Online travel time estimation method for road segments based on multi-source data fusion

Abstract

This application discloses a method for estimating online travel time of road segments based on multi-source data fusion, including: Step 1, generating corresponding road network topology relationships based on the composition and spatial distribution of elements within the studied road network; Step 2, preprocessing checkpoint and radar data to filter out anomalies and duplicate records; Step 3, setting time window rules for upstream and downstream license plate matching within a section, matching license plates and calculating the travel time of individual vehicles, and extracting the free-flow travel time of each section; dividing the individual vehicle status of the current calculation period section, statistically analyzing the sample size of the current calculation period, and judging the operating status of the current calculation period section; Step 4, calculating the travel time of the current period section; Step 5, splitting the section travel time into road segments to obtain the online travel time estimate corresponding to the road segment. This application achieves full mining and utilization of data resources in cases of partial missing checkpoint data and small sample sizes, improving the spatiotemporal completeness, effectiveness, and real-time performance of travel time for each road segment within the road network.

Description

Technical Field

[0001] This application relates to a method for estimating online travel time for road segments based on multi-source data fusion, belonging to the field of traffic condition prediction technology. Background Technology

[0002] Travel time is one of the most important and closely watched indicators in traffic flow information. It evaluates road traffic flow and reflects road transport efficiency, playing a crucial role in traffic planning, design, management, and travel information services. Planning departments can use travel time to assess the service level and operational efficiency of the transportation system, serving as a basis for upgrading and transforming the system. Road management departments can use travel time to identify traffic bottlenecks and analyze their causes, providing quantitative support for traffic facility improvements. Individual travelers can use travel time to choose their mode of transportation, determine their departure time, and adjust their routes to avoid congestion.

[0003] Travel time is a complex dynamic parameter influenced by numerous fixed and unforeseen factors, including road geometry, road conditions, weather, traffic accidents, and the traffic behavior of drivers and pedestrians. Travel time varies dynamically across different road segments and at different times, making it a challenging aspect of traffic information to collect and process. However, advancements in information collection and communication technologies have enabled large-scale and real-time traffic information collection and acquisition, providing favorable conditions for calculating real-time dynamic road segment travel times. Currently, the main information collection methods for travel time estimation fall into three categories: the first is location and segment (coverage range 150 meters) traffic flow parameter collection, represented by loop detectors and radar; the second is vehicle trajectory collection, represented by floating car detection; and the third is location vehicle feature collection and upstream / downstream matching, represented by checkpoints. Due to the influence of signal control at urban road intersections and the location and interval layout of detection points, fixed detection cannot accurately estimate urban road traffic flow delays and signal-controlled stopping delays. The occupancy and departure rates of floating cars vary significantly at different times, resulting in insufficient and unevenly distributed data samples. While checkpoints can obtain direct matching interval travel times, local equipment failures and small sample sizes can significantly impact the real-time performance, completeness, and stability of road network travel times. Summary of the Invention

[0004] The technical problem this application aims to solve is how to estimate the online travel time of a road segment.

[0005] To address the aforementioned technical problems, the technical solution of this application provides a method for estimating online travel time for road segments based on multi-source data fusion, comprising the following steps:

[0006] Step 1: Generate the corresponding road network topology relationship for the elements and their spatial distribution within the studied road network, including node order, interval order, road segment and node relationship, checkpoint and road segment relationship, and radar and road segment relationship; acquire historical checkpoint and radar data;

[0007] Step 2: Preprocess the checkpoint and radar data to filter out anomalies and duplicate records;

[0008] Step 3: Set up time window rules for matching license plates upstream and downstream of the interval, match license plates and calculate the single vehicle travel time, extract the free flow travel time of each interval based on the historical single vehicle travel time, divide the single vehicle status of the interval in the current calculation period using the interval free flow time, and statistically analyze the sample size of the current calculation period by status group to determine the operating status of the interval in the current calculation period.

[0009] Step 4: Based on the state discrimination results, filter the single-vehicle travel time samples in the current calculation period interval and judge the reliability of the interval travel time sample size; for the interval that meets the reliability calculation results, calculate the travel time of the current period interval.

[0010] Step 5: Break down the interval travel time into road segments to obtain the online travel time estimate for the corresponding road segment.

[0011] Preferably, in step two, the preprocessing of checkpoint and radar data includes license plate data filtering and deduplication, and radar data validity judgment. Radar data validity judgment is based on overall judgment and single record judgment, identifying suspicious faulty radar equipment and single invalid records. Overall judgment is based on statistical rules, and suspicious fault modes include all traffic parameters being zero, occupancy rate being 100%, and abnormal traffic flow parameters being identified. Based on the statistics of suspicious fault records and the distribution of statistical data within a day, a threshold is set for identification, with the default threshold being that abnormal records exceed half of the total number of records for the day. Single record judgment includes traffic flow combination patterns and threshold ranges for each parameter.

[0012] Preferably, step three specifically includes:

[0013] S3.1 Time Window Rule Settings: Set the interval formation time calculation period, and set the maximum time window for matching license plate sections upstream and downstream of the interval;

[0014] S3.2 Interval travel time matching and calculation: Filter the vehicle passage records of downstream license plate sections and upstream license plate sections by calculation period and maximum matching time window, match the upstream and downstream license plates of the current period, and calculate the single vehicle travel time.

[0015] S3.3 Free-flow travel time calculation: Based on the historical single-vehicle travel time of each section, the single-vehicle travel time of each section is sorted from high to low on a daily basis, and the 85th percentile travel time is taken as the free-flow travel time of the section.

[0016] S3.4 Sectional Vehicle Status Classification: The status of vehicles within a section is classified based on the free-flow travel time. The classification rules are shown in the table below:

[0017] Intersection travel time distribution Single vehicle status classification (0, 0, 5 times free-flow travel time) abnormally small [0.5 times free-flow time, 2 times free-flow travel time] Smooth [2 times free flow time, 4 times free flow travel time] Crowded [4 times the free-flow time, 8 times the free-flow travel time] block Greater than or equal to 8 times the free-flow travel time High guidance

[0018] S3.5 Current Period Interval Status Classification: Calculate the ratio of single-vehicle travel time to free-flow speed within the current period. Classify single-vehicle status according to the interval single-vehicle status classification rules, then count the number of vehicles grouped by single-vehicle status. Determine the current calculation period interval operation status according to the interval status period classification rules, as shown in the table below:

[0019] Interval state Alternative Judgment Rules unknown During the calculation period, the number of unobstructed instances is zero, and the number of congested and blocked instances are both less than 8. Congestion Within the calculation period, the number of congested or blocked items is greater than or equal to 8. Smooth During the calculation period, the number of unobstructed instances is greater than zero, while the number of congested instances and the number of blocked instances are both less than 8.

[0020] Preferably, the current period interval travel time in step four is specifically calculated as follows: when the interval period status is smooth, the interval travel time is calculated based on the smooth single-vehicle travel time sample; when the interval period status is congested, the travel time is calculated based on the smooth, congested, and blocked single-vehicle travel time samples; when the interval period status is unknown, the interval travel time sample does not meet the calculation conditions, and the interval travel time is -1.

[0021] Preferably, step five specifically involves: dividing the travel time of intervals with a travel time greater than zero and spanning multiple road segments into segments, and decomposing the interval travel time into a free-flow portion and a delay portion; the free-flow portion is divided proportionally according to the free-flow time of the segments that make up the interval, and the delay portion is divided proportionally according to the queue length collected by radar for the segments that make up the interval; finally, the online travel time estimate corresponding to the segment is obtained.

[0022] The advantage of this application lies in its ability to determine the reliability of travel time within a continuous road section simultaneously equipped with checkpoints and radar detection devices. This is achieved by leveraging the upstream and downstream topological relationships of the road network and the statistical sample size of single-vehicle travel time status grouped and matched upstream and downstream within the section. The travel time can be adaptively adjusted to calculate the section and its travel time. By collecting radar data before the stop line within each section, the weighting of travel time distribution for each road segment within the section is estimated. This application enables the full mining and utilization of data resources, particularly in cases of missing checkpoint data and small sample sizes, thereby improving the spatiotemporal completeness, effectiveness, and real-time performance of travel time for each road segment within the road network. Attached Figure Description

[0023] Figure 1 A flowchart illustrating the road network segment travel time estimation method of this application;

[0024] Figure 2-1 and Figure 2-2 This is a schematic diagram showing the composition of road network-related elements, where... Figure 2-1 Includes road segments, nodes, checkpoints, and their location relationships. Figure 2-2 Includes radar and road segment location relationships;

[0025] Figure 3-1 and Figure 3-2 This represents the results of the interval travel time matching, where... Figure 3-1 This is the result of matching bicycles at 5-minute intervals. Figure 3-2 The results are synthesized over a 5-minute interval;

[0026] Figure 4-1 and Figure 4-2 The weights and results are broken down into the travel times of each segment within the interval. Figure 4-1 This refers to the radar queue length for each section within the interval. Figure 4-2 This is the 5-minute breakdown result for each road segment within the interval. Detailed Implementation

[0027] To make this application more apparent and understandable, preferred embodiments are described in detail below with reference to the accompanying drawings.

[0028] This embodiment proposes an online travel time estimation method for road segments based on multi-source data fusion. It can determine the reliability of interval travel time based on the upstream and downstream topological relationships of the road network and the statistical sample size of the matched single-vehicle travel time status between upstream and downstream sections. It adaptively adjusts travel time to calculate intervals and travel times. By collecting data from radar before the stop line in each interval, it estimates the weight of travel time allocation for each road segment within the interval, thereby fully mining and utilizing data resources in cases of missing information, ensuring the spatiotemporal integrity and effectiveness of travel time for each road segment within the road network. (Refer to...) Figure 1 The method includes the following steps:

[0029] I. Acquire and organize historical checkpoint and radar data for several urban roads. The historical checkpoint data includes checkpoint number, capture time, lane, and license plate number; the radar data includes radar number, processing cycle, traffic flow, average speed, time occupancy, spatial occupancy, and queue length. The data organization relationship is designed based on data processing requirements, considering the composition and spatial distribution of elements within the studied road network, and generating corresponding road network topology relationships, including node order, interval order, road segment and node relationships, checkpoint and road segment relationships, and radar and road segment relationships, such as... Figure 2-1 , Figure 2-2 As shown.

[0030] Step 1-1: Data Acquisition

[0031] The data used in this embodiment was collected from urban road checkpoints and radar data in the main urban area of ​​a certain city on July 17 and July 18, 2022.

[0032] Steps 1-2: Design of data organization and generation of basic information

[0033] The node sequence includes the node number and the node's position on the road along the direction of travel.

[0034] Table 1 Example of node order format

[0035] Node number Node order A 1 B 2 C 3 D 4 E 5

[0036] The interval sequence includes upstream nodes and downstream nodes. The downstream node is the sequential number of the same upstream node in the road according to the direction of travel, as shown in Table 2.

[0037] Table 2 Example of Interval Order Format

[0038] Upstream node number Downstream node number Interval order A B 1 A C 2 A D 3 A E 4

[0039] The relationship between road segments and nodes includes the road segment number, the upstream node number, and the downstream node number.

[0040] Table 3. Examples of Road Segment and Node Formats

[0041] Road section number Upstream node number Downstream node number A->B A B B->C B C C->D C D D->E D E

[0042] The relationship between checkpoints and road sections includes checkpoint number and road section number.

[0043] Table 4. Examples of checkpoint and road segment formats

[0044] checkpoint number Road section number VD_B1 A->B VD_B2 A->B VD_B3 A->B

[0045] The relationship between radar and road segment includes radar number and road segment number.

[0046] Table 5. Examples of Radar and Road Segment Formats

[0047] Radar number Road section number LD_B1 A->B LD_B2 A->B LD_B3 A->B

[0048] 2. Import the original checkpoint and radar files into the database, then filter out all data within the study area based on the checkpoint and radar numbers. Based on the characteristics of the checkpoint and radar data, set data quality filtering and discrimination rules respectively, clean the data, and delete duplicate, erroneous, and abnormal data.

[0049] Step 2-1: License Plate Data Filtering and Deduplication

[0050] Based on license plate rules, filtering conditions are set to filter abnormal license plates and remove duplicate license plates within a 5-minute timeframe. For identical license plates, the record with the longest time frame is retained. Filtering conditions include: the first digit of the license plate is zero; and the license plate number is less than seven digits.

[0051] Step 2-2: Radar data validity assessment

[0052] Based on overall and individual record discrimination, suspicious faulty radar equipment and individual invalid records are identified. Overall discrimination is based on statistical rules; suspicious fault patterns include all traffic parameters being zero, occupancy being 100%, and abnormal traffic flow parameters. Based on the statistics of suspicious fault records and the distribution of statistical data within a day, a threshold is set for identification; the default threshold is when abnormal records exceed half of the total number of records for the day. Individual record discrimination includes traffic flow combination patterns and threshold ranges for each parameter.

[0053] Third, based on the filtered license plate records, set up time window rules for matching license plates upstream and downstream within an interval, match license plates, and calculate the single-vehicle travel time. Extract the free-flow travel time for each interval based on historical single-vehicle travel times. Use the interval free-flow time to divide the single-vehicle status of the current calculation period interval, and statistically analyze the sample size for the current calculation period grouping to determine the interval's operational status. Based on the status determination results, filter the single-vehicle travel time samples for the current calculation period interval and assess the reliability of the interval travel time sample size. For intervals that meet the reliability calculation results, calculate the current period interval travel time.

[0054] 3-1: Setting Time Window Rules

[0055] Set the interval travel time calculation period, with the start time representing the period. The default is 5 minutes. For example, "2022-7-18 9:15:00" represents the time period (2022-7-18 9:15:00, 2022-7-18 9:20:00). Set the maximum time window for matching upstream and downstream license plate sections of the interval. The default is 1800 seconds.

[0056] 3-2: Intersection Travel Time Matching and Calculation

[0057] The system filters vehicle passage records at downstream and upstream license plate sections within a given interval using the calculation period and the maximum matching time window. It matches upstream and downstream license plates within the current period's filtered interval and calculates the single-vehicle travel time, in seconds.

[0058] Table 6. Examples of Inter-regional Travel Time Matching and Calculation

[0059] Interval numbering Calculation cycle License plate code Upstream node number When the vehicle passes the upstream section Downstream node number When the vehicle passes the downstream section Trip Time B->D 2022 / 7 / 18 9:15:00 1 B 2022 / 7 / 18 9:08:12 D 2022 / 7 / 18 9:10:18 126 B->D 2022 / 7 / 18 9:15:00 2 B 2022 / 7 / 18 9:09:29 D 2022 / 7 / 18 9:12:10 161 B->D 2022 / 7 / 18 9:15:00 3 B 2022 / 7 / 18 9:09:42 D 2022 / 7 / 18 9:11:46 124 B->D 2022 / 7 / 18 9:15:00 4 B 2022 / 7 / 18 9:07:37 D 2022 / 7 / 18 9:10:08 151 B->D 2022 / 7 / 18 9:15:00 5 B 2022 / 7 / 18 9:12:32 D 2022 / 7 / 18 9:14:28 116 B->D 2022 / 7 / 18 9:15:00 6 B 2022 / 7 / 18 9:08:17 D 2022 / 7 / 18 9:10:17 120 B->D 2022 / 7 / 18 9:15:00 7 B 2022 / 7 / 18 9:09:28 D 2022 / 7 / 18 9:11:52 144 B->D 2022 / 7 / 18 9:15:00 8 B 2022 / 7 / 18 9:07:45 D 2022 / 7 / 18 9:10:20 155 B->D 2022 / 7 / 18 9:15:00 9 B 2022 / 7 / 18 9:09:28 D 2022 / 7 / 18 9:11:57 149 B->D 2022 / 7 / 18 9:15:00 10 B 2022 / 7 / 18 9:11:19 D 2022 / 7 / 18 9:14:08 169 B->D 2022 / 7 / 18 9:15:00 11 B 2022 / 7 / 18 9:11:16 D 2022 / 7 / 18 9:14:18 182

[0060] 3-3: Calculation of Free Flow Travel Time

[0061] Based on historical cycling trip time records for each section, the cycling trip times for each section are sorted from highest to lowest on a daily basis. The 85th percentile trip time is taken as the free-flow trip time for that section. The historical results of cycling trip times for each section are as follows: Figure 3-1 As shown.

[0062] Table 7 Examples of Free Flow Travel Time in Intersections

[0063] Interval numbering date Free Flow Travel Time B->C 2022 / 7 / 17 50 B->D 2022 / 7 / 17 132 C->D 2022 / 7 / 17 71

[0064] 3-4: Division of Single-Vehicle Status within a Section

[0065] The status of bicycles in a given section is classified based on the free-flow travel time within that section, and the classification rules are shown in Table 8.

[0066] Table 8 Classification Rules for Intersection Travel Time

[0067] Intersection travel time distribution Single vehicle status classification (0, 0, 5 times free-flow travel time) abnormally small [0.5 times free-flow time, 2 times free-flow travel time] Smooth [2 times free flow time, 4 times free flow travel time] Crowded [4 times the free-flow time, 8 times the free-flow travel time] block Greater than or equal to 8 times the free-flow travel time abnormally high

[0068] 3-5: Classification of current period interval states and judgment of sample credibility

[0069] The ratio of single-vehicle travel time to free-flow speed within the current period is calculated. The single-vehicle status is then determined according to the interval single-vehicle status classification rules, and the number of vehicles grouped by status is counted. Based on the interval single-vehicle status grouping statistics, the current period status and reliability are determined. The classification rules are shown in Table 9, and the determination results are shown in Table 10.

[0070] Table 9 Classification Rules for Interval State Periods

[0071] Interval state Judgment rules unknown During the calculation period, the number of unobstructed instances is zero, and the number of congested and blocked instances are both less than 8. Congestion Within the calculation period, the number of congested or blocked items is greater than or equal to 8. Smooth During the calculation period, the number of unobstructed instances is greater than zero, while the number of congested instances and the number of blocked instances are both less than 8.

[0072] Table 10 Examples of Interval State Period Classification Calculation

[0073] Interval numbering Calculation cycle Number of vehicles in smooth traffic Number of congested vehicles Number of vehicles blocked Abnormally high number of vehicles Periodic state classification B->D 2022 / 7 / 18 7:35 83 5 1 0 Smooth B->D 2022 / 7 / 18 7:50 60 31 0 0 Congestion B->D 2022 / 7 / 18 19:55 2 1 0 0 Smooth B->D 2022 / 7 / 18 23:05 0 2 3 1 unknown

[0074] 3-6: Calculation of travel time within the current period

[0075] Based on the current period interval state classification and sample reliability judgment, the current period interval travel time is synthesized. When the interval period state is "smooth," the interval travel time is calculated using the travel time samples of "smooth" single-vehicles; when the interval period state is "congested," the travel time is calculated using the travel time samples of "smooth," "congested," and "blocked" single-vehicles; when the interval period state is unknown, the interval travel time sample does not meet the calculation conditions, and the interval travel time is output as -1. The interval period calculation result is as follows: Figure 3-2 As shown.

[0076] IV. Breakdown of Travel Time by Route

[0077] For interval travel times greater than zero that cross road segments, the interval travel time is broken down into its constituent road segments. This breakdown further divides the interval travel time into free-flow and delay components. The free-flow component is proportionally allocated based on the free-flow time of each road segment within the interval, while the delay component is proportionally allocated based on the queue lengths collected by radar for each road segment within the interval. This yields the final online travel time estimate for each corresponding road segment.

[0078] 4-1: Current Period Interval Travel Time Splitting Rules

[0079] Obtain the travel time of intervals with a travel time greater than zero that cross road segments, as well as the free-flow travel time of the intervals. Then, split the current period's interval travel time according to the following rules:

[0080] If: The current period's interval travel time is less than or equal to the interval's free-flow travel time

[0081]

[0082] Else if: The current period interval travel time is greater than the interval free-flow travel time

[0083]

[0084] Where: tt(i,t,free) represents the free-flow portion of the travel time within the current period interval.

[0085] tt(i,t,delay) represents the portion of the travel time delayed within the current period.

[0086] tt(i,t) represents the actual travel time within the current period interval.

[0087] tt(i,freedom) represents the free-flow travel time within the interval.

[0088] 4-2: Breakdown of travel time for each segment within the interval

[0089] Read the free-flow travel time of each road segment within the interval, and then use radar data that has undergone data quality assessment. Based on the free-flow travel time of each road segment, proportionally divide the free-flow travel time of the interval; based on the radar queue length, proportionally divide the delay travel time of the interval. Each radar within the interval collects the interval queue data as follows: Figure 4-1 As shown. The travel time for each segment within the interval is obtained by merging the results of the free-flow portion and the delay portion. The travel time breakdown results for each segment within the interval are as follows: Figure 4-2 As shown.

Claims

1. A method for estimating online travel time of road segments based on multi-source data fusion, characterized in that, Includes the following steps: Step 1: Generate the corresponding road network topology relationship for the elements and their spatial distribution within the studied road network, including node order, interval order, road segment and node relationship, checkpoint and road segment relationship, and radar and road segment relationship; acquire historical checkpoint data and historical radar data; Step 2: Preprocess the checkpoint and radar data to filter out anomalies and duplicate records; Step 3: Set up time window rules for matching license plates upstream and downstream of the interval, match license plates and calculate the single vehicle travel time, extract the free flow travel time of each interval based on the historical single vehicle travel time, divide the single vehicle status of the interval in the current calculation period using the interval free flow time, and statistically analyze the sample size of the current calculation period by status group to determine the operating status of the interval in the current calculation period. Step 4: Based on the state discrimination results, filter the single-vehicle travel time samples in the current calculation period interval and judge the reliability of the interval travel time sample size; for the interval that meets the reliability calculation results, calculate the travel time of the current period interval. Step 5: Divide the current period interval travel time into the segments that make up the interval to obtain the online travel time estimate for the corresponding segments; wherein, during the division, the interval travel time is decomposed into a free-flow part and a delay part; the free-flow part is divided according to the proportion of free-flow time of each segment that make up the interval; the delay part is divided according to the proportion of queue length collected by radar in real time on each segment that make up the interval.

2. The method for estimating online travel time of road segments based on multi-source data fusion as described in claim 1, characterized in that, In step two, the preprocessing of checkpoint and radar data includes license plate data filtering and deduplication, and radar data validity determination. Radar data validity assessment is based on overall assessment and single record assessment. It identifies suspicious faulty radar equipment and single invalid records. Overall assessment is based on statistical rules. Suspicious fault modes include all traffic parameters being zero, occupancy rate being 100%, and abnormal traffic flow parameters. Based on the statistics of suspicious fault records and the distribution of statistical data within a day, a threshold is set for identification. The default threshold is that abnormal records exceed half of the total number of records in a day. Single record assessment includes traffic flow combination patterns and threshold ranges for each parameter.

3. The method for estimating online travel time of road segments based on multi-source data fusion as described in claim 1, characterized in that, Step three specifically includes: S3.1 Time Window Rule Settings: Set the interval formation time calculation period, and set the maximum time window for matching license plate sections upstream and downstream of the interval; S3.2 Interval travel time matching and calculation: Filter the vehicle passage records of downstream license plate sections and upstream license plate sections by calculation period and maximum matching time window, match the upstream and downstream license plates of the current period, and calculate the single vehicle travel time. S3.3 Free-flow travel time calculation: Based on the historical single-vehicle travel time of each section, the single-vehicle travel time of each section is sorted from high to low on a daily basis, and the 85th percentile travel time is taken as the free-flow travel time of the section; S3.4 Section single-vehicle status classification: The single-vehicle status of the section is classified according to the free-flow travel time of the section. The classification rules are shown in the table below: S3.5 Current Period Interval Status Classification: Calculate the ratio of single-vehicle travel time to free-flow speed within the current period. Classify single-vehicle status according to the interval single-vehicle status classification rules, then count the number of vehicles grouped by single-vehicle status. Determine the current calculation period interval operation status according to the interval status period classification rules, as shown in the table below: 。 4. The method for estimating online travel time of road segments based on multi-source data fusion as described in claim 3, characterized in that, The current cycle interval travel time in step four is specifically calculated as follows: when the interval cycle status is unobstructed, the interval travel time is calculated based on the unobstructed single vehicle travel time sample. When the interval period status is congested, the travel time is calculated using the travel time samples of unobstructed, crowded, and blocked single vehicles; when the interval period status is unknown, the interval travel time sample does not meet the calculation conditions, and the interval travel time is -1.

Images