the current vehicle's position point C) is used for calculation. When the conditions are met, it is determined whether the target location C is located in the middle area of the road segment, thereby achieving the initial matching of the current vehicle and the lane. Then, the vertical distance between the current vehicle and the center line of each lane that meets the initial matching conditions is calculated, the lane with the closest vertical distance is determined as the lane where the current vehicle is located, the lane ID of the lane where the current vehicle is located and the minimum vertical distance from the center line of the lane are recorded, and the first location information is generated.
[0060] Similar to the current vehicle's lane matching method, the second lane information of the target vehicle is determined based on high-precision map information and second location information. This includes: determining the vertical distance between the target vehicle and the center lines of each lane in the target area based on the second location information corresponding to the target sampling time in the high-precision map information, and determining the vertical distance with the minimum value among all vertical distances corresponding to the target vehicle as the target vertical distance. The difference from the current vehicle's lane matching method is that, when determining the lane where the target vehicle is located, if the target vertical distance is not greater than a first distance threshold (e.g., 100m), the lane corresponding to the target vertical distance is determined as the lane where the target vehicle is located. The lane ID of the current vehicle's lane and the minimum vertical distance from the lane center line are recorded, and the second lane information of the target vehicle is generated based on the lane where the target vehicle is located. It is understandable that when the calculated target vertical distance of the target vehicle is greater than the first distance threshold (e.g., 100m), it indicates that the lateral distance between the target vehicle and the current vehicle is very far, and the target vehicle's driving state will not affect the current vehicle.
[0061] To further optimize the data samples and improve the accuracy of offline prediction training, this application further determines the target scene in which the target vehicle is located and writes the scene into the second vehicle state information at the corresponding sampling time, such as... Figure 2 As shown, the target scene determination methods include:
[0062] S1. Determine if the target vehicle is facing an intersection. Specifically, first, obtain all lane information within the target area using a high-precision map. Determine the lane closest to the target vehicle's current lane, such as the lane following the target vehicle. Then, determine if the target vehicle is facing an intersection by analyzing the attributes of the following lane. If the lane attribute is a virtual lane or marked as an intersection, it indicates that the target vehicle's current lane is at an intersection; otherwise, it is not. The intersection attribute of the lane is typically obtained through high-precision mapping or perception fusion.
[0063] S2. Determine if the target vehicle is facing a fork in the road. Specifically, determine whether it is facing a fork in the road by counting the number of lanes following the target vehicle's lane. If the number of lanes following the target vehicle is greater than 1, it is a fork in the road; otherwise, it is not a fork in the road.
[0064] S3. Determine if the target vehicle is facing a merging intersection. Specifically, determine whether it is facing a merging intersection by judging the number of preceding lanes of the lane following the target vehicle. If the preceding lane size of the following lane is greater than 1, it is a merging intersection; otherwise, it is not a merging intersection.
[0065] S4. Determine if the target vehicle is facing a straight road or a curve. Specifically, determine whether the target vehicle is facing a merging intersection by judging the curvature of the lane in which it is located. By performing cubic curve fitting on the trajectory points of the current lane, the curvature and rate of change of curvature of the current lane are obtained. If the curvature is >0.002, it is a curve; otherwise, it is a straight road.
[0066] like Figure 3 As shown, before determining the driving status of the target vehicle, the interaction relationship between the target vehicle and the current vehicle is first determined, i.e., whether the target vehicle is a cutting-in target, a merging target, or a following target. Specifically, a high-precision map is first used to search for the current vehicle's lane and all its subsequent lanes. A depth-first search can be used to find all subsequent lanes of the current vehicle, using the current vehicle's current lane as the root node and traversing backwards by accumulating distances. All subsequent lanes within a set distance, such as 250m backwards, are found, and the lane IDs, including the current lane, are stored in a search container. Therefore, in step S300 of this application, the relationship between the target vehicle's lane and the current vehicle's lane is used to determine whether the target vehicle is a candidate cutting-in target. If the target vehicle's lane is the same as the current vehicle's lane, or if the target vehicle's lane is an adjacent lane to a subsequent lane of the current vehicle's lane, and the vertical distance between the target vehicle and the centerline of the current vehicle's lane is between 1.0 and 4.75m, the target vehicle is determined to be in the first target scenario, i.e., the target vehicle is determined to be a candidate cutting-in target.
[0067] Similarly, in this application, the relationship between the lane where the target vehicle is located and the lane where the current vehicle is located is determined to determine whether it is a candidate merging target. If the lane following the lane where the target vehicle is located is the lane where the current vehicle is located, or the lane following the lane where the target vehicle is located is the lane following the lane where the current vehicle is located, the target vehicle is determined to be in the second target scenario, that is, the target vehicle is determined to be a candidate merging target.
[0068] Similarly, if the target vehicle is in the same lane as the current vehicle, or the target vehicle is in the lane following the current vehicle, and the vertical distance between the target vehicle and the center line of its lane is within 1 meter, the target vehicle is identified as a candidate vehicle to follow.
[0069] After determining the candidate target category of the target vehicle, it is further marked according to the future trajectory of the target vehicle, that is, the second vehicle status information obtained within the future observation period. The future trajectory of the target vehicle is used to determine whether the target vehicle has cut into or merged into the lane, thereby determining the driving status of the target vehicle and establishing corresponding data.
[0070] When determining the driving state of a target vehicle, the first step is to filter the identified candidate targets, such as candidate entry targets. Based on the second vehicle state information of each candidate entry target for its future observation period, the future trajectory of the corresponding target vehicle is determined. Then, according to the first preset condition, the driving state of the target vehicle is judged as follows: Combining high-precision map information, if there is a point in the future trajectory indicating that the target vehicle is within the lane currently occupied by the current vehicle, and the vertical distance between the target vehicle and the centerline of the current vehicle's lane is less than a second distance threshold, then the target vehicle is determined to be in the first driving state, i.e., the entry state. The second vehicle state information corresponding to this target vehicle is stored in the entry scenario database, establishing a database for target vehicle entry behavior. This provides better training samples for models or algorithms that predict the entry behavior of target vehicles. The second distance threshold can be 1 / 4 of the lane width.
[0071] When determining whether the target vehicle's driving state is in the second driving state, i.e., the merging state, candidate merging targets are first screened. Based on the second vehicle state information of each candidate merging target in the future observation period, the future trajectory of the corresponding target vehicle is determined. Then, the driving state of the target vehicle is judged according to the second preset condition as follows: Combining high-precision map information, if the future trajectory indicates that the target vehicle is in the successor lane of the current vehicle's lane, and it is determined that in the future observation period, the sampling time of the target vehicle entering the successor lane of the current vehicle's lane is earlier than the sampling time of the current vehicle entering the successor lane of the current vehicle's lane, the target vehicle is determined to be in the second driving state, i.e., the merging state. The second vehicle state information corresponding to the target vehicle is stored in the merging scene database to establish a database for the merging behavior of the target vehicle, so as to provide better training samples for models or algorithms that predict the merging behavior of the target vehicle. Understandably, when determining whether a target vehicle entered the lane following the current vehicle's lane before the current vehicle, the sampling time of the current vehicle's entry into the lane can be determined using the current vehicle's first time-series information. For example, after determining the starting point of the lane following the current vehicle's lane based on high-precision map information, the first vehicle status information indicating that the vehicle is in the lane is queried from the first time-series information. The sampling time of the obtained first vehicle status information is then determined. Since the first vehicle status information in the first time-series information is arranged in ascending order of sampling time, the sampling time with the minimum value can be determined as the time when the current vehicle entered the lane. Similarly, the time when the target vehicle entered the lane can be determined. By comparing the sampling times of the current vehicle and the target vehicle entering the lane, it can be determined whether the target vehicle entered the lane before the current vehicle.
[0072] like Figure 4As shown, an autonomous driving target scene recognition device includes: a data acquisition module configured to acquire, upon recognizing a target vehicle, first temporal information of the current vehicle and second temporal information of the target vehicle, wherein the first temporal information includes first vehicle state information of the current vehicle at each consecutive sampling moment within a preset sampling period, and the second temporal information includes second vehicle state information of the target vehicle at each consecutive sampling moment within the preset sampling period, wherein the first vehicle state information includes first position information of the current vehicle, and the second vehicle state information includes second position information of the target vehicle; and a lane information determination module configured to acquire high-precision map information of the target area including the current vehicle and the target vehicle. The system determines the first lane information of the current vehicle based on high-precision map information and first location information, and determines the second lane information of the target vehicle based on high-precision map information and second location information. The target vehicle driving state determination module is configured to determine the target vehicle as a first driving state if, when the target vehicle is determined to be in a first target scenario based on the first lane information and second lane information, there is second vehicle state information in the second time sequence information that meets the first preset condition; and to determine the target vehicle as a second driving state if, when the target vehicle is determined to be in a second target scenario based on the first lane information and second lane information, there is second vehicle state information in the second time sequence information that meets the second preset condition.
[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0074] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned method for recognizing autonomous driving target scenes.
[0075] like Figure 5 The diagram shown is a schematic representation of a terminal device provided in an embodiment of this application. Figure 5As shown, the terminal device 10 of this embodiment includes a processor 100, a memory 101, and a computer program 102 stored in the memory 101 and executable on the processor 100. When the processor 100 executes the computer program 102, it implements the steps in the above method embodiments. Alternatively, when the processor 100 executes the computer program 102, it implements the functions of each module / unit in the above device embodiments.
[0076] For example, computer program 102 may be divided into one or more modules / units, one or more of which are stored in memory 101 and executed by processor 100 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 102 in terminal device 10.
[0077] Terminal device 10 may be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. Terminal device 10 may include, but is not limited to, a processor 100 and a memory 101. Those skilled in the art will understand that... Figure 5 This is merely an example of terminal device 10 and does not constitute a limitation on terminal device 10. It may include more or fewer components than shown, or combine certain components, or different components. For example, terminal device may also include input / output devices, network access devices, buses, etc.
[0078] The processor 100 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.
[0079] The memory 101 can be an internal storage unit of the terminal device 10, such as a hard disk or RAM of the terminal device 10. The memory 101 can also be an external storage device of the terminal device 10, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device 10. Furthermore, the memory 101 can include both internal and external storage units of the terminal device 10. The memory 101 is used to store computer programs and other programs and data required by the terminal device 10. The memory 101 can also be used to temporarily store data that has been output or will be output.
[0080] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0081] A vehicle includes: the aforementioned autonomous driving target scene recognition device.
[0082] In summary, this application is applicable to hierarchical scene recognition for target prediction in urban autonomous driving, effectively improving the accuracy of target vehicle driving status recognition. It can provide target screening information for the online prediction process based on the interpretation of the target's surrounding environment and relative position to the vehicle, and provide data sample labeling for offline prediction training based on the observation of the target's driving trajectory over a period of time. Thus, it can establish a target database for different scenarios required for target prediction to meet different target prediction needs, enabling target screening, strategy activation, and performance optimization under complex and diverse working conditions.
[0083] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0084] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention.
Claims
1. A method for recognizing target scenes in autonomous driving, characterized in that, include: When a target vehicle is identified, the first time-series information of the current vehicle and the second time-series information of the target vehicle are obtained. The first time-series information includes the first vehicle status information of the current vehicle at each consecutive sampling moment within a preset sampling period. The second time-series information includes the second vehicle status information of the target vehicle at each consecutive sampling moment within the preset sampling period. The first vehicle status information includes the first location information of the current vehicle. The second vehicle status information includes the second location information of the target vehicle. Determine the target sampling time; The set of first vehicle state information with a sampling time less than the target sampling time in the first time sequence information is determined as the first sub-time sequence information, and the set of first vehicle state information with a sampling time greater than the target sampling time in the first time sequence information is determined as the second sub-time sequence information. The first sub-time sequence information is used as the historical time sequence information of the current vehicle, and the second sub-time sequence information is used as the future time sequence information of the current vehicle. as well as The set of second vehicle state information with a sampling time less than the target sampling time in the second time series information is determined as the third sub-time series information, and the set of second vehicle state information with a sampling time greater than the target sampling time in the second time series information is determined as the fourth sub-time series information. The third sub-time series information is used as the historical time series information of the target vehicle, and the fourth sub-time series information is used as the future time series information of the target vehicle. Acquire high-precision map information of the target area including the current vehicle and the target vehicle; determine the first lane information of the current vehicle based on the high-precision map information and the first location information; determine the second lane information of the target vehicle based on the high-precision map information and the second location information. If, based on the first lane information and the second lane information, it is determined that the target vehicle is in the first target scenario, and if the fourth sub-time sequence information contains second vehicle status information that meets the first preset condition, the target vehicle is determined to be in the first driving state. as well as If the target vehicle is determined to be in the second target scenario based on the first lane information and the second lane information, and if the fourth sub-time sequence information contains second vehicle status information that meets the second preset condition, the target vehicle is determined to be in the second driving state.
2. The method for recognizing target scenes for autonomous driving according to claim 1, characterized in that, Based on the high-precision map information and the first location information, the first lane information of the current vehicle is determined, including: Based on the high-precision map information and the first location information corresponding to the target sampling time, the lane with the shortest vertical distance between the current vehicle and the center line of each lane in the target area is determined as the lane where the current vehicle is located, and the first lane information of the current vehicle is generated based on the lane where the current vehicle is located. Determining the second lane information of the target vehicle based on the high-precision map information and the second location information includes: Based on the second location information corresponding to the target sampling time in the high-precision map information, the vertical distance between the target vehicle and the center line of each lane in the target area is determined; The vertical distance with the minimum value among all vertical distances corresponding to the target vehicle is determined as the target vertical distance. If the vertical distance to the target is not greater than the first distance threshold, the lane corresponding to the vertical distance to the target is determined as the lane where the target vehicle is located, and the second lane information of the target vehicle is generated based on the lane where the target vehicle is located.
3. The method for recognizing target scenes for autonomous driving according to claim 2, characterized in that, Determining that the target vehicle is in the first target scenario based on the first lane information and the second lane information includes: If the lane where the target vehicle is located is the same lane as the current vehicle, or if the lane where the target vehicle is located is an adjacent lane to the lane following the current vehicle, the target vehicle is determined to be in the first target scenario. Determining that the target vehicle is in the second target scenario based on the first lane information and the second lane information includes: If the lane following the target vehicle is the same as the lane currently occupied by the vehicle, or if the lane following the target vehicle is the same as the lane currently occupied by the vehicle, the target vehicle is determined to be in the second target scenario.
4. The method for recognizing target scenes for autonomous driving according to claim 2, characterized in that, The first preset condition includes: Based on the second location information and the high-precision map information, it is determined that the target vehicle is in the lane where the current vehicle is located, and the vertical distance between the target vehicle and the lane centerline of the lane where the current vehicle is located is lower than the second distance threshold.
5. The method for recognizing target scenes for autonomous driving according to claim 2, characterized in that, The second preset condition includes: Based on the second location information and the high-precision map information, it is determined that the target vehicle is in the lane following the current vehicle's lane, and that the target vehicle entered the lane following the current vehicle's lane before the current vehicle.
6. The method for recognizing target scenes for autonomous driving according to claim 5, characterized in that, Determine the successor lane to the lane where the target vehicle enters before the current vehicle, including: The sampling time in the fourth sub-time series information that represents the target vehicle entering the lane of the current vehicle is determined as the first merging time, and the sampling time in the second sub-time series information that represents the current vehicle entering the lane of the current vehicle is determined as the second merging time. If the first merging time is less than the second merging time, it is determined that the target vehicle enters the next lane of the lane where the current vehicle is located before the current vehicle.
7. An autonomous driving target scene recognition device, employing the autonomous driving target scene recognition method as described in any one of claims 1-6, characterized in that, include: The data acquisition module is configured to acquire first time-series information of the current vehicle and second time-series information of the target vehicle when the target vehicle is identified. The first time-series information includes first vehicle status information of the current vehicle at each consecutive sampling moment within a preset sampling period. The second time-series information includes second vehicle status information of the target vehicle at each consecutive sampling moment within the preset sampling period. The first vehicle status information includes first location information of the current vehicle and the second vehicle status information includes second location information of the target vehicle. The lane information determination module is configured to acquire high-precision map information of the target area including the current vehicle and the target vehicle, determine the first lane information of the current vehicle based on the high-precision map information and the first location information, and determine the second lane information of the target vehicle based on the high-precision map information and the second location information. The target vehicle driving status determination module is configured to determine the target vehicle as the first driving status if, when the target vehicle is determined to be in the first target scenario based on the first lane information and the second lane information, there is second vehicle status information in the second time sequence information that meets the first preset condition. as well as If the target vehicle is determined to be in the second target scenario based on the first lane information and the second lane information, and if the second time sequence information contains second vehicle status information that meets the second preset conditions, the target vehicle is determined to be in the second driving state.
8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the autonomous driving target scene recognition method according to any one of claims 1 to 6.
9. A car, characterized in that, include: The autonomous driving target scene recognition device as described in claim 7.
Citation Information
Patent Citations
CN115837919A