Automatic driving vehicle management method and system, electronic equipment and storage medium

Abstract

The invention relates to the technical field of automatic driving, in particular to an automatic driving vehicle management method and system, electronic equipment and a storage medium. The method comprises the following steps: firstly, automatically identifying and classifying a current driving scene based on vehicle real-time data, and selecting a most suitable driving strategy from a preset strategy library; then, map data are optimized in real time to reflect the current road condition; then predicting future trajectories of surrounding dynamic objects; planning a safe bypass path according to a prediction result; and finally, generating a corresponding vehicle control instruction. The system can adapt to complex and changeable driving environments, can improve the driving efficiency on the premise of guaranteeing the safety, and remarkably improves the intelligent level and practicability of an automatic driving system.

Description

Technical Field

[0001] The present application relates to the technical field of autonomous driving, and in particular to an autonomous driving vehicle management method, system, electronic device and storage medium. Background Art

[0002] With the rapid development of artificial intelligence and sensor technology, autonomous driving technology is being applied to the automotive industry. By integrating advanced perception, decision-making and control systems, autonomous vehicles can drive autonomously in complex and changing traffic environments, which is expected to significantly improve road safety, reduce traffic accidents, and improve travel efficiency and riding experience.

[0003] At present, autonomous driving systems usually adopt rule-based decision-making methods and preset driving strategies. These systems use on-board sensors to collect information about the surrounding environment and combine it with high-precision map data for positioning and path planning. At the same time, by perceiving and predicting the behavior of other traffic participants in real time, the system can make corresponding driving decisions, such as accelerating, decelerating, changing lanes or stopping.

[0004] However, when faced with complex and ever-changing driving scenarios, the preset driving strategies of existing autonomous driving systems are difficult to adapt to all possible driving situations, resulting in the system being less flexible in some special scenarios. This situation needs to be further improved. Summary of the invention

[0005] In order to solve the problem that the existing autonomous driving system is not flexible enough when facing complex and changeable driving scenarios, the present application provides an autonomous driving vehicle management method, system, electronic device and storage medium, which adopt the following technical solutions: In a first aspect, the present application provides an autonomous driving vehicle management method, comprising the following steps: Acquire real-time data of the autonomous driving vehicle, and automatically identify and classify the current driving scene based on the real-time data; According to the current driving scenario, selecting an automatic driving strategy suitable for the current driving scenario from a preset strategy library; Call and optimize map data in real time based on current scene features and vehicle real-time location; Based on the current driving scenario, the optimized map data and the surrounding environment information, predicting the future movement trajectory of the surrounding dynamic objects; Plan a safe detour route based on the prediction results, optimized map data and current vehicle status; Generate vehicle control instructions based on the autonomous driving strategy and the safe detour path.

[0006] By adopting the above technical solution, in congested urban roads, traditional autonomous driving systems may not be able to flexibly respond to emergencies, such as road construction or temporary diversions, resulting in wrong decisions or low driving efficiency; the present application first automatically identifies and classifies the current driving scenario based on the real-time vehicle data, and selects the most suitable driving strategy from the preset strategy library; then, the map data is optimized in real time to reflect the current road conditions; then the future trajectory of the surrounding dynamic objects is predicted; then a safe detour route is planned based on the prediction results; and finally, corresponding vehicle control instructions are generated; it can not only adapt to complex and changeable driving environments, but also improve driving efficiency while ensuring safety, significantly improving the intelligence level and practicality of the autonomous driving system.

[0007] Optionally, real-time data of the autonomous driving vehicle is obtained, and the current driving scene is automatically identified and classified according to the real-time data, specifically including the following steps: Acquire real-time data of the autonomous driving vehicle and pre-process the real-time data; Input the preprocessed real-time data into the pre-trained classification model, and output the preliminary classification results and confidence of the driving scene; When the confidence level is higher than a preset threshold, taking the preliminary classification result as the current driving scene type; When the confidence level is lower than a preset threshold, feature extraction is performed on the real-time data to obtain complexity features and traffic flow features of the driving environment; A secondary judgment is performed based on the classification result, confidence, complexity characteristics and traffic flow characteristics, and the current driving scene is determined based on the secondary judgment result.

[0008] By adopting the above technical solution, the present application first pre-processes the acquired real-time data and performs preliminary classification using a pre-trained classification model; then decides whether further analysis is needed based on the confidence level; if necessary, extracts environmental complexity and traffic flow characteristics; finally, performs a secondary judgment to determine the final scene type. Through the confidence threshold and secondary judgment mechanism, not only can common scenes be processed quickly, but also in-depth analysis can be performed on complex or ambiguous situations, greatly improving the accuracy and adaptability of scene recognition.

[0009] Optionally, according to the current driving scenario, selecting an automatic driving strategy suitable for the current driving scenario from a preset strategy library specifically includes the following steps: Query the preset strategy library to obtain a candidate strategy set that matches the current driving scenario; Score each strategy in the candidate strategy set, with the scoring criteria including safety, efficiency and adaptability; Select the strategy with the highest score as the current autonomous driving strategy; According to the characteristics of the current driving scene, the strategy parameters of the current autonomous driving strategy are dynamically fine-tuned using preset parameter adjustment rules, where the strategy parameters include a safety distance, an acceleration limit, and a steering angle.

[0010] By adopting the above technical solutions, autonomous driving vehicles need to adopt highly adaptable driving strategies in different driving scenarios, while traditional fixed strategy selection methods often fail to meet this requirement. For example, on a highway on a rainy day, simply applying the driving strategy for normal weather may ignore the safety risks caused by slippery road surfaces. The present application first selects a set of candidate strategies that match the current scenario from a preset strategy library, and comprehensively scores each candidate strategy based on safety, efficiency, and adaptability. The strategy with the highest score is then selected as the initial driving strategy. Finally, according to the current scenario characteristics, the preset parameter adjustment rules are used to dynamically fine-tune the strategy parameters. The application is able to select the driving strategy that best suits the current scenario, and can also fine-tune the strategy according to the specific situation, greatly improving the adaptability of the autonomous driving system to different driving environments.

[0011] Optionally, based on the current scene features and the real-time position of the vehicle, map data is called and optimized in real time, specifically including the following steps: Based on the real-time location of the vehicle, map data of the corresponding area is retrieved from the map database; Compare the current scene features with the retrieved map data to identify and mark dynamic changes in the map data; Based on the identified dynamic changes, the map update priority is dynamically evaluated to generate a map update strategy; According to the map update strategy, local map information is optimized in real time; The optimized map data is shared with surrounding vehicles and uploaded to the cloud server.

[0012] By adopting the above technical solution, in the case of urban road construction or temporary control, the use of outdated map data may cause the autonomous driving vehicle to make incorrect path planning, and even cause safety hazards; the present application first retrieves the map data of the corresponding area based on the real-time position of the vehicle, compares the current scene features with the map data, and identifies dynamic changes; then evaluates the map update priority and generates an update strategy, and optimizes the local map in real time according to the update strategy; finally, the optimized data is shared with surrounding vehicles and uploaded to the cloud; it can quickly respond to environmental changes and perform targeted updates according to importance, greatly improving the real-time and accuracy of map data.

[0013] Optionally, based on the current driving scene, the optimized map data and the surrounding environment information, predicting the future movement trajectory of the surrounding dynamic objects specifically includes the following steps: Based on the current driving scenario, determining a dynamic object type and a predicted time range; Obtain real-time status information of surrounding dynamic objects and prioritize the real-time status information according to the current driving scenario; Generate multiple possible trajectories of each dynamic object based on the sorted real-time state information; Combined with the optimized map data, a comprehensive trajectory prediction result reflecting the mutual influence of multiple dynamic objects is generated; Based on the comprehensive trajectory prediction results and the risk level of the current driving scenario, a dynamically adjusted safety envelope area is generated for each dynamic object.

[0014] By adopting the above technical solution, at busy intersections, traditional trajectory prediction algorithms may find it difficult to accurately predict the complex interactive behaviors of pedestrians, bicycles and other vehicles, resulting in errors in autonomous driving decisions or excessive conservatism; the present application first determines the type of dynamic object and the prediction time range based on the current driving scenario, obtains and prioritizes the real-time status information of the surrounding dynamic objects; then generates multiple possible trajectories for each object, and combines the optimized map data to generate a comprehensive prediction result that takes into account the mutual influence of multiple objects; finally, a dynamically adjusted safety envelope area is generated based on the prediction results and the risk level of the scenario; it can more accurately predict the trajectory of a single object and take into account the mutual influence between multiple objects; by combining priority sorting and dynamic safety envelope areas, the decision-making accuracy and safety of the autonomous driving system are greatly improved.

[0015] Optionally, the dynamic object types include pedestrians, non-motor vehicles, small motor vehicles, large motor vehicles, special vehicles and animals. Based on the current driving scene, determining the dynamic object type and the prediction time range specifically includes the following steps: Detect and classify dynamic objects in the surrounding environment in real time to obtain the type of dynamic objects; Based on the current driving scene characteristics, the dynamic object type is confirmed; According to the confirmed dynamic object type, extract corresponding behavior features from the preset object behavior pattern library; Set basic prediction time ranges for different types of dynamic objects based on the complexity and risk level of the current driving scenario; According to the motion state of the dynamic object, the basic prediction time range is adjusted to obtain a final prediction time range.

[0016] By adopting the above technical solution, the present application first detects and classifies dynamic objects in the surrounding environment in real time, and confirms the object type based on the current scene characteristics; then extracts the corresponding behavior characteristics from the preset behavior pattern library, and sets the basic prediction time range according to the scene complexity and risk level; finally, the prediction time range is dynamically adjusted according to the object motion state; it can more accurately identify the type of dynamic objects, and dynamically adjust the prediction strategy according to the scene characteristics and object behavior, thereby realizing the recognition and personalized prediction of various dynamic objects, and improving the accuracy and adaptability of the prediction.

[0017] In a second aspect, the present application provides an autonomous driving vehicle management system, comprising: A driving scene recognition module, used to obtain real-time data of the autonomous driving vehicle, and automatically recognize and classify the current driving scene based on the real-time data; A driving strategy determination module, configured to select an automatic driving strategy suitable for the current driving scenario from a preset strategy library according to the current driving scenario; A map data optimization module is used to call and optimize map data in real time according to the current scene characteristics and the real-time position of the vehicle; An action trajectory prediction module, used to predict the future action trajectory of surrounding dynamic objects based on the current driving scene, the optimized map data and the surrounding environment information; The detour path planning module is used to plan a safe detour path based on the prediction results, optimized map data and current vehicle status; A control instruction generation module is used to generate vehicle control instructions based on the automatic driving strategy and the safe detour path.

[0018] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned autonomous driving vehicle management method when executing the computer program.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned autonomous driving vehicle management method.

[0020] In summary, the present application includes at least one of the following beneficial technical effects: 1. This application first automatically identifies and classifies the current driving scene based on the real-time data of the vehicle, and selects the most suitable driving strategy from the preset strategy library; then, optimizes the map data in real time to reflect the current road conditions; then predicts the future trajectory of the surrounding dynamic objects; then plans a safe detour path based on the prediction results; and finally generates corresponding vehicle control instructions; it can not only adapt to complex and changeable driving environments, but also improve driving efficiency while ensuring safety, significantly improving the intelligence level and practicality of the autonomous driving system; 2. This application first pre-processes the acquired real-time data and uses the pre-trained classification model for preliminary classification; then decides whether further analysis is needed based on the confidence level; if necessary, extracts the environmental complexity and traffic flow characteristics; finally, performs a secondary judgment to determine the final scene type. Through the confidence threshold and secondary judgment mechanism, it can not only quickly process common scenes, but also conduct in-depth analysis for complex or ambiguous situations, greatly improving the accuracy and adaptability of scene recognition; 3. At busy intersections, traditional trajectory prediction algorithms may have difficulty accurately predicting the complex interactive behaviors of pedestrians, bicycles and other vehicles, resulting in errors or over-conservative decisions in autonomous driving; this application first determines the type of dynamic object and the prediction time range based on the current driving scenario, obtains and prioritizes the real-time status information of the surrounding dynamic objects; then generates multiple possible trajectories for each object, and combines the optimized map data to generate a comprehensive prediction result that takes into account the mutual influence of multiple objects; finally, a dynamically adjusted safety envelope area is generated based on the prediction results and the risk level of the scenario; it can more accurately predict the trajectory of a single object and take into account the mutual influence between multiple objects; by combining priority sorting and dynamic safety envelope areas, the decision-making accuracy and safety of the autonomous driving system are greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of an autonomous driving vehicle management method according to an embodiment of the present application; Figure 2 It is a flowchart of step S100 in an autonomous driving vehicle management method according to an embodiment of the present application; Figure 3 It is a flowchart of step S200 in an automatic driving vehicle management method according to an embodiment of the present application; Figure 4 It is a flowchart of step S300 in an automatic driving vehicle management method according to an embodiment of the present application; Figure 5 It is a flowchart of step S400 in an automatic driving vehicle management method according to an embodiment of the present application; Figure 6 It is a flowchart of step S410 in an autonomous driving vehicle management method according to an embodiment of the present application; Figure 7 This is a module diagram of an autonomous driving vehicle management system according to an embodiment of the present application; Figure 8 It is a diagram of the internal structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification and appended claims of the present application, the singular expressions "one", "a kind of", "said", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more listed items.

[0023] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0024] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0025] In the first aspect, the present application provides an autonomous driving vehicle management method, system, electronic device and storage medium, referring to Figure 1 , including the following steps: S100: Acquire real-time data of the autonomous driving vehicle, and automatically identify and classify the current driving scene based on the real-time data.

[0026] Among them, the real-time data of autonomous driving vehicles includes visual data, lidar data, millimeter-wave radar data, GPS positioning data, vehicle motion status data, etc.

[0027] Specifically, autonomous vehicles collect real-time data about the surrounding environment through multiple cameras, lidar, millimeter-wave radar and other sensors. After preprocessing, the data is input into the pre-trained scene classification model, which outputs the type of the current driving scene, such as urban roads, highways, school areas, etc., and the corresponding confidence level.

[0028] S200. Select an automatic driving strategy suitable for the current driving scenario from a preset strategy library according to the current driving scenario.

[0029] The preset strategy library stores autonomous driving strategies for different driving scenarios. Each strategy contains a series of parameters and rules to guide the vehicle's behavior decisions in specific scenarios.

[0030] Specifically, the system first selects a set of candidate strategies that match the current driving scenario from the preset strategy library based on the identified current driving scenario. Then, each candidate strategy is scored based on safety, efficiency, and adaptability. For example, in a congested urban road scenario, the system may choose a strategy that focuses more on safety and flexibility, while in a smooth highway scenario, it may choose a strategy that focuses more on efficiency. The strategy with the highest score will be selected as the current autonomous driving strategy.

[0031] S300: According to the current scene characteristics and the real-time position of the vehicle, the map data is called and optimized in real time.

[0032] Among them, map data is an important part of the autonomous driving system, which contains static information such as road networks, traffic signs, buildings, as well as dynamic information such as real-time traffic conditions.

[0033] Specifically, the system first retrieves the map data of the corresponding area from the map database based on the vehicle's GPS location. Then, it compares the current scene features with the retrieved map data to identify possible changes or errors in the map. For example, the system may find that a road is under construction or a new traffic sign has been added. Based on these findings, the system will update and optimize the map data in real time.

[0034] S400: predicting future motion trajectories of surrounding dynamic objects based on the current driving scenario, the optimized map data, and the surrounding environment information.

[0035] Among them, dynamic objects include other vehicles, pedestrians, cyclists and other road users.

[0036] Specifically, the system first detects and classifies the surrounding dynamic objects to determine the type and current state of each object. Then, it combines the characteristics of the current driving scene (such as road type, traffic rules) and the optimized map data to generate multiple possible future trajectories for each dynamic object. Trajectory prediction is based on probabilistic models such as Bayesian networks or recursive neural networks. For example, for a vehicle approaching an intersection, the system predicts the probability of it going straight, turning left, or turning right. Finally, the system comprehensively considers the predicted trajectories of all dynamic objects and generates a comprehensive prediction result that reflects the mutual influence of multiple objects.

[0037] S500: Plan a safe detour route based on the prediction results, the optimized map data and the current vehicle status.

[0038] Among them, the safe detour path refers to a driving path that can effectively avoid potential obstacles or dangerous situations while ensuring safety.

[0039] Specifically, the system first generates an initial path based on the optimized map data. The predicted trajectory of the dynamic object is then superimposed on this initial path to identify possible conflict points. For each potential conflict point, the system calculates multiple possible detour solutions, such as slowing down, changing lanes, or stopping. These solutions are evaluated for their safety and efficiency, and the optimal solution is selected as the final safe detour path. For example, if it is predicted that a car ahead may suddenly change lanes, the system will choose to slow down in advance or change lanes to an adjacent lane.

[0040] S600: Generate vehicle control instructions based on the autonomous driving strategy and the safe detour path.

[0041] Among them, vehicle control instructions refer to specific commands that directly control the vehicle's actuators, including but not limited to steering, acceleration, braking and other operations.

[0042] In one embodiment, referring to Figure 2 In step S100, real-time data of the autonomous driving vehicle is obtained, and the current driving scene is automatically identified and classified according to the real-time data, which specifically includes the following steps: S110: Acquire real-time data of the autonomous driving vehicle and pre-process the real-time data.

[0043] Specifically, the system first collects real-time data from various sensors of the vehicle. For visual data, preprocessing may include image denoising, adjusting brightness and contrast, image cropping and other operations. For lidar data, point cloud filtering, ground point removal and other processing are required. Millimeter wave radar data requires clutter suppression and target extraction. GPS data needs to be combined with inertial measurement unit data for Kalman filtering to improve positioning accuracy. Vehicle status data such as speed, acceleration, etc. may require unit conversion and outlier detection.

[0044] S120: input the preprocessed real-time data into the pre-trained classification model, and output the preliminary classification result and confidence of the driving scene.

[0045] Specifically, the pre-trained classification model is a deep neural network that receives pre-processed multimodal data as input. For example, visual data is input into CNN to extract spatial features, while time series data such as vehicle speed is input into RNN to extract temporal features. The features are then combined and passed through a fully connected layer for final classification. The model output may include probability distributions of multiple scene types, and the final confidence takes the highest probability value.

[0046] S130: When the confidence level is higher than a preset threshold, the preliminary classification result is used as the current driving scene type.

[0047] The preset threshold is a value pre-set according to system requirements and actual application scenarios, which is used to determine whether the classification result is sufficiently reliable. When the confidence exceeds this threshold, the system considers the classification result to be sufficiently accurate and directly adopts it.

[0048] Furthermore, in actual applications, the confidence threshold varies according to different scenario types. For example, for scenarios with higher security requirements, a higher threshold will be set.

[0049] S140: When the confidence level is lower than a preset threshold, feature extraction is performed on the real-time data to obtain complexity features of the driving environment and traffic flow features.

[0050] Among them, the complexity feature reflects the complexity of the current driving environment, including the road structure and obstacle distribution, and the traffic flow feature describes the density and movement pattern of vehicles and pedestrians on the current road.

[0051] S150, performing a secondary judgment based on the classification result, confidence, complexity characteristics and traffic flow characteristics, and determining the current driving scene based on the secondary judgment result.

[0052] Specifically, the secondary judgment uses a decision tree model with preliminary classification results, confidence, complexity characteristics and traffic flow characteristics as input. For example, if the preliminary classification result is "urban road", but the confidence is low, the system will further analyze the complexity and traffic flow characteristics. If it is found that the environment complexity is high (such as multiple intersections) and the traffic flow is large and unstable, the system will reclassify the scene as "complex urban traffic". On the contrary, if the complexity is low and the traffic flow is small and stable, the scene may be classified as "urban residential area". In the end, the system will output a more accurate scene classification result, including some subcategories or special tags, such as "urban roads-school areas-morning rush hour".

[0053] In one embodiment, referring to Figure 3 In step S200, according to the current driving scene, an automatic driving strategy suitable for the current driving scene is selected from a preset strategy library, which specifically includes the following steps: S210: Query a preset strategy library to obtain a candidate strategy set that matches the current driving scenario.

[0054] The preset strategy library is a database containing strategies for various driving scenarios. Each strategy is optimized for a specific type of driving scenario and contains a series of parameters and rules to guide the behavior decisions of the autonomous driving vehicle in that scenario.

[0055] S220. Score each strategy in the candidate strategy set, where the scoring criteria include security, efficiency, and adaptability.

[0056] Specifically, the system scores each candidate strategy in three dimensions: safety, efficiency, and adaptability. The safety score takes into account the safety distance setting and emergency braking rules in the strategy. For example, in congested sections, a strategy that maintains a larger safety distance may receive a higher safety score. The efficiency score takes into account the path selection and speed control in the strategy. For example, a strategy that can flexibly adjust the path to avoid congested sections may receive a higher efficiency score. The adaptability score considers the strategy's ability to respond to specific scenario characteristics. For example, a strategy that can automatically reduce speed in a school area will receive a higher adaptability score. The weights of the three dimensions may vary in different scenarios, and a comprehensive score is finally obtained.

[0057] S230: Select the strategy with the highest score as the current autonomous driving strategy.

[0058] Specifically, the system sorts the comprehensive scores of all candidate strategies obtained in step S220 and selects the strategy with the highest score as the current autonomous driving strategy. If multiple strategies have the same and highest scores, the system gives priority to the strategy with lower computational complexity to improve the system response speed. Furthermore, in order to ensure the stability of the system, the strategy switching will adopt a smooth transition method to avoid the impact of sudden behavior changes on passenger comfort and surrounding traffic.

[0059] S240. According to the characteristics of the current driving scene, using preset parameter adjustment rules, dynamically fine-tune the strategy parameters of the current autonomous driving strategy, where the strategy parameters include a safety distance, acceleration limit, and steering angle.

[0060] Among them, strategy parameters are the numerical values ​​that determine the specific characteristics of autonomous driving behavior. By adjusting these parameters, the strategy can be better adapted to the current specific situation.

[0061] Specifically, the system first analyzes the characteristics of the current driving scene, such as road width, traffic density, and weather conditions. Then, based on these characteristics and preset parameter adjustment rules, the parameters of the selected strategy are fine-tuned. For example, if the current road is slippery, the system may increase the value of the safety distance parameter and reduce the acceleration limit to make the vehicle drive more cautiously. If you are passing a sharp turn, the system will adjust the steering angle parameter to make the steering smoother. In this embodiment, the parameter adjustment rules are formulated based on expert knowledge.

[0062] In one embodiment, referring to Figure 4In step S300, according to the current scene characteristics and the real-time position of the vehicle, the map data is called and optimized in real time, which specifically includes the following steps: S310. Based on the real-time position of the vehicle, retrieve map data of a corresponding area from a map database.

[0063] Among them, the map database is a data storage system containing detailed geographic information.

[0064] Specifically, the system first obtains the precise coordinates of the vehicle, and then uses this as the center to retrieve map data within a certain range. The size of the retrieved range is dynamically adjusted according to the speed of the vehicle and the complexity of the surrounding environment. For example, when driving on a highway, a larger range of map data may be retrieved to plan the route in advance; while when driving on city streets, the retrieved range may be relatively small, but contain more detailed information.

[0065] S320: Compare the current scene features with the retrieved map data, and identify and mark dynamic changes in the map data.

[0066] Specifically, the system aligns the data collected by the on-board sensors with the map data, uses a feature point matching algorithm to compare the camera image with the street view image in the map, and detects and marks the differences between the actual environment and the map data, including newly appeared roadblocks, disappeared traffic signs, changed lane lines, etc.

[0067] S330: Based on the identified dynamic changes, dynamically evaluate the priority of map update and generate a map update strategy.

[0068] Among them, the priority of map updates reflects the impact of a certain change on safe driving and navigation accuracy.

[0069] Specifically, the system evaluates each identified change, taking into account factors such as the type, scale, duration, and impact on traffic flow. For example, a newly emerged temporary construction zone will be given a higher priority because it directly affects driving safety, while a newly opened coffee shop will be given a lower priority because it mainly affects the point of interest information.

[0070] S340: Optimize local map information in real time according to the map update strategy.

[0071] S350: Share the optimized map data with surrounding vehicles and upload it to a cloud server.

[0072] The purpose of sharing and uploading map data is to achieve collaborative perception among vehicles and crowdsourced map updates. As the central data processing and storage center, the cloud server is responsible for aggregating map update information from multiple vehicles and pushing verified updates to other vehicles.

[0073] Specifically, the system will first package the optimized map data, including updated content, timestamp, confidence level and other information. Then, the data will be broadcast to surrounding vehicles through the vehicle communication system. The receiving vehicle verifies these updates based on its own observations. If any inconsistency is found, it can be fed back to the sender or cloud server. At the same time, the system will upload the updated data to the cloud server.

[0074] In one embodiment, referring to Figure 5 In step S400, based on the current driving scene, the optimized map data and the surrounding environment information, the future movement trajectory of the surrounding dynamic objects is predicted, which specifically includes the following steps: S410: Determine a dynamic object type and a prediction time range based on a current driving scenario.

[0075] Among them, dynamic object types may include vehicles, pedestrians, cyclists, etc. The prediction time range refers to how long the system needs to predict the trajectory in the future, which will be dynamically adjusted according to the complexity and risk level of the current driving scenario.

[0076] S420: Acquire real-time status information of surrounding dynamic objects, and prioritize the real-time status information according to the current driving scenario.

[0077] The real-time status information includes the position, speed, acceleration, direction and other data of the dynamic objects. Prioritization is to ensure more accurate trajectory prediction for the most important dynamic objects when computing resources are limited.

[0078] Specifically, the system uses a variety of sensor fusion technologies to obtain real-time status information of surrounding dynamic objects. After obtaining this information, the system will prioritize the dynamic objects according to the characteristics of the current driving scene. The sorting criteria include the distance between the object and the vehicle, relative speed, and direction of movement.

[0079] S430 : Generate multiple possible trajectories of each dynamic object according to the sorted real-time status information.

[0080] Among them, multiple possible trajectories reflect the uncertainty of the future behavior of the dynamic object.

[0081] Specifically, the system generates multiple possible future trajectories for each dynamic object. For example, for vehicles, the system may consider multiple possibilities such as going straight, changing lanes, and turning; for pedestrians, it may consider continuing to go straight, stopping suddenly, changing direction, etc. The system may use the Monte Carlo simulation method to generate a large number of possible trajectories through multiple sampling, and then cluster and screen these trajectories to finally retain several of the most representative trajectories. For dynamic objects with higher priorities, the system will generate more and more detailed trajectories; for objects with lower priorities, only a few rough trajectories are generated. Each generated trajectory is assigned a probability value that reflects the likelihood of the trajectory occurring.

[0082] S440 , combining the optimized map data to generate a comprehensive trajectory prediction result reflecting the mutual influence of multiple dynamic objects.

[0083] Among them, the comprehensive trajectory prediction results take into account the interactions between dynamic objects and their interactions with the static environment.

[0084] Specifically, the system first projects multiple possible trajectories of each dynamic object onto the optimized map. Then, an interaction model is used to simulate the mutual influence between multiple dynamic objects. Using the concepts of game theory, for example, the system may consider the following behavior between vehicles, the yielding behavior when changing lanes, the avoidance behavior between pedestrians and vehicles, etc.

[0085] S450: Generate a dynamically adjusted safety envelope area for each dynamic object according to the comprehensive trajectory prediction result and the risk level of the current driving scene.

[0086] Among them, the safety envelope area refers to the spatial range that dynamic objects may occupy. Taking into account the uncertainty of the prediction and potential risks, this area is dynamically adjusted.

[0087] Specifically, the system generates an initial safety envelope for each dynamic object. The shape of this area may not be a simple geometric shape, but an irregular probability distribution area. The size and shape of the area will expand as the prediction time increases, reflecting the increase in prediction uncertainty. For scenarios with higher risk levels, the system will expand the safety envelope to reserve more buffer space for possible emergencies.

[0088] In one embodiment, referring to Figure 6 In step S410, the dynamic object types include pedestrians, non-motor vehicles, small motor vehicles, large motor vehicles, special vehicles and animals. Based on the current driving scene, the dynamic object type and the prediction time range are determined, which specifically includes the following steps: S411 , detecting and classifying dynamic objects in the surrounding environment in real time to obtain dynamic object types.

[0089] S412: Confirm the type of dynamic object based on the current driving scene characteristics.

[0090] S413: extract corresponding behavior features from a preset object behavior pattern library according to the confirmed dynamic object type.

[0091] Among them, the preset object behavior pattern library is a database containing typical behavior patterns of various dynamic objects. It contains behavioral characteristics summarized based on a large amount of historical data and expert knowledge, and provides a basic model for behavior prediction for different types of dynamic objects.

[0092] Specifically, the system first locates the corresponding entry in the preset object behavior pattern library based on the confirmed dynamic object type. The object behavior pattern library is a structured database that contains typical behavior patterns of various dynamic objects in different scenarios. For example, for pedestrians, the library may include behavior patterns such as normal walking, sudden stopping, fast running, and crossing the road; for vehicles, it may include straight driving, lane changing, turning, acceleration, deceleration, and other modes. Each behavior pattern will have a corresponding feature description, such as speed range, acceleration characteristics, motion trajectory characteristics, etc. The system will extract the most relevant behavior features based on the specific type of the current object (such as adult pedestrians, children, cyclists, etc.).

[0093] S414: Based on the complexity and risk level of the current driving scenario, a basic prediction time range is set for different types of dynamic objects.

[0094] The basic prediction time range refers to the length of time that the system preliminarily determines is required to predict the trajectory of a dynamic object.

[0095] Specifically, the system first evaluates the complexity and risk level of the current driving scenario. The complexity assessment may consider factors such as road structure (such as simple straight roads, complex intersections), traffic density, and line of sight conditions. The risk level assessment may consider factors such as road type (such as school areas, highways), weather conditions, time periods (such as nighttime, rush hour), etc. Based on these evaluation results, the system will set preliminary prediction time ranges for different types of dynamic objects.

[0096] S415: Adjust the basic prediction time range according to the motion state of the dynamic object to obtain a final prediction time range.

[0097] Specifically, the system first analyzes the current motion state of each dynamic object, including calculating the instantaneous speed and acceleration of the object, determining whether its motion direction is stable, and evaluating its relative position relationship with the surrounding environment. Based on these analysis results, the basic prediction time range is adjusted. For example, for a vehicle that is traveling at high speed and shows no signs of deceleration, the system will extend its prediction time range because the vehicle may cover a larger distance in a shorter time in this state. If it is found that the actual trajectory of an object often deviates from the predicted trajectory, the system will shorten its prediction time range, but increase the frequency of prediction updates. Through this adjustment based on real-time motion status, a more personalized and accurate prediction time range can be generated for each dynamic object, thereby improving the accuracy and reliability of the overall prediction.

[0098] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0099] On the second aspect, the present application provides an autonomous driving vehicle management system. The autonomous driving vehicle management system of the present application is described below in combination with the above-mentioned autonomous driving vehicle management method.

[0100] Reference Figure 7 , an automatic driving vehicle management system, comprising: The driving scene recognition module is used to obtain real-time data of the autonomous driving vehicle and automatically identify and classify the current driving scene based on the real-time data; A driving strategy determination module is used to select an automatic driving strategy suitable for the current driving scenario from a preset strategy library according to the current driving scenario; A map data optimization module is used to call and optimize map data in real time according to the current scene characteristics and the real-time position of the vehicle; The motion trajectory prediction module is used to predict the future motion trajectory of surrounding dynamic objects based on the current driving scene, optimized map data and surrounding environment information; The detour path planning module is used to plan a safe detour path based on the prediction results, optimized map data and current vehicle status; The control instruction generation module is used to generate vehicle control instructions based on the autonomous driving strategy and safe detour path.

[0101] In one embodiment, the present application provides an electronic device, which may be a server, and its internal structure diagram may be as follows: Figure 8As shown. The electronic device includes a processor, a memory and a network interface connected via a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the electronic device is used to store data. The network interface of the electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for managing an autonomous driving vehicle is implemented.

[0102] Those skilled in the art will understand that Figure 8 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0103] In one embodiment, an electronic device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0104] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the above-mentioned computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0105] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for managing an autonomous driving vehicle, characterized in that: The steps include: Acquire real-time data of the autonomous driving vehicle, and automatically identify and classify the current driving scene based on the real-time data; According to the current driving scenario, selecting an automatic driving strategy suitable for the current driving scenario from a preset strategy library; Call and optimize map data in real time based on current scene features and vehicle real-time location; Based on the current driving scenario, the optimized map data and the surrounding environment information, predicting the future movement trajectory of the surrounding dynamic objects; Plan a safe detour route based on the prediction results, optimized map data and current vehicle status; Generate vehicle control instructions based on the autonomous driving strategy and the safe detour path.

2. The method for managing an autonomous driving vehicle according to claim 1, characterized in that: Acquiring real-time data of the autonomous driving vehicle, and automatically identifying and classifying the current driving scene based on the real-time data, specifically comprising the following steps: Acquire real-time data of the autonomous driving vehicle and pre-process the real-time data; Input the preprocessed real-time data into the pre-trained classification model, and output the preliminary classification results and confidence of the driving scene; When the confidence level is higher than a preset threshold, taking the preliminary classification result as the current driving scene type; When the confidence level is lower than a preset threshold, feature extraction is performed on the real-time data to obtain complexity features and traffic flow features of the driving environment; A secondary judgment is performed based on the classification result, confidence, complexity characteristics and traffic flow characteristics, and the current driving scene is determined based on the secondary judgment result.

3. The method for managing an autonomous driving vehicle according to claim 1, characterized in that: According to the current driving scenario, selecting an automatic driving strategy suitable for the current driving scenario from a preset strategy library specifically includes the following steps: Query the preset strategy library to obtain a candidate strategy set that matches the current driving scenario; Score each strategy in the candidate strategy set, with the scoring criteria including safety, efficiency and adaptability; Select the strategy with the highest score as the current autonomous driving strategy; According to the characteristics of the current driving scene, the strategy parameters of the current autonomous driving strategy are dynamically fine-tuned using preset parameter adjustment rules, where the strategy parameters include a safety distance, an acceleration limit, and a steering angle.

4. The method for managing an autonomous driving vehicle according to claim 1, characterized in that: According to the current scene characteristics and the real-time position of the vehicle, the map data is called and optimized in real time, which includes the following steps: Based on the real-time location of the vehicle, map data of the corresponding area is retrieved from the map database; Compare the current scene features with the retrieved map data to identify and mark dynamic changes in the map data; Based on the identified dynamic changes, the map update priority is dynamically evaluated to generate a map update strategy; According to the map update strategy, local map information is optimized in real time; The optimized map data is shared with surrounding vehicles and uploaded to the cloud server.

5. The method for managing an autonomous driving vehicle according to claim 1, characterized in that: Based on the current driving scene, the optimized map data and the surrounding environment information, predicting the future movement trajectory of the surrounding dynamic objects specifically includes the following steps: Based on the current driving scenario, determining a dynamic object type and a predicted time range; Obtain real-time status information of surrounding dynamic objects and prioritize the real-time status information according to the current driving scenario; Generate multiple possible trajectories of each dynamic object based on the sorted real-time state information; Combined with the optimized map data, a comprehensive trajectory prediction result reflecting the mutual influence of multiple dynamic objects is generated; Based on the comprehensive trajectory prediction results and the risk level of the current driving scenario, a dynamically adjusted safety envelope area is generated for each dynamic object.

6. The method for managing an autonomous driving vehicle according to claim 5, characterized in that: The dynamic object types include pedestrians, non-motor vehicles, small motor vehicles, large motor vehicles, special vehicles and animals. Based on the current driving scene, determining the dynamic object type and the prediction time range specifically includes the following steps: Detect and classify dynamic objects in the surrounding environment in real time to obtain the type of dynamic objects; Based on the current driving scene characteristics, the dynamic object type is confirmed; According to the confirmed dynamic object type, extract corresponding behavior features from the preset object behavior pattern library; Set basic prediction time ranges for different types of dynamic objects based on the complexity and risk level of the current driving scenario; According to the motion state of the dynamic object, the basic prediction time range is adjusted to obtain a final prediction time range.

7. An automatic driving vehicle management system, characterized in that: include: A driving scene recognition module, used to obtain real-time data of the autonomous driving vehicle, and automatically recognize and classify the current driving scene based on the real-time data; A driving strategy determination module, configured to select an automatic driving strategy suitable for the current driving scenario from a preset strategy library according to the current driving scenario; A map data optimization module is used to call and optimize map data in real time according to the current scene characteristics and the real-time position of the vehicle; An action trajectory prediction module, used to predict the future action trajectory of surrounding dynamic objects based on the current driving scene, the optimized map data and the surrounding environment information; The detour path planning module is used to plan a safe detour path based on the prediction results, optimized map data and current vehicle status; A control instruction generation module is used to generate vehicle control instructions based on the automatic driving strategy and the safe detour path.

8. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the autonomous driving vehicle management method described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the autonomous driving vehicle management method described in any one of claims 1-6 are implemented.

Images