A method for estimating traffic flow of all regional road sections based on probe vehicle data

Abstract

The present application belongs to the field of traffic flow estimation, and relates to a full-area road section traffic flow estimation method based on floating car data. The present application firstly classifies full-area road sections and determines reference road sections and non-reference road sections based on road section information investigation; then collects the flow of the reference road sections using radar, and collects the free flow speed and real-time average speed of the full-area road sections using the floating car method; finally, selects each reference road section, calibrates the jam density on the single lane of the road section based on the collected flow data and floating car data, and estimates the flow of the remaining non-reference road sections using the jam density and floating car data. The present application reduces the cost of flow data acquisition, reduces the workload and data storage of manual work, classifies the road sections in the research area according to the urban built environment, classifies and estimates the flow of the full-area road sections according to the classification results, and significantly improves the accuracy of flow estimation.

Description

Technical Field

[0001] This invention belongs to the field of traffic flow estimation and relates to a method for estimating traffic flow across a whole area road segment based on floating car data. Background Technology

[0002] With urban traffic congestion becoming increasingly serious, accurately estimating traffic flow on roads within a given area to provide a reference for traffic management departments in formulating traffic plans and policies has always been a crucial issue in the transportation field. Existing methods for estimating road segment flow can be mainly divided into three categories: the first is the direct observation method, which generally relies on roadside sensor equipment or traffic detectors such as intersection radar and cameras to directly measure road segment flow; the second is the traffic assignment method, which is generally based on a four-stage model. Given the known OD demand of users, it allocates traffic demand to the actual road network according to a certain path selection method. For example, Shao Hu et al., in "Operations Research Methods for Urban Traffic Flow Estimation," reviewed existing operations research-based methods for estimating road segment flow from seven aspects: linear programming, integer programming, dynamic programming, graph theory, statistical methods, heuristic methods, and machine learning. The third category is methods for estimating the flow of unknown road segments based on the flow of known road segments. For example, in "Real-time Estimation Method of Flow of Undetected Road Segments in Urban Road Network Based on Transfer Learning" by Xing Jiping et al., the road network with detectors found by the similarity evaluation system is used as the source domain. The fine-tuning method is used to transfer learning the flow estimation model, thereby completing the flow estimation of undetected road segments.

[0003] In summary, existing methods for estimating traffic flow on road segments have the following problems: First, obtaining traffic flow data often relies on static sensor data, such as traffic cameras and traffic flow sensors, which suffers from high equipment installation costs, high maintenance difficulties, and low coverage. Second, existing methods often require complete user travel demand data within the study area, which is often not readily available. Finally, existing studies only consider the impact of user choices on traffic flow, without considering the impact of the urban built environment on traffic flow. Summary of the Invention

[0004] To address the aforementioned problems, this invention proposes a method for estimating traffic flow across a full-area road segment based on floating car data. This method primarily considers the impact of the urban built environment on road segment traffic flow. When it is impossible to collect traffic flow data for all road segments within the study area, it can estimate the traffic flow data for remaining unknown road segments using known road segment traffic flow data and floating car data. This invention is based on the following assumption: road segments with similar urban built environments have similar traffic flow characteristics, i.e., similar congestion densities. Therefore, known road segment traffic flow data can be used to determine the congestion density of such road segments, and the traffic flow of unknown road segments can be estimated based on this congestion density.

[0005] Based on the above assumptions, this invention classifies all road segments within the study area according to the urban built environment attributes surrounding the road segments. This invention defines all road segments within the study area as follows: among the classified road segments, those with good traffic conditions and open visibility at intersection entrances are selected as reference road segments; the remaining road segments within the study area other than the reference road segments are called non-reference road segments; and all road segments within the study area are referred to as the entire area road segments.

[0006] This invention first classifies all road segments in the region based on road segment information surveys and determines reference road segments and non-reference road segments; then, it uses radar to collect traffic flow data of the reference road segments and uses the floating car method to collect the free-flow velocity and real-time average velocity of all road segments in the region; finally, it selects each reference road segment, and based on the collected traffic flow data and floating car data, it calibrates the congestion density on a single lane of that type of road segment, and uses the congestion density and floating car data to estimate the traffic flow of the remaining non-reference road segments.

[0007] The technical solution of the present invention:

[0008] A method for estimating traffic flow across a whole area based on floating car data includes three steps: road segment information survey and classification, basic data collection and processing, and traffic flow estimation. First, road segment information survey and classification are conducted, investigating information on the entrance roads of all road segments in the entire area and the built-up environment attributes of the surrounding cities, and classifying all road segments in the area. Next, basic data collection and processing are performed, including using radar to collect traffic flow data for reference road segments, using the floating car method to collect the free-flow velocity and actual average speed of vehicles on all road segments in the entire area, and preprocessing the collected raw data. Finally, traffic flow estimation is performed, based on the preprocessed raw data, calibrating the congestion density of each type of road segment, and estimating the traffic flow of the remaining non-reference road segments.

[0009] The specific steps are as follows:

[0010] Step 1, Road segment information survey and classification:

[0011] To facilitate the classification, collection, and estimation of traffic flow data, considering the characteristics of each road segment and the surrounding urban built environment, the road segments within the study area are classified. The specific steps are as follows:

[0012] Step 1.1, Road Section Information Survey:

[0013] Traffic flow on a road segment is related to the characteristics of the segment itself and the surrounding urban built environment. Therefore, information on all road segments within the study area should be investigated before classification. Since traffic flow is generally similar on the same road, this invention investigates the approach lane information at intersections where each road segment is located, using this information as the data for that road segment. The information to be investigated includes whether the approach lane has a dedicated left-turn lane, the number of lanes on the approach lane, and the number of traffic hotspots around the approach lane.

[0014] If the approach lane is widened, the number of lanes on the approach lane is counted according to the number of lanes before widening. In this invention, traffic hotspots refer to areas or road sections with high traffic volume and severe congestion. These areas are typically commercial districts, schools, shopping malls, transportation hubs, entertainment venues, residential areas, etc.

[0015] Step 1.2, Road segment classification and selection of reference road segments:

[0016] Based on the road segment information survey in step 1.1, all road segments in the area are classified according to the following standards: A road segment is classified as a Class A intersection if it has more than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is greater than the average number of traffic hotspots in the study area; a road segment is classified as a Class B intersection if it has more than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is less than the average number of traffic hotspots in the study area; a road segment is classified as a Class C intersection if it has fewer than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is greater than the average number of traffic hotspots in the study area; and a road segment is classified as a Class C intersection if it has fewer than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is less than the average number of traffic hotspots in the study area. The intersection with the average number of traffic hotspots in the study area is called a Class D intersection. An intersection without a left-turn approach lane, with more than 3 approach lanes, and a traffic hotspot number in the surrounding area greater than the average number of traffic hotspots in the study area is called a Class E intersection. An intersection without a left-turn approach lane, with more than 3 approach lanes, and a traffic hotspot number in the surrounding area less than the average number of traffic hotspots in the study area is called a Class F intersection. An intersection without a left-turn approach lane, with fewer than 3 approach lanes, and a traffic hotspot number in the surrounding area greater than the average number of traffic hotspots in the study area is called a Class G intersection. An intersection without a left-turn approach lane, with fewer than 3 approach lanes, and a traffic hotspot number in the surrounding area less than the average number of traffic hotspots in the study area is called a Class H intersection.

[0017] After classification, to facilitate subsequent traffic flow data collection, road segments with good traffic conditions and open visibility at intersection entrances should be selected as reference segments within each type of road segment. The remaining road segments within the study area are referred to as non-reference segments. At least one road segment should be selected as a reference segment for each type of intersection.

[0018] Step 2, Basic Data Collection and Processing:

[0019] The basic data collection and processing section, based on road segment classification, collects the necessary foundational data for traffic flow estimation, as detailed below:

[0020] Step 2.1, Traffic Data Collection:

[0021] Traffic flow data for a reference road segment is collected using radar. Radar is installed at the intersection where the reference road segment is located, and the peak-hour traffic flow at the intersection's approach lanes is measured, distinguishing between weekdays and non-weekdays, to serve as the traffic flow data for the reference road segment.

[0022] Step 2.2, Floating car data acquisition:

[0023] The floating car method was used to collect the free-flow speed and real-time average speed of vehicles across the entire road area, as detailed below:

[0024] Step 2.2.1, Free-flow velocity acquisition. In the early morning, the test vehicle is not disturbed by external vehicles and can be approximated as being in a free-flow state. The average speed measured by the test vehicle traveling 12 times in one direction on the test section is taken as the free-flow velocity of that section.

[0025] Step 2.2.2, Real-time average speed acquisition. During peak hours, the average speed measured by the test vehicle traveling 12 times in one direction on the test section is taken as the real-time average speed of that section.

[0026] Step 2.3, Data Preprocessing:

[0027] First, check if there are any missing data during the collection process. If there are any missing data, they need to be collected and supplemented. After ensuring the integrity of the dataset, the traffic flow data of the reference road segment collected by the radar is aggregated every hour and converted into the hourly traffic flow of the road segment. Then, the traffic flow data of the same road segment at the same time is matched with the free flow speed data and real-time average speed data collected by the floating car.

[0028] Step 3, Traffic Flow Estimation:

[0029] Traffic flow estimation is based on preprocessed reference road segment traffic flow data and floating car data. According to the road segment classification results, the congestion density, a characteristic parameter of traffic flow for a certain type of road segment, is calibrated, and the traffic flow of other road segments is estimated based on the calibration results. The specific steps are as follows:

[0030] Step 3.1, Single-lane traffic flow calculation. Calculate the average single-lane traffic flow of the reference road segment during peak hours on a weekday:

[0031]

[0032] Among them, Q SL Q is the average traffic flow of a single lane in the reference road segment. FL N represents the total traffic flow of the reference road segment, where N is the number of lanes in the reference road segment.

[0033] Step 3.2, calibrate congestion density. Based on the average single-lane flow rate Q obtained in Step 3.1... SL And the actual average speed v and free-flow speed v of the vehicles in the floating car data obtained in step 2.2. f The single-lane congestion density K corresponding to the calibrated reference road segment is determined. j Then the congestion density of this type of road segment can be approximated as K. j During the calibration process, based on the Greenhill model, the velocity-flow relationship is as follows:

[0034]

[0035] Among them, Q SL K is used as a reference for the average traffic flow of a single lane in the road segment. j Let v be the congestion density of a certain type of road segment, and v be the actual average speed of vehicles. f The free-flow speed of the vehicle.

[0036] Step 3.3, Estimate the flow rate of non-reference road segments. Considering road segments of the same type, which have similar urban built environment attributes and therefore similar congestion densities, use the calibrated congestion density obtained in Step 3.2, and estimate the single-lane flow rate Q of the non-reference road segments according to the Greenhill model in Step 3.2. SL_est The final traffic flow on non-reference road sections can be estimated based on the number of lanes and the flow rate per lane.

[0037] Q FL_est =N·Q SL_est

[0038] Among them, Q FL_est Q is an estimate of the traffic flow on the entire road segment. SL_est This is an estimate of the traffic flow on a single road segment.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] (1) Compared with the traditional method of estimating road segment traffic based on traffic assignment, this invention does not require origin-destination travel data for all users in the entire region. It can estimate the traffic flow of road segments with missing data in the study area.

[0041] (2) This invention reduces the cost of traffic flow data acquisition. Compared with traditional methods that use radar or other traffic detection equipment to measure each intersection, this invention reduces the amount of manual work and data storage, and is easy to operate and implement in actual traffic flow estimation. In addition, this invention does not require the large-scale installation of sensors, cameras and other traffic detection equipment on road sections during the traffic flow estimation process, which greatly reduces the cost of data acquisition.

[0042] (3) In the process of estimating traffic flow on road segments, this invention classifies road segments within the study area according to the urban built environment, and estimates the traffic flow of all road segments in the entire area based on the classification results, which significantly improves the accuracy of traffic flow estimation. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for estimating traffic flow across a whole area road segment based on floating car data, according to the present invention.

[0044] Figure 2 This is a schematic diagram of the study area in the embodiment. Detailed Implementation

[0045] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings and technical solutions.

[0046] This case study selects Chengyang District of Qingdao City as the research area, bounded by Ningcheng Road to the east, Xiucheng Road to the west, Xingyang Road to the south, and Chongyang Road to the north. Within this area, 55 intersections were selected, forming 200 road segments. A schematic diagram of the research area is shown below. Figure 2 As shown. This study area faces the following problems: the number of radar deployment points within the area is very limited; floating car data for road segments can be obtained, but traffic signal control requires the traffic flow of the entire road segment in the area. If the traffic flow data is measured sequentially using radar at each intersection entrance in this study area, it will consume a huge amount of manpower and resources, and the collection period will be long. Therefore, this example uses the above method to estimate the traffic flow of the entire road segment in Chengyang District. This example first classifies the road segments in Chengyang District, Qingdao City, selects reference road segments to measure peak hour traffic flow, uses the floating car method to measure the free flow velocity and actual average velocity of the road segments in the entire area, and uses linear regression to calibrate the congestion density K based on the above data. j Finally, the traffic flow of non-reference road sections in Chengyang District was estimated, and the traffic flow of all road sections in the entire area was obtained.

[0047] Basic process as follows Figure 1 As shown, the specific implementation steps are as follows:

[0048] Step 1: Road Segment Information Survey and Classification. The survey investigates the urban built environment attributes of the approach lanes for each road segment within the study area, including whether a dedicated left-turn lane is provided, the number of lanes on the approach lane, and the number of traffic hotspots around the intersection where the approach lane is located. The survey results are compiled and all road segments are classified according to the aforementioned classification criteria. Ultimately, there are eight categories of road segments in this study area: A, B, C, D, E, F, G, and H. Within each category, road segments with radar installed at intersections are selected as reference road segments.

[0049] Step 2: Basic data collection and processing, collecting the data needed for traffic flow estimation of road sections.

[0050] Step 2.1, Traffic Flow Data Collection. Using radar at intersections along each reference road segment, and distinguishing between weekdays and non-weekdays, traffic flow data is collected for the entrance lanes of each segment during the morning rush hour (7-8 AM). The average of the acquired 30 consecutive days of traffic flow data, averaging the weekday and non-weekday data, is used as the traffic flow for that reference road segment.

[0051] Step 2.2, Free-flow velocity acquisition. In the early morning, the maximum safe speed that the test vehicle can reach while driving undisturbed on various road sections can be approximated as the vehicle's free-flow velocity. In the early morning, the test vehicle is driven 12 times on various road sections within the study area, and the length of each road section and the free-flow time taken for the test vehicle to traverse the entire section are recorded. The free-flow velocity of a given road section can be calculated using the following formula:

[0052]

[0053] Where v f Let L be the free-flow velocity of a certain road segment, and t be the length of that road segment. ft Let be the free passage time when passing through a certain road segment for the i-th time.

[0054] Step 2.3, Real-time Average Speed ​​Acquisition. During peak hours, the test vehicle was driven 12 times across various road sections within the study area, and the time taken to traverse the entire road section was recorded again. The actual average speed of a particular road section can be calculated using the following formula:

[0055]

[0056] Where v is the free-flow velocity of a certain road segment, L is the length of the road segment, and t i Let be the actual travel time when passing through a certain road segment for the i-th time.

[0057] Step 2.4, Data Preprocessing. The traffic flow data for the reference road segment collected by radar is aggregated according to peak hours (7:00 AM - 8:00 AM) and converted into hourly traffic flow for the road segment. The traffic flow data for the obtained 30 consecutive days is averaged between weekdays and non-weekdays to obtain the weekday or non-weekday full-lane traffic flow Q for that reference road segment. FL .according to Full lane traffic flow Q FL Converted to single-lane traffic flow Q SL The traffic flow data for each reference road segment is then matched with the floating car data.

[0058] Step 3, Traffic Flow Estimation for the Entire Area. Based on the data collected above, estimate the traffic flow for all road segments in the entire area.

[0059] Step 3.1, based on the average single-lane flow Q obtained in the above steps SL The actual average speed of the vehicle, v; the free-flow speed of the vehicle, vfree f Calibrate the single-lane congestion density K corresponding to a certain type of entrance lane. j This invention employs a novel calibration scheme. Based on the Greenhill model, it first calculates the formula... Partial. Thus Q SL and A linear relationship is then established between them, therefore linear regression can be used to determine the undetermined parameter K in the model. j Calibration was performed, and the K values ​​presented by each representative intersection on weekdays and non-weekdays were determined. j The data is calibrated to determine the congestion density of this type of intersection on weekdays and weekends.

[0060] Step 3.2, based on K in step 3.1 j The v and v obtained from floating car data calculation f Based on the Greenhill model, estimate the flow Q on the non-reference road segment. SL_est According to Q FL_est =N·Q SL_est This allows for the estimation of traffic flow on non-reference road segments throughout the region. Allocating the corresponding traffic flow to specific road networks completes the estimation of traffic flow for all road segments in the entire region.

[0061] Table 2 Residual of Traffic Flow Estimation for Road Sections

[0062]

[0063]

[0064] This example selects a portion of the road segment for estimation results and residual analysis, as shown in Table 2. In the table, Q... FL_obsFor the observed flow data obtained using the floating car method, Q FL_est The estimated traffic flow data obtained using this method is shown. The estimation results show that the residual rate of the estimated results is less than 10%, which is close to the actual traffic flow on the road segments. Therefore, the method of this invention has good application and implementation effects in the estimation of traffic flow on actual road segments across the entire region.

Claims

1. A method for estimating traffic flow across a whole area road segment based on floating car data, characterized in that, The method for estimating traffic flow across a whole area based on floating car data includes three steps: road segment information survey and classification, basic data collection and processing, and traffic flow estimation. First, road segment information survey and classification are conducted, investigating the information of the entrance roads to all road segments in the whole area and the built-up environment attributes of the surrounding cities, and classifying all road segments in the whole area. Then, basic data collection and processing are performed, including using radar to collect traffic flow data for reference road segments, using the floating car method to collect the free-flow speed and actual average speed of vehicles on all road segments in the whole area, and preprocessing the collected raw data. Finally, traffic flow estimation is performed, based on the preprocessed raw data, calibrating the congestion density of various road segments, and estimating the traffic flow of the remaining non-reference road segments. The specific steps are as follows: Step 1, Road segment information survey and classification: Step 1.1, Road Section Information Survey: Traffic flow on a road segment is related to the characteristics of the road segment itself and the attributes of the urban built environment surrounding the road segment. Before classification, information on all road segments in the study area should be investigated. The information on the approach lanes of the intersections where each road segment is located should be investigated as the information for that road segment. The information to be investigated includes whether the approach lane has a separate dedicated left-turn lane, the number of lanes of the approach lane, and the number of traffic hotspots around the approach lane. If the approach lane is widened, the number of lanes in the approach lane will be counted according to the number of lanes before the widening. Step 1.2, Road segment classification and selection of reference road segments: Based on the road segment information survey in step 1.1, all road segments in the area are classified according to the following standards: A road segment is classified as a Class A intersection if it has more than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is greater than the average number of traffic hotspots in the study area; a road segment is classified as a Class B intersection if it has more than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is less than the average number of traffic hotspots in the study area; a road segment is classified as a Class C intersection if it has fewer than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is greater than the average number of traffic hotspots in the study area; and a road segment is classified as a Class C intersection if it has fewer than 3 left-turn lanes at the intersection and the number of traffic hotspots around the intersection is less than the average number of traffic hotspots in the study area. The intersection with the average number of traffic hotspots in the surrounding area is called a Class D intersection. An intersection without a left-turn approach lane, with more than 3 approach lanes, and a traffic hotspot number in the surrounding area greater than the average number of traffic hotspots in the study area is called a Class E intersection. An intersection without a left-turn approach lane, with more than 3 approach lanes, and a traffic hotspot number in the surrounding area less than the average number of traffic hotspots in the study area is called a Class F intersection. An intersection without a left-turn approach lane, with fewer than 3 approach lanes, and a traffic hotspot number in the surrounding area greater than the average number of traffic hotspots in the study area is called a Class G intersection. An intersection without a left-turn approach lane, with fewer than 3 approach lanes, and a traffic hotspot number in the surrounding area less than the average number of traffic hotspots in the study area is called a Class H intersection. After classification, in order to facilitate subsequent traffic flow data collection, road segments with good traffic conditions and open visibility at intersection entrances should be selected as reference road segments in each type of road segment. The remaining road segments in the study area are called non-reference road segments. At least one road segment should be selected as a reference road segment in each type of intersection. Step 2, Basic Data Collection and Processing: Step 2.1, Traffic Data Collection: Use radar to collect traffic flow data for reference road sections; install radar at the intersection where the reference road section is located, distinguish between weekdays and non-weekdays, and measure the peak traffic flow of the intersection's approach lanes as the traffic flow data for the reference road section. Step 2.2, Floating car data acquisition: The floating car method was used to collect the free-flow speed and real-time average speed of vehicles across the entire road area, as detailed below: Step 2.2.1, Free-flow velocity acquisition; In the early morning, the test vehicle is not disturbed by external vehicles and is regarded as being in a state of free-flow. The average speed measured by the test vehicle traveling 12 times in one direction on the test section is taken as the free-flow velocity of the test section. Step 2.2.2, Real-time average speed acquisition; During peak hours, the average speed measured by the test vehicle traveling 12 times in one direction on the road segment under test is taken as the real-time average speed of the road segment. Step 2.3, Data Preprocessing: First, check if there is any missing data during the collection process. If there is any missing data, it is necessary to re-collect and supplement the missing data. After ensuring the completeness of the dataset, the traffic flow data of the reference road segment collected by the radar is aggregated every hour and converted into the hourly traffic flow of the road segment. Then, the traffic flow data of the same road segment at the same time is matched with the free flow speed data and real-time average speed data collected by the floating car. Step 3, Traffic Flow Estimation: Step 3.1, Single-lane traffic flow calculation; Calculate the average single-lane traffic flow of the reference road segment during peak hours on a weekday: Among them, Q SL Q is the average traffic flow of a single lane in the reference road segment. FL The reference road segment represents the total traffic flow for all lanes, and N represents the number of lanes in the reference road segment. Step 3.2, calibrate the congestion density; based on the average single-lane flow rate Q obtained in Step 3.

1. SL And the actual average speed v and free-flow speed v of the vehicles in the floating car data obtained in step 2.

2. f The single-lane congestion density K corresponding to the calibrated reference road segment is determined. j Then the congestion density of this type of road segment can be approximated as K. j During the calibration process, based on the Greenhill model, the velocity-flow relationship is as follows: Among them, Q SL K is used as a reference for the average traffic flow of a single lane in the road segment. j Let v be the congestion density of a certain type of road segment, and v be the actual average speed of vehicles. f The free-flow velocity of the vehicle; Step 3.3, Estimate the flow rate of the non-reference road segment; using the calibrated congestion density obtained in Step 3.2, estimate the single-lane flow rate Q of the non-reference road segment according to the Greenhill model in Step 3.

2. SL_est The final traffic flow on non-reference road sections can be estimated based on the number of lanes and the flow rate per lane. Q FL_est =N·Q SL_est Among them, Q FL_est Q is an estimate of the traffic flow on the entire road segment. SL_est This is an estimate of the traffic flow on a single road segment.

Images