Parking path generation method and system

Abstract

The invention provides a parking path generation method and system, and relates to the technical field of parking control, and the method comprises the steps: responding to a parking request from a target vehicle in the parking process of the target vehicle according to a specified parking path, and generating N alternative paths; for each alternative path in the N alternative paths, deconstructing scene information included in the alternative paths through a scene recognition model to obtain scene scores corresponding to the alternative paths, and determining predicted passing duration of the target vehicle driving the alternative paths according to the scene information included in the alternative paths through a duration prediction model, according to the predicted passing duration and the vehicle state of the target vehicle, determining the predicted energy consumption of the alternative driving path of the target vehicle, and according to the scene score corresponding to the alternative path, the predicted passing duration and the predicted energy consumption, determining the total cost of the alternative path; and determining a target parking path from the N alternative paths according to the N total costs corresponding to the N alternative paths.

Description

Technical Field

[0001] This invention relates to the field of parking control technology, and in particular to a method and system for generating parking paths. Background Technology

[0002] With the rapid development and widespread adoption of autonomous driving technology, automatic parking has become a key feature for enhancing vehicle intelligence and user experience. Currently, automatic parking systems typically perform path planning and decision-making locally within the vehicle, relying on onboard sensors and limited computing resources. However, this vehicle-driven approach has inherent limitations: in the face of complex and ever-changing parking environments, a single vehicle's computing power struggles to deeply integrate global information and make long-term predictions; pre-planned single paths lack sufficient flexibility and robustness when encountering sudden obstacles or changes in the scenario, potentially leading to parking interruptions, excessively long parking times, or a surge in energy consumption, impacting parking efficiency and user experience. Summary of the Invention

[0003] This invention provides a method and system for generating parking paths, which solves the problem in related technologies where the target vehicle determines the parking path based on real-time perception by onboard sensors, resulting in the path being unable to adapt to environmental changes.

[0004] In a first aspect, embodiments of the present invention provide a method for generating parking paths, applied in the cloud, the method comprising: During the process of the target vehicle parking according to the specified parking path, in response to the parking request from the target vehicle, N alternative paths from the current location of the target vehicle to the target parking space are generated. For each of the N alternative paths, the scene information included in the alternative path is deconstructed using a scene recognition model to obtain the scene score corresponding to the alternative path. The predicted travel time of the target vehicle on the alternative path is determined using a time prediction model based on the scene information included in the alternative path. The predicted energy consumption of the target vehicle on the alternative path is determined based on the predicted travel time and the vehicle status of the target vehicle. Finally, the total cost of the alternative path is determined based on the scene score corresponding to the alternative path, the predicted travel time, and the predicted energy consumption. Based on the N total costs corresponding to the N alternative paths, a target parking path is determined from the N alternative paths, and the target parking path is issued to the target vehicle so that the target vehicle updates the designated parking path according to the target parking path.

[0005] Optionally, the target parking space is located within the target parking lot indicated by the parking request; the step of generating N alternative paths from the current location of the target vehicle to the target parking space in response to the parking request from the target vehicle includes: in response to the parking request, obtaining available parking space data sent by the parking lot terminal equipment of the target parking lot from the target vehicle, the available parking space data including the number and / or location of unoccupied parking spaces in the target parking lot; and determining a target parking space for parking the target vehicle from the target parking lot based on the current location and the available parking space data. The system then sends a scheduling instruction to the site control unit, which instructs the site control unit to modify the status of the target parking space; it also sends an entrance gate coordination instruction to the M entrance gates of the target parking lot to obtain the waiting traffic flow at the M entrance gates; based on the waiting traffic flow at the M entrance gates, it determines the target entrance gate from the M entrance gates; and it generates N alternative routes based on the current location, the target entrance gate, and the target parking space, with each alternative route including the target entrance gate and the target parking space.

[0006] Optionally, the step of deconstructing the scene information included in the alternative paths through the scene recognition model to obtain the scene score corresponding to the alternative paths includes: identifying the alternative paths through the scene recognition model to determine Y target scene regions included on the alternative paths; labeling the Y target scene regions with Y target scene tags on the alternative paths, wherein the scene recognition model is trained based on X preset scene tags, and the Y target scene tags belong to the X scene tags; issuing a site-end collaboration command to the site-end device to obtain parking lot video of the target parking lot from the site-end device; and obtaining data from the site-end device. The system uses sensors on the target vehicle to collect road condition videos. For each of the Y target scene areas, it determines the target scene video corresponding to the target scene area from the parking lot video and / or the road condition video. It then injects the target scene tags corresponding to the target scene areas into the Y target scene videos. The system analyzes the Y target scene videos using the scene recognition model to determine the vehicle congestion and / or pedestrian density within the Y target scene areas, obtaining risk scores corresponding to the Y target scene areas. Based on the risk scores corresponding to the Y target scene areas, it determines the scene score corresponding to the alternative path.

[0007] Optionally, determining the predicted travel time of the target vehicle on the alternative routes based on the scene information included in the alternative routes using the time prediction model includes: obtaining the vehicle status of the target vehicle, the vehicle status including vehicle hardware parameters, the vehicle hardware parameters including at least the length, width, height, wheelbase, and minimum turning radius of the vehicle; predicting the travel time of the target vehicle in Y target scene areas based on the vehicle hardware parameters using the time prediction model; and obtaining the predicted travel time of the alternative routes based on the Y travel times corresponding to the Y target scene areas.

[0008] Optionally, the vehicle status further includes the energy consumption type and energy consumption per unit time of the target vehicle; determining the predicted energy consumption of the target vehicle traveling the alternative route based on the predicted travel time and the vehicle status of the target vehicle includes: determining the predicted energy consumption based on the energy consumption type and energy consumption per unit time of the target vehicle and the predicted travel time.

[0009] Optionally, determining the total cost of the alternative path based on the scenario score, predicted travel time, and predicted energy consumption of the alternative path includes: determining the driving style of the target vehicle based on its vehicle hardware parameters, energy consumption type, and energy consumption per unit time; determining the weight ratio corresponding to the target vehicle based on its driving style; and determining the total cost of the alternative path based on the weight ratio, the scenario score, predicted travel time, and predicted energy consumption of the alternative path.

[0010] Secondly, embodiments of the present invention provide a parking path generation system, the system comprising: The target vehicle is used to park according to a specified parking path, and during the parking process of the target vehicle, a parking request is sent to the cloud. In the cloud, in response to a parking request from a target vehicle, N alternative routes are generated from the target vehicle's current location to the target parking space. For each of the N alternative routes, a scene recognition model is used to deconstruct the scene information included in the alternative route to obtain a scene score corresponding to the alternative route. A duration prediction model is used to determine the predicted travel time of the target vehicle traveling on the alternative route based on the scene information included in the alternative route. Based on the predicted travel time and the vehicle status of the target vehicle, the predicted energy consumption of the target vehicle traveling on the alternative route is determined. Based on the scene score, predicted travel time, and predicted energy consumption corresponding to the alternative route, the total cost of the alternative route is determined. Based on the N total costs corresponding to the N alternative routes, the target parking route is determined from the N alternative routes and the target parking route is issued to the target vehicle. The target vehicle is further configured to, upon receiving the target parking path from the cloud, update the designated parking path according to the target parking path and park according to the updated designated parking path.

[0011] Optionally, the system further includes: a parking lot terminal device, configured at the target parking lot indicated by the parking request, wherein the parking lot terminal device is configured to: send the available parking space data to the target vehicle when the target vehicle retrieves the available parking space data of the target parking lot, the available parking space data including the number and / or location of the unoccupied parking spaces in the target parking lot; the target vehicle is further configured to retrieve the available parking space data from the parking lot terminal device in response to the parking request; and generate an initial parking path from the target vehicle to the target parking lot when the number of parking spaces is not zero, and use the initial parking path as the designated parking path.

[0012] Optionally, the site-end device is further configured to: in response to a site-end collaboration instruction from the cloud, send a parking lot video of the target parking lot to the cloud; the cloud is further configured to, upon receiving the parking lot video, determine the target scene video corresponding to the target scene area from the parking lot video.

[0013] Optionally, the system further includes: M entrance gates, installed in the target parking lot, wherein the M entrance gates are used to: in response to an entrance gate coordination instruction from the cloud, send the waiting traffic flow at the M entrance gates of the target parking lot to the cloud; the cloud is also used to determine the target entrance gate from the M entrance gates based on the waiting traffic flow at the M entrance gates.

[0014] A parking path generation method according to an embodiment of the present invention, applied in the cloud, includes: during the process of a target vehicle parking according to a specified parking path, in response to a parking request from the target vehicle, generating N alternative paths from the current location of the target vehicle to the target parking space; for each of the N alternative paths, deconstructing the scene information included in the alternative path using a scene recognition model to obtain a scene score corresponding to the alternative path; and determining the target vehicle based on the scene information included in the alternative path using a duration prediction model. The system calculates the predicted travel time for the alternative routes, determines the predicted energy consumption of the target vehicle traveling the alternative routes based on the predicted travel time and the vehicle status, and determines the total cost of the alternative routes based on the scene score, predicted travel time, and predicted energy consumption. Based on the N total costs corresponding to the N alternative routes, a target parking route is determined from the N alternative routes and sent to the target vehicle so that the target vehicle can update its designated parking route accordingly. During the target vehicle's journey, the cloud evaluates the alternative routes from multiple perspectives, including scene score, predicted travel time, and predicted energy consumption, and calculates the comprehensive cost of each route by integrating multiple factors. Based on the total cost of the N alternative routes, the cloud can determine the most suitable target parking route for the target vehicle from multiple alternative routes and send it to the target vehicle to update its designated parking route. This achieves dynamic, real-time, and global optimization of parking routes, significantly improving the adaptability, efficiency, and economy of the parking process. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 The flowchart of a parking path generation method provided by an embodiment of the present invention is shown; Figure 2 This illustrates the technical concept of selecting a target entrance gate according to an embodiment of the present invention; Figure 3 This illustrates a scene recognition concept for alternative paths provided by an embodiment of the present invention; Figure 4 The diagram illustrates the deconstruction of a parking path generation system provided in one embodiment of the present invention; Figure 5 This illustrates the concept of a parking path generation method provided by one embodiment of the present invention. Detailed Implementation

[0017] As described in the background section, with the rapid development and popularization of autonomous driving technology, automatic parking has become a key feature for improving vehicle intelligence and user experience. Currently, automatic parking systems typically perform path planning and decision-making locally within the vehicle, relying on onboard sensors and limited computing resources. However, this vehicle-driven approach has inherent limitations: facing complex and ever-changing parking environments, the computing power of a single vehicle is insufficient for deep fusion and long-term prediction of global information; pre-planned single paths lack sufficient flexibility and robustness when encountering sudden obstacles or changes in the scenario, potentially leading to parking interruptions, excessively long parking times, or a surge in energy consumption, affecting parking efficiency and user experience.

[0018] A parking path generation method according to an embodiment of the present invention, applied in the cloud, includes: during the process of a target vehicle parking according to a specified parking path, in response to a parking request from the target vehicle, generating N alternative paths from the current location of the target vehicle to the target parking space; for each of the N alternative paths, deconstructing the scene information included in the alternative path using a scene recognition model to obtain a scene score corresponding to the alternative path; and determining the target vehicle based on the scene information included in the alternative path using a duration prediction model. The system calculates the predicted travel time for the alternative routes, determines the predicted energy consumption of the target vehicle traveling the alternative routes based on the predicted travel time and the vehicle status, and determines the total cost of the alternative routes based on the scene score, predicted travel time, and predicted energy consumption. Based on the N total costs corresponding to the N alternative routes, a target parking route is determined from the N alternative routes and sent to the target vehicle so that the target vehicle can update its designated parking route accordingly. During the target vehicle's journey, the cloud evaluates the alternative routes from multiple perspectives, including scene score, predicted travel time, and predicted energy consumption, and calculates the comprehensive cost of each route by integrating multiple factors. Based on the total cost of the N alternative routes, the cloud can determine the most suitable target parking route for the target vehicle from multiple alternative routes and send it to the target vehicle to update its designated parking route. This achieves dynamic, real-time, and global optimization of parking routes, significantly improving the adaptability, efficiency, and economy of the parking process.

[0019] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0020] Figure 1 The flowchart illustrates a parking path generation method according to an embodiment of the present invention. It should be understood that, as... Figure 1 The method for generating the parking path shown can be executed in the cloud. The cloud can be a cloud server capable of cloud computing. For example... Figure 1 As shown, an embodiment of the present invention provides a parking path generation method including steps 110 to 130.

[0021] Step 110: During the process of the target vehicle parking according to the specified parking path, in response to the parking request from the target vehicle, generate N alternative paths from the current position of the target vehicle to the target parking space, where N is any positive integer.

[0022] In this embodiment of the invention, the target vehicle can be any vehicle including an in-vehicle infotainment system. The target vehicle can park according to a designated parking path under the control of the in-vehicle infotainment system or under the manual control of the driver. The designated parking path is the parking path that the target vehicle is currently tracking.

[0023] In this embodiment of the invention, during the parking process of the target vehicle according to the specified parking path, the target vehicle can send a parking request to the cloud. After receiving the parking request, the cloud executes step 110. The parking request may include the target vehicle's parking intention and its current Global Positioning System (GPS) information. During the process of the target vehicle sending the parking request to the cloud, it may also include the target vehicle's sensor data and vehicle status. The sensor data may include Inertial Measurement Unit (IMU) information, the target vehicle's current speed, and the target vehicle's current heading angle. The vehicle status may include the vehicle's hardware status, such as whether the sensors collecting the sensor data are functioning properly, to indicate whether the sensor data is available. After receiving the parking request, the cloud can fuse the target vehicle's GPS information, IMU information from the sensor data, and the target vehicle's current heading angle to determine the target vehicle's current position, calibrate the target vehicle's GPS information, and improve the accuracy of the target vehicle's current position. The current location of the target vehicle can be any location, such as the starting parking space of the target vehicle, any location on the specified parking path, or within the target parking lot to which the target parking space belongs.

[0024] In this embodiment of the invention, the target parking space can be a parking space for the target vehicle, located within the target parking lot. The target parking space can be selected by the target vehicle, and the cloud can obtain the target parking space by receiving the selection from the target vehicle; alternatively, the cloud can automatically allocate a parking space based on the target vehicle's parking request.

[0025] In this embodiment of the invention, after determining the current location of the target vehicle and the target parking space, the cloud can generate N alternative routes based on city map information and the parking lot's internal information. Each alternative route starts at the current location of the target vehicle and ends at the target parking space, facilitating parking for the target vehicle. One of the N alternative routes is the route the target vehicle is currently following. During the generation of alternative routes in the cloud, not only the vehicle's current location and the target parking space are considered, but also the vehicle's status, which may include vehicle hardware parameters such as length, width, height, wheelbase, and minimum turning radius. Alternative routes that better match the target vehicle's hardware parameters can be generated based on these parameters. After generating the N alternative routes, the cloud can execute step 120 for each of the N alternative routes.

[0026] Step 120: For each of the N alternative paths, the scene information included in the alternative path is deconstructed using a scene recognition model to obtain the scene score corresponding to the alternative path. The predicted travel time of the target vehicle on the alternative path is determined using a duration prediction model based on the scene information included in the alternative path. The predicted energy consumption of the target vehicle on the alternative path is determined based on the predicted travel time and the vehicle status of the target vehicle. Finally, the total cost of the alternative path is determined based on the scene score corresponding to the alternative path, the predicted travel time, and the predicted energy consumption.

[0027] In this embodiment of the invention, during the process of determining the scene score corresponding to each candidate path, the scene information included in the candidate path can be deconstructed using a scene recognition model to obtain the scene score corresponding to the candidate path. The scene score can characterize the traffic efficiency of the candidate path; the higher the scene score, the higher the traffic efficiency of the candidate path. The scene recognition model can be a machine learning-based semantic model, which deconstructs the candidate path into multiple road segments, analyzes the traffic efficiency of each road segment based on the semantics in the scene information of each road segment, and obtains the scene score of the candidate path. The scene recognition model can also be a logical matching model, which presets basic scene regions, deconstructs the scene information in the candidate path, determines the scene region corresponding to the scene information in the preset scene as the scene deconstruction result of the candidate path, and further matches the scene score of the candidate path according to the correspondence between the preset scene regions and scene scores.

[0028] In this embodiment of the invention, during the process of determining the predicted travel time of alternative routes, a time prediction model can be used to determine the predicted travel time of the target vehicle traveling on the alternative routes based on the scene information included in the alternative routes. The time prediction model can be a machine learning model trained on a large number of historical parking events of vehicles, such as a gradient boosting decision tree model or a temporal convolutional network model. The historical parking events of a vehicle can include the target key area in the scene information that the vehicle passes through sequentially during parking, the start event of the parking event, the actual parking trajectory of the vehicle, and the deviation between the actual parking trajectory and the planned parking trajectory. The time prediction model learns the actual travel time patterns of vehicles in different time periods and scene areas based on historical parking events, thereby predicting the travel time required for the target vehicle to travel on the alternative routes in the corresponding time periods based on the scene information included in the alternative routes, thus obtaining the predicted travel time corresponding to the alternative routes.

[0029] In this embodiment of the invention, during the process of determining the predicted energy consumption of alternative routes, the predicted energy consumption of the target vehicle traveling the alternative routes can be determined based on the predicted travel time and the vehicle status of the target vehicle. The vehicle status may further include the target vehicle's system status, such as charging status or automatic parking status, and may also include the target vehicle's energy consumption type and energy consumption per unit time. Based on the target vehicle's system status, energy consumption type, and energy consumption per unit time, the predicted energy consumption of the target vehicle within the predicted travel time can be obtained as the predicted energy consumption of the alternative routes.

[0030] In this embodiment of the invention, for each of the N alternative paths, after determining the scenario score, predicted travel time, and predicted energy consumption, a comprehensive quantitative evaluation index, namely the total cost, can be used to characterize the overall driving cost of each alternative path based on the scenario score, predicted travel time, and predicted energy consumption corresponding to the alternative path. The lower the total cost value, the better the alternative path is overall and the more suitable it is for the target vehicle to travel. After obtaining the N total costs corresponding to the N alternative paths, step 130 can be executed.

[0031] Step 130: Based on the N total costs corresponding to the N alternative paths, determine the target parking path from the N alternative paths, and issue the target parking path to the target vehicle so that the target vehicle updates the designated parking path according to the target parking path.

[0032] In this embodiment of the invention, after obtaining N total costs corresponding to N alternative paths, a target parking path can be determined from the N alternative paths based on the N total costs. For example, the alternative path with the lowest total cost can be selected as the target parking path. The target parking path is then sent to the target vehicle so that the target vehicle updates its designated parking path according to the target parking path and parks according to the updated designated parking path (i.e., the target parking path).

[0033] During the target vehicle's journey, the cloud-based system evaluates each alternative path based on scenario scoring, predicted travel time, and predicted energy consumption, thus integrating multiple factors to determine the total cost of each alternative path. Based on the total cost of each of the N alternative paths, the cloud can determine the most suitable target parking path for the target vehicle and distribute it to the target vehicle, enabling it to update its designated parking path. This achieves dynamic, real-time, and global optimization of parking paths, significantly improving the adaptability, efficiency, and economy of the parking process.

[0034] In this embodiment of the invention, the target parking space is located within the target parking lot indicated by the parking request. The target parking lot can be obtained by analyzing the parking destination and navigation information near the parking destination in the parking request, or it can be the parking destination itself in the parking request. To better reduce the time and energy consumption of the target vehicle during the parking process, this embodiment of the invention also proposes a linkage between the cloud and the entrance gate of the target parking lot during the generation of alternative routes. The target entrance gate is determined based on the traffic flow at the entrance gate of the target parking lot, and alternative routes are actively generated based on the target entrance gate. This not only reduces the waiting time of the target vehicle at the entrance gate, but also achieves a balanced distribution of traffic flow throughout the entire target parking lot from the source, avoiding local congestion and improving the operating efficiency of the target parking lot system.

[0035] In embodiments of the present invention, such as Figure 1Step 110, as shown, in response to a parking request from a target vehicle, generates N alternative paths from the current location of the target vehicle to a target parking space. This may include: in response to the parking request, obtaining available parking space data sent by the parking lot's terminal equipment from the target vehicle; the available parking space data includes the number and / or location of unoccupied parking spaces in the target parking lot; based on the current location and the available parking space data, determining a target parking space for the target vehicle from the target parking lot, and sending a scheduling instruction to the terminal equipment. The dispatch instruction instructs the site-end equipment to modify the status of the target parking space; sends an entrance gate coordination instruction to the M entrance gates of the target parking lot to obtain the waiting traffic flow at the M entrance gates of the target parking lot; determines the target entrance gate from the M entrance gates based on the waiting traffic flow at the M entrance gates; and generates N alternative paths based on the current location, the target entrance gate, and the target parking space, each alternative path including the target entrance gate and the target parking space, where M is any positive integer.

[0036] In this embodiment of the invention, in response to a parking request, the cloud obtains the available parking space data sent to the target vehicle by the parking lot's terminal equipment. The terminal equipment may include the target parking lot's central management equipment, recording the status and quantity of all parking spaces within the target parking lot. The available parking space data sent to the target vehicle by the terminal equipment may include the number and / or location of unoccupied parking spaces within the target parking lot, representing the current number and / or location of available parking spaces. If the number of parking spaces in the available parking space data is not zero, the cloud can select the parking space closest to the target vehicle as the target parking space for parking the target vehicle, based on the target vehicle's current location and the parking space locations in the available parking space data. After determining the target parking space, the cloud can send a scheduling command to the terminal equipment, instructing the terminal equipment to modify the status of the target parking space from unoccupied to occupied, preventing the target parking space from being prematurely occupied by other vehicles.

[0037] In this embodiment of the invention, after determining the target parking space, an entrance gate coordination command can be sent to the M entrance gates of the target parking lot. This command wakes up the M entrance gates from a waiting state to an active cooperation state, enabling them to actively cooperate with the cloud and send the waiting traffic flow at the M entrance gates of the target parking lot to the cloud. The waiting traffic flow is the number of vehicles currently queuing at the entrance gates. The cloud can also obtain the current queue length and entrance status from the M entrance gates. The current queue length can be determined by the number of queuing vehicles and the length of vehicles currently queuing. The entrance status can also be determined by the number of queuing vehicles; different numbers of queuing vehicles correspond to different entrance statuses. After obtaining the waiting traffic flow at the M entrance gates, the cloud can execute the following... Figure 2 The concept shown is to determine the target entrance gate from M entrance gates.

[0038] In an embodiment of the present invention, Figure 2 This illustrates the technical concept of selecting a target entrance gate according to one embodiment of the present invention. For example... Figure 2 As shown, when M is 2, the waiting traffic flow at entrance gate 1 is 5 vehicles, so the entrance status of entrance gate 1 can be "busy". At entrance gate 2, the waiting traffic flow is 2 vehicles, so the entrance status of entrance gate 2 can be "unobstructed". Entrance gates 1 and 2 respond to the entrance gate coordination command from the cloud and send their waiting traffic flow to the cloud. After receiving the waiting traffic flow from each entrance gate, the cloud can determine the target entrance gate according to a preset entrance gate selection strategy. Table 1 below shows an example of selecting an entrance gate based on a preset entrance gate selection strategy according to an embodiment of the present invention.

[0039] Table 1

[0040] like Figure 2 As shown in Table 1, the entrance gate with the smallest waiting traffic flow is selected from the M entrance gates as the target entrance gate. If multiple entrance gates have the same waiting traffic flow (number of vehicles in queue), they can be randomly assigned or assigned based on the selection data of entrance gates in historical parking events. Figure 2 As shown, if the queue size of entrance gate 1 is greater than that of entrance gate 2, the target vehicle is recommended to use entrance gate 1; if the queue size of entrance gate 1 is less than that of entrance gate 2, the target vehicle is recommended to use entrance gate 2; if the queue size of entrance gate 1 and entrance gate 2 are the same, the vehicle can be randomly assigned or assigned based on historical data (entrance gate selection data in historical parking events). After determining the target entrance gate, the cloud can also issue a recommendation command "Please enter through the target entrance gate" to the target vehicle, and the corresponding content can be displayed on the vehicle's in-vehicle screen.

[0041] In this embodiment of the invention, after determining the target entrance gate and the target parking space, navigation information can be used to generate N alternative routes based on the current location, the target entrance gate, and the target parking space. Each alternative route includes the target entrance gate and the target parking space. During the generation of alternative routes, road condition video uploaded by the vehicle can also be combined. The road condition video may include real-time environmental perception data, such as radar data and / or point cloud images. The road condition video is fused with parking lot video from the parking lot itself, using the road condition video to compensate for obstacles not captured by the parking lot equipment, thereby better analyzing the obstacle situation of the alternative routes and generating alternative routes that are more conducive to the passage of the target vehicles.

[0042] In this embodiment of the invention, in step 120, during the process of deconstructing the scene information included in each candidate path using a scene recognition model to obtain the scene score corresponding to the candidate path, the process may include: identifying the candidate path using the scene recognition model to determine Y target scene regions included on the candidate path; labeling the Y target scene regions with Y target scene tags on the candidate path, wherein the scene recognition model is trained based on X preset scene tags, and the Y target scene tags belong to the X scene tags; issuing a field-end collaboration command to the field-end device to obtain parking lot video of the target parking lot from the field-end device; and obtaining video from the field-end device. The system collects road condition videos from the sensors of the target vehicle; for each of the Y target scene areas, it determines the target scene video corresponding to the target scene area from the parking lot video and / or the road condition video; it injects the target scene tags corresponding to the target scene areas into the Y target scene videos; it analyzes the Y target scene videos through the scene recognition model to determine the vehicle congestion and / or pedestrian density in the Y target scene areas, and obtains the risk scores corresponding to the Y target scene areas; based on the risk scores corresponding to the Y target scene areas, it determines the scene score corresponding to the alternative path, where X and Y are both arbitrary positive integers, and X is not less than Y.

[0043] In this embodiment of the invention, the scene recognition model can be trained using X preset scene labels as semantic labels. During use, the scene recognition model can perform refined scene deconstruction on each candidate path, using computer vision technology to identify Y target scene regions on the candidate paths. Target scene regions can be key nodes on the candidate paths involving the separation of pedestrians and / or vehicles, such as elevator entrances in a target parking lot or intersections on the candidate paths. Target scene regions can also be scheduling nodes on the candidate paths involving no-passage restrictions, such as temporary detours on road maintenance sections. After the scene recognition model identifies Y target scene regions, the Y target scene labels corresponding to these Y target scene regions can be used as semantic labels to annotate the candidate paths.

[0044] In this embodiment of the invention, the parking lot-side equipment may include cameras installed inside and outside the parking lot to collect parking lot video within a radius of the parking lot. During the scene scoring process for alternative routes, a parking lot-side coordination command can be issued to the parking lot-side equipment, facilitating its transition from a waiting state to a working state cooperating with the cloud. This allows the equipment to send parking lot video within a radius of the target parking lot to the cloud and fuse road condition video collected by sensors from the target vehicle, thus obtaining better real-time video data on the alternative routes. Based on feature points and / or location information from navigation data, the target scene video corresponding to the target scene area is determined from the parking lot video and / or the road condition video. Specifically, if the target scene area is located within the target parking lot, the target scene video can be determined from the parking lot video; if the target scene area is located outside the target parking lot, the target scene video can be determined from the road condition video or navigation data sent by the target vehicle, compensating for the lack of information about congestion near the target parking lot due to the target vehicle's lack of awareness of the real-time situation within the parking lot.

[0045] In this embodiment of the invention, after the scene recognition model labels Y target scene tags, the target scene video corresponding to each target scene tag can be injected into the scene recognition model. This allows the scene recognition model to evaluate the current status of alternative routes (such as pedestrian density and vehicle congestion) based on traffic flow data and / or pedestrian flow data in the target scene video. Furthermore, target scene areas corresponding to "high-risk areas" are assigned higher risk scores. The sum of the risk scores of the Y target scene areas is used as the scene score corresponding to the alternative route. This encourages the selection of "quieter" routes that bypass these areas during planning, improving traffic efficiency and reducing energy consumption from sudden stops and starts.

[0046] In this embodiment of the invention, after determining the risk score corresponding to each target scene area, the target scene area information with a risk score higher than the risk score threshold can also be sent as scene warning information to the target vehicle, so that the target vehicle can respond in a timely manner.

[0047] Figure 3 This illustrates a scene recognition concept for alternative paths provided by an embodiment of the present invention. For example... Figure 3 As shown, when N is 3, there are three alternative routes at the parking lot entrance: alternative route 1 (dashed line), alternative route 2 (dashed dot), and alternative route 3 (solid line). The area marked B-xxx on the map represents the parking space area. The scene recognition model can identify that alternative route 1 passes through a construction area where passage is prohibited and the elevator lobby, as shown by the on-site video, has a large flow of people. Therefore, alternative route 1 has a high scene score. For alternative route 2, the scene recognition model can identify that alternative route 2 passes through an intersection area, and the on-site video shows merging traffic, making it a medium-risk area with a medium scene score. Alternative route 3 is a clear passage, a low-risk area, and has a high scene score. Therefore, the scene recognition model prioritizes alternative route 3, which passes through parking spaces B-209 and B-206, and can also select the target parking space from within it.

[0048] In this embodiment of the invention, step 120, in which the predicted travel time of the target vehicle on the alternative path is determined by the time prediction model based on the scene information included in the alternative path, may include: obtaining the vehicle status of the target vehicle, the vehicle status including vehicle hardware parameters, the vehicle hardware parameters including at least the length, width, height, wheelbase, and minimum turning radius of the vehicle; predicting the travel time of the target vehicle in Y target scene areas based on the vehicle hardware parameters using the time prediction model; and obtaining the predicted travel time of the alternative path based on the Y travel times corresponding to the Y target scene areas.

[0049] In this embodiment of the invention, the duration prediction model can be a machine learning model trained based on a large number of historical parking events of vehicles, such as a gradient boosting decision tree model or a temporal convolutional network model, which transforms the implicit correlation between vehicle driving behavior and travel time into an identifiable explicit correlation. The parking event data on which the duration prediction model is trained contains a complete sequence of each parking event. Each sequence records the target scene area that the vehicle passes through in sequence, the parking start time, the environmental state shown by the vehicle sensor data, the deviation between the actual driving trajectory and the planned trajectory, the vehicle hardware parameters, the difference between the actual parking space and the recommended parking space, and the influence of the time environment. These data atomically record the identifier, travel time, and timestamp of each target scene area, which together constitute a set of decision sequences with spatiotemporal context and efficiency labels. Thus, real human driving data feeds back into the duration prediction model, enabling the duration prediction model to predict the estimated travel time of a certain route at any given time (such as a weekend evening), thereby proactively avoiding routes that may become congested in the future by predicting the travel time. The time-based prediction model learns the historical travel times of different vehicles' hardware parameters under different time periods and scenario labels, thus mastering the mapping relationship between key factors such as driving trajectory, node type, vehicle status, and travel time period and traffic efficiency. When determining the predicted travel time for alternative routes, the model first obtains the target vehicle's status, including its hardware parameters such as length, width, height, wheelbase, and minimum turning radius. Based on these hardware parameters and the target scenario areas included in the current alternative routes, the model predicts the travel time of the target vehicle in each target scenario area within the alternative routes during the current time period. Finally, the predicted travel times for all scenario areas are summed to obtain the predicted travel time for the entire alternative route.

[0050] In this embodiment of the invention, the vehicle status also includes the energy consumption type and energy consumption per unit time of the target vehicle. In step 120, determining the predicted energy consumption of the target vehicle traveling the alternative route based on the predicted travel time and the vehicle status of the target vehicle may include: determining the predicted energy consumption based on the energy consumption type and energy consumption per unit time of the target vehicle and the predicted travel time.

[0051] In this embodiment of the invention, the energy consumption type of the vehicle includes fuel type, pure electric type, and hybrid type, and the energy consumption per unit time represents the energy consumption required for the target vehicle to travel per unit time. The predicted energy consumption is further determined based on the energy consumption type of the target vehicle, the energy consumption per unit time, and the number of unit times included in the predicted travel time.

[0052] In this embodiment of the invention, the process of determining the total cost of the alternative path in step 120 based on the scene score, predicted travel time, and predicted energy consumption of the alternative path may include: determining the driving style of the target vehicle based on the vehicle hardware parameters, energy consumption type, and energy consumption per unit time of the target vehicle; determining the weight ratio corresponding to the target vehicle based on the driving style of the target vehicle; and determining the total cost of the alternative path based on the weight ratio, the scene score, predicted travel time, and predicted energy consumption of the alternative path.

[0053] In this embodiment of the invention, the weighting ratio includes at least the weight corresponding to the scene score, the weight corresponding to the predicted travel time, and the weight corresponding to the predicted energy consumption. Based on the target vehicle's hardware parameters, energy consumption type, energy consumption per unit time, and driving data from sensor data, such as the number of braking actions and steering wheel adjustments during historical driving, the driving style of the target vehicle is determined. For example, if the target vehicle's hardware parameters indicate it is a small sedan with a pure electric energy consumption type, high energy consumption per unit time, and frequent braking, its driving style can be determined to be aggressive. To avoid safety accidents, the weighting ratio for the target vehicle includes higher weights for the scene score and the predicted travel time, while the weight for the predicted energy consumption can be lower. Finally, a weighted sum is performed based on the weighting ratio, the scene score of the alternative path, the predicted travel time, and the predicted energy consumption to determine the total cost of the alternative path.

[0054] In this embodiment of the invention, the cloud quantifies the path selection for parking of the target vehicle into a total cost based on the scenario score representing scenario risk, the predicted passage time representing passage duration, the predicted energy consumption representing energy consumption, and the selection of the target entrance gate. The cloud also determines the weighting ratio in the total cost calculation process based on the driving style of the target vehicle, which facilitates the determination of a suitable target parking path for the target vehicle and improves the accuracy and adaptability of the target parking path.

[0055] In this embodiment of the invention, if the vehicle is currently located in the initial parking lot, the cloud can also obtain the waiting traffic flow at the exit gate of the initial parking lot from the exit gate of the initial parking lot, select the target exit gate, and generate alternative routes for the target vehicle from its parking space to the target exit gate based on the location of the target vehicle. Then, based on the parking lot video collected by the site equipment of the initial parking lot, the cloud can analyze the designated parking route of the target vehicle in the initial parking lot.

[0056] In an embodiment of the present invention, Figure 4 This illustrates a deconstructed concept of a parking path generation system provided by an embodiment of the present invention. For example... Figure 4As shown, the parking path generation system provided in this embodiment of the invention includes: The target vehicle is used to park according to a specified parking path, and during the parking process of the target vehicle, a parking request is sent to the cloud. In the cloud, in response to a parking request from a target vehicle, N alternative routes are generated from the target vehicle's current location to the target parking space. For each of the N alternative routes, a scene recognition model is used to deconstruct the scene information included in the alternative route to obtain a scene score corresponding to the alternative route. A duration prediction model is used to determine the predicted travel time of the target vehicle traveling on the alternative route based on the scene information included in the alternative route. Based on the predicted travel time and the vehicle status of the target vehicle, the predicted energy consumption of the target vehicle traveling on the alternative route is determined. Based on the scene score, predicted travel time, and predicted energy consumption corresponding to the alternative route, the total cost of the alternative route is determined. Based on the N total costs corresponding to the N alternative routes, the target parking route is determined from the N alternative routes and the target parking route is issued to the target vehicle. The target vehicle is further configured to, upon receiving the target parking path from the cloud, update the designated parking path according to the target parking path and park according to the updated designated parking path.

[0057] In this embodiment of the invention, the target vehicle may be as follows: Figure 4 As shown, the vehicle is equipped with onboard sensors, such as cameras, radar, and GPS, to collect sensor data. The target vehicle may also include a communication module for communicating with cloud and / or site-side equipment, and a vehicle control unit for controlling the target vehicle to park according to a specified parking path. The cloud may be configured with a cloud server and include a path planning algorithm to generate N alternative paths from the current location of the target vehicle to the target parking space. The algorithm then uses a big data analysis platform to determine the total cost of each alternative path and thus the target parking path. The cloud may also include a data storage center to store all data generated during the implementation of the parking path generation method provided in this embodiment of the invention.

[0058] In this embodiment of the invention, the target vehicle sends a parking request, its real-time location, sensor data, and vehicle status to the cloud. The cloud can then send the target parking path and scene warnings for high-risk areas along that path to the target vehicle.

[0059] In embodiments of the present invention, such as Figure 4The system further includes: a parking lot terminal device, installed at the target parking lot indicated by the parking request, wherein the parking lot terminal device is configured to: send the available parking space data to the target vehicle when the target vehicle retrieves the available parking space data of the target parking lot, the available parking space data including the number and / or location of the unoccupied parking spaces in the target parking lot; the target vehicle is further configured to: retrieve the available parking space data from the parking lot terminal device in response to the parking request; and generate an initial parking path from the target vehicle to the target parking lot when the number of parking spaces is not zero, and use the initial parking path as the designated parking path.

[0060] In this embodiment of the invention, the parking lot terminal device includes sensors inside the parking lot, such as cameras, geomagnetic sensors, or infrared sensors of various forms. The terminal device may also include a terminal server for processing information and communication equipment for communicating with other devices. In response to a parking request, the target vehicle retrieves available parking space data from the terminal device. If the target vehicle retrieves available parking space data from the target parking lot, the terminal device sends the available parking space data to the target vehicle. If the number of available parking spaces is not zero, the target vehicle generates an initial parking path from the target vehicle to the target parking lot and uses this initial parking path as the designated parking path for parking. Simultaneously, the parking request and the available parking space data from the terminal device are sent to the cloud.

[0061] In this embodiment of the invention, the field terminal device is further configured to: in response to a field terminal coordination instruction from the cloud, send a parking lot video of the target parking lot to the cloud; the cloud is further configured to, upon receiving the parking lot video, determine the target scene video corresponding to the target scene area from the parking lot video.

[0062] In this embodiment of the invention, the field-end device, responding to field-end collaboration commands from the cloud, can send parking lot video to the cloud. This allows the cloud to determine the target scene video corresponding to the target scene area if the target scene area is located within the target parking lot upon receiving the video. The field-end device can also send a panoramic view of the parking spaces within the parking lot to the cloud, including the status, location, and number of all parking spaces in the target parking lot. Furthermore, the field-end device can send historical parking events within the target parking lot to the cloud for use in training a duration prediction model.

[0063] In this embodiment of the invention, the cloud can also send the target parking path to the parking lot equipment, allowing the parking lot equipment to pre-plan the traffic conditions of the target vehicles during subsequent calculations. After confirming the target parking space in the cloud, a scheduling command can also be issued to the parking lot equipment, allowing the parking lot equipment to modify the status of the target parking space.

[0064] In embodiments of the present invention, such as Figure 4 As shown, the system further includes: M entrance gates, installed in the target parking lot, wherein the M entrance gates are used to: respond to an entrance gate coordination instruction from the cloud, send the waiting traffic flow at the M entrance gates of the target parking lot to the cloud; the cloud is also used to determine the target entrance gate from the M entrance gates based on the waiting traffic flow at the M entrance gates.

[0065] In this embodiment of the invention, the entrance gate may include a vehicle detection sensor for detecting the number of waiting vehicles at the entrance gate. The entrance gate may also include a communication module for communicating with other devices in the parking path generation system. Among the M entrance gates, each gate can perform vehicle flow counting and queue detection to obtain the waiting vehicle flow, and send the waiting vehicle flow, queue length, and entrance status to the cloud, so that the cloud can perform actions such as... Figure 2 The steps shown identify the target entrance gate from among M entrance gates. After confirming the target entrance gate, the cloud can send entrance recommendation instructions and traffic control signals to the entrance gate, allowing the entrance gate to preset the waiting position for the target vehicle and ensure its passage.

[0066] To better understand the parking path generation method provided in this embodiment of the invention, examples are given below. It should be understood that these examples are not intended to be limiting. Figure 5 This illustrates the concept of a parking path generation method provided by one embodiment of the present invention. For example... Figure 5 As shown, in the parking path generation method provided in this embodiment of the invention, the target vehicle first responds to the parking request by retrieving available parking space data from the parking lot equipment. If the number of parking spaces in the available parking space data is not zero and there are available parking spaces in the target parking lot, the target vehicle generates an initial parking path as the designated parking path and parks according to the designated parking path. Subsequently, the cloud receives the parking request, obtains the waiting traffic flow of M entrance gates, and executes the following based on the waiting traffic flow of the M entrance gates: Figure 2The steps shown involve identifying the target entrance gate from among M entrance gates and generating N alternative routes from the current location of the target vehicle, through the target entrance gate, to the target parking space. Each alternative route is then processed in parallel. The parallel processing of alternative routes in the cloud primarily includes scene recognition, duration prediction, and energy consumption prediction. During scene recognition in the cloud, a scene recognition model can identify target scene areas on the alternative routes. If the target scene area is located within the target parking lot, risk identification can be performed using parking lot video transmitted by the parking lot equipment. If the target scene area is outside the target parking lot, risk identification can be performed using navigation data and sensor data transmitted by the target vehicle, thus obtaining a risk score for each target scene area and a scene score for each alternative route. During duration prediction in the cloud, a duration prediction model can predict the passage time for each target scene area, thus obtaining the predicted passage time for each alternative route. In the process of predicting energy consumption for alternative routes in the cloud, the predicted energy consumption can be obtained by considering the target vehicle's status, energy consumption type, and energy consumption per unit time, combined with the predicted travel time. After analyzing the alternative routes from multiple dimensions, the total cost is obtained by quantifying and scoring the alternative routes through a comprehensive optimization objective function. The objective function includes a weight W1 corresponding to the risk score, a weight W2 corresponding to the predicted travel time, and a weight W3 corresponding to the predicted energy consumption. W1, W2, and W3 are weight ratios determined based on the target vehicle's driving style. Subsequently, the alternative route with the minimum total cost is selected from N alternative routes as the target parking route, and the target parking route is sent to the target vehicle. The target vehicle updates its designated parking route according to the target parking route, and this process is repeated to ensure that the target vehicle drives on the designated parking route with the minimum total cost until the target vehicle parks in the target parking space.

[0067] Specifically, for example, suppose the scenario is a weekend evening, and a vehicle with its valet parking function enabled drives into the underground parking lot of a large shopping mall.

[0068] During the request and initialization phase, the vehicle sends a parking request. The cloud system receives the request, confirms the target parking lot, and obtains real-time parking space information (such as...). Figure 3 Parking space B-205 in section B shown is vacant.

[0069] During the data fusion and route generation phase, the entrance gates coordinate, and the cloud checks the real-time data of all entrance gates, finding that there are 5 vehicles queuing at entrance 1 and no queue at entrance 2. The system immediately recommends to the driver via the vehicle's screen or app: "Please enter through entrance 2." Subsequently, the cloud generates alternative routes. Assuming the vehicle enters through entrance 2, the system generates 3 alternative routes to B-205 based on the parking lot map of the target parking lot (which can be as follows: Figure 3 The alternative paths 1, 2, and 3 are shown.

[0070] In the scene classification and prediction stage, analysis of field-end cameras reveals, for example... Figure 3 Alternative route 1, as shown, requires passing through elevator hall number 3. Current parking lot video image recognition indicates a large crowd gathering at this location, resulting in a high risk score for the scenario. Figure 3 Alternative path 2 shown is a remote loop, with unobstructed traffic and a low risk score. For example... Figure 3 Alternative route 3, as shown, requires traversing a narrow two-way road with oncoming traffic, and has a medium risk score. The features of the three alternative routes (length, number of turns, scene label, and current time "weekend evening") are input into a pre-trained duration prediction model. The output of the duration prediction model is as follows: Figure 3 The estimated travel time for alternative route 1 shown is 180 seconds. Figure 3 The estimated travel time for alternative route 2 shown is 150 seconds. Figure 3 The estimated travel time for alternative route 3 shown is 200 seconds.

[0071] In the decision-making and planning phase, the objective function is calculated by comprehensively optimizing it. Total cost ( Figure 3 Alternative path 1) = W1 * 180 + W2 * (high risk) + W3 * (high energy consumption) = high cost; Total cost ( Figure 3 Alternative path 2) = W1*150 + W2*(low risk) + W3*(medium energy consumption) = low cost; Total cost ( Figure 3 Alternative path 3) = W1*200 + W2*(medium risk) + W3*(high energy consumption) = medium to high cost.

[0072] Therefore, the cloud was ultimately selected. Figure 3 Alternative route 2 is the target parking route. Although Figure 3 Alternative path 2 has the lowest geometric distance, but its total cost is the lowest. Ultimately... Figure 3 Alternative route 2 is sent to the target vehicle, which then automatically drives along this route to the target parking space. The entire process is efficient, smooth, and energy-saving.

[0073] The parking path generation method provided in this invention dynamically avoids high-risk scenarios in the target area by fusing multi-source data from site-end equipment and target vehicles, and conducting real-time risk assessment to ensure the driving safety of target vehicles. Simultaneously, it predicts traffic conditions at different times based on a duration prediction model trained on historical data, achieving forward-looking planning. Furthermore, the cloud can personalize paths based on vehicle performance and driving style, and optimize global traffic flow distribution from the entrance end by combining entrance gate coordination and traffic monitoring. Through smooth path design and energy consumption estimation models, unnecessary starts, stops, and turns are reduced, significantly lowering vehicle energy consumption while improving safety.

[0074] In an exemplary embodiment, the cloud may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0075] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory including instructions that can be executed by a processor of a device to perform the described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. This non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of an electronic device, enables the cloud to execute... Figure 1 The method shown.

[0076] This application also provides a computer program product, including a computer program, which, when executed by a processor, performs... Figure 1 The method shown.

[0077] The above description does not provide detailed technical specifications regarding the structure of each layer. However, those skilled in the art should understand that various technical means can be used to form layers and regions of the desired shape. Furthermore, to achieve the same structure, those skilled in the art can design methods that are not entirely identical to those described above. Additionally, although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be advantageously combined.

[0078] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0079] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating a parking path, characterized in that, Applied to the cloud, the method includes: During the process of the target vehicle parking according to the specified parking path, in response to the parking request from the target vehicle, N alternative paths from the current location of the target vehicle to the target parking space are generated. For each of the N alternative paths, the scene information included in the alternative path is deconstructed using a scene recognition model to obtain the scene score corresponding to the alternative path. The predicted travel time of the target vehicle on the alternative path is determined using a time prediction model based on the scene information included in the alternative path. The predicted energy consumption of the target vehicle on the alternative path is determined based on the predicted travel time and the vehicle status of the target vehicle. Finally, the total cost of the alternative path is determined based on the scene score corresponding to the alternative path, the predicted travel time, and the predicted energy consumption. Based on the N total costs corresponding to the N alternative paths, a target parking path is determined from the N alternative paths, and the target parking path is issued to the target vehicle so that the target vehicle updates the designated parking path according to the target parking path.

2. The method as described in claim 1, characterized in that, The target parking space is located within the target parking lot indicated by the parking request; In response to a parking request from the target vehicle, N alternative paths are generated from the current location of the target vehicle to the target parking space, including: In response to the parking request, the system obtains available parking space data sent by the parking lot terminal equipment of the target parking lot from the target vehicle. The available parking space data includes the number of unoccupied parking spaces and / or the location of the parking spaces in the target parking lot. Based on the current location and the available parking space data, a target parking space for parking the target vehicle is determined from the target parking lot, and a dispatch instruction is sent to the parking lot equipment, which instructs the parking lot equipment to modify the status of the target parking space. Send entrance gate coordination instructions to the M entrance gates of the target parking lot to obtain the waiting traffic flow at the M entrance gates of the target parking lot; Based on the waiting traffic flow at the M entrance gates, determine the target entrance gate from the M entrance gates; Based on the current location, the target entrance gate, and the target parking space, N alternative routes are generated, and each alternative route includes the target entrance gate and the target parking space.

3. The method as described in claim 2, characterized in that, The step of deconstructing the scene information included in the alternative paths through a scene recognition model to obtain the scene score corresponding to the alternative paths includes: The alternative paths are identified using the scene recognition model, and Y target scene regions included on the alternative paths are determined. The Y target scene labels corresponding to the Y target scene regions are marked on the candidate path, wherein the scene recognition model is trained based on X preset scene labels, and the Y target scene labels belong to the X scene labels; Send a field-end collaboration command to the field-end device to obtain the parking video of the target parking lot from the field-end device; Acquire road condition videos collected by sensors from the target vehicle; For each of the Y target scene regions, determine the target scene video corresponding to the target scene region from the parking lot video and / or the road condition video; Inject the Y target scene videos into the target scene tags corresponding to the target scene regions; By analyzing the Y target scene videos using the scene recognition model, the vehicle congestion and / or pedestrian density within the Y target scene areas are determined, and risk scores are obtained for the Y target scene areas. Based on the risk scores corresponding to the Y target scene areas, the scene score corresponding to the alternative path is determined.

4. The method as described in claim 3, characterized in that, The step of determining the predicted travel time of the target vehicle on the alternative routes based on the scenario information included in the alternative routes using the time prediction model includes: The vehicle status of the target vehicle is obtained, and the vehicle status includes vehicle hardware parameters, which include at least the vehicle's length, width, height, wheelbase, and minimum turning radius. The time prediction model predicts the travel time of the target vehicle in Y target scene areas based on the vehicle hardware parameters; Based on the Y travel times corresponding to the Y target scene areas, the predicted travel time of the alternative paths is obtained.

5. The method as described in claim 4, characterized in that, The vehicle status also includes the target vehicle's energy consumption type and energy consumption per unit time. Determining the predicted energy consumption of the target vehicle traveling the alternative route based on the predicted travel time and the vehicle status of the target vehicle includes: The predicted energy consumption is determined based on the energy consumption type of the target vehicle, the energy consumption per unit time, and the predicted travel time.

6. The method as described in claim 5, characterized in that, The step of determining the total cost of the alternative paths based on the scenario score, predicted travel time, and predicted energy consumption corresponding to the alternative paths includes: The driving style of the target vehicle is determined based on its vehicle hardware parameters, energy consumption type, and energy consumption per unit time. Based on the driving style of the target vehicle, determine the corresponding weight ratio for the target vehicle; The total cost of the alternative path is determined based on the weighting ratio, the scene score corresponding to the alternative path, the predicted travel time, and the predicted energy consumption.

7. A parking path generation system, characterized in that, The system includes: The target vehicle is used to park according to a specified parking path, and during the parking process of the target vehicle, a parking request is sent to the cloud. In the cloud, in response to a parking request from a target vehicle, N alternative routes are generated from the target vehicle's current location to the target parking space. For each of the N alternative routes, a scene recognition model is used to deconstruct the scene information included in the alternative route to obtain a scene score corresponding to the alternative route. A duration prediction model is used to determine the predicted travel time of the target vehicle traveling on the alternative route based on the scene information included in the alternative route. Based on the predicted travel time and the vehicle status of the target vehicle, the predicted energy consumption of the target vehicle traveling on the alternative route is determined. Based on the scene score, predicted travel time, and predicted energy consumption corresponding to the alternative route, the total cost of the alternative route is determined. Based on the N total costs corresponding to the N alternative routes, the target parking route is determined from the N alternative routes and the target parking route is issued to the target vehicle. The target vehicle is further configured to, upon receiving the target parking path from the cloud, update the designated parking path according to the target parking path and park according to the updated designated parking path.

8. The system as described in claim 7, characterized in that, The system further includes: a parking lot terminal device, installed at the target parking lot indicated by the parking request, wherein the parking lot terminal device is used for: When the target vehicle retrieves the available parking space data of the target parking lot, the available parking space data is sent to the target vehicle. The available parking space data includes the number and / or location of the parking spaces that are not occupied in the target parking lot. The target vehicle is also configured to, in response to a parking request, retrieve the available parking space data from the parking lot equipment; and, if the number of parking spaces is not zero, generate an initial parking path from the target vehicle to the target parking lot, and use the initial parking path as the designated parking path.

9. The system as described in claim 8, characterized in that, The field-end equipment is also used for: In response to a site-to-site coordination command from the cloud, the parking lot video of the target parking lot is sent to the cloud. The cloud platform is also used to determine the target scene video corresponding to the target scene area from the parking lot video when the parking lot video is received.

10. The system as described in claim 8, characterized in that, The system further includes: M entrance gates, installed in the target parking lot, wherein the M entrance gates are used for: In response to the entrance gate coordination command from the cloud, the waiting traffic flow at the M entrance gates of the target parking lot is sent to the cloud. The cloud platform is also used to determine the target entrance gate from the M entrance gates based on the waiting traffic flow at the M entrance gates.

Images