A highway traffic status prediction method based on behavior feature capture

By defining detection sections and sub-sections on the highway, capturing vehicle behavior characteristics, classifying vehicles and establishing speed prediction models, the problem of insufficient prediction accuracy based on time series in the prior art is solved, and the accuracy of highway traffic state prediction is improved and Fine-grained, supporting refined and proactive management.

CN118397831BActive Publication Date: 2024-11-15UNIV OF SHANGHAI FOR SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410473617.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-19
Publication Date
2024-11-15
Estimated Expiration
2044-04-19

AI Technical Summary

Technical Problem

The existing highway traffic state prediction method is based on time series and fails to effectively capture the heterogeneity of vehicle behavior, resulting in insufficient prediction accuracy and cannot support the refined and active management of highways.

Method used

By defining detection sections and sub-sections, traffic status data are obtained, data preprocessing and vehicle behavior characteristics are captured, and vehicles are classified as low-speed trucks, conservative buses, radical buses and conventional buses, and vehicle behavior characteristics are used to establish a speed prediction model. and parameter adjustment to improve prediction accuracy.

Benefits of technology

It improves the accuracy and fine-grainedness of highway traffic status prediction, supports refined management and active traffic safety management, and provides targeted support for the management of vehicles of different types and driving characteristics.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118397831B_ABST
    Figure CN118397831B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for predicting the traffic state of a highway based on behavior feature capture, including: S1, defining a detection section according to the layout of detection equipment, and dividing the highway section into sub-sections and lanes; S2, obtaining a set of traffic state data of each detection section, and performing data preprocessing; S3, splicing the data of continuous sections, and capturing vehicle behavior features; S4, classifying vehicles by vehicle behavior features; S5, selecting a speed prediction model to predict traffic speed; S6, verifying subsequent section prediction results and correcting parameters in the speed prediction model; S7, using the corrected speed prediction model to predict the traffic state of the highway section to be predicted. Through accurate vehicle behavior feature capture and speed prediction based on this, the present invention can improve the accuracy and granularity of traffic state prediction, thereby effectively improving the refined and intelligent management level of highways, and providing support for targeted management of vehicles of different types and driving characteristics, and providing support for active traffic safety management.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of information prediction, and in particular to a method for predicting highway traffic status based on capturing behavior characteristics. Background Art

[0002] Today's highway management is increasingly becoming more intelligent, refined and proactive. The realization of these goals is inseparable from the accurate perception and prediction of highway traffic conditions. Accurate prediction of highway traffic conditions plays a vital role in highway traffic management, anomaly detection, congestion control and proactive safety management.

[0003] In the field of short-term prediction of highway traffic flow status, there are mainly parametric methods represented by Kalman filtering, ARIMA, multivariate regression theory, exponential smoothing, etc., and non-parametric methods represented by deep learning, machine learning, etc., but no matter what method is used, it is generally based on time series for prediction. This type of method uses the traffic status data of a certain section of road to learn the traffic status data in the past period of time to predict the traffic status of the section in a specific time in the future. However, the traffic flow characteristics of highways are different from commuting traffic on urban roads. Most of the trips of traffic participants are for business or tourism, and do not have obvious periodicity or time regularity. Therefore, the prediction method based on time series may not be applicable. In addition, compared with urban expressways, the heterogeneity of vehicles on highways is more significant. This heterogeneity is reflected in both vehicle types - not only passenger cars, but also trucks of various sizes; and vehicle behavior - for example, heavy trucks travel at low speeds due to insufficient speed maintenance ability, becoming bottlenecks on the road; some inexperienced drivers choose to drive in the innermost lane and drive at a relatively low constant speed to stay away from trucks; and more aggressive drivers often change lanes and overtake to maintain a higher driving speed. This heterogeneity has a significant impact on the evolution and operation speed of traffic flow states.

[0004] In past studies, although a few studies have noticed the heterogeneity of vehicle types, the heterogeneity of vehicle behavior has been neglected to a certain extent. In fact, the heterogeneity of vehicle types does not have a greater impact on traffic flow than the heterogeneity of vehicle behavior. Compared with the size differences of different types of vehicles, the speed differences caused by the heterogeneity of vehicle behavior have a more obvious impact on traffic flow. Intuitively, a truck traveling at a speed of 80 km / h does not have a greater impact on traffic flow than a bus traveling at a speed of 60 km / h. Therefore, the heterogeneity of vehicle behavior is the key factor affecting traffic state prediction. Moreover, both active traffic safety management and refined traffic management rely on the accurate prediction of the distribution of vehicles with different driving behaviors. The previous traffic state prediction methods ignored the heterogeneity of vehicle behavior, resulting in insufficient prediction accuracy and failure to provide sufficient support for the refined and active management of highways. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for predicting highway traffic status based on behavioral feature capture, so as to improve the accuracy and granularity of highway traffic status prediction, thereby effectively improving the level of refined and intelligent management of highways, providing support for targeted management of vehicles of different types and driving characteristics, and also providing support for active traffic safety management such as active safety warning. In order to achieve the above purpose, the present invention provides a method for predicting highway traffic status based on behavioral feature capture, comprising:

[0006] S1. Define the detection section according to the layout of the detection equipment, and divide the highway section into sub-sections and lanes;

[0007] S2, obtaining a traffic status data set of each detection section and performing data preprocessing;

[0008] S3, splicing the data of continuous sections and capturing the vehicle behavior characteristics;

[0009] S4, classifying vehicles according to vehicle behavior characteristics, the vehicle classification includes low-speed trucks, conservative buses, aggressive buses and regular buses;

[0010] S5, predicting the traffic speed by using the speed prediction model based on the vehicle category and traffic parameters;

[0011] S6. Verify the subsequent cross-section prediction results and correct the parameters in the velocity prediction model;

[0012] S7. Use the revised speed prediction model to predict the highway traffic status of the highway section to be predicted.

[0013] Preferably, in step S1, according to the vehicle's driving direction, the detection sections are defined in sequence as the first detection section, the second detection section, the third detection section...; and the section between the first detection section and the second detection section is defined as sub-section 1, the section between the second detection section and the third detection section is defined as sub-section 2, and so on.

[0014] Preferably, in step S2, the traffic status data includes vehicle ID, and the time, speed, headway, headway and lane information of each ID vehicle passing through the detection section;

[0015] Data preprocessing includes data desensitization and abnormal data removal.

[0016] Preferably, the continuous sections in step S3 refer to two or more sections with consecutive numbers, specifically: the data of detection section 1 and detection section 2 are spliced, and the change in the relative position of the vehicle in the traffic flow, the average speed of the traffic flow, and the speed of each vehicle relative to the traffic flow are determined based on the lane where the vehicle with the same ID is located at the two sections, the section speed, the time of passing the section, etc., so as to obtain the lane preference, speed preference and aggressive preference behavior characteristics of vehicles with different IDs.

[0017] Preferably, the vehicle behavior characteristics in step S3 need to be mapped with the surrounding traffic flow status. Specifically, the vehicle density and flow rate of sub-section 1 are obtained by calculating the difference between the number of vehicles passing through detection section 1 and detection section 2 within a certain time interval, and a mapping is established with the behavior characteristics of the vehicles on sub-section 1 during this period of time.

[0018] Preferably, in step S5, the speeds of the low-speed truck, conservative bus, conventional bus, and aggressive bus are recorded as v1, v2, v3, and v4, respectively, wherein the speed function of the low-speed truck is: v1 is the predicted speed of the low-speed truck, v 1e is the expected speed of the low-speed truck, v a is the average speed of the traffic flow;

[0019] The speed function of the conservative bus is:

[0020] v2 is the predicted speed of the conservative passenger car, v 2e is the expected speed of the conservative passenger car;

[0021] The speed function of a conventional passenger car is:

[0022] v3 is the predicted speed of a conventional passenger car, v 3e is the expected speed of conventional buses, β is the coefficient of the influence of low-speed trucks on conventional buses, which is related to the flow rate of low-speed trucks and the overall flow rate and can be calibrated by the data of the detection section;

[0023] The speed function of the aggressive bus is:

[0024] v4=γv 4e , v4 is the predicted speed of the aggressive bus, v 4e is the expected speed of the aggressive bus, γ is the coefficient of traffic flow affecting the aggressive bus, which is related to traffic flow and density and can be calibrated by the data of the detection section.

[0025] Preferably, in step S6, the speed and time of different types of vehicles arriving at subsequent detection sections are predicted through a vehicle speed prediction model, and the prediction results are verified through detection data of subsequent sections, and the parameters in the prediction model are further corrected, and this process is repeated until the parameters are stable and the prediction accuracy meets the requirements.

[0026] Compared with the prior art, the present invention has the following beneficial effects: the present invention captures the spatiotemporal continuity of vehicle driving characteristics, and through vehicle classification based on driving characteristics capture, and a vehicle speed prediction model based on this, it can reflect the impact of vehicle type and behavior heterogeneity on traffic flow evolution, thereby improving the accuracy and granularity of the prediction, thereby effectively improving the level of refined and intelligent management of highways.

[0027] The traffic state prediction result proposed in the present invention can not only reflect the speed characteristics, but also reproduce the dynamic evolution of traffic flow distribution, thereby providing support for the targeted management of vehicles of different types and driving characteristics by the traffic management department.

[0028] The capture of vehicle behavior characteristics and the speed prediction based on this can provide support for active traffic safety management, especially the management of vehicles with aggressive driving behaviors, as well as active safety warnings for downstream vehicles and vehicles about to merge. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flowchart of a highway traffic status prediction method based on behavior feature capture according to the present invention;

[0030] Figure 2 It is a schematic diagram of the detection section definition and sub-segment division of the highway traffic state prediction method based on behavior feature capture according to the present invention. DETAILED DESCRIPTION

[0031] 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.

[0032] Refer to the attached Figure 1-2 , a highway traffic state prediction method based on behavior feature capture, comprising the following steps:

[0033] S1. According to the layout of the detection equipment, the detection section is defined, and the highway section is divided into sub-sections and lanes. According to the direction of vehicle travel, the detection sections are defined as the first detection section, the second detection section, the third detection section, etc., and the section between the first detection section and the second detection section is defined as sub-section 1, the section between the second detection section and the third detection section is defined as sub-section 2, and so on.

[0034] S2. Obtain the traffic status data set of each detection section and perform data preprocessing, wherein the detection equipment must be able to obtain the vehicle ID, the time when the vehicle passes through the detection section, the speed, the headway, the headway and the lane information, etc. The equipment can be the ETC gantry that is currently widely used on highways, or it can be any detection equipment or combination of detection equipment that can obtain the above information, such as video image processing equipment. After obtaining the traffic status data set of each detection section, data preprocessing must be performed first. Data preprocessing includes data desensitization and abnormal data elimination. Data desensitization refers to vehicle information desensitization, converting vehicle identity information into vehicle ID number to ensure user information security. Abnormal data processing mainly refers to assigning Null to erroneous, repeated, and fluctuating data, and then completing missing data. In this step of processing, special attention should be paid to the processing of abnormal speed data. It is necessary to combine multiple section data to determine whether the speed is continuously high or low due to driver factors, and it cannot be simply eliminated;

[0035] S3. The data of continuous sections are spliced ​​and the vehicle behavior characteristics are captured. The vehicle behavior characteristics captured include lane preference, speed preference and aggressiveness preference. A continuous section refers to two or more sections with consecutive numbers. Taking two continuous sections as an example, specifically, the data of detection section 1 and detection section 2 are spliced. According to the lane where the vehicle with the same ID is located at the two sections, the section speed, the time of passing the section and other information, the change of the relative position of the vehicle in the traffic flow, the average speed of the traffic flow on the section, the speed of each vehicle relative to the traffic flow, etc. are determined, so as to obtain the lane preference, speed preference, aggressiveness preference and other behavioral characteristics of vehicles with different IDs.

[0036] The details are as follows: Assuming that vehicle 1 is a truck, it is located in the outermost lane in both detection section 1 and detection section 2, its speed is relatively constant, and its speed is significantly lower than the entire traffic flow, then it can be preliminarily judged as a low-speed truck.

[0037] Assuming that vehicle 2 is a passenger car, it is located in the innermost lane in both detection sections 1 and 2, its speed is relatively constant, and its speed is significantly lower than that of the entire passenger car traffic flow. It can be preliminarily judged as a conservative passenger car.

[0038] Assuming that vehicle 3 is a passenger car, the lane distribution in different detection sections is not constant, and the vehicle speed is not much different from the entire passenger car traffic flow, it can be preliminarily determined to be a regular passenger car.

[0039] Assuming that vehicle 4 is a passenger car, the lane distribution in different detection sections is not constant, but the vehicle speed is significantly higher than the passenger car traffic flow, it can be preliminarily judged as an aggressive passenger car.

[0040] The above judgment is made based on the traffic information of two consecutive sections as an example. It should be noted that more consecutive sections can make the judgment result more accurate.

[0041] The capture of behavioral characteristics needs to be associated with the flow and traffic composition of the sub-section, because the gap between overtaking demand and supply is different under different flow and traffic composition, so the speed will be affected differently. For example, in the case of extremely low flow, there is little demand for overtaking, and the speed difference between aggressive buses and regular buses may not be obvious; in the case of extremely high flow, although the demand for overtaking is large, due to the high overall traffic density, there are few overtaking opportunities, and the speed difference between aggressive buses and regular buses may not be obvious. Therefore, the capture of behavioral characteristics requires as many sections as possible on the one hand, and on the other hand, it needs to be associated with traffic flow macro parameters such as flow. Relatively speaking, under medium flow, its behavioral characteristics will be more obvious.

[0042] S4. Classify vehicles by their behavior characteristics, including low-speed trucks, conservative buses, aggressive buses and regular buses. Low-speed trucks are usually located in the outermost lane due to regulations, and their speed is usually lower than that of buses. Conservative buses are often located in the innermost lane, and they drive at a relatively constant speed lower than that of other buses. Some drivers with less driving experience usually adopt this driving style. Regular buses are the main body of traffic flow and have their own expected speed, but they will be significantly affected by traffic flow, and they will overtake and change lanes under the premise of ensuring safety. The drivers of aggressive buses are more aggressive, and in order to pursue higher speeds, they usually take very risky overtaking methods, and even take shoulder overtaking or S-shaped overtaking methods. These vehicles are also one of the vehicles that highway managers need to pay attention to, because their strong aggressiveness can easily cause traffic safety problems and put surrounding vehicles in danger. One of the tasks of active traffic safety management is to provide early warnings for such vehicles, which relies on accurate prediction of vehicle speed based on behavioral characteristics. Here, vehicles are divided into four categories, which does not mean that they can only be divided into four categories. It is just that these four categories of behavioral characteristics are quite representative. In implementation, the classification can be further refined, but it should be noted that finer classification will also lead to more complex speed prediction models and longer response times. The specific classification can be selected according to the prediction accuracy requirements.

[0043] S5. The vehicle type and traffic parameters are used to predict the traffic speed through a speed prediction model; the speed prediction model can be a mathematical model containing parameters, or a deep learning model or other models, which is not limited in the present invention. An embodiment of a mathematical model containing parameters is given here.

[0044] The speeds of the low-speed truck, conservative bus, conventional bus, and aggressive bus are recorded as v1, v2, v3, and v4, respectively. The speed function of the low-speed truck is:

[0045] v1 is the predicted speed of the low-speed truck, v 1e is the expected speed of the low-speed truck, v a is the average speed of the traffic flow;

[0046] The speed function of the conservative bus is:

[0047] v2 is the predicted speed of the conservative passenger car, v 2e is the expected speed of the conservative passenger car;

[0048] The speed function of a conventional passenger car is:

[0049] v3 is the predicted speed of a conventional passenger car, v 3eis the expected speed of conventional buses, β is the influence coefficient of conventional buses on low-speed trucks, and the influence of the flow rate and overall flow rate of low-speed trucks, which can be calibrated by the data of the detection section;

[0050] The speed function of the aggressive bus is:

[0051] v4=γv 4e , v4 is the predicted speed of the aggressive bus, v 4e is the expected speed of the aggressive bus, γ is the coefficient of traffic flow affecting the aggressive bus, which is related to traffic flow and density and can be calibrated by the data of the detection section.

[0052] S6. Verify the prediction results of subsequent sections and correct the parameters in the speed prediction model; obtain the speed and time of different types of vehicles passing through the subsequent sections according to the speed prediction model, and verify them with the speed and time of the vehicles actually passing through the sections. Through verification, on the one hand, it is determined whether the vehicle behavior characteristics are captured accurately, so as to correct the classification. On the other hand, some parameters in step 5, such as β and γ, are related to multiple factors such as traffic flow and traffic composition. The more detection sections and sub-sections, the more accurate their calibration will be. Therefore, using subsequent sections to correct them can further improve the subsequent prediction accuracy.

[0053] S7. Use the modified speed prediction model to predict the highway traffic status of the highway section to be predicted

[0054] The number of devices and processing scales described here are used to simplify the description of the present invention, and the application, modification and variation of the present invention will be obvious to those skilled in the art.

[0055] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and implementation modes. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A highway traffic status prediction method based on behavior feature capture, characterized in that: The following steps are involved: S1. Define the detection section according to the layout of the detection equipment, and divide the highway section into sub-sections and lanes; S2, obtaining a traffic status data set of each detection section and performing data preprocessing; S3, splicing the data of continuous sections and capturing the vehicle behavior characteristics; S4, classifying vehicles according to vehicle behavior characteristics, the vehicle classification includes low-speed trucks, conservative buses, aggressive buses and regular buses; S5. Select a speed prediction model based on vehicle classification and traffic parameters to predict traffic speed; S6, verify the subsequent cross-section prediction results and calibrate the parameters in the velocity prediction model; S7, using the corrected speed prediction model to predict the highway traffic state for the highway section to be predicted; Through the speed prediction model, the speed and time of different types of vehicles arriving at the subsequent predicted section are predicted; In step S5, the speeds of the low-speed truck, conservative bus, conventional bus, and aggressive bus are recorded as , , and , where the speed function of the low-speed truck is: , is the predicted speed of the low-speed truck, is the expected speed of the low-speed truck, is the average speed of the traffic flow; The speed function of the conservative bus is: , is the predicted speed of the conservative passenger car, is the expected speed of the conservative passenger car; The speed function of a conventional passenger car is: , is the predicted speed of a conventional passenger car, is the expected speed of conventional buses, β is the coefficient of the influence of low-speed trucks on conventional buses, which is related to the flow rate of low-speed trucks and the overall flow rate and is calibrated by the data of the detection section; The speed function of the aggressive bus is: , is the predicted speed of the aggressive coach, is the expected speed of the aggressive bus, It is the coefficient of influence of traffic flow on aggressive buses, which is related to traffic flow and density and is calibrated by the data of the detection section.

2. A highway traffic status prediction method based on behavior feature capture as claimed in claim 1, characterized in that: In step S1, according to the vehicle's driving direction, the detection sections are defined in sequence as the first detection section, the second detection section, the third detection section...; and the section between the first detection section and the second detection section is defined as sub-section 1, the section between the second detection section and the third detection section is defined as sub-section 2, and so on.

3. A highway traffic status prediction method based on behavior feature capture as claimed in claim 1, characterized in that: In step S2, the traffic status data includes vehicle ID, and the time, speed, headway, headway and lane information of each vehicle with ID passing through the detection section; Data preprocessing includes data desensitization and abnormal data removal.

4. A highway traffic status prediction method based on behavior feature capture as claimed in claim 1, characterized in that: The continuous sections in step S3 refer to two or more sections with consecutive numbers. Specifically, the data of detection section 1 and detection section 2 are spliced ​​and processed, and the change of the relative position of the vehicle in the traffic flow and the speed of the vehicle relative to the traffic flow are determined according to the lane, section speed and time information of the vehicle passing through the section with the same ID in the two sections, so as to obtain the lane preference, speed preference and aggressive preference behavior characteristics of vehicles with different IDs.

5. A highway traffic status prediction method based on behavior feature capture as claimed in claim 4, characterized in that: In step S3, a mapping is established between the vehicle behavior characteristics and the surrounding traffic flow status. Specifically, the vehicle density and flow rate of sub-section 1 are obtained by calculating the difference between the number of vehicles passing through detection section 1 and detection section 2 within a certain time interval, and a mapping is established with the behavior characteristics of the vehicles on sub-section 1 during this period of time.

6. A highway traffic status prediction method based on behavior feature capture as claimed in claim 1, characterized in that: In step S6, the vehicle speed prediction model is used to predict the speed and time of different types of vehicles arriving at subsequent detection sections. The prediction results are verified by the detection data of subsequent sections, and the parameters in the prediction model are further corrected. This process is repeated until the parameters are stable and the prediction accuracy meets the requirements.