A dynamic electronic lane planning method and device based on dynamic traffic flow

Abstract

The application discloses a dynamic electronic lane planning method and device based on dynamic traffic flow of V2X, and relates to the technical field of intelligent management and control systems.The method comprises the following steps: constructing an input feature vector based on the average number of vehicles and the average vehicle speed of each lane in each time slice; constructing an output feature vector based on the width proportion of artificially divided lanes; constructing a training set, a verification set and a test set based on historical traffic data of the preset period of the artificially divided lanes; constructing a shallow neural network model, including an activation function Softmax and a loss function D(P, Y); iteratively training parameters of the shallow neural network model; testing the neural network model after the parameter training is completed; and if the test reaches a preset correct rate, deploying the neural network model to a dynamic electronic lane planning environment.The application dynamically adjusts electronic lane planning of a current traffic environment in real time based on high-precision GNSS positioning and roadside equipment information, and realizes flexible and easy-to-maintain electronic lane effects in a variety of available scenarios.

Description

Technical Field

[0001] This application relates to the field of intelligent control system application technology, and in particular to a V2X dynamic electronic lane planning method and device based on dynamic traffic flow. Background Technology

[0002] To ensure smooth road traffic flow, traffic management departments divide roads into lanes, commonly by direction of travel, speed, and other specific requirements, such as opposing lanes and fast / slow lanes. Traditionally, traffic management departments plan lanes in advance and use special paint to demarcate lanes on the physical road surface, distinguishing the functions and usage rights of different areas on the same road. Lane division must comply with traffic regulations and standards while also meeting the actual traffic flow needs of the road segment. When vehicles travel according to traffic regulations on roads with fixed lanes divided according to these principles, the predetermined design goals of the lanes (restrictions on travel direction and speed) are indirectly achieved, simplifying traffic flow and preventing congestion and accidents.

[0003] While traditional fixed-lane systems have achieved significant results, they still have shortcomings. These mainly include: 1) Lane markings are printed on the road surface, requiring regular inspection and maintenance by designated personnel to ensure their usability, resulting in high maintenance costs; 2) In reality, the optimal lane marking method varies depending on factors such as weather conditions and traffic flow timeliness. Traditional fixed-lane systems are not conducive to dynamically adjusting lane marking methods and positions to adapt to short-term changes in road conditions; 3) In some special situations where painting fixed lane markings is unsuitable, but where fixed lanes are necessary to assist drivers in actual use, such as grasslands and rural roads, traditional lane markings are also ineffective.

[0004] Therefore, it is desirable to provide a V2X dynamic electronic lane planning method and device based on dynamic traffic flow. This method integrates high-precision GNSS satellite positioning information, lane marker information described by V2X communication methods, and real-time traffic conditions near the lanes perceived by roadside equipment, such as the driving status of all traffic participants, average vehicle speed, and average number of vehicles. This provides an automated dynamic electronic lane division method and device. The method automatically and dynamically adjusts the GNSS coordinates of the lane centerlines in the MAP messages of the V2X communication mechanism according to road conditions and traffic conditions, and broadcasts this information to surrounding vehicles with V2X communication capabilities, thereby presenting them with the effect of dynamic electronic lanes. Summary of the Invention

[0005] According to a first aspect of some embodiments of this application, a V2X dynamic electronic lane planning method based on dynamic traffic flow is provided, applied in a terminal (e.g., connected vehicles). The method may include S1: obtaining the total road width X0 and the number of lanes m, and setting a time slice interval n; constructing an input feature vector based on X0 and the average number of vehicles and average speed of each lane in each time slice. Includes 2m+1 attribute values; constructs the output feature vector based on the width ratio of manually divided lanes. Includes m attribute values; S2: Based on historical traffic data for a preset period of manually divided lanes, extract information from each row of fields in S1 to construct training, validation, and test sets; S3: Construct a shallow neural network model, where the activation function Softmax is used to determine that the sum of lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. S4: Based on the training set constructed in S2 and the loss function in S3, iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br, and perform overfit verification based on the validation set constructed in S2; S5: Based on the test set constructed in S2, test the neural network model after parameter training. If the test does not reach the preset accuracy, return to S4 for training; S6: If the test reaches the preset accuracy, deploy to the dynamic electronic lane planning environment.

[0006] In some embodiments, based on real-time road traffic data collected by roadside sensing devices, an input feature vector is constructed and input into a trained neural network model to obtain a prediction of the real-time division ratio of electronic lanes; the current total road width is allocated to each lane according to the predicted ratio, and the centerline coordinates of each lane are calculated based on the GNSS road centerline coordinates; the MAP message of the V2X communication message is updated according to the width and centerline coordinate information of each lane, and the encoded message is broadcast as dynamic electronic lane information to connected vehicles within a preset distance.

[0007] In some embodiments, when the number of lanes m = 3, the specific steps include obtaining the average number of vehicles passing through lane 1 (X1), the average speed of lane 1 (X2), the average number of vehicles passing through lane 2 (X3), the average speed of lane 2 (X4), the average number of vehicles passing through lane 3 (X5), and the average speed of lane 3 (X6) within each n-second time slice, and constructing an input feature vector X = [X0, X1, X2, X3, X4, X5, X6]; and constructing an output feature vector Y = [Y0, Y1, Y2] based on the manually divided width ratios Y0, Y1, and Y2 of lane 1, lane 2, and lane 3, where...

[0008] In some embodiments, constructing a shallow neural network model specifically includes:

[0009]

[0010] Where X0 to X6 are the input feature vectors defined in S1 that are strongly correlated with the lane width division ratio, P0, P1, and P2 are the target lane width distributions, and w ij To use real-time traffic features X j Calculate the width P of the i-th lane. i The required weighting factor, b i The offset required to calculate the width of the i-th lane is w. ij b i Determined through training.

[0011] In some embodiments, the activation function Softmax specifically includes:

[0012]

[0013] The activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, Z. i This is the output value of the i-th node in the last layer of the neural network.

[0014] In some embodiments, based on the characteristic that the model output has a probability distribution form P0+P1+P2=1, the loss function D(P,Y) specifically includes:

[0015]

[0016] The loss function D(P,Y) is based on lane width prediction. Cross-entropy with the proportion of manual division, Y i P represents the actual width ratio of road i. i The width ratio of road i is predicted by the model.

[0017] In some embodiments, the iterative training of the shallow neural network model parameters specifically includes using the backpropagation (BP) algorithm to iteratively train the neural network model parameters on a high-performance server based on the training set and the loss function.

[0018] In some embodiments, testing the neural network model that has completed parameter training to determine whether it has reached a preset accuracy rate specifically includes randomly sampling data from the test set for a preset period and dividing it into four groups according to traffic flow conditions: no congestion, slight congestion, moderate congestion, and heavy congestion, with 1000 data points in each group; using the neural network model that has completed parameter training to predict the accuracy of each data point in the four groups; if the highest probability item in the predicted lane usage probability distribution is the same as the lane usage actually marked manually, then the prediction is determined to be correct; otherwise, the prediction is determined to be incorrect; the accuracy rate of the prediction results for the four groups of data is calculated separately, where the accuracy rate is the ratio of the number of correct predictions in each group to the total number of test data in each group; if the accuracy rate is greater than the preset threshold for each group, then the actual usage requirements are determined to be met; otherwise, the actual usage requirements are determined not to be met.

[0019] In some embodiments, the accuracy rate is greater than the preset threshold corresponding to each group, specifically including 70% for no congestion, 75% for slight congestion, 80% for moderate congestion, and 85% for severe congestion.

[0020] According to a second aspect of some embodiments of this application, a V2X dynamic electronic lane planning device based on dynamic traffic flow is provided. The V2X dynamic electronic lane planning device includes a memory configured to store data and instructions; and a processor communicating with the memory. When executing instructions from the memory, the processor is configured to: S1: acquire the total road width X0 and the number of lanes m, and set a time slice interval n; construct an input feature vector based on X0 and the average number of vehicles and average speed of each lane in each time slice. Includes 2m+1 attribute values; constructs the output feature vector based on the width ratio of manually divided lanes. Includes m attribute values; S2: Based on historical traffic data for a preset period of manually divided lanes, extract information from each row of fields in S1 to construct training, validation, and test sets; S3: Construct a shallow neural network model, where the activation function Softmax is used to determine that the sum of lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. S4: Based on the training set constructed in S2 and the loss function in S3, iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br, and perform overfit verification based on the validation set constructed in S2; S5: Based on the test set constructed in S2, test the neural network model after parameter training. If the test does not reach the preset accuracy, return to S4 for training; S6: If the test reaches the preset accuracy, deploy to the dynamic electronic lane planning environment.

[0021] Therefore, the V2X dynamic electronic lane planning method and apparatus based on dynamic traffic flow according to some embodiments of this application, by fusing high-precision satellite positioning information (GNSS), lane marker information described by V2X communication methods, and traffic conditions near the lanes as perceived in real time by roadside equipment, such as the driving status of all traffic participants, average vehicle speed, and average number of vehicles, provides an automated dynamic electronic lane division method and apparatus. This method automatically and dynamically adjusts the GNSS coordinates of the markers used for the center lines of multiple lanes within the road in the MAP message of the V2X communication mechanism according to road conditions and traffic conditions, and broadcasts them to surrounding vehicles with V2X communication capabilities to achieve the effect of presenting dynamic electronic lanes. Attached Figure Description

[0022] To better understand and illustrate some embodiments of this application, the following description of the embodiments will be made in conjunction with the accompanying drawings, in which the same numerical designations indicate corresponding parts.

[0023] Figure 1 This is an exemplary schematic diagram of a V2X dynamic electronic lane planning system based on dynamic traffic flow, provided according to some embodiments of this application.

[0024] Figure 2 This is an exemplary flowchart of a V2X dynamic electronic lane planning method based on dynamic traffic flow, provided according to some embodiments of this application. Detailed Implementation

[0025] The following description with reference to the accompanying drawings provides an illustrative understanding of various embodiments of the present application as defined by the claims and their equivalents. These embodiments include various specific details for ease of understanding, but are to be considered exemplary only. Therefore, those skilled in the art will understand that various changes and modifications can be made to the various embodiments described herein without departing from the scope and spirit of the present application. Furthermore, for the sake of brevity and clarity, descriptions of well-known functions and structures will be omitted.

[0026] The terms and phrases used in the following description and claims are not limited to their literal meaning, but are intended only to enable a clear and consistent understanding of this application. Therefore, it will be understood by those skilled in the art that the description of various embodiments of this application is provided for illustrative purposes only and is not intended to limit the application as defined in the appended claims and their equivalents.

[0027] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of some embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0028] It should be noted that the terminology used in the embodiments of this application is for the purpose of describing specific embodiments only and is not intended to limit the application. The singular forms “a,” “an,” “an,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the linked listed items. The expressions “first,” “second,” “the first,” and “the second” are used to modify the corresponding elements without regard to order or importance, and are merely used to distinguish one element from another, without limiting the corresponding elements.

[0029] According to some embodiments of this application, the terminal may be a smart terminal, platform, equipment, and / or electronic device, etc.; the smart terminal may include a positioning device, etc. The platform may include a cloud platform, etc., and the platform may include a system platform composed of one or more electronic devices; the equipment may include an intelligent connected vehicle (ICV); the electronic device may include one or more of the following: personal computer (PC, such as tablet computer, desktop computer, laptop, netbook, PDA), client device, virtual reality device (VR), augmented reality device (AR), mixed reality device (MR), XR device, renderer, smartphone, mobile phone, e-book reader, portable multimedia player (PMP), audio / video player (MP3 / MP4), camera, and wearable device, etc. According to some embodiments of this application, the wearable device may include accessory type (e.g., watch, ring, bracelet, glasses, or head-mounted device (HMD)), integrated type (e.g., electronic clothing), decorative type (e.g., skin pad, tattoo, or built-in electronic device), etc., or a combination of several. In some embodiments of this application, the electronic device may be flexible and not limited to the above-mentioned devices, or may be one or more of the various devices mentioned above. In this application, the term "user" may refer to a person using an electronic device or a device using an electronic device (e.g., an artificial intelligence electronic device).

[0030] This application provides a V2X dynamic electronic lane planning method and apparatus based on dynamic traffic flow. To facilitate understanding of this application, the embodiments will be described in detail below with reference to the accompanying drawings.

[0031] Figure 1 This is an exemplary schematic diagram of a V2X dynamic electronic lane planning system based on dynamic traffic flow, provided according to some embodiments of this application. Figure 1 The V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may include a network 110, an information terminal 120, a user terminal 130, and a server 140. Specifically, the information terminal 120 and the user terminal 130 establish communication through the network; for example, the information terminal 120 and the user terminal 130 can communicate in the same local area network (e.g., a network environment with the same router). Furthermore, the information terminal 120 can connect to the network 110 via wired (e.g., network cable) or wireless (e.g., cloud server), and the user terminal 130 can establish a communication connection to the network 110 via wired or wireless (e.g., Wi-Fi). In some embodiments, the user terminal 130 can send high-precision satellite positioning information (GNSS) to the information terminal 120 and the server 140. Furthermore, the information terminal 120 and the server 140 can provide feedback on dynamic electronic lane information to the user terminal 130. As an example, server 140 and / or information terminal 120 can acquire data such as GNSS road centerline coordinates based on roadside equipment, or collect road traffic data in real time based on roadside sensing equipment. The roadside equipment (RSU) can sense road environment information, obstacle information, etc. through sensors, and the sensors may include, but are not limited to, cameras, lidar, millimeter-wave radar, etc.

[0032] According to some embodiments of this application, the information terminal 120 and the user terminal 130 can be the same or different terminal devices. The terminal devices may include, but are not limited to, smart terminals, cloud platforms, mobile terminals, and computers. In a dynamic electronic lane scenario, the information terminal 120 may include roadside equipment, and the user terminal 130 may include a positioning device. In some embodiments, the information terminal 120 and the user terminal 130 can be integrated into one device, for example, a positioning device integrating sensors. In some embodiments, the server 140 is a type of computer, possessing advantages such as faster operation and higher load capacity than ordinary computers, but correspondingly, a higher price. In a network environment, the server can provide computing or application services to other clients (e.g., PCs, smartphones, ATMs, and large equipment such as transportation systems). The server has high-speed CPU computing power, long-term reliable operation, powerful I / O external data throughput capabilities, and better scalability. The services that the server can provide include, but are not limited to, the ability to respond to service requests, provide services, and ensure service availability. As an electronic device, the server has an extremely complex internal structure, including an internal structure similar to that of a regular computer. For example, the internal structure of the server may include a central processing unit (CPU), hard disk, memory, system, system bus, etc.

[0033] In some embodiments of this application, the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may omit one or more components, or may further include one or more other components. For example, the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may include multiple user terminals 130, such as multiple connected vehicles. As another example, the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may include one or more information terminals 120. Yet another example, the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may include multiple servers 140, etc. In some embodiments, the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow may include, but is not limited to, a system based on dynamic electronic lane scenario planning. The network 110 can be any type of communication network, which may include computer networks (e.g., local area networks (LANs) or wide area networks (WANs)), the Internet, and / or telephone networks, or a combination thereof. In some embodiments, the network 110 can be other types of wireless communication networks. The wireless communication may include microwave communication and / or satellite communication, etc. The wireless communication may include cellular communication, such as Global System for Mobile Communications (GSM), Code Division Multiple Access (CDMA), 3rd Generation Telecommunication (3G), 4th Generation Telecommunication (4G), 5th Generation Telecommunication (5G), 6th Generation Telecommunication (6G), Long Term Evolution (LTE), LTE-Advanced (LTE-A), Wideband Code Division Multiple Access (WCDMA), Universal Mobile Telecommunications System (UMTS), Wireless Broadband (WiBro), etc., or a combination of several of these.In some embodiments, the user terminal 130 may be other equipment and / or electronic devices with equivalent functional modules, which may include one or a combination of virtual reality devices (VR), rendering machines, personal computers (PCs, such as tablets, desktop computers, laptops, netbooks, PDAs), smartphones, mobile phones, e-book readers, portable multimedia players (PMPs), audio / video players (MP3 / MP4), cameras, and wearable devices.

[0034] In some embodiments, the WIFI can be other types of wireless communication technologies. According to some embodiments of this application, the wireless communication may include Wireless Fidelity (WiFi), Bluetooth, Bluetooth Low Energy (BLE), ZigBee, Near Field Communication (NFC), Magnetic Secure Transmission, Radio Frequency and Body Area Network (BAN), or a combination thereof. According to some embodiments of this application, the wired communication may include Global Navigation Satellite System (GNSS), Global Positioning System (GPS), BeiDou Navigation Satellite System, or Galileo (European Global Navigation Satellite System), etc. The wired communication may include Universal Serial Bus (USB), High-Definition Multimedia Interface (HDMI), Recommended Standard 232 (RS-232), and / or Plain Old Telephone Service (POTS), or a combination thereof.

[0035] It should be noted that the above description of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow is for convenience only and should not be construed as limiting this application to the scope of the embodiments described. It is understood that those skilled in the art, based on the principles of this system, may arbitrarily combine the various components or connect them to other components to form subsystems without departing from these principles, making various modifications and changes in form and detail to the application areas of the above methods and systems. For example, the server 140 and / or the information terminal 120 can collect road traffic data in real time through roadside sensing devices, etc. Another example is that the information terminal 120 / user terminal 130 can be integrated into connected vehicles, etc. Such modifications are all within the protection scope of this application.

[0036] Figure 2 This is an exemplary flowchart of a V2X dynamic electronic lane planning method based on dynamic traffic flow, provided according to some embodiments of this application. Figure 2 The process 200 can be implemented using a V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the V2X dynamic electronic lane planning method 200 based on dynamic traffic flow can be started automatically or by instruction. The instruction may include system instructions, device instructions, user instructions, action instructions, or a combination of several.

[0037] In S1, 201: Obtain the total road width X0 and the number of lanes m, and set the time slice interval n; construct the input feature vector based on X0 and the average number of vehicles and average speed of each lane in each time slice. Includes 2m+1 attribute values; constructs the output feature vector based on the width ratio of manually divided lanes. It includes m attribute values. Operation 201 can be implemented by the server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the user terminal 130 can send high-precision satellite positioning information (GNSS) based on high-precision GNSS positioning. In some embodiments, the server 140 can construct input feature vectors and output feature vectors, etc. For example, the number of lanes m is an integer greater than or equal to 2; this embodiment only takes m = 3 as an example. As an example, when the number of lanes m = 3, it can include:

[0038] Obtain the average number of vehicles passing through lane 1 (X1), average vehicle speed of lane 1 (X2), average number of vehicles passing through lane 2 (X3), average vehicle speed of lane 2 (X4), average number of vehicles passing through lane 3 (X5), and average vehicle speed of lane 3 (X6) within each n-second time slice, and construct the input feature vector X = [X0, X1, X2, X3, X4, X5, X6]. The time slice interval n can be set according to different traffic flow, for example, including time slice intervals from 10 seconds to 240 seconds, such as 20 seconds, 60 seconds, etc.

[0039] Based on the manually defined width ratios Y0, Y1, and Y2 of lanes 1, 2, and 3, an output feature vector Y = [Y0, Y1, Y2] is constructed, where...

[0040] In step 202, S2: Based on historical traffic data during a preset period of manually divided lanes, information for each row is extracted according to the S1 field to construct a training set, a validation set, and a test set. Operation 202 can be implemented through the information terminal 120 and server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the information terminal 120 and server 140 can extract fields such as total road width, average number of vehicles per lane, average vehicle speed, and width ratio of manually divided lanes to extract historical traffic data and construct training sets, validation sets, and test sets.

[0041] In some embodiments, the preset period may include a time range of 1 to 24 months, such as 6 months, 12 months, etc. For example, based on 12 months of historical traffic data with manually divided lanes, information for each row is extracted according to the S1 field to construct training sets, validation sets, test sets, etc.

[0042] In section 203, S3: A shallow neural network model is constructed, where the activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. Cross-entropy with the manually assigned proportion. Operation 203 can be implemented by server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, server 140 can construct a shallow neural network model.

[0043] For example, constructing a shallow neural network model may include:

[0044]

[0045] Where X0 to X6 are the input feature vectors defined in S1 that are strongly correlated with the lane width division ratio, P0, P1, and P2 are the target lane width distributions, and wij To use real-time traffic features X j Calculate the width P of the i-th lane. i The required weighting factor, b i The offset required to calculate the width of the i-th lane is w. ij b i Determined through training.

[0046] According to some embodiments of this application, the activation function Softmax may include:

[0047]

[0048] The activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, Z. i This is the output value of the i-th node in the last layer of the neural network.

[0049] According to some embodiments of this application, based on the characteristic that the model output has a probability distribution form P0+P1+P2=1, the loss function D(P,Y) may include:

[0050]

[0051] The loss function D(P,Y) is based on lane width prediction. Cross-entropy with the proportion of manual division, Y i P represents the actual width ratio of road i. i The width ratio of road i is predicted by the model.

[0052] In step 204, S4: Based on the training set constructed in S2 and the loss function in S3, the parameters of the shallow neural network model are iteratively trained to determine the optimal weights and offset parameters of Wr and br, and overfitting verification is performed based on the validation set constructed in S2. Operation 204 can be implemented by the server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the server 140 can iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br.

[0053] According to some embodiments of this application, the iterative training of the shallow neural network model parameters may include using the BP algorithm to iteratively train the neural network model parameters on a high-performance server based on the training set and the loss function.

[0054] In step 205, S5: Based on the test set constructed in S2, the neural network model that has completed parameter training is tested. If the test does not reach the preset accuracy, the process returns to S4 for further training. Operation 205 can be implemented through the information terminal 120 and server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the server 140 can test the neural network model that has completed parameter training.

[0055] As an example, testing whether the neural network model that has completed parameter training reaches a preset accuracy rate can include randomly sampling data from the test set for a preset period and dividing it into four groups according to traffic flow conditions: no congestion, slight congestion, moderate congestion, and heavy congestion, with 1000 data points in each group. The neural network model that has completed parameter training is used to predict the accuracy of each data point in the four groups. If the highest probability item in the predicted lane usage probability distribution matches the lane usage actually designated manually, the prediction is considered correct; otherwise, it is considered incorrect. The accuracy rate of the prediction results for each of the four groups is calculated. The accuracy rate is the ratio of the number of correct predictions in each group to the total number of test data points in each group. If the accuracy rate is greater than a preset threshold for each group, the actual usage requirements are met; otherwise, the actual usage requirements are not met. The accuracy rate being greater than the preset threshold for each group can include 70% for no congestion, 75% for slight congestion, 80% for moderate congestion, and 85% for heavy congestion.

[0056] In steps 206 and S6: If the test achieves a preset accuracy rate, the system is deployed to the dynamic electronic lane planning environment. Operation 205 can be implemented through the information terminal 120 and server 140 of the V2X dynamic electronic lane planning system 100 based on dynamic traffic flow. In some embodiments, the information terminal 120 and server 140 can deploy the trained neural network model to the dynamic electronic lane planning environment.

[0057] According to some embodiments of this application, process 200 may further include constructing an input feature vector based on real-time road traffic data collected by roadside sensing devices. Input the trained neural network model to obtain a prediction of the real-time division ratio of electronic lanes. The current total road width is allocated to each lane according to the predicted ratio. Based on the GNSS road centerline coordinates, the centerline coordinates of each lane are calculated. According to the width and centerline coordinates of each lane, the MAP message of the V2X communication message is updated, and the encoded message is broadcast as dynamic electronic lane information to connected vehicles within a preset distance.

[0058] Specifically, in practical applications, roadside sensing devices (such as cameras and lidar) connected to the roadside device collect real-time road traffic information. The dynamic electronic lane planning device constructs an input feature vector from the acquired data based on the fields listed in S1, and inputs it into the trained model algorithm in the device via S7 to obtain a prediction of the real-time lane division ratio. Further, the current total road width M is allocated to lanes 1, 2, and 3 according to the ratio, and the centerline coordinates of lanes 1, 2, and 3 are calculated with reference to the known GNSS centerline coordinates. Based on the calculated width information and centerline coordinate information of each lane, the MAP message in the V2X communication message is updated, and the encoded message is broadcast to surrounding connected vehicles via the PC5 interface V2X communication channel as dynamic electronic lane information, ultimately realizing the function of dynamic electronic lane planning information service.

[0059] According to some embodiments of this application, the dynamic electronic lane information can be displayed in the user interface (UI) of the user terminal 130. The display scenario of the dynamic electronic lane information can include, but is not limited to, scene display through any or a combination of VR, AR, MR, and XR. As an example, connected vehicle users can obtain electronic lane information based on any or a combination of VR, AR, MR, and XR.

[0060] According to some embodiments of this application, this application only takes a road with 3 lanes as an example, but it can also be applied to other numbers of lanes. By adjusting the corresponding feature vector form and output vector form, and constructing and training the corresponding dedicated neural network model for the adjusted input and output vectors.

[0061] It should be noted that the above description of process 200 is for convenience only and should not limit this application to the scope of the embodiments described. It is understood that those skilled in the art, based on the principles of this system, may arbitrarily combine the various operations without departing from these principles, or combine sub-processes with other operations, making various formal and detailed modifications and changes to the functions implementing the above processes and operations. For example, process 200 may further include operations such as allocating the current total road width to each lane according to a predicted ratio, and calculating the centerline coordinates of each lane based on the GNSS road centerline coordinates. As another example, process 200 may further include operations such as updating the MAP message of the V2X communication message according to the width and centerline coordinate information of each lane, and broadcasting the encoded message as dynamic electronic lane information to connected vehicles within a preset distance. Such variations are all within the protection scope of this application.

[0062] According to some embodiments of this application, a V2X dynamic electronic lane planning device based on dynamic traffic flow is provided. The V2X dynamic electronic lane planning device includes a memory configured to store data and instructions; and a processor communicating with the memory. When executing instructions from the memory, the processor is configured to: S1: obtain the total road width X0 and the number of lanes m, and set a time slice interval n; construct an input feature vector based on X0 and the average number of vehicles and average speed of each lane in each time slice. It includes 2m+1 attribute values; based on the width ratio of manually divided lanes, an output feature vector is constructed. S1 includes m attribute values; S2: Based on historical traffic data for a preset period of manually divided lanes, extract information from each row of fields in S1 to construct training, validation, and test sets; S3: Construct a shallow neural network model, where the activation function Softmax is used to determine that the sum of lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. S4: Based on the training set constructed in S2 and the loss function in S3, iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br, and perform overfit verification based on the validation set constructed in S2; S5: Based on the test set constructed in S2, test the neural network model after parameter training. If the test does not reach the preset accuracy, return to S4 for training; S6: If the test reaches the preset accuracy, deploy to the dynamic electronic lane planning environment.

[0063] In some embodiments, this application obtains the spatial location information of connected vehicle users through a high-precision positioning scheme. Other information can be obtained from the roadside unit (RSU) via short-range V2X communication based on the PC5 interface. The RSU acts as a broadcast sharing device, providing information data services to a large number of positioning devices described in this application within a radius of several hundred meters. The beneficial effects of this application include applying the current V2X technology's support for map (MAP) messages to dynamic electronic lanes. In some embodiments, the wireless communication between the information terminal 120 and the user terminal 130 may include end-to-end short-range wireless direct communication based on the PC5 interface, which may differ from existing 4G-based long-range wireless communication methods (including 4G modules, 4G SIM cards, and 4G antennas).

[0064] In summary, the V2X dynamic electronic lane planning method and apparatus based on dynamic traffic flow according to the embodiments of this application provides an automated dynamic electronic lane division method and apparatus by fusing high-precision satellite positioning information (GNSS), lane marker information described by V2X communication methods, and real-time traffic conditions near the lanes perceived by roadside equipment, such as the driving status of all traffic participants, average vehicle speed, and average number of vehicles. This method automatically and dynamically adjusts the GNSS coordinates of the markers used for the center lines of multiple lanes within the road in the MAP message of the V2X communication mechanism according to road conditions and traffic conditions, and broadcasts this information to surrounding vehicles with V2X communication capabilities to achieve the effect of presenting dynamic electronic lanes.

[0065] It should be noted that the above embodiments are merely examples, and this application is not limited to such examples, but various changes can be made.

[0066] It should be noted that, in this specification, the terms "comprising," "including," or any other variations thereof are intended to cover a 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 limitation, 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.

[0067] Finally, it should be noted that the above series of processes includes not only processes executed in the order described herein in a time sequence, but also processes executed in parallel or separately, rather than in a time sequence.

[0068] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0069] The above-disclosed embodiments are merely some preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of this application still fall within the scope of the invention.

Claims

1. A V2X dynamic electronic lane planning method based on dynamic traffic flow, characterized in that, include: S1: Obtain the total road width X0 and the number of lanes m, and set the time slice interval n; Based on X0 and the average number of vehicles and average speed of each lane in each time slice, construct the input feature vector. This includes 2m+1 attribute values; Based on the width ratio of manually divided lanes, an output feature vector is constructed. , including m attribute values; S2: Based on historical traffic data for a preset period of manually divided lanes, extract information from each row of the S1 field to construct training set, validation set, and test set; S3: Construct a shallow neural network model, where the activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. Cross-entropy with the proportion of manual division; S4: Based on the training set constructed in S2 and the loss function in S3, iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br, and perform overfit verification based on the validation set constructed in S2. S5: Based on the test set built in S2, test the neural network model that has completed parameter training. If the test does not reach the preset accuracy, return to S4 for training. S6: If the test achieves the preset accuracy rate, deploy to the dynamic electronic lane planning environment; The method further includes: Based on real-time road traffic data collected by roadside sensing devices, an input feature vector is constructed. Input the trained neural network model to obtain a prediction of the real-time division ratio of electronic lanes. ; The current total road width is allocated to each lane according to the predicted ratio, and the centerline coordinates of each lane are calculated based on the GNSS road centerline point coordinates. Based on the width and centerline coordinates of each lane, the MAP message of the V2X communication message is updated, and the encoded message is broadcast as dynamic electronic lane information to connected vehicles within a preset distance.

2. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 1, characterized in that, When the number of lanes m=3, the specific contents include: Obtain the average number of vehicles passing through lane 1 in each n-second time slice X1, the average speed of lane 1 in each n-second time slice X2, the average number of vehicles passing through lane 2 in each n-second time slice X3, the average speed of lane 2 in each n-second time slice X4, the average number of vehicles passing through lane 3 in each n-second time slice X5, and the average speed of lane 3 in each n-second time slice X6. Construct the input feature vector X=[X0,X1,X2,X3,X4,X5,X6]. Based on the manually defined width ratios Y0, Y1, and Y2 of lanes 1, 2, and 3, an output feature vector Y = [Y0, Y1, Y2] is constructed, where... =1.

3. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 2, characterized in that, The construction of the shallow neural network model specifically includes: ， Where X0 to X6 are the input feature vectors defined in S1 that are strongly correlated with the lane width division ratio, P0, P1, and P2 are the target lane width distributions, and w ij To use real-time traffic features X j Calculate the width P of the i-th lane. i The required weighting factor, b i To calculate the offset required when calculating the width of the i-th lane, the w ij b i Determined through training.

4. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 3, characterized in that, The activation function Softmax specifically includes: ， The activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, Z. i This is the output value of the i-th node in the last layer of the neural network.

5. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 4, characterized in that, Based on the characteristic that the model output has a probability distribution form P0+P1+P2=1, the loss function D(P,Y) specifically includes: ， The loss function D(P,Y) is based on lane width prediction. Cross-entropy with the proportion of manual division, Y i P represents the actual width ratio of road i. i The width ratio of road i is predicted by the model.

6. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 5, characterized in that, The iterative training of the shallow neural network model parameters specifically includes: Using the backpropagation (BP) algorithm, the parameters of the neural network model are iteratively trained on a high-performance server based on the training set and the loss function.

7. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 6, characterized in that, The step of testing the neural network model that has completed parameter training to see if it has reached a preset accuracy rate specifically includes: Data from the test set for a preset period was randomly selected and divided into four groups based on traffic flow conditions: no congestion, slight congestion, moderate congestion, and severe congestion. Each group had a fixed 1,000 data points. For each data point in the four sets of data, a neural network model with completed parameter training is used to make a prediction. If the highest probability term in the predicted lane usage probability distribution is the same as the lane usage actually marked by humans, then the prediction is determined to be correct; otherwise, the prediction is determined to be incorrect. The accuracy rate of the prediction results for the four sets of data is calculated. The accuracy rate is the ratio of the number of correct predictions for each set to the total number of test data for each set. If the accuracy rate is greater than the preset threshold for each set, it is determined that the actual use requirements have been met; otherwise, it is determined that the actual use requirements have not been met.

8. The V2X dynamic electronic lane planning method based on dynamic traffic flow according to claim 7, characterized in that, The accuracy rates are respectively greater than the preset thresholds for each group, specifically including 70% for no congestion, 75% for slight congestion, 80% for moderate congestion, and 85% for heavy congestion.

9. A V2X dynamic electronic lane planning device based on dynamic traffic flow, characterized in that, Specifically, it includes: A memory, configured to store data and instructions; A processor that communicates with memory, wherein, when executing instructions in memory, the processor is configured to: S1: Obtain the total road width X0 and the number of lanes m, and set the time slice interval n; construct the input feature vector based on X0 and the average number of vehicles and average speed of each lane in each time slice. It includes 2m+1 attribute values; based on the width ratio of manually divided lanes, an output feature vector is constructed. , including m attribute values; S2: Based on historical traffic data for a preset period of manually divided lanes, extract information from each row of the S1 field to construct training set, validation set, and test set; S3: Construct a shallow neural network model, where the activation function Softmax is used to determine that the sum of the lane width allocation ratios is 1, and the loss function D(P,Y) is based on lane width prediction. Cross-entropy with the proportion of manual division; S4: Based on the training set constructed in S2 and the loss function in S3, iteratively train the parameters of the shallow neural network model to determine the optimal weights and offset parameters of Wr and br, and perform overfit verification based on the validation set constructed in S2. S5: Based on the test set built in S2, test the neural network model that has completed parameter training. If the test does not reach the preset accuracy, return to S4 for training. S6: If the test achieves the preset accuracy rate, deploy to the dynamic electronic lane planning environment; It is also configured to: construct input feature vectors based on real-time road traffic data collected by roadside sensing devices. Input the trained neural network model to obtain a prediction of the real-time division ratio of electronic lanes. ; The current total road width is allocated to each lane according to the predicted ratio, and the centerline coordinates of each lane are calculated based on the GNSS road centerline point coordinates. Based on the width and centerline coordinates of each lane, the MAP message of the V2X communication message is updated, and the encoded message is broadcast as dynamic electronic lane information to connected vehicles within a preset distance.

Images