An aerial vehicle based vehicle path planning method and data processing system

Abstract

Embodiments of the present application provide a kind of vehicle path planning method and data processing system based on aircraft, method includes: obtaining the planning path of target vehicle to destination;When determining that target sub-path occurs congestion, extract spatial feature, dynamic feature, environmental feature and time feature from the road environment data of target sub-path;At least part of road environment data is obtained by aircraft sensing;Characteristic set is input into congestion diffusion model, and congestion diffusion prediction result is obtained;Congestion diffusion model includes first congestion diffusion model, is used to determine free flow speed and traffic capacity based on characteristic set, and the congestion diffusion prediction result of target sub-path is predicted based on free flow speed and traffic capacity;If congestion diffusion prediction result indicates congestion mitigation, determine that target vehicle keeps planning path and travels;If congestion diffusion prediction result indicates congestion maintenance, re-planning the planning path of target vehicle to destination and not containing target sub-path.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and more specifically, to a vehicle path planning method and data processing system based on an aircraft. Background Technology

[0002] Traditional vehicle navigation systems often rely on image data collected by onboard GPS and cameras for navigation, which suffers from data latency and limited coverage, making it impossible to respond to sudden traffic congestion events in real time. Therefore, there is an urgent need for a solution that can detect congestion events in real time while the vehicle is in motion, in order to help the vehicle avoid congestion. Summary of the Invention

[0003] The purpose of this application is to provide a vehicle path planning method and data processing system based on an aircraft, so as to achieve the technical effect of real-time detection of congestion events during vehicle operation, thereby assisting vehicles in avoiding congestion.

[0004] A first aspect of this application provides a vehicle path planning method based on an aircraft, the method comprising:

[0005] Obtain the planned route from the target vehicle to the destination, wherein the planned route includes multiple sub-paths;

[0006] If it is determined that a target sub-path in the multi-segment sub-path is congested, one or more of the spatial features, dynamic features, environmental features, and temporal features of the target sub-path are extracted from the road environment data corresponding to the target sub-path to obtain a feature set; wherein, at least part of the road environment data is obtained by sensing the target sub-path by an aircraft.

[0007] The feature set is input into the congestion diffusion model to obtain the congestion diffusion prediction result of the target sub-path; wherein, the congestion diffusion model includes a first congestion diffusion model, which is used to determine the free flow speed and capacity of the target sub-path based on the feature set, and predict the congestion diffusion prediction result of the target sub-path based on the free flow speed and the capacity.

[0008] If the congestion spread prediction result indicates that the congestion on the target sub-path is alleviated, then the target vehicle is determined to continue traveling along the planned path.

[0009] If the congestion spread prediction result indicates that the target sub-path remains congested, a new planned route from the target vehicle to the destination that does not include the target sub-path is replanned.

[0010] In the aforementioned implementation process, the aircraft arrives ahead of the target sub-path in advance to sense the road environment data of the target sub-path. If congestion is detected on the target sub-path, the feature set extracted from the road environment data is used to predict the congestion spread trend. Due to the aircraft's excellent scheduling flexibility, even if there is sudden congestion on the target sub-path, the congestion spread trend can be detected and predicted in a timely manner. Therefore, if the predicted congestion fails to be alleviated, the route can be replanned in a timely manner to prevent vehicles from entering congested sections. This achieves the technical effect of real-time detection of congestion events during vehicle travel, assisting vehicles in avoiding congestion.

[0011] Furthermore, the method also includes:

[0012] Based on the current sub-path where the target vehicle is located, the next sub-path that the target vehicle will reach from the multiple sub-paths connected to the current sub-path is determined as the target sub-path.

[0013] In the above implementation process, by setting the next sub-path to which the target vehicle is connected to the current sub-path of the target vehicle as the target sub-path, the control aircraft senses the target sub-path before the target vehicle arrives at the target sub-path, determines whether the target sub-path is congested and predicts the congestion spread trend of the target sub-path, so that congestion events on the road ahead can be detected in advance and vehicles can be prevented from entering congested road sections.

[0014] Furthermore, the method also includes:

[0015] If it is determined that the target sub-path is unobstructed, the target vehicle will continue to travel along the planned path.

[0016] In the above implementation process, the control aircraft senses the target sub-path before the target vehicle arrives to determine whether the sub-path is congested. The target vehicle is then allowed to enter the target sub-path only if the sub-path is clear, thus avoiding entering congested sections.

[0017] Further, the replanning of the planned route from the target vehicle to the destination, excluding the target sub-path, includes:

[0018] Based on road topology, candidate sub-paths other than the target sub-path that remains congested are identified that are connected to the current sub-path where the target vehicle is located.

[0019] Based on the road environment data of each candidate sub-path, a passable sub-path with normal road conditions is determined from all candidate sub-paths; wherein, normal road conditions include smooth traffic or reduced congestion.

[0020] The plan is a planned path that starts from the current sub-path, ends at the destination, and passes through the accessible sub-paths.

[0021] In the above implementation process, the highly flexible aircraft is used to pre-select passable sub-paths with normal road conditions from multiple candidate sub-paths, and then the path planning is re-performed, which improves the efficiency of path planning and avoids passing through another severely congested sub-path again when replanning the path.

[0022] Further, the road environment data includes prior road data and sensing data obtained by the aircraft sensing the target sub-path; the spatial features include one or more of vehicle density, accident information, and lane structure, wherein the accident information includes accident type and / or accident location; the spatial features are obtained through the following steps:

[0023] The traffic images from the sensed data are input into a trained vehicle recognition model to obtain the output vehicle recognition results. Based on the number of vehicles and the road size of the target sub-path, the vehicle density is determined.

[0024] Accident features are extracted from the traffic images to characterize the accident scene, and the traffic images carrying the accident features are input into a trained accident classification model to obtain the output accident information; and / or the accident information is extracted from the accident reporting data included in the road prior data;

[0025] The traffic image is input into a trained lane recognition model to obtain the output lane structure recognition result; and / or the lane structure is extracted from the road information included in the road prior data.

[0026] In the above implementation process, one or more of the following are extracted from the target sub-path: vehicle density, accident information and lane structure, using sensing data and / or road prior data, so as to obtain the spatial characteristics of the target sub-path, which allows the impact on congestion spread to be considered from the spatial location dimension.

[0027] Furthermore, the road environment data includes prior road data and sensing data obtained by the aircraft sensing the target sub-path; the dynamic features include the driving direction and / or speed distribution of the vehicles; the dynamic features are obtained through the following steps:

[0028] Based on the traffic image sequence in the sensed data, determine the driving direction of the vehicle; and / or extract the driving direction from the road information included in the road prior data;

[0029] The vehicle speed distribution is determined based on the traffic image sequence.

[0030] In the above implementation process, the driving direction and / or speed distribution of the target sub-path are extracted using sensing data and / or road prior data, thereby obtaining the dynamic characteristics of the target sub-path, so that the impact on congestion spread can be considered from the dimension of dynamic characteristics.

[0031] Furthermore, the road environment data includes prior road data and sensing data obtained by the aircraft sensing the target sub-path; the environmental features include meteorological features and / or visibility features; the environmental features are obtained through the following steps:

[0032] Input the traffic images from the sensed data into a trained meteorological classification model to obtain the output meteorological features and / or visibility features; or

[0033] Extract the meteorological features and / or visibility features from the meteorological sensing data in the sensing data; or

[0034] The meteorological features and / or visibility features are extracted from the meteorological information included in the prior road data.

[0035] In the above implementation process, meteorological and / or visibility features of the target sub-path are extracted using sensing data and / or road prior data, thereby obtaining the environmental features of the target sub-path, so that the impact on congestion spread can be considered from an environmental perspective.

[0036] Furthermore, the time characteristics include the current time period and historical congestion patterns; the historical congestion patterns are used to indicate the statistical congestion patterns of the target sub-path in different historical time periods.

[0037] In the above implementation process, based on the current time period and historical congestion patterns, it can be determined whether the congestion occurring in the current time period is consistent with the statistical congestion pattern of a certain historical time period. Thus, the statistical congestion pattern can be referenced to predict the spread of congestion.

[0038] Furthermore, the congestion diffusion model includes a second congestion diffusion model, which comprises a spatiotemporal graph neural network module, a time series module, and a reinforcement learning module; the time series module includes a Transformer sub-model and a temporal convolution sub-model; the step of inputting the feature set into the congestion diffusion model to obtain the output congestion diffusion prediction result of the target sub-path includes:

[0039] The feature set is input into the spatiotemporal graph neural network module, the time series module, and the reinforcement learning module, respectively, to obtain the congestion propagation simulation results output by the spatiotemporal graph neural network module, the traffic flow change prediction results output by the time series module, and the vehicle diversion impact prediction results output by the reinforcement learning module.

[0040] The congestion propagation simulation results, the traffic flow change prediction results, and the prediction results of the impact of vehicle diversion on congestion are used to generate the congestion diffusion prediction results; wherein, the congestion diffusion prediction results include one or more of the following: a predicted congestion heat map for characterizing the degree of congestion at different locations, the direction of congestion diffusion, and the predicted congestion mitigation time.

[0041] In the above implementation process, the input feature combinations are processed by a spatiotemporal graph neural network module, a time series module, and a reinforcement learning module, respectively, to obtain congestion propagation simulation results, traffic flow change prediction results, and prediction results of the impact of vehicle diversion on congestion. Based on these three prediction results, a congestion diffusion prediction result is generated, enabling timely rerouting of routes to avoid vehicles entering congested sections if the predicted congestion fails to ease. This achieves the technical effect of real-time detection of congestion events during vehicle travel, assisting vehicles in avoiding congestion.

[0042] A second aspect of this application provides a data processing system, the system comprising: a vehicle, an aircraft, and a control platform;

[0043] The control platform includes: a processor; and a memory for storing processor-executable instructions; wherein, when the processor invokes the executable instructions, it implements the operation of any of the methods described in the first aspect. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating a vehicle path planning method based on an aircraft, provided for an embodiment of this application;

[0046] Figure 2 A flowchart illustrating another vehicle path planning method based on an aircraft provided in this application embodiment;

[0047] Figure 3 This is a hardware structure diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0049] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0050] Traditional vehicle navigation systems rely on image data collected by in-vehicle GPS and cameras for navigation. However, in the face of sudden traffic congestion, relying solely on GPS and cameras is often insufficient to detect it promptly. The congestion detection process is as follows: Navigation service providers analyze the GPS position change speed of traffic flow on a specific road segment to determine if slow-moving traffic is occurring. If so, they send accident inquiries to multiple vehicles on that segment, asking drivers if an accident has occurred and what type of accident it was. Based on these manually reported accident information, the navigation service provider confirms the accident has occurred and then marks the segment as an accident on the map to warn other vehicles to avoid that segment or plan alternative routes. Clearly, relying on in-vehicle GPS for congestion detection suffers from data latency. Furthermore, in-vehicle cameras have limited data collection range, making it impossible to detect congestion occurring at a certain distance from vehicles. Therefore, current vehicle GPS and vehicle cameras are unable to detect congestion in a timely manner, let alone deal with sudden congestion events, causing vehicles to inevitably enter congested sections of road, exacerbating congestion and delaying travel.

[0051] The ability to detect traffic congestion in real time during vehicle operation to assist vehicles in avoiding congestion is of great significance. This is especially true in emergency scenarios such as medical and fire emergencies, where emergency vehicles like ambulances, fire trucks, and roadside assistance vehicles need to reach the scene promptly. Delays caused by traffic congestion can lead to irreparable losses. Therefore, a solution specifically designed for emergency vehicles is needed to detect traffic congestion in real time during their journey, thus assisting them in avoiding congestion.

[0052] Therefore, this application provides a vehicle path planning method based on aircraft, which can be used for path planning of the aforementioned emergency vehicles. Of course, the application scenarios of this solution are not limited to emergency vehicles, but are also applicable to path planning of general social vehicles. The aircraft refers to a device with flight capabilities, including but not limited to unmanned aerial vehicles (UAVs), unmanned aerial vehicles, manned aircraft, flying cars, civil aircraft, and other types of air transportation. The aircraft is equipped with sensors for collecting sensing data. These sensors include various types, such as but not limited to image sensors, lidar, millimeter-wave radar, and meteorological sensors. The image sensors include multispectral cameras, such as visible light cameras and infrared cameras. The meteorological sensors include humidity sensors and visibility sensors, etc. Furthermore, the aircraft may also be equipped with a communication module for communicating with the vehicle or control platform. This communication module includes, but is not limited to, 5G communication modules, image transmission protocol communication modules, and IoT communication modules, etc.

[0053] As an example, the vehicle could be an onboard hangar for the aircraft. That is, the aircraft can be stored in the vehicle, and when stored, the aircraft is mechanically connected to the vehicle to secure and protect it. When the aircraft needs to take off, the mechanical connection between the vehicle and the aircraft is released, allowing the aircraft to take off from the vehicle.

[0054] As another example, hangars and garages can be deployed adjacent to each other. Taking emergency vehicles as an example, a fixed hangar with an aircraft can be deployed near the emergency vehicle's garage. When the vehicle drives out of the garage to its destination, the aircraft can take off from the adjacent fixed hangar to achieve collaborative operation.

[0055] Based on this, the vehicle path planning method based on an aircraft provided in this application can be executed by a control platform. The control platform can be a platform for controlling vehicles, a platform for controlling aircraft, or a platform integrating vehicle control and aircraft control functions. The control platform can establish communication links with both the vehicle and the aircraft separately, thereby communicating with each separately. Alternatively, the control platform can establish a communication link with the vehicle, and the vehicle can establish a communication link with the aircraft, allowing the control platform to communicate directly with the vehicle and use the vehicle as a communication intermediary to communicate with the aircraft. Or, the control platform can establish a communication link with the aircraft, and the vehicle can establish a communication link with the aircraft, allowing the control platform to communicate directly with the aircraft and use the aircraft as a communication intermediary to communicate with the vehicle. This application does not limit the communication methods between the control platform, the aircraft, and the vehicle. Figure 1 As shown, the method includes steps 110-150.

[0056] Step 110: Obtain the planned route from the target vehicle to the destination; wherein the planned route includes multiple sub-paths.

[0057] The target vehicle refers to any vehicle that requires real-time route planning and adjustment. Besides the emergency vehicles mentioned above, it can also be any other vehicle.

[0058] The planned route refers to the planned path for the target vehicle from the starting point to the destination. The control platform can interface with the navigation service provider's API (Application Programming Interface). As an example, the control platform can obtain a map from the starting point to the destination from the navigation service provider, and then plan the target vehicle's path from the starting point to the destination based on the road network in the map. Optionally, it can also plan the path from the starting point to the destination based on filtering conditions input by the target vehicle (including shortest route, fewest traffic lights, avoid highways, take highways, etc.). As yet another example, the navigation service provider can directly plan the path from the starting point to the destination for the target vehicle. Optionally, the navigation service provider can plan the path from the starting point to the destination based on filtering conditions input by the target vehicle. Subsequently, the control platform can directly obtain the planned path from the navigation server.

[0059] The planned path can be divided into multiple sub-paths. For example, key nodes can be identified from the planned path. The road segment between two adjacent key nodes is determined as a sub-path according to the travel direction of the planned path. The key nodes include the starting point, destination, and intersection on the planned path. Thus, if N key nodes are identified on the planned path, the planned path is divided into N-1 sub-paths.

[0060] Optionally, after dividing the sub-paths based on key nodes, it can be further checked whether the length of each sub-path segment is greater than the communication distance between the vehicle and the aircraft. The communication distance refers to the distance at which the vehicle and the aircraft can communicate directly without relying on third-party communication equipment. If the length of a sub-path exceeds the communication distance, the sub-path is split such that the length of each resulting sub-path segment is less than the communication distance. That is, the planned path can be divided into multiple sub-paths based on key nodes and the communication distance between the vehicle and the aircraft, and the length of each sub-path segment is less than or equal to the communication distance.

[0061] Step 120: If it is determined that the target sub-path in the multi-segment sub-path is congested, extract one or more of the spatial features, dynamic features, environmental features and temporal features of the target sub-path from the road environment data corresponding to the target sub-path to obtain a feature set; wherein, at least part of the road environment data is obtained by the aircraft sensing the target sub-path.

[0062] The target sub-path can be any sub-path that the target vehicle will travel to in the future, excluding the sub-path the target vehicle is currently on. Regarding whether the target sub-path is congested, as an example, an aircraft can be controlled to acquire traffic sequence images of the target sub-path, and the traffic image sequence can be input into a trained congestion recognition model to obtain the model's output congestion recognition result. The congestion recognition model can be a binary classification model. The output congestion recognition result includes whether congestion has occurred or not. Since vehicle speed, vehicle density, and traffic flow over a continuous period can be extracted from the traffic image sequence, the congestion recognition model can be used to identify whether the target sub-path is currently congested from the traffic sequence.

[0063] As another example, it's possible to determine whether congestion has occurred on the target sub-path from the acquired prior data. For instance, vehicles currently traveling on the target sub-path can report congestion events to the navigation service provider. The management platform can then obtain the first prior data, including the congestion events, from the navigation service provider to determine whether congestion has occurred on the target sub-path.

[0064] When congestion is determined to occur on the target sub-path, one or more of the following features—spatial features, dynamic features, environmental features, and temporal features—are extracted from the road environment data corresponding to the target sub-path to obtain a feature set. At least a portion of the road environment data is obtained through sensing of the target sub-path by an aircraft.

[0065] Specifically, the aircraft can be controlled to collect first sensing data of the target sub-path and determine the road environment data of the target sub-path from the first sensing data. Before controlling the aircraft to collect the first sensing data, the waypoints corresponding to the aircraft in each sub-path segment can be determined first. For example, for each sub-path segment, the waypoints corresponding to the aircraft in that sub-path can be determined based on the location of the sub-path and the location of high-altitude obstacles in the sub-path, so that the waypoints corresponding to each sub-path segment avoid high-altitude obstacles and are located above each sub-path segment. The high-altitude obstacles include, but are not limited to, buildings, bridges, and high-altitude lines. After obtaining the waypoints corresponding to each sub-path segment, multiple waypoints can be connected sequentially, and the flight path of the aircraft can be generated based on safety planning conditions. The safety planning conditions include, but are not limited to, flight obstacle avoidance conditions and no-fly zone conditions. Based on this, the aircraft can be controlled to fly to the waypoints corresponding to the target sub-path according to the flight path, and collect the first sensing data of the target sub-path at the waypoints. As can be seen, since the target sub-path is the sub-path that the target vehicle will travel to in the future, the control aircraft collects the first sensing data of the target sub-path. That is, the control aircraft arrives at the sub-path that the target vehicle will travel to in advance to collect data, and then determines the road environment data of the target sub-path and analyzes the road conditions of the target sub-path.

[0066] The first sensing data refers to the sensing data collected by the aircraft's onboard sensors along the target sub-path. For example, if the aircraft is equipped with an image sensor, the first sensing data includes traffic images and / or traffic videos collected by the image sensor along the target sub-path. Multiple traffic images may be included. If the traffic images are arranged chronologically, a traffic image sequence can be obtained. Alternatively, the traffic video may include multiple frames of traffic images, and consecutive frames of traffic images in the video can form a traffic image sequence. The frame rate of the traffic video can be adjusted according to computing resources, such as 1-5 frames per second. The traffic images can be pre-corrected, including eliminating camera distortion and calibrating the viewing angle, to accurately estimate distance and area. As another example, if the aircraft is equipped with a lidar, the first sensing data includes point cloud data collected by the lidar along the target sub-path. As yet another example, if the aircraft is equipped with a weather sensor, the first sensing data includes weather data collected by the weather sensor along the target sub-path, such as humidity and visibility.

[0067] Road environment data can be obtained not only through aircraft sensing the target path, but also from prior information. Prior information refers to information about the target sub-path obtained by the control platform from other platforms and channels. Prior information includes one or more types, such as, but not limited to, weather data, map data, and accident reporting data. Since prior information is related to the road environment data of the target sub-path, road environment data can be determined from it.

[0068] Subsequently, one or more of the spatial features, dynamic features, environmental features, and temporal features of the target sub-path can be extracted from the road environment data to obtain a feature set.

[0069] The spatial features characterize the spatial location of the target sub-path. The dynamic features indicate the driving characteristics of vehicles traveling on the target sub-path. The environmental features indicate the environment of the target sub-path. The temporal features indicate the temporal changes in road conditions along the target sub-path.

[0070] Step 130: Input the feature set into the congestion diffusion model to obtain the output congestion diffusion prediction result of the target sub-path.

[0071] Understandably, spatial features characterize the spatial location of the target sub-path, and these spatial features affect the congestion dissipation speed. For example, if the target sub-path is narrow and has few lanes, the congestion dissipation speed will be slower. Therefore, the spatial features of the target sub-path are one of the influencing factors of congestion diffusion prediction results.

[0072] Furthermore, dynamic characteristics indicate the driving patterns of vehicles on the target sub-path. For example, the congestion dispersal speed differs significantly between vehicles moving slowly on the target sub-path and vehicles standing still on the target sub-path. Therefore, the dynamic characteristics of the target sub-path are also one of the influencing features of the congestion diffusion prediction results.

[0073] Furthermore, environmental characteristics indicate the environment in which the target sub-path is located, such as visibility and weather conditions. For example, in rainy or foggy weather, low visibility leads to slower vehicle speeds. Therefore, compared to sunny weather, congestion dispersal is slower in rainy or foggy weather. Thus, the environmental characteristics of the target sub-path are also one of the influencing factors of congestion diffusion prediction results.

[0074] Furthermore, temporal characteristics indicate the temporal variation of traffic conditions along the target sub-path. For example, the target sub-path may experience congestion during daily rush hours, with prolonged congestion periods. Introducing temporal characteristics allows us to consider the impact on congestion spread prediction results from a temporal perspective. Therefore, the temporal characteristics of the target sub-path are also one of the influencing features of congestion spread prediction results.

[0075] Thus, by inputting one or more of the spatial, dynamic, environmental, and temporal characteristics of the target sub-path into the congestion diffusion model, congestion diffusion can be predicted based on these multiple characteristics.

[0076] A congestion diffusion model is used to predict the congestion diffusion trend of a target sub-path based on a feature set, and outputs the congestion diffusion prediction result. The congestion diffusion model can include an algorithmic model or a trained deep learning-based neural network model. For example, a congestion diffusion model includes a first congestion diffusion model, which is an algorithmic model used to determine the free-flow speed and capacity of the target sub-path based on the feature set, and predict the congestion diffusion result of the target sub-path based on the free-flow speed and capacity. Here, the free-flow speed of the target sub-path characterizes the vehicle speed of the target sub-path when there is no congestion; the capacity characterizes the number of vehicles that can pass through the target sub-path.

[0077] For example, the congestion diffusion model includes a second congestion diffusion model, which is a trained neural network model. The second congestion diffusion model learns the relationship between historical feature sets and congestion diffusion trends based on training data, thus outputting congestion diffusion prediction results for the target sub-path based on the input feature set.

[0078] For example, the congestion diffusion model includes a first congestion diffusion model and a second congestion diffusion model. Thus, the first congestion diffusion prediction result output by the first congestion diffusion model and the second congestion diffusion prediction result output by the second congestion diffusion model can be weighted to obtain a weighted congestion diffusion prediction result, which is the congestion diffusion prediction result described in step 130. The specific prediction process of the first and second congestion diffusion models will be explained below. Based on the congestion diffusion prediction result, step 140 or step 150 is selected to be executed.

[0079] Step 140: If the congestion diffusion prediction result indicates that the congestion on the target sub-path is alleviated, determine that the target vehicle continues to travel along the planned path.

[0080] If the congestion propagation prediction indicates that congestion on the target sub-path will be alleviated, it means that the congestion on the target sub-path is not severe, and therefore the target vehicle can continue to travel along its original planned route. Furthermore, after completing the congestion propagation prediction for the current target sub-path, the next target sub-path can be determined, and the process returns to step 120 until the target vehicle reaches its destination.

[0081] As an example, the current target sub-path can be designated as the next target sub-path. That is, after predicting congestion spread for the current target sub-path, congestion assessment and spread prediction continue to be performed on the current target sub-path to update the congestion spread prediction results. It is understandable that, because road conditions are dynamically changing, even if the current road environment data predicts that congestion on the current target sub-path will soon ease, congestion may occur again in the future due to other reasons. Therefore, periodic congestion monitoring of the target sub-path can be performed, including periodically determining whether congestion has occurred, and, if congestion is confirmed, periodically estimating the congestion spread trend, i.e., periodically updating the congestion spread prediction results. For example, if the aircraft's remaining battery power is greater than a preset battery threshold, the aircraft can hover at the waypoints corresponding to the target sub-path to periodically sense the target sub-path, obtain road environment data, extract a feature set from the road environment data, and predict the congestion spread trend based on the feature set, thereby achieving periodic congestion monitoring of the target sub-path. If the remaining battery power of the aircraft is less than the preset battery power threshold, the aircraft can return to the hangar for battery swapping or charging, and then continue to sense the target sub-path to obtain road environment data.

[0082] As another example, the next sub-path of the current target sub-path can be determined according to the direction of travel in the planned path. In this way, the aircraft can be controlled to fly along the flight path to the waypoint corresponding to the next target sub-path, and at the waypoint, the next target sub-path can be sensed to obtain road environment data.

[0083] Step 150: If the congestion spread prediction result indicates that the target sub-path is still congested, replan the planned route from the target vehicle to the destination that does not include the target sub-path.

[0084] If the congestion propagation prediction indicates that congestion on the target sub-path will persist, it means that congestion has not been alleviated or has even worsened. In this case, the planned route from the target vehicle to its destination can be replanned, and the replanned route will not include the target sub-path where congestion persists. That is, the planned route from the target vehicle's current location to its destination is replanned, avoiding the target sub-path.

[0085] As an example, if the control platform obtains a map from the starting point to the destination from the navigation service provider, it can replan a route that does not include the target sub-path based on the target vehicle's current location and the destination. As another example, if the control platform does not obtain the map, it can instruct the navigation service provider to replan a route that does not include the target sub-path based on the target vehicle's current location and the destination. The control platform then obtains the replanned route from the navigation service provider. After obtaining the replanned route, it can return to step 110 until the target vehicle reaches its destination.

[0086] Optionally, the congestion spread prediction result may also include the congestion mitigation prediction time. Based on this, the congestion spread prediction result indicating congestion mitigation in step 140 specifically means that the congestion spread prediction result indicates that when the target vehicle arrives at the target sub-path, the congestion is expected to have eased. The congestion spread prediction result indicating congestion maintenance in step 150 specifically means that when the congestion spread prediction result indicates that when the target vehicle arrives at the target sub-path, the congestion is expected to remain (including congestion not easing or congestion worsening). Specifically, the estimated travel time required for the target vehicle to travel from its current location to the target sub-path can be determined, and based on the congestion mitigation prediction time and the estimated travel time, it can be predicted whether the congestion has eased when the target vehicle arrives at the target sub-path.

[0087] As can be seen, the vehicle routing method based on an aircraft provided in this application obtains road environment data of the target sub-path by sensing it in advance. When congestion is detected on the target sub-path, the method uses a feature set extracted from the road environment data to predict the congestion spread trend. Due to the excellent scheduling flexibility of aircraft, even if there is sudden congestion on the target sub-path, the congestion spread trend can be detected and predicted in a timely manner. Therefore, if the predicted congestion fails to be alleviated, the route can be replanned in a timely manner to avoid vehicles entering congested sections. This achieves the technical effect of real-time detection of congestion events during vehicle travel, assisting vehicles in avoiding congestion.

[0088] The following sections will elaborate on steps 110-150.

[0089] According to some embodiments of this application, the target sub-path can be determined through the following steps, that is, before executing step 120, the following steps can be performed first:

[0090] Based on the current sub-path where the target vehicle is located, the next sub-path that the target vehicle will reach from the multiple sub-paths connected to the current sub-path is determined as the target sub-path.

[0091] For example, if the planned path is divided into sub-path a-sub-path b-sub-path c, when the target vehicle is currently on sub-path a, the next sub-path b that the target vehicle connected to the current sub-path a can be determined as the target sub-path; when the target vehicle travels to sub-path b, the sub-path c that the target vehicle connected to the current sub-path b can be determined as the target sub-path.

[0092] Furthermore, as mentioned above, the aircraft can hover at waypoints corresponding to the same target sub-path to periodically sense that sub-path, thereby periodically assessing congestion and predicting congestion spread trends. Specifically, when a target vehicle is traveling on the current sub-path (e.g., path a), and the aircraft first arrives at the target sub-path (e.g., path b) to sense and obtain road environment data, and if the target sub-path (e.g., path b) is not congested or the congestion spread prediction indicates that congestion has eased, if the target vehicle has not yet arrived at the target sub-path (e.g., path b), the aircraft continues to hover at the current waypoint to periodically sense the target sub-path (e.g., path b) and periodically assess congestion and predict congestion spread trends based on the obtained road environment data. Only after the target vehicle arrives at the current target sub-path (e.g., path b) does the aircraft fly to the next target sub-path (e.g., path c) to collect data and perform congestion assessment and congestion spread trend prediction.

[0093] As can be seen, this embodiment sets the next sub-path that the target vehicle will reach, which is connected to the current sub-path of the target vehicle, as the target sub-path. The control aircraft senses the target sub-path before the target vehicle arrives at the target sub-path, determines whether the target sub-path is congested, and predicts the congestion spread trend of the target sub-path. This allows for the early detection of congestion events on the road ahead, preventing vehicles from entering congested sections.

[0094] According to some embodiments of this application, based on any of the above embodiments, the method further includes the following steps:

[0095] If it is determined that the target sub-path is unobstructed, the target vehicle will continue to travel along the planned path.

[0096] As described in the above embodiments, it can be determined whether the target sub-path is congested using a congestion identification model and / or first prior data obtained from the navigation service provider. If it is determined that the target sub-path is not congested, i.e., traffic is smooth, there is no need to replan the route, and the target vehicle can continue to travel along the original planned route. Optionally, the next target sub-path can be determined and the process can return to step 120 until the target vehicle reaches its destination.

[0097] In this embodiment, the control aircraft senses the target sub-path before the target vehicle arrives to determine if congestion has occurred. The target vehicle is then allowed to enter the target sub-path only if smooth traffic flow is ensured, thus preventing the vehicle from entering congested sections.

[0098] According to some embodiments of this application, based on any of the above embodiments, the replanning of the planned route from the target vehicle to the destination in step 150, excluding the target sub-path, specifically includes, as follows: Figure 2 Steps 151-153 are shown.

[0099] Step 151: Based on the road topology, determine the candidate sub-paths connected to the current sub-path where the target vehicle is located, excluding the target sub-path that is maintaining congestion.

[0100] Road topology refers to the connection between different roads in a transportation network. From the road topology, all sub-paths connected to the current sub-path of the target vehicle can be found. Subsequently, from all sub-paths, the remaining sub-paths other than the target sub-path that maintains congestion can be identified as candidate sub-paths, which include one or more segments.

[0101] Step 152: Based on the road environment data of each candidate sub-path, determine the passable sub-path with normal road conditions from all candidate sub-paths; wherein, normal road conditions include smooth traffic or reduced congestion.

[0102] Similarly, at least a portion of the road environment data for each candidate sub-path is obtained through sensing of each candidate sub-path by the aircraft. Specifically, for each candidate sub-path, the aircraft can be controlled to collect second sensing data for the candidate sub-path, resulting in second sensing data corresponding to each sub-path. The specific implementation of the second sensing data can be found in the description of the first sensing data above, and will not be repeated here. Subsequently, the road environment data for each candidate sub-path is determined from the second sensing data of each candidate sub-path.

[0103] Similarly, in addition to obtaining road environment data for candidate sub-paths through aircraft sensing, it can also be acquired from the second prior data. The specific implementation of the second prior data can be found in the description of the first prior data above, and will not be repeated here.

[0104] For each candidate sub-path, the first step is to determine whether the sub-path is unobstructed, i.e., whether it is congested. The specific process for determining congestion is described above for the congestion determination process of the target sub-path, and will not be repeated here. If the candidate sub-path is determined to be unobstructed, i.e., not congested, it can be identified as a passable sub-path with normal road conditions. If the candidate sub-path is determined to be congested, one or more of the spatial, dynamic, environmental, and temporal features of the candidate sub-path are further extracted from the road environment data corresponding to the candidate sub-path to obtain a feature set. This feature set is then input into the congestion diffusion model to obtain the congestion diffusion prediction result for the candidate sub-path. If the congestion diffusion prediction result indicates that the congestion on the candidate sub-path has eased, it can be identified as a passable sub-path with normal road conditions. If the congestion diffusion prediction result indicates that the congestion on the candidate sub-path remains, it is not a passable sub-path with normal road conditions.

[0105] Step 153: Plan a route that starts from the current sub-path, ends at the destination, and passes through the passable sub-path.

[0106] If there are multiple candidate sub-paths, a planned route can be designed that starts from the current sub-path, ends at the destination, and passes through at least one accessible sub-path. In this way, the replanned route reaches the destination via accessible sub-paths, effectively avoiding congested sections of road.

[0107] As can be seen in this embodiment, by using a highly flexible aircraft to pre-select passable sub-paths with normal road conditions from multiple candidate sub-paths, and then re-planning the path, the efficiency of path planning is improved, and the process of re-planning the path is avoided by passing through another severely congested sub-path again.

[0108] Of course, besides such Figure 2 In addition to the illustrated embodiment, a candidate sub-path can be selected from multiple candidate sub-paths for path planning, and then the aircraft can be controlled to collect sensing data and monitor the road conditions of that candidate sub-path. If the candidate sub-path becomes congested or the congestion persists, another candidate sub-path can be selected and the next planned path can be replanned.

[0109] According to some embodiments of this application, based on any of the above embodiments, the road environment data of the target sub-path includes road prior data and sensing data obtained by the aircraft sensing the target sub-path. The road prior data is the first prior data mentioned above, and the sensing data is the first sensing data mentioned above.

[0110] Spatial features extracted from road environment data include one or more of vehicle density, accident information, and lane structure. Accident information includes accident type and / or accident location. Therefore, step 120 extracts spatial features from the road environment data, specifically including steps S11, S121, S122, S131, and S132. These five steps do not have a specific execution order and can be performed simultaneously. Depending on the specific types of features included in the spatial features, one or more of these five steps can be selected for execution.

[0111] Step S11 is used to detect the vehicle density in the target sub-path.

[0112] Step S11: Input the traffic image from the sensing data into the trained vehicle recognition model to obtain the output vehicle recognition result, and determine the vehicle density based on the number of vehicles and the road size of the target sub-path.

[0113] For example, the vehicle recognition model is used to identify vehicles traveling on a target sub-path from a traffic image; therefore, the vehicle recognition model is a visual recognition model. Optionally, the vehicle recognition model can be a target detection model such as the YOLO model (e.g., the YOLOv8 model) or the EfficientDet model. Inputting the traffic image of the target sub-path into the vehicle recognition model yields the output vehicle recognition results. These results include, but are not limited to, the positions of the vehicles in the traffic image and the number of vehicles. The positions of the vehicles can be marked using bounding boxes.

[0114] Subsequently, the vehicle density in the target sub-path can be determined based on the number of vehicles and the road size of the target sub-path. As an example, the number of vehicles can be represented by the first coverage area of ​​the detection boxes of all vehicles in the traffic image, and the road size of the target sub-path can be the second coverage area of ​​the target sub-path in the traffic image. Therefore, the vehicle density can be the ratio of the first coverage area to the second coverage area. As yet another example, the road size of the target sub-path can be the length of the target sub-path in the traffic image. Therefore, the vehicle density can be the number of vehicles per unit length of the target sub-path, that is, the ratio of the number of vehicles to the length of the target sub-path in the traffic image.

[0115] Optionally, to improve the accuracy of vehicle density detection, depth estimation techniques can be combined to enhance the spatial accuracy of traffic images. Specifically, the traffic image can first be input into a monocular depth estimation model, and feature fusion and detail enhancement can be used to improve the spatial accuracy of the traffic image, making the outlines and details of objects in the traffic image more prominent. The depth-estimated traffic image is then input into the vehicle recognition model for vehicle identification.

[0116] Steps S121 and S122 are used to detect incident information. Steps S121 and / or S122 can be selected to obtain incident information of the target sub-path.

[0117] Step S121: Extract accident features from the traffic image to characterize the accident scene, and input the traffic image carrying the accident features into the trained accident classification model to obtain the output accident information.

[0118] If an accident occurs on the target sub-path, traffic images can capture the accident scene, allowing for the extraction of accident features. Accident features are the image content within the traffic image, including but not limited to the overlapping positions of multiple vehicles involved in the accident, debris, hazard lights, and accident markings. For example, if multiple vehicles are involved in a collision, their positions will overlap. Therefore, the overlapping positions of multiple vehicles can be identified from the traffic image to assist in subsequent accident type and location identification. Similarly, debris from the collision may be scattered at the accident scene; the image content corresponding to this debris can be identified from the traffic image to assist in subsequent accident type and location identification. Furthermore, vehicles at accident scenes typically use hazard lights to warn following vehicles to avoid them, or police vehicles may illuminate their sirens to alert other drivers. Therefore, the image content of hazard lights can be identified from the traffic image to assist in subsequent accident type and location identification. Finally, traffic cones or warning tape are often placed at accident scenes for accident marking. Therefore, the image content of emergency lights can be identified from traffic images to assist in subsequent identification of accident type and accident location.

[0119] After extracting accident features, the traffic image carrying these features can be input into a trained accident classification model. The accident classification model can be a multi-classification model; for example, a residual network model or a Vision Transformer (ViT) can be selected as the accident classification model to obtain the output accident type and / or accident location.

[0120] Step S122: Extract accident information from the accident reporting data included in the road prior data.

[0121] As mentioned above, vehicles currently traveling on the target sub-path can report accidents to the navigation service provider. Therefore, the prior road data obtained by the control platform from the navigation service provider can include accident reporting data. Typically, accident reporting data describes the type and location of the accident. Therefore, the accident type and / or accident location can be extracted from the accident reporting data.

[0122] Optionally, after obtaining the accident location through step S121 and / or step S122, it can also be determined whether a critical lane or intersection is blocked based on the accident location. If the accident blocks a critical lane or intersection, the affected critical lane and intersection need to be avoided in the route planning of subsequent vehicles.

[0123] Steps S131 and S132 are used to identify lane structure. Steps S131 and / or S132 can be selected to obtain the lane structure of the target sub-path.

[0124] Step S131: Input the traffic image into the trained lane recognition model to obtain the output lane structure recognition result.

[0125] For example, a lane recognition model is used to identify the lane structure of a target sub-path from a traffic image; therefore, a lane recognition model is a visual recognition model. Inputting the traffic image of the target sub-path into the lane recognition model yields the output lane structure recognition result. The lane structure includes the number of lanes and lane types, etc. Lane types include fast lanes, slow lanes, reversible lanes, non-reversible lanes, etc.

[0126] Step S132: Extract the lane structure from the road information included in the road prior data.

[0127] For example, the control platform can obtain map data from a navigation service provider. This map data is the road prior data mentioned in step S132. The map data records road information for each road, including lane structure, driving direction, road dimensions, etc. Therefore, the lane structure of the target sub-path can be extracted from the road information.

[0128] As can be seen in this embodiment, one or more of the following are extracted from the target sub-path: vehicle density, accident information, and lane structure, using sensing data and / or road prior data, thereby obtaining the spatial characteristics of the target sub-path, so that the impact on congestion spread can be considered from the spatial location dimension.

[0129] According to some embodiments of this application, based on any of the above embodiments, the dynamic features include the driving direction and / or speed distribution of vehicles on the target sub-path. Based on this, step 120 extracts the dynamic features of the target sub-path from the road environment data, specifically including steps S211, S212, and S22. These three steps do not have a specific execution order and can be executed simultaneously. Depending on the specific types of features included in the dynamic features, one or more of these three steps can be selected for execution.

[0130] Steps S211 and S212 are used to detect the driving direction of the vehicles. Steps S211 and / or S212 can be selected to obtain the driving direction of the vehicles on the target sub-path.

[0131] Step S211: Determine the driving direction of the vehicle based on the traffic image sequence in the sensing data.

[0132] For example, a tracking algorithm can be used to determine the direction of travel of a vehicle from a sequence of traffic images. Specifically, the traffic image sequence can be used as input to the tracking algorithm, which can generate the trajectory of the vehicle from the sequence. Subsequently, the displacement of the vehicle is calculated based on the trajectory, and the direction of travel of the vehicle is determined. The tracking algorithm may include, but is not limited to, the DeepSORT algorithm or the FairMOT algorithm.

[0133] Step S212: Extract the driving direction from the road information included in the road prior data.

[0134] As described above, the control platform can obtain map data from the navigation service provider. This map data is the road prior data mentioned in step S212. The map data records road information for each road, including lane structure, driving direction, and road dimensions. Therefore, the driving direction of the target sub-path can be extracted from the road information.

[0135] Step S22 is used to obtain the vehicle speed distribution in the target sub-path.

[0136] Step S22: Determine the vehicle speed distribution based on the traffic image sequence.

[0137] It is understandable that vehicle speeds may differ in different regions of the target sub-path. For example, if the target sub-path includes multiple lanes, the speed in each lane may be different. Therefore, the speed distribution refers to the distribution of vehicle speeds across different regions of the target sub-path. For instance, a tracking algorithm can be used to extract the motion trajectories of multiple vehicles from a traffic image sequence, and then the speed of each vehicle can be determined using its trajectory. Finally, based on the position of each vehicle within the target sub-path, the speed distribution of the target sub-path is determined.

[0138] As can be seen in this embodiment, the driving direction and / or speed distribution of the target sub-path are extracted using sensing data and / or road prior data, thereby obtaining the dynamic characteristics of the target sub-path, so that the impact on congestion spread can be considered from the dimension of dynamic characteristics.

[0139] According to some embodiments of this application, based on any of the above embodiments, the environmental features include meteorological features and / or visibility features. The meteorological features can be weather types, such as rain, hail, or fog / haze. The visibility features can be the value of visibility. Based on this, step 120 extracts the environmental features of the target sub-path from the road environment data, specifically including steps S31-S33. One of these three steps may be performed.

[0140] Step S31: Input the traffic image from the sensed data into the trained meteorological classification model to obtain the output meteorological features and / or visibility features.

[0141] For example, the meteorological classification model is used to identify meteorological features and / or visibility features of a target sub-path from traffic images; therefore, the meteorological classification model is a visual recognition model. Optionally, the meteorological classification model can be a CNN (Convolutional Neural Network) model or a ViT model. Since traffic images may capture some meteorological information about the target sub-path, and the visibility of the target sub-path can be intuitively perceived from traffic images, inputting traffic images into the meteorological classification model can yield the output meteorological features and / or visibility features.

[0142] Step S32: Extract the meteorological features and / or visibility features from the meteorological sensing data in the sensing data.

[0143] As mentioned above, the aircraft can be equipped with weather sensors, including humidity sensors and visibility sensors. Therefore, the weather sensing data can include humidity data sensed by the humidity sensor and visibility data sensed by the visibility sensor. Furthermore, the accuracy of extracting weather features and visibility features can be further improved by combining the weather classification model from step S31.

[0144] Step S33: Extract the meteorological features and / or visibility features from the meteorological information included in the prior road data.

[0145] The control platform can also obtain meteorological information from the meteorological platform. The meteorological information records the weather conditions of each region, including weather type and visibility. Therefore, meteorological features and / or visibility features can be extracted from the meteorological information included in the prior road data.

[0146] As can be seen in this embodiment, the meteorological features and / or visibility features of the target sub-path are extracted using sensing data and / or road prior data, thereby obtaining the environmental features of the target sub-path, so that the impact on congestion spread can be considered from the perspective of the environment.

[0147] According to some embodiments of this application, based on any of the above embodiments, the time feature includes the current time period and historical congestion patterns. The historical congestion pattern refers to the statistical congestion pattern of the target sub-path in different historical time periods.

[0148] It is understandable that some road sections will experience congestion during specific time periods, making congestion strongly correlated with time. For example, main roads in a city may experience congestion during rush hour. Furthermore, congestion dissipation patterns are similar within the same time period. The statistical congestion pattern refers to the start time, cause, and end time of congestion on a target sub-path within multiple identical historical time periods. Based on the current time period and historical congestion patterns, it can be determined whether the congestion occurring in the current time period matches the statistical congestion pattern of a certain historical time period. Therefore, the statistical congestion pattern can be referenced to predict congestion spread.

[0149] Based on this, and in accordance with any of the above embodiments, the congestion diffusion model includes a first congestion diffusion model. Besides predicting the congestion diffusion prediction result of the target sub-path based on the feature set, the first congestion diffusion model can also be used to determine whether congestion has occurred on the target sub-path. The specific determination process includes steps S41-S44.

[0150] Step S41: Based on the feature set, determine the current traffic capacity impact factor and the current visibility impact factor affecting vehicle speed for the target sub-path. The traffic capacity impact factor includes one or more of the following: accident impact factor, weather impact factor, and time impact factor.

[0151] For example, the accident impact factor f can be determined based on accident information included in spatial features. acc The accident impact factor f acc The accident impact factor f is used to characterize the severity of an accident. accThe value of f is negatively correlated with the severity of the accident. That is, the accident impact factor f corresponds to a severe accident. acc The value is less than the accident impact factor f corresponding to a minor accident. acc As an example, the accident impact factor f corresponds to minor accidents (such as minor scratches between vehicles). acc The accident impact factor f is 0.7; for serious accidents (such as multi-vehicle collisions), the factor is 0.7. acc The accident impact factor f is 0.4; there is no accident. acc The value is 1.

[0152] Furthermore, meteorological influencing factors f can be determined based on environmental characteristics, including meteorological characteristics. weather The meteorological influencing factor f weather Used to characterize the severity of weather, and meteorological impact factor f weather It has a negative correlation with the severity of weather. As an example, the meteorological impact factor f corresponds to sunny days. weather The meteorological impact factor f for rainy days is 1. weather The meteorological impact factor f for snow or fog weather is 0.8. weather It is 0.6.

[0153] In addition, the time influence factor f can be determined based on time characteristics. time The time-related factor is used to characterize the similarity between the current level of road congestion and historical congestion patterns. As an example, the time-related factor f... time It can be represented as f time =1-α H ·I peak Among them, α H This represents a parameter indicating the historical congestion pattern, used to characterize the severity of the historical congestion pattern. The historical congestion pattern parameter α... H The value of α is positively correlated with the severity of historical congestion patterns; that is, the more severe the historical congestion pattern, the higher the value of the historical congestion pattern parameter α. H The higher the value of f, the greater the time influence factor. time The smaller the value of I. peak This represents the peak period indicator, indicating how close the current time is to the peak period. The closer the current time is to the peak period, the higher the peak period indicator I becomes. peak The larger the value of f, the greater the time influence factor. time The smaller the value, the better.

[0154] Furthermore, the visibility impact factor f, which affects vehicle speed, can be determined based on environmental characteristics, including visibility features. vis The visibility impact factor f vis Used to characterize environmental visibility. Environmental visibility affects vehicle speed. Visibility influence factor f.vis The value of f is positively correlated with environmental visibility, i.e., the visibility influence factor f vis The smaller the value, the lower the environmental visibility.

[0155] Step S42: Determine the current traffic capacity of the target sub-path based on the basic traffic capacity of the target sub-path and the current traffic capacity influence factor, and determine the current free-flow speed of the target sub-path based on the basic free-flow speed of the target sub-path and the current visibility influence factor.

[0156] Among them, the basic traffic capacity C 0,i Used to indicate the maximum number of vehicles that can pass through the target sub-path i. Basic capacity C 0,i It represents the maximum number of vehicles that a road can handle under ideal conditions (no weather, accidents, or time-related interference). If the target sub-path includes one or more lanes, then the basic capacity C... 0,i This can be expressed as the product of the maximum number of vehicles that can pass through each lane per unit time (e.g., 1 hour) and the number of lanes, i.e., C. 0,i =L i ·A. Among them, L i The number of lanes can be obtained from the lane structure included in the spatial features; A represents the maximum number of vehicles that can pass through each lane per unit time, which can be obtained based on prior road data.

[0157] Current traversability of target subpath i It is composed of basic traffic capacity C 0,i This is obtained after adjusting for the current traffic capacity impact factor. Therefore, the current traffic capacity... This is used to characterize the number of vehicles that can currently pass through target sub-path i. As an example, the current capacity of target sub-path i... It can be represented as =C 0,i ·f acc ·f weather ·f time It can be seen that, since the capacity influencing factor may change at different times, the capacity of the target sub-path i may be the same or different at different times.

[0158] Furthermore, the basic free-flow velocity v f,0 This indicates the vehicle speed of target subpath i under ideal conditions (no congestion, no interference, and no weather effects). This vehicle speed should refer to the highest sustainable speed at which the vehicle can safely travel. Basic free-flow speed v f,0 It can be obtained based on prior road data.

[0159] Current free-flow velocity of target sub-path i It is determined by the basic free-flow velocity v f,0 Based on the current visibility impact factor f vis This is the corrected value. Therefore, the current free-flow velocity... This characterizes the current vehicle speed of target subpath i when there is no congestion. As an example, the current free-flow speed of target subpath i. It can be represented as =v f,0 ·f vis It can be seen that due to the visibility influencing factor f vis The free-flow velocity of the target sub-path i may be the same or different at different times.

[0160] Step S43: Determine the current critical density, which is used to characterize the congestion threshold of the target sub-path, based on the current capacity and the current free-flow velocity.

[0161] For example, the current critical density The congestion threshold used to characterize target sub-path i, that is, the critical point at which target sub-path i becomes congested. It can determine the current traffic capacity. With the current free flow velocity The ratio is the current critical density. ,Right now = / Therefore, the critical density is a time-dependent variable, meaning that the critical density of the target sub-path i may change at different times.

[0162] Step S44: Obtain the current vehicle density of the target sub-path from the spatial features, and determine whether the target sub-path is congested based on the current vehicle density and the current critical density.

[0163] For example, the calculation process for vehicle density is described above and will not be repeated here. If the current vehicle density... Greater than the current critical density If the target sub-path i is congested, then it is determined that congestion has occurred; if the preceding vehicle density... Less than or equal to the current critical density If so, then it is determined that the target sub-path i is not congested.

[0164] Thus, by executing steps S41-S44, congestion can be determined for the target sub-path.

[0165] Furthermore, the first congestion diffusion model determines the free-flow velocity and capacity of the target sub-path based on a feature set, and predicts the congestion diffusion prediction result of the target sub-path based on the free-flow velocity and capacity, specifically including steps S45-S48. It is important to note that steps S45-S48 are not necessarily executed after steps S41-S44. For example, congestion can be determined to have occurred on the target sub-path using other methods provided above, and then steps S45-S48 can be executed to obtain the congestion diffusion prediction result.

[0166] Step S45: If it is determined that the target sub-path is congested, predict the maximum outflow of the target sub-path and the maximum inflow of the downstream sub-path connected to the target sub-path at a future target time.

[0167] For example, a target time refers to any future time relative to the current time. As an example, a target time could be the predicted time when a target vehicle travels from its current location to the target sub-segment.

[0168] Predicting the maximum outflow of the target sub-path at the target time refers to predicting the maximum flow that the target sub-path can output to its connected downstream sub-paths at the target time. The prediction process specifically includes steps S451-S452.

[0169] Step S451: Predict the maximum number of outgoing vehicles and the theoretical maximum number of passing vehicles on the target sub-path at the target time; wherein, the maximum number of outgoing vehicles is related to the free flow velocity and vehicle density of the target sub-path at the target time; the theoretical maximum number of passing vehicles is related to the traffic capacity of the target sub-path at the target time.

[0170] Wherein, the free-flow velocity of target sub-path i at target time T It can be determined by the basic free-flow velocity v f,0 Visibility impact factor f at target time T vis After correction, the specific implementation process is similar to that of step S42, and will not be repeated here.

[0171] Vehicle density of target sub-path i at target time T It can be based on the current vehicle density The time difference between the current time t and the target time T Length of the target sub-path The net inflow of the target sub-path is obtained from the target sub-path's upstream sub-path flow into the target sub-path and the outflow of the target sub-path flow into the downstream sub-path. As an example, the target sub-path at target time T (T=t+ Vehicle density under ) It can be as shown in Formula 1.

[0172] Formula 1

[0173] in, The inflow of traffic from the upstream subpath to the target subpath; This represents the outflow from the target sub-path to the downstream sub-path.

[0174] The maximum number of outflowing vehicles on the target subpath and the free-flow velocity of the target subpath at the target time. and vehicle density Related. As an example, the maximum number of outgoing vehicles on the target subpath at the target time can be expressed as: It can be seen that the maximum number of vehicles flowing out is limited by vehicle density and vehicle free-flow velocity.

[0175] The traversability of target sub-path i at target time T It can be determined by the basic accessibility C. 0,i The result is obtained after correcting the traffic capacity influence factor at the target time. The specific implementation process is similar to that of step S42, and will not be described in detail here.

[0176] The theoretical maximum number of vehicles that can pass through the target sub-path at the target time and the traffic capacity of the target sub-path at the target time. Related. As an example, the theoretical maximum number of vehicles that can pass through the target sub-path at the target time can be expressed as: It can be seen that the theoretical maximum number of vehicles that can pass is limited by the traffic capacity of the target sub-path, that is, by the road structure of the target sub-path.

[0177] Step S452: Determine the minimum of the maximum outflow number of vehicles and the theoretical maximum number of vehicles that can pass through as the maximum outflow flow of the target sub-path.

[0178] For example, the maximum outflow of the target sub-path It can be represented as .

[0179] Predicting the maximum received traffic of the downstream sub-path at the target time means predicting the maximum traffic that the downstream sub-path can receive from the target sub-path at the target time. The prediction process specifically includes steps S453-S454.

[0180] Step S453: Predict the maximum remaining received traffic and the theoretical maximum number of vehicles that can pass through the downstream sub-path at the target time; wherein, the maximum remaining received traffic is related to the vehicle density of the downstream sub-path at the target time; and the theoretical maximum number of vehicles that can pass through is related to the traffic capacity of the downstream sub-path at the target time.

[0181] Vehicle density of downstream sub-path j at target time T The calculation process can be found in step S451, which mentions the vehicle density of the target sub-path at the target time. The calculation process, and the throughput capacity of downstream sub-path j at target time T. The calculation process can be found in step S451, which mentions the traffic capacity of the target sub-path at the target time. The calculation process will not be elaborated here.

[0182] The maximum remaining received traffic of the downstream subpath and the vehicle density of the downstream subpath at the target time. Related. As an example, the maximum remaining received traffic on the downstream subpath at the target time can be expressed as: Among them, K jam For congestion density, it is a constant term; This represents the backward wave velocity, i.e., the propagation speed of congestion. Wherein, the congestion density K... jam It refers to the traffic density under congested road conditions, that is, the number of vehicles per unit length of road segment under congested road conditions.

[0183] The theoretical maximum number of vehicles that can pass through the downstream sub-path at the target time and the traffic capacity of the downstream sub-path at the target time. Related. As an example, the theoretical maximum number of vehicles that can pass through the downstream sub-path at the target time can be expressed as: It can be seen that the theoretical maximum number of vehicles that can pass through is limited by the capacity of the downstream sub-path, that is, by the road structure of the downstream sub-path.

[0184] Step S454: Determine the minimum of the maximum remaining received traffic and the theoretical maximum number of vehicles that can pass as the maximum received traffic for the downstream sub-path.

[0185] For example, the maximum received traffic of the downstream sub-path It can be represented as .

[0186] Thus, steps S451-S454 complete the prediction of the maximum outflow of the target sub-path and the maximum inflow of the downstream sub-path at the target time, as described in step S45. Then, steps S46-S48 are executed.

[0187] Step S46: Determine the predicted flow from the target sub-path to the downstream sub-path based on the maximum outflow flow and the maximum inflow flow.

[0188] For example, the minimum of the maximum outflow and the maximum inflow can be determined as the predicted flow from the target sub-path to the downstream sub-path at the target time. Predicted flow It can be represented as .

[0189] Understandably, if the downstream subpath is congested (manifested as maximum received traffic) Smaller, predicted flow Maximum received traffic If restrictions are imposed, the outflow of traffic from the target sub-path will decrease, thereby exacerbating the congestion of the target sub-path.

[0190] Step S47: Predict the congestion spread rate of the target sub-path at the target time based on the predicted flow and the density margin of the downstream sub-path.

[0191] For example, the density margin of the downstream sub-path This refers to the vehicle density of the downstream sub-path at the target time. With critical density The difference between them. Wherein, the critical density of downstream sub-paths at the target time. The calculation process and the current critical density of the target sub-path in step S43. The calculation process is similar and will not be repeated here. The density margin of the downstream sub-path at the target time. It can be represented as = - .

[0192] As an example, the congestion spread rate can be predicted based on the difference between predicted and normal traffic flow, and the density margin. Here, normal traffic flow refers to the basic capacity C of the target sub-path. 0,i For example, congestion diffusion velocity is used to characterize the speed at which congestion is alleviated. A higher congestion diffusion velocity indicates that congestion on the target subpath is alleviated more quickly, and vice versa. As an example, the difference between predicted traffic and normal traffic can be determined. and the density margin The ratio between these two values ​​represents the congestion spread rate. Thus, the congestion spread rate c... diff It can be represented as .

[0193] Step S48: If the congestion spread rate is less than the preset speed threshold, then the target sub-path is determined to maintain congestion; if the congestion spread rate is greater than the speed threshold, then the target sub-path is determined to alleviate congestion.

[0194] As mentioned above, the congestion spread rate c diff Used to characterize the speed at which congestion is alleviated. Therefore, if the congestion diffusion speed c diff A speed below the preset threshold indicates slow congestion relief on the target sub-path, which will persist for an extended period. If the target time is the predicted time for the target vehicle to travel from its current location to the target sub-path, then this means that by the time the target vehicle arrives on the target sub-path, the congestion on that sub-path has not yet been relieved.

[0195] Conversely, if the congestion spreads at a rate of c diff A speed greater than the stated speed threshold indicates that congestion on the target sub-path will be quickly alleviated, and traffic will return to normal within a short period. If the target time is the predicted time when the target vehicle arrives at the target sub-path from its current location, then this means that congestion on the target sub-path has already eased by the time the target vehicle arrives.

[0196] As can be seen, this embodiment realizes the congestion judgment and congestion spread prediction of the target sub-path through the first congestion diffusion model, which can effectively prevent the target vehicle from entering the congested road section.

[0197] According to some embodiments of this application, based on any of the above embodiments, the congestion diffusion model includes a second congestion diffusion model. The second congestion diffusion model is a trained neural network model. The second congestion diffusion model includes a Spatio-Temporal Graph Neural Network (ST-GNN) module, a time series module, and a Reinforcement Learning (RL) module. The time series module includes a Transformer sub-model and a temporal convolutional sub-model. The training data for the second congestion diffusion model includes historical congestion diffusion data for various historical congestion scenarios under different meteorological conditions. The loss function of the second congestion diffusion model may include: the difference between the congestion level predicted by the model and the actual congestion level extracted from the historical congestion diffusion data, and the difference between the spatial range of congestion diffusion predicted by the model and the actual spatial range of congestion diffusion extracted from the historical congestion diffusion data. Furthermore, deploying a lightweight model can meet the low-latency requirement for congestion diffusion prediction, thereby enabling real-time inference. Based on this, in step 140, the feature set is input into the congestion diffusion model to obtain the output congestion diffusion prediction result, specifically including steps 141-142.

[0198] Step 141: Input the feature set into the spatiotemporal graph neural network module, the time series module, and the reinforcement learning module respectively to obtain the congestion propagation simulation results output by the spatiotemporal graph neural network module, the traffic flow change prediction results output by the time series module, and the vehicle diversion impact prediction results output by the reinforcement learning module.

[0199] As mentioned above, the feature set includes one or more of spatial features, dynamic features, environmental features, and temporal features. The feature set is input into a spatiotemporal graph neural network module, which can divide the target sub-path into multiple nodes with traffic flow as edges, and simulate congestion propagation based on the input feature set, thereby obtaining the output congestion propagation simulation results.

[0200] The time series module consists of a Transformer sub-model and a temporal convolution sub-model. It leverages the Transformer sub-model's advantage in capturing long-range dependencies and the temporal convolution sub-model's ability to efficiently process local features within the network. Therefore, it is suitable for processing time series data with complex spatiotemporal dependencies. By inputting the feature set into the time series module, it can capture the dynamic changes of the target sub-path over time, thereby predicting traffic flow changes along the target sub-path and obtaining the output traffic flow prediction result.

[0201] The feature set is input into the reinforcement learning module, which can simulate the impact of vehicle diversion on congestion and thus output the prediction results of the impact of vehicle diversion on congestion.

[0202] It can be seen that the spatiotemporal graph neural network module, time series module, and reinforcement learning module in the second congestion diffusion prediction model are three independent modules. The feature set is input into each of the three modules, and the three modules process the feature set independently and output different prediction results.

[0203] Step 142: Generate the congestion diffusion prediction result based on the congestion propagation simulation result, the traffic flow change prediction result, and the prediction result of the impact of vehicle diversion on congestion; wherein, the congestion diffusion prediction result includes one or more of the following: a predicted congestion heat map for characterizing the degree of congestion at different locations, the direction of congestion diffusion, and the predicted congestion mitigation time.

[0204] For example, after obtaining the three prediction results, a congestion spread prediction result can be generated by combining the three prediction results, including a predicted congestion heatmap, the direction of congestion spread, and the predicted congestion mitigation time. The predicted congestion heatmap characterizes the degree of congestion at different locations, visually displaying the degree and distribution of traffic congestion in each area. Typically, different colors are used to represent changes in vehicle density or speed on the road, thus helping to quickly identify areas with severe congestion problems. The predicted congestion heatmap can include predicted congestion heatmaps for multiple future time periods, such as predicted congestion heatmaps for the next 15 minutes, 30 minutes, and 60 minutes. The direction of congestion spread indicates the areas where the current congestion will subsequently spread, thus allowing for the estimation of congestion in the spread areas. The predicted congestion mitigation time indicates when the congestion will end. Based on the predicted congestion mitigation time, it can be determined when the road will return to smooth traffic flow.

[0205] For example, if the target sub-path is a section of highway, the aircraft collects the first sensing data for that highway section. The real-time traffic image shows high-density traffic flow from east to west (travel direction), a rear-end collision in the left lane (accident information), and moderate rain (environmental information). Through feature extraction, spatial features can be obtained, including: vehicle density of 0.2 vehicles / meter, and one lane blocked by the accident; dynamic features, including 80% of vehicles traveling from east to west; and environmental features, including low visibility due to rain. By inputting the extracted features into the congestion spread prediction model, the output congestion spread prediction result can be obtained, specifically including: the congestion is expected to spread upstream by 3 kilometers in the next 30 minutes.

[0206] Optionally, the congestion spread prediction results can also be output to the target vehicle so that the driver of the target vehicle can know the congestion situation ahead and the expected congestion spread.

[0207] In this embodiment, the spatiotemporal graph neural network module, time series module, and reinforcement learning module process the input feature combinations to obtain congestion propagation simulation results, traffic flow change prediction results, and vehicle diversion impact prediction results. Based on these three prediction results, a congestion diffusion prediction result is generated, enabling timely route replanning to avoid vehicles entering congested sections if the predicted congestion fails to ease. This achieves the technical effect of real-time detection of congestion events during vehicle travel, assisting vehicles in avoiding congestion.

[0208] Furthermore, in some embodiments, congestion root cause analysis is also performed. It is understood that congestion can be caused by a variety of factors, such as waiting at traffic lights, signal light malfunctions, accidents, and road construction. Different causes of congestion will have different impacts on the route planning of subsequent vehicles. For example, if the cause of congestion is a signal light malfunction, since signal light malfunctions generally have a quick repair response, they will not cause widespread congestion. Subsequent vehicles can still consider including the road where the signal light malfunction occurred in their route planning. On the other hand, if the cause of congestion is a serious rear-end collision or road repairs, it will require a longer period of traffic management. Therefore, subsequent vehicles can avoid the congested section when planning their routes.

[0209] Based on the vehicle path planning method based on an aircraft described in any of the above embodiments, this application also provides a data processing system, the system comprising: a vehicle, an aircraft, and a control platform. The control platform includes: a processor; and a memory for storing processor-executable instructions; wherein, when the processor invokes the executable instructions, it implements the operation of the vehicle path planning method based on an aircraft described in any of the above embodiments.

[0210] Based on the vehicle path planning method for aircraft described in any of the above embodiments, this application also provides a computer program product, which includes one or more computer programs or instructions. The computer programs or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. When the computer program is executed by a processor, it implements the vehicle path planning method for aircraft described in any of the above embodiments.

[0211] Based on the vehicle path planning method for aircraft described in any of the above embodiments, this application also provides, as well as... Figure 3 The diagram shows the structure of an electronic device. Figure 3 At the hardware level, the electronic device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the vehicle path planning method based on an aircraft as described in any of the above embodiments. As an example, the electronic device may be equipped with a control platform to implement the vehicle path planning method based on an aircraft as described in any of the above embodiments.

[0212] This application also provides a computer storage medium storing a computer program, which, when executed by a processor, can be used to perform a vehicle path planning method based on an aircraft as described in any of the above embodiments.

[0213] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0214] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0215] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0216] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0217] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0218] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 a process, method, article, or apparatus. Without further limitations, 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 said element.

Claims

1. An aerial vehicle based vehicle path planning method, characterized in that, The method comprises: acquiring a planned path of a target vehicle to a destination, wherein the planned path comprises a plurality of sub-paths; in a case where it is determined that a target sub-path in the plurality of sub-paths is congested, extracting one or more of a spatial feature, a dynamic feature, an environmental feature and a time feature of the target sub-path from road environment data corresponding to the target sub-path, to obtain a feature set; wherein at least part of the road environment data is obtained by sensing the target sub-path by an aerial vehicle; the spatial feature comprises one or more of vehicle density, accident information and lane structure, the accident information comprises accident type and / or accident location; the dynamic feature comprises driving direction and / or vehicle speed distribution of a driving vehicle; the environmental feature comprises meteorological feature and / or visibility feature; the time feature comprises current time period and historical congestion pattern; the historical congestion pattern is used to indicate statistical congestion pattern of the target sub-path in different historical time periods; inputting the feature set into a congestion diffusion model to obtain an output congestion diffusion prediction result of the target sub-path; wherein the congestion diffusion model comprises a first congestion diffusion model, the first congestion diffusion model is used to determine free flow speed and traffic capacity of the target sub-path based on the feature set, and predict the congestion diffusion prediction result of the target sub-path based on the free flow speed and the traffic capacity; if the congestion diffusion prediction result indicates that the target sub-path is congested, determining that the target vehicle keeps driving along the planned path; if the congestion diffusion prediction result indicates that the target sub-path is congested, re-planning a planned path of the target vehicle to the destination and not containing the target sub-path.

2. The vehicle path planning method according to claim 1, characterized by, The method further comprises: based on a current sub-path where the target vehicle is located, determining, from the plurality of sub-paths, that a sub-path reached by the target vehicle next is the target sub-path connected with the current sub-path.

3. The vehicle path planning method according to claim 1, characterized by, The method further comprises: in a case where it is determined that the target sub-path is not congested, determining that the target vehicle keeps driving along the planned path.

4. The vehicle path planning method according to claim 1, characterized by, The re-planning of the planned path of the target vehicle to the destination and not containing the target sub-path comprises: based on road topological relationship, determining candidate sub-paths connected with a current sub-path where the target vehicle is located, except for the target sub-path maintaining congestion; based on road environment data of each of the candidate sub-paths, determining a passable sub-path with normal road conditions from all candidate sub-paths; wherein the normal road conditions comprise smooth traffic or congestion alleviation; planning a planned path with the current sub-path as a starting point, the destination as an end point, and passing through the passable sub-path.

5. The vehicle path planning method according to any one of claims 1-4, characterized in that, The road environment data comprises road priori data and sensing data obtained by sensing the target sub-path by the aerial vehicle; the spatial feature is obtained by the following steps: inputting a traffic image in the sensing data into a trained vehicle recognition model to obtain an outputted vehicle recognition result of a traveling vehicle, and determining a vehicle density based on a quantity of the traveling vehicles and a road size of the target sub-path; extracting an accident feature for representing an accident scene from the traffic image, and inputting the traffic image carrying the accident feature into a trained accident classification model to obtain the outputted accident information; and / or extracting the accident information from accident report data included in the road priori data; inputting the traffic image into a trained lane recognition model to obtain an outputted lane structure recognition result; and / or extracting the lane structure from road information included in the road priori data.

6. The vehicle path planning method according to any one of claims 1 to 4, characterized by, The road environment data includes road priori data and sensing data obtained by sensing the target sub-path by the aerial vehicle; and the dynamic feature is obtained by: determining a traveling direction of the traveling vehicle based on a traffic image sequence in the sensing data; and / or extracting the traveling direction from road information included in the road priori data; determining a vehicle speed distribution of the traveling vehicle based on the traffic image sequence.

7. The vehicle path planning method according to any one of claims 1-4, characterized by, The road environment data includes road priori data and sensing data obtained by sensing the target sub-path by the aerial vehicle; and the environmental feature is obtained by: inputting a traffic image in the sensing data into a trained weather classification model to obtain an outputted weather feature and / or visibility feature; or extracting the weather feature and / or visibility feature from weather sensing data in the sensing data; or extracting the weather feature and / or visibility feature from weather information included in the road priori data.

8. The vehicle path planning method according to any one of claims 1 to 4, characterized by, The congestion diffusion model includes a second congestion diffusion model, the second congestion diffusion model includes a spatio-temporal graph neural network module, a time series module and a reinforcement learning module; the time series module includes a Transformer sub-model and a time convolution sub-model; and the inputting the feature set into the congestion diffusion model to obtain an outputted congestion diffusion prediction result of the target sub-path includes: inputting the feature set into the spatio-temporal graph neural network module, the time series module and the reinforcement learning module respectively to obtain a congestion propagation simulation result outputted by the spatio-temporal graph neural network module, a traffic flow change prediction result outputted by the time series module, and a prediction result of an influence of vehicle shunting on congestion outputted by the reinforcement learning module; generating the congestion diffusion prediction result based on the congestion propagation simulation result, the traffic flow change prediction result and the prediction result of the influence of vehicle shunting on congestion; wherein the congestion diffusion prediction result includes one or more of a predicted congestion heat map for representing a congestion degree at different positions, a congestion diffusion direction, and a congestion prediction relief time.

9. A data processing system, characterized by The system includes a vehicle, an aerial vehicle and a management and control platform. The management platform comprises: a processor; a memory for storing processor-executable instructions; wherein the processor invokes the executable instructions to implement the operations of the vehicle path planning method of any one of claims 1-8.

Images