Vehicle guiding method and system for unattended highway

Abstract

The invention relates to the technical field of intelligent traffic, in particular to a vehicle guiding method and system for an unattended highway. The invention discloses a vehicle guiding method for an unattended highway. The method comprises the following steps: S1, acquiring road image information and vehicle information on the highway in real time; s2, performing road section division according to the road image information and a historical congestion area, and analyzing each branch road section by using a mobile calculation formula to obtain an abnormal road section; s3, analyzing the behaviors of vehicles in each abnormal road section to obtain the lane changing times of the expressway entrance; and S4, performing planning guidance on the buffer road section according to the road image information of the expressway entrance. According to the method, the vehicle behavior in the abnormal road section is analyzed, the switching frequency of the expressway entrance is obtained, the buffer road section is further planned and guided, and the vehicle flow in the buffer area can be efficiently guided to avoid congestion through the dynamic analysis of the vehicle behavior and the intelligent planning of the divided road sections.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a vehicle guiding method and system for an unmanned highway. Background Art

[0002] With the rapid expansion of the global highway network, traffic volume is also growing. In order to ensure the safe and orderly passage of vehicles on highways, traditional vehicle guidance usually relies on manual command and management, especially in high-traffic areas such as congested sections and service areas. However, manual guidance has many defects, such as slow response speed, many monitoring blind spots, and difficulty in achieving full-time and all-round vehicle guidance, which leads to frequent accidents and serious traffic jams.

[0003] In recent years, with the advancement of intelligent transportation technology, unmanned technology has begun to be applied to the field of traffic management, especially in highway scenarios, through real-time monitoring of roads and vehicle behavior analysis through cameras, radars and other equipment, and intelligent regulation of traffic flow by combining big data and artificial intelligence technology.

[0004] However, existing monitoring systems mostly rely on image recognition, and have low accuracy in identifying vehicle behavior in complex scenarios. Near some key nodes, existing division methods easily lead to concentrated traffic pressure in certain areas, making it impossible to achieve precise diversion. Summary of the invention

[0005] In order to overcome the shortcoming of difficulty in accurate traffic diversion on expressways, the present invention provides a vehicle guidance method and system for unmanned expressways.

[0006] Technical solution: A vehicle guidance method for an unmanned highway, comprising the following steps:

[0007] S1: Real-time collection of road image information and vehicle information on highways;

[0008] S2: Divide the road sections according to the road image information and the historical congestion areas, and analyze each section using a movement calculation formula to obtain abnormal sections;

[0009] S3: Analyze the behavior of vehicles in each abnormal road section and obtain the number of lane changes at the highway entrance;

[0010] S4: Plan and guide the buffer section based on the road image information at the highway entrance.

[0011] Preferably, the real-time collection of road image information and vehicle information on the highway includes: obtaining road image information and vehicle images based on highway cameras, and performing image enhancement, scale scaling, and normalization preprocessing on the images, and using Gaussian filtering to perform image noise reduction; and obtaining vehicle speed information through radar speed measurement.

[0012] Preferably, the road sections are divided according to the road image information and the historical congested areas, and each sub-section is analyzed using a mobile calculation formula to obtain abnormal sections, including: dividing the sections of the expressway, dividing the risk areas of the road into sections according to the division nodes of the expressway entrances and exits, forks, service areas, and historical congested areas, using each expressway entrance and exit and each fork as a division node, using the boundary nodes of the first preset distance between each service area and each historical congested area as a division node, and dividing the expressway according to the division nodes to obtain each sub-section.

[0013] Preferably, the method of using the boundary nodes of each service area and each historical congestion area at a first preset distance as the dividing nodes includes: selecting a service area or a historical congestion area, taking each selected service area or historical congestion area as a center point, setting a first preset distance around each center point, identifying all nodes located on the boundary of the first preset distance, and defining the nodes on the boundary as dividing nodes; wherein the historical congestion area refers to an area where the number of traffic congestion events within a preset time period exceeds a first preset threshold.

[0014] Preferably, the use of a movement calculation formula to analyze each sub-section to obtain an abnormal section includes:

[0015] Identify the turning nodes in each sub-section and obtain the angle of the turning node. When the turning node angle is greater than the angle threshold, use the turning node as the starting point, follow the route direction until a turning node with an angle less than or equal to the angle threshold is found, use the turning node as the end point, and obtain the turning section and turning section ratio of each sub-section according to the starting point and end point of different sub-sections.

[0016] According to the ratio of turning sections of each sub-section and the historical traffic volume of each lane, the movement value of each sub-section is obtained using a movement calculation formula, and the abnormal section is obtained according to the vehicle density and the movement value.

[0017] Preferably, the movement value of each sub-section is obtained using a movement calculation formula according to the turning section ratio of each sub-section and the historical traffic volume of each lane, and the abnormal section is obtained according to the vehicle density and the movement value, including: when the vehicle density of the sub-section is greater than or equal to a preset density threshold or the total movement value of the vehicles in the sub-section is less than the preset movement value, the sub-section is regarded as an abnormal section; wherein the movement calculation formula is:

[0018]

[0019] Where L is the moving value, m is the number of lanes, and n is i is the number of vehicles on the i-th lane, V ij is the speed of the jth vehicle on the i-th lane, w i is the weight of the i-th lane, and the weight is allocated according to the traffic volume of the lane. σ is the speed influence coefficient of the turning section of the road, and R represents the proportion of the turning section of each sub-section.

[0020] Preferably, the analyzing the behavior of vehicles in each abnormal road section to obtain the number of lane changes at the highway entrance includes: inputting the vehicle information in each abnormal road section collected in real time into a vehicle analysis module to obtain the vehicle behavior of each vehicle, and obtaining the number of lane changes at the highway entrance based on the historical vehicle behavior number of each vehicle and the behavior correspondence table.

[0021] Preferably, the vehicle information in each abnormal road section collected in real time is input into the vehicle analysis module to obtain the vehicle behavior of each vehicle, including: the vehicle analysis module adopts a CNN combined with LSTM model, extracts spatial features from the input vehicle trajectory image or radar image through multi-layer convolution operations, performs convolution operations on the vehicle's position information at different times, generates a high-level feature map, and inputs the high-level feature map into the pooling layer to obtain spatial features, combines the spatial features with the dynamic parameter data of the vehicle to obtain a time series, processes the input sequence through multiple LSTM levels to obtain vehicle behavior; and puts the vehicle behavior into the high-speed behavior sequence of the corresponding vehicle; wherein, a Softmax classifier is used as a classification function, and the model parameters are optimized based on a gradient descent algorithm and a back propagation algorithm, and the model is evaluated using a verification data set; wherein, vehicle behavior includes active avoidance, active queue jumping or normal driving behavior.

[0022] Preferably, the planning and guiding of the buffer road section according to the road image information of the expressway entrance includes: dividing the expressway entrance area to obtain a first buffer zone and a second buffer zone, and dividing the first buffer zone to obtain an ETC toll road, a manual toll road and a mixed toll road;

[0023] Guide the vehicles in the second buffer zone. If the vehicle is an ETC toll collection vehicle, guide the vehicle to the ETC toll road and the mixed toll road; if the vehicle is a manual toll collection vehicle, guide the vehicle to the manual toll road and the mixed toll road;

[0024] Guiding the vehicles in the first buffer zone, obtaining the number of lane changes at the highway entrance of each vehicle in the first buffer zone; and guiding each vehicle to change lanes in the first buffer zone according to the number of lane changes at the highway entrance of each vehicle;

[0025] When congestion occurs, the number of lane changes at the highway entrance of each vehicle in the first buffer zone is obtained, and the vehicles whose number of lane changes at the highway entrance is greater than or equal to N times are guided to change lanes.

[0026] Preferably, a vehicle guidance system for an unmanned highway comprises:

[0027] The data acquisition module is used to obtain road image information and vehicle images based on the cameras and radars on the highway and pre-process the data;

[0028] Section division module: divide the highway into sections according to highway entrances and exits, forks, service areas and historical congestion areas, and generate sub-sections;

[0029] Anomaly detection module: uses the mobile calculation formula, combined with vehicle density and turning section ratio, to analyze the anomalies of each section;

[0030] Vehicle behavior detection module: uses a CNN combined with LSTM model to identify vehicle behaviors, including active avoidance, active queue cutting, and normal driving, through convolution and time series analysis;

[0031] Vehicle behavior classification module: Based on the historical behavior data of vehicles, the number of switching behaviors of vehicles at highway entrances is classified;

[0032] Buffer zone division module: divides the expressway entrance area into the first buffer zone and the second buffer zone, and further divides the first buffer zone into ETC toll roads, manual toll roads and mixed toll roads;

[0033] Vehicle guidance module: guides the vehicle to the ETC toll road, manual toll road or mixed toll road according to the vehicle's ETC toll status or manual toll vehicle status;

[0034] Switching times monitoring module: monitors the number of vehicle switches in the buffer zone. When congestion occurs, vehicles with a switching times greater than N times are guided to switch roads.

[0035] Beneficial effects: The present invention uses intelligent transportation technology, combined with real-time collected road and vehicle data, to automatically analyze highway conditions and vehicle behavior, thereby achieving unmanned highway vehicle guidance, reducing dependence on manual operations, and greatly improving the level of intelligent guidance;

[0036] By combining road image information with historical congestion area data, the present invention can effectively identify abnormal traffic sections. In particular, by combining the mobile calculation formula with vehicle density, it can accurately lock the traffic bottleneck area, adjust the guidance strategy in real time, and improve the efficiency of abnormal processing.

[0037] A deep learning model combining CNN and LSTM is introduced to dynamically analyze the behavior of vehicles in abnormal sections of road. This model can not only efficiently identify the trajectory and behavior patterns of vehicles, but also combine them with historical behavior records, further improving the accuracy of abnormal vehicle identification and ensuring the reasonable guidance of vehicles;

[0038] Based on the real-time division of vehicle switching times and buffer sections, the present invention can dynamically adjust the vehicle's driving route in the buffer zone according to the actual road conditions, effectively avoiding congestion and waste of resources. When the number of vehicle switching times at the highway entrance is high, the system can give priority to guiding these vehicles to reduce traffic pressure in the buffer zone. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A flow chart of a vehicle guidance method for an unmanned highway according to the present invention;

[0040] Figure 2 A schematic diagram of the division of the expressway of the present invention;

[0041] Figure 3 The following is a diagram showing an example of the behavior correspondence table of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Embodiment 1: A vehicle guidance method for an unmanned highway, such as Figure 1-Figure 3 As shown, the following steps are included:

[0044] S1: Real-time collection of road image information and vehicle information on highways;

[0045] Based on highway cameras, road image information and vehicle images are obtained, and the images are pre-processed by image enhancement, scale scaling, and normalization, and Gaussian filtering is used for image noise reduction; radar speed measurement is used to obtain vehicle speed information.

[0046] It should be noted that in the specific implementation, the camera is installed above the highway to ensure that the acquisition angle covers the vehicle image of the entire lane. The acquired image is first improved in clarity through the image enhancement algorithm, and then the image size is scaled and normalized to make it suitable for subsequent analysis steps. Since the noise in the traffic environment (such as dust, light fluctuations, etc.) affects the image quality, the system uses a Gaussian filtering algorithm to reduce the noise of the image to ensure the removal of useless interference information and enhance the clarity of the image, especially in the case of fast-moving vehicles. The accuracy of image processing is crucial. In addition, the vehicle's driving speed is obtained through the radar speed measurement equipment to form a complete data set for subsequent analysis.

[0047] S2: Divide the road sections according to the road image information and the historical congestion areas, and analyze each section using a movement calculation formula to obtain abnormal sections;

[0048] The sections of the expressway are divided, and the risk areas of the road are divided into sections according to the division nodes of the expressway entrances and exits, forks, service areas, and historical congested areas. Each expressway entrance and exit and each fork are used as division nodes, and the boundary nodes of the first preset distance between each service area and each historical congested area are used as division nodes. The expressway is divided according to the division nodes to obtain each sub-section.

[0049] It should be noted that the system identifies areas with frequent congestion based on historical data (such as the number of traffic jams that occurred within a preset time period) and sets these areas as risk areas. Then, the system divides the highway into regions based on the locations of entrances and exits, service areas, and forks. Specifically, the system sets highway entrances and exits and forks as primary division nodes, and service areas and historical congestion areas as secondary division nodes, thus forming multiple highway sub-sections. The definition of congested areas is based on historical data analysis. The system sets a predetermined time period (such as 6 months or 1 year) and counts the frequency of traffic congestion during this period. If the number of congestion times in a certain area exceeds the preset threshold, it will be marked as a "historical congestion area", and this information will be used for further section division and traffic analysis.

[0050] A service area or a historical congestion area is selected, and each selected service area or historical congestion area is taken as a center point. A first preset distance is set around each center point, and all nodes located on the boundary of the first preset distance are identified, and the nodes on the boundary are defined as dividing nodes; wherein the historical congestion area refers to an area where the number of traffic congestion events within a preset time period exceeds a first preset threshold.

[0051] It should be noted that the system selects a service area or a historical congestion area as the center and sets a "first preset distance" around it, which is usually set according to specific traffic management needs. For example, the preset distance can be determined to be 5 kilometers or 10 kilometers based on the average distance traveled by vehicles and traffic flow density. Within this distance range, the system identifies the boundary nodes of the service area and the congestion area and uses these boundary points as division nodes.

[0052] Identify the turning nodes in each sub-section and obtain the angle of the turning node. When the turning node angle is greater than the angle threshold, use the turning node as the starting point, follow the route direction until a turning node with an angle less than or equal to the angle threshold is found, use the turning node as the end point, and obtain the turning section and turning section ratio of each sub-section according to the starting point and end point of different sub-sections.

[0053] According to the ratio of turning sections of each sub-section and the historical traffic volume of each lane, the movement value of each sub-section is obtained using a movement calculation formula, and the abnormal section is obtained according to the vehicle density and the movement value.

[0054] It should be noted that the movement calculation formula is used to calculate the road sections to determine which road sections have abnormalities (such as traffic jams, abnormal vehicle driving). The movement calculation formula calculates the movement value of the road section by considering the number of lanes, vehicle speed, and vehicle density, so as to identify the problem sections in a timely manner.

[0055] When the vehicle density of a road section is greater than or equal to a preset density threshold or the total movement value of vehicles in the road section is less than a preset movement value, the road section is regarded as an abnormal road section; the movement calculation formula is:

[0056]

[0057] Where L is the moving value, m is the number of lanes, and n is i is the number of vehicles on the i-th lane, V ij is the speed of the jth vehicle on the i-th lane, w i is the weight of the i-th lane, and the weight is allocated according to the traffic volume of the lane. σ is the speed influence coefficient of the turning section of the road, and R represents the proportion of the turning section of each sub-section.

[0058] It should be noted that the system first calculates the average movement value based on the actual speed of the vehicles in each sub-section. If the movement value of a sub-section is lower than the preset value (for example, due to high vehicle density or too sharp turning angle resulting in slow speed), the system marks the section as an abnormal section.

[0059] S3: Analyze the behavior of vehicles in each abnormal road section and obtain the number of lane changes at the highway entrance;

[0060] The vehicle information collected in real time in each abnormal road section is input into the vehicle analysis module to obtain the vehicle behavior of each vehicle, and the number of lane changes at the highway entrance is obtained based on the historical vehicle behavior times of each vehicle and the behavior correspondence table.

[0061] It should be noted that after the system identifies an abnormal road section, it combines the camera and speed measurement data to analyze the vehicle behavior through a deep learning model. The model inputs the vehicle's driving trajectory, speed change and other data, and outputs a judgment on the vehicle's behavior, such as whether it frequently changes lanes on abnormal roads or frequently changes lanes at highway entrances. The behavior correspondence table is a correspondence table between active avoidance, active queue jumping, normal driving behavior and the number of lane changes at highway entrances, such as Figure 3 shown.

[0062] The vehicle analysis module adopts a CNN combined with LSTM model. It extracts spatial features from the input vehicle trajectory image or radar image through multi-layer convolution operations, performs convolution operations on the vehicle's position information at different times, generates a high-level feature map, and inputs the high-level feature map into the pooling layer to obtain spatial features. The spatial features are combined with the dynamic parameter data of the vehicle to obtain a time series, and the input sequence is processed through multiple LSTM layers to obtain vehicle behavior; and the vehicle behavior is placed in the high-speed behavior sequence of the corresponding vehicle; wherein, the Softmax classifier is used as the classification function, and the model parameters are optimized based on the gradient descent algorithm and the back propagation algorithm, and the model is evaluated using a verification data set; wherein, vehicle behavior includes active avoidance, active queue jumping or normal driving behavior.

[0063] It should be noted that, first of all, the system obtains real-time vehicle images through cameras installed on the highway. The CNN network will extract relevant feature information of the vehicle from these images, such as the body outline and lane position, so as to achieve a preliminary judgment of the vehicle's position. This image feature extraction is the basis for the system to identify vehicle behavior; next, based on the time series data of the vehicle's driving trajectory, speed change, and lane change frequency, the LSTM network is used for analysis. LSTM can capture the trend of changes in the time dimension, such as the speed change and driving route of the vehicle in the past few minutes, to help identify potential abnormal driving behavior. By combining CNN and LSTM, the system can analyze vehicle behavior more comprehensively. For example, CNN is responsible for image feature extraction, and LSTM is responsible for behavior trend analysis. Finally, the driving behavior of the vehicle is comprehensively judged through the output results of the two to ensure the accuracy and real-time nature of the analysis.

[0064] S4: Plan and guide the buffer section based on the road image information at the highway entrance.

[0065] The expressway entrance area is divided into a first buffer zone and a second buffer zone, and the first buffer zone is divided into an ETC toll road, a manual toll road and a mixed toll road;

[0066] Guide the vehicles in the second buffer zone. If the vehicle is an ETC toll collection vehicle, guide the vehicle to the ETC toll road and the mixed toll road; if the vehicle is a manual toll collection vehicle, guide the vehicle to the manual toll road and the mixed toll road;

[0067] Guiding the vehicles in the first buffer zone, obtaining the number of lane changes at the highway entrance of each vehicle in the first buffer zone; and guiding each vehicle to change lanes between roads in the first buffer zone according to the number of lane changes at the highway entrance of each vehicle;

[0068] When congestion occurs, the number of lane changes at the highway entrance of each vehicle in the first buffer zone is obtained, and the vehicles whose number of lane changes at the highway entrance is greater than or equal to N times are guided to change lanes.

[0069] It should be noted that by dividing the highway entrance area into the first buffer zone (ETC toll roads, manual toll roads, mixed toll roads) and the second buffer zone, since ETC toll vehicles are more efficient than manual toll vehicles at the highway entrance, different types of vehicles (such as ETC toll vehicles and manual toll vehicles) can be diverted. Among them, ETC toll vehicles need to be guided to ETC toll roads in advance, and manual toll vehicles need to be guided to manual toll roads in advance. If the number of highway roads is greater than the number of ETC toll roads and manual toll roads, the extra intermediate roads will be set as mixed toll roads, such as Figure 2 As shown. Therefore, guiding vehicles to appropriate roads in advance can reduce cross-interference and avoid unnecessary congestion caused by mixed traffic. By pre-dividing and guiding, the vehicle flow path is optimized, which helps to improve traffic efficiency; vehicles that actively cut in line often switch roads frequently for their own interests. This behavior not only disrupts normal traffic order and increases road complexity, but also easily causes accidents. In the buffer zone of the highway, frequent lane switching will lead to the risk of rear-end collision or collision. By reducing the number of switches, accidents can be avoided.

[0070] Embodiment 2: Based on Embodiment 1, a vehicle guidance system for an unmanned highway includes:

[0071] The data acquisition module is used to obtain road image information and vehicle images based on the cameras and radars on the highway and pre-process the data;

[0072] Section division module: divide the highway into sections according to highway entrances and exits, forks, service areas and historical congestion areas, and generate sub-sections;

[0073] Anomaly detection module: uses the mobile calculation formula, combined with vehicle density and turning section ratio, to analyze the anomalies of each section;

[0074] Vehicle behavior detection module: uses a CNN combined with LSTM model to identify vehicle behaviors, including active avoidance, active queue cutting, and normal driving, through convolution and time series analysis;

[0075] Vehicle behavior classification module: Based on the historical behavior data of vehicles, the number of switching behaviors of vehicles at highway entrances is classified;

[0076] Buffer zone division module: divides the expressway entrance area into the first buffer zone and the second buffer zone, and further divides the first buffer zone into ETC toll roads, manual toll roads and mixed toll roads;

[0077] Vehicle guidance module: guides the vehicle to the ETC toll road, manual toll road or mixed toll road according to the vehicle's ETC toll status or manual toll vehicle status;

[0078] Switching times monitoring module: monitors the number of vehicle switches in the buffer zone. When congestion occurs, vehicles with a switching times greater than N times are guided to switch roads.

[0079] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A vehicle guidance method for an unmanned highway, characterized in that: The following steps are involved: S1: Real-time collection of road image information and vehicle information on highways; S2: Divide the road sections according to the road image information and the historical congestion areas, and analyze each section using a movement calculation formula to obtain abnormal sections; S3: Analyze the behavior of vehicles in each abnormal road section and obtain the number of lane changes at the highway entrance; S4: Plan and guide the buffer section based on the road image information at the highway entrance.

2. The vehicle guidance method for unmanned highway according to claim 1, characterized in that: The real-time collection of road image information and vehicle information on the highway includes: obtaining road image information and vehicle images based on highway cameras, performing image enhancement, scale scaling, and normalization preprocessing on the images, and using Gaussian filtering to reduce image noise; and obtaining vehicle speed information through radar speed measurement.

3. The vehicle guidance method for unmanned highway according to claim 1, characterized in that: The method divides the road sections according to the road image information and the historical congested areas, and uses a mobile calculation formula to analyze each sub-section to obtain abnormal sections, including: dividing the sections of the expressway, dividing the risk areas of the road according to the division nodes of the expressway entrances and exits, bifurcations, service areas, and historical congested areas, using each expressway entrance and exit and each bifurcations as a division node, using the boundary nodes of the first preset distance between each service area and each historical congested area as a division node, and dividing the expressway according to the division nodes to obtain each sub-section.

4. The vehicle guidance method for unmanned highway according to claim 3 is characterized in that: The method of using the boundary nodes of each service area and each historical congestion area at a first preset distance as the dividing nodes includes: selecting a service area or a historical congestion area, taking each selected service area or historical congestion area as a center point, setting a first preset distance around each center point, identifying all nodes located on the boundary of the first preset distance, and defining the nodes on the boundary as dividing nodes; wherein the historical congestion area refers to an area where the number of traffic congestion events within a preset time period exceeds a first preset threshold.

5. The vehicle guidance method for unmanned highway according to claim 1, characterized in that: The use of the mobile calculation formula to analyze each sub-section to obtain abnormal sections includes: Identify the turning nodes in each sub-section and obtain the angle of the turning node. When the turning node angle is greater than the angle threshold, use the turning node as the starting point, follow the route direction until a turning node with an angle less than or equal to the angle threshold is found, use the turning node as the end point, and obtain the turning section and turning section ratio of each sub-section according to the starting point and end point of different sub-sections. According to the ratio of turning sections of each sub-section and the historical traffic volume of each lane, the movement value of each sub-section is obtained using a movement calculation formula, and the abnormal section is obtained according to the vehicle density and the movement value.

6. The vehicle guidance method for unmanned highway according to claim 5, characterized in that: The method uses a movement calculation formula to obtain the movement value of each sub-section according to the ratio of turning sections of each sub-section and the historical traffic volume of each lane, and obtains the abnormal section according to the vehicle density and the movement value, including: when the vehicle density of the sub-section is greater than or equal to a preset density threshold or the total movement value of the vehicles in the sub-section is less than the preset movement value, the sub-section is regarded as an abnormal section; wherein the movement calculation formula is: Where L is the moving value, m is the number of lanes, and n is i is the number of vehicles on the i-th lane, V ij is the speed of the jth vehicle on the i-th lane, w i is the weight of the i-th lane, and the weight is allocated according to the traffic volume of the lane. σ is the speed influence coefficient of the turning section of the road, and R represents the proportion of the turning section of each sub-section.

7. The vehicle guidance method for unmanned highway according to claim 1, characterized in that: The analysis of the behaviors of vehicles in each abnormal road section to obtain the number of lane changes at the highway entrance includes: The vehicle information collected in each abnormal road section in real time is input into the vehicle analysis module to obtain the vehicle behavior of each vehicle, and the number of lane changes at the highway entrance is obtained based on the historical vehicle behavior times of each vehicle and the behavior correspondence table.

8. The vehicle guidance method for unmanned highway according to claim 7, characterized in that: The vehicle information in each abnormal road section collected in real time is input into the vehicle analysis module to obtain the vehicle behavior of each vehicle, including: the vehicle analysis module adopts a CNN combined with LSTM model, extracts spatial features from the input vehicle trajectory image or radar image through multi-layer convolution operations, performs convolution operations on the position information of the vehicle at different times, generates a high-level feature map, and inputs the high-level feature map into the pooling layer to obtain spatial features, combines the spatial features with the dynamic parameter data of the vehicle to obtain a time series, processes the input sequence through multiple LSTM levels to obtain vehicle behavior; and puts the vehicle behavior into the high-speed behavior sequence of the corresponding vehicle; wherein, a Softmax classifier is used as a classification function, and the model parameters are optimized based on a gradient descent algorithm and a back propagation algorithm, and the model is evaluated using a verification data set; wherein the vehicle behavior includes active avoidance, active queue jumping or normal driving behavior.

9. The vehicle guidance method for unmanned highway according to claim 1, characterized in that: The planning and guiding of the buffer road section according to the road image information of the expressway entrance includes: dividing the expressway entrance area to obtain a first buffer zone and a second buffer zone, and dividing the first buffer zone to obtain an ETC toll road, a manual toll road and a mixed toll road; Guide the vehicles in the second buffer zone. If the vehicle is an ETC toll collection vehicle, guide the vehicle to the ETC toll road and the mixed toll road; if the vehicle is a manual toll collection vehicle, guide the vehicle to the manual toll road and the mixed toll road; Guiding the vehicles in the first buffer zone, obtaining the number of lane changes at the highway entrance of each vehicle in the first buffer zone; and guiding each vehicle to change lanes in the first buffer zone according to the number of lane changes at the highway entrance of each vehicle; When congestion occurs, the number of lane changes at the highway entrance of each vehicle in the first buffer zone is obtained, and the vehicles whose number of lane changes at the highway entrance is greater than or equal to N times are guided to change lanes.

10. A vehicle guidance system for an unmanned highway, according to a vehicle guidance method for an unmanned highway according to any one of claims 1 to 9, characterized in that: include: The data acquisition module is used to obtain road image information and vehicle images based on the cameras and radars on the highway and pre-process the data; Road segmentation module: divide the highway into sections according to highway entrances and exits, forks, service areas and historical congestion areas, and generate sub-sections; Anomaly detection module: uses the mobile calculation formula, combined with vehicle density and turning section ratio, to analyze the anomalies of each section; Vehicle behavior detection module: uses a CNN combined with LSTM model to identify vehicle behaviors, including active avoidance, active queue cutting, and normal driving, through convolution and time series analysis; Vehicle behavior classification module: Based on the historical behavior data of vehicles, the number of switching behaviors of vehicles at highway entrances is classified; Buffer zone division module: divides the expressway entrance area into the first buffer zone and the second buffer zone, and further divides the first buffer zone into ETC toll roads, manual toll roads and mixed toll roads; Vehicle guidance module: guides the vehicle to the ETC toll road, manual toll road or mixed toll road according to the vehicle's ETC toll status or manual toll vehicle status; Switching times monitoring module: monitors the number of vehicle switches in the buffer zone. When congestion occurs, vehicles with a switching times greater than N times are guided to switch roads.

Images