A method for estimating urban road travel time based on complete and incomplete trajectories

By using data cleaning and KNN models of complete and incomplete trajectories in the spatiotemporal and spatially temporal areas of floating vehicles, the problem of low accuracy of travel time estimation is solved, and the efficiency and service level of the intelligent traffic information system are improved.

CN114202022BActive Publication Date: 2024-11-05KUNMING UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111474431.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-03
Publication Date
2024-11-05
Estimated Expiration
2041-12-03

Smart Images

  • Figure CN114202022B_ABST
    Figure CN114202022B_ABST
Patent Text Reader

Abstract

The present invention relates to a method for estimating the travel time of urban roads based on complete trajectories and incomplete trajectories, and belongs to the field of intelligent transportation technology. The present invention first performs data cleaning and map matching on floating vehicle data, then constructs floating vehicle trajectories, classifies the current trajectories in combination with the spatial coverage relationship between the trajectories and the target road sections, and implements data repair in the case of loss of global positioning system points in the floating vehicle trajectories, and then estimates the road section travel time of the complete trajectory, constructs a K nearest neighbor model, estimates the road section travel time of the incomplete trajectory, and finally averages the travel time of the complete trajectory and the incomplete trajectory to obtain the average travel time of the road section. Under the condition of a low floating vehicle penetration rate, the present invention makes full use of the incomplete trajectory information widely present in the road section, can help improve the accuracy of the average travel time estimation of the road section, and has high reliability and accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for estimating urban road travel time based on a complete trajectory and an incomplete trajectory, and belongs to the technical field of intelligent transportation. Background Art

[0002] Real-time and accurate travel time estimation is the foundation of intelligent traffic management and information service systems. Existing estimation methods are mostly based on fixed detectors, which are expensive to install and maintain and have limited coverage. Therefore, the scenarios are mostly limited to specific sections in urban road networks. With the rapid development of network communication technology and satellite positioning technology, massive vehicle trajectory data with larger wide-area space and continuous time has gradually become an important data source in the field of intelligent transportation. This type of data describes the spatiotemporal "imprints" left by floating vehicles equipped with GPS devices, such as taxis, buses, and Didi online ride-hailing cars, when they move in the road network, which can be called floating vehicle data.

[0003] At present, the floating car application systems in most cities in my country use the spatial average method based on large samples to calculate road travel time. However, due to the volatility and imbalance of the spatiotemporal distribution of floating cars in the urban road network, the number of floating cars passing through different sections at different times is too different, and the number of observed samples is too small or zero, resulting in the traditional large sample-based travel time estimation method being unable to achieve satisfactory results.

[0004] (1) The patent "Intermittent flow section travel time estimation method based on floating vehicle data and coil flow fusion (CN102930718A)" discloses a method for estimating the average travel time of a section in an intermittent flow environment. This method effectively integrates floating vehicle data and coil flow data, and estimates the section travel time based on the maximum floating vehicle coverage rate in the same period of history. Compared with the traditional floating vehicle travel time estimation, this method can achieve better results. However, this method requires a large sample size as input, and it is difficult to ensure high accuracy under the sparse spatiotemporal conditions of FCD.

[0005] (2) The patent "Road section travel time estimation method based on built environment and low-frequency floating vehicle data (CN106781468 A)" adds the built environment as an explanatory variable for the road section running time, and proves the explanatory power of the built environment for the road section running time through examples. This method can reflect the difference in running speed between different parts of the road section, effectively improving the accuracy of the road section travel time. Similarly, the spatiotemporal sparsity problem of FCD is likely to affect the estimation accuracy of this type of method.

[0006] (3) The patent "Travel time estimation method based on low-sampling historical trajectories (CN 106156531 B)" discloses a travel time estimation method based on low-sampling historical trajectories. The method is insensitive to the sampling rate of historical data and can achieve good results under low sampling rate data. However, the method uses a large number of complete trajectories in the road section. When the complete trajectory coverage rate is low, the method cannot show its superiority.

[0007] Based on the limitations of the above patents, it can be found that most of the existing patented travel time estimation methods estimate the travel time of road sections based on FCDs with a large sample size, which fails to effectively overcome the spatiotemporal sparsity problem of existing FCDs, and thus directly affects the accuracy of travel time estimation in intermittent flow environments. Therefore, how to estimate travel time and obtain highly reliable travel time information using sparse FCDs in areas with limited spatiotemporal coverage of floating vehicles is a core issue that needs to be solved urgently, and it has important practical significance for improving the utilization efficiency of existing traffic information resources and the level of traffic information services. Summary of the invention

[0008] The technical problem to be solved by the present invention is to propose a method for estimating urban road travel time based on complete trajectories and incomplete trajectories in view of the limitations of existing travel time estimation methods. Based on sparse floating vehicle trajectory data, the spatiotemporal K nearest neighbor algorithm is used to expand the sample size of travel time on the road section, and then the average travel time of the road section is estimated to overcome the shortcomings of the prior art.

[0009] The technical scheme of the present invention is: a method for estimating the travel time of urban roads based on complete trajectories and incomplete trajectories, firstly, data cleaning and map matching are performed on floating car data (FCD), then floating car trajectories are constructed, and the spatial coverage relationship between the trajectory and the target road section is combined to classify the current trajectory, and data repair is implemented when the global positioning system (GPS) point in the floating car trajectory is lost, and then the road section travel time of the complete trajectory is estimated, a K-nearest neighbor (KNN) model is constructed, and the road section travel time of the incomplete trajectory is estimated, and finally the travel time of the complete trajectory and the incomplete trajectory is averaged to obtain the average travel time of the road section.

[0010] The specific steps are:

[0011] Step 1: Use data preprocessing methods to improve the quality of the original floating car data. Use a topological weighted map matching algorithm to map and associate the floating car's global positioning system points to the urban road network.

[0012] The map matching algorithm in Step 1 is specifically as follows:

[0013] Step 1.1: Input trajectory data, including timestamp t, longitude x, latitude y, speed spe, and direction angle dir.

[0014] Step 1.2: Input the road network data, including the link ID number, the starting node SN and the ending node EN and their x, y coordinates.

[0015] Step 1.3: Calculate the proximity score and heading score of the candidate road segments, calculate the matching score, and select the candidate road segment with the highest matching score as the correct road segment for GPS positioning.

[0016] The specific steps for calculating the matching score in Step 1.3 are:

[0017]

[0018] Where P is the vertical distance between a GPS point to be matched and a candidate road section, B is the angle between the vehicle driving direction provided by the GPS point and the direction of the candidate road section, and w p is the weight of the impact of matching factor p on the matching score, P err The maximum error allowed for GPS positioning.

[0019] Step 2: Combined with the spatial coverage relationship between the trajectory and the target section, the current trajectory is classified into complete trajectory and incomplete trajectory, and data repair is implemented when the GPS point in the floating vehicle trajectory is lost.

[0020] The data repair in the case of GPS point loss in Step 2 is specifically as follows:

[0021] Step 2.1: Input trajectory Calculate neighboring points and The sampling interval is Δt seconds, 1 <n≤N。

[0022] Step 2.2: Define the triggering conditions of the data repair program as: 60 seconds ≤ Δt ≤ 300 seconds.

[0023] Step 2.3: If 60 seconds ≤ Δt ≤ 300 seconds, divide the time into H time slices according to the average sampling interval, and Among them, the last time slice dt H ≤30 seconds.

[0024] If the remainder of Δt / 30 is 0, then H=Δt / 30, otherwise, H=H+1.

[0025] Step 2.4: Set the initial cumulative distance and timestamp: S = 0 meters, seconds, in the time range of h = 1 to H-1, search for a set of GPS points from the trajectory database, with a time range of [T, T+30] and a spatial range of the current trajectory chain. and

[0026] Step 2.5: Calculate the average speed using the instantaneous speed of the selected GPS point set, labeled as aveSpeed(dt h ), and update the cumulative distance and timestamp:

[0027] S=S+avespeed(dt h )×30

[0028] T=T+30

[0029] Step 2.6: If Then add a new GPS point

[0030] If not, stop the calculation and return to Step 2.4 to find a new GPS point set until the H-1th time slice ends.

[0031] Step 3: Calculate the travel time of a single complete trajectory data segment as follows:

[0032] TT(Tr i )=Δt 2,3 +TT2(Tr i )+TT3(Tr i )

[0033] In the formula, TT(Tr i ) is the travel time of a bicycle section, TT2(Tr i ) is the travel time from the upstream intersection to the first GPS point on the road section, TT3(Tr i ) is the travel time from the last GPS point on the road section to the downstream intersection, Δt 2,3 is the travel time difference between the first GPS point and the last GPS point of the road section, that is, Δt 2,3 =t3-t2.

[0034] Step 4: Construct a KNN model for the travel time of incomplete trajectory data and obtain the travel time TT(Tr0) of a single road section based on incomplete trajectory data.

[0035] The Step 4 is specifically as follows:

[0036] According to the currently input incomplete trajectory data, find the K nearest complete trajectories closest to the incomplete trajectory from the spatiotemporal complete trajectory dataset, weight the travel time of the K complete trajectories, and finally obtain the travel time of a single section based on the incomplete trajectory data, as follows:

[0037]

[0038] Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory data set, TT(Tr0) is the travel time of the incomplete trajectory, and TT(Tr k ) is the travel time of the kth complete trajectory, w k is the weight of the kth complete trajectory.

[0039] The weight w of the kth complete trajectory k for:

[0040]

[0041] The method of finding the K nearest complete trajectories closest to the incomplete trajectory from the spatiotemporal complete trajectory dataset is implemented by quantifying the familiarity of the two trajectories, specifically:

[0042] The length of the spatial overlap between the incomplete trajectory and the complete trajectory is set to L meters, and divided into F equally spaced sub-segments. The similarity between the two trajectories is quantified by calculating the difference in travel time between the incomplete trajectory and the complete trajectory in multiple sub-segments, which can be expressed as:

[0043]

[0044] Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory dataset, TT f (Tr0) is the travel time spent by the incomplete trajectory on the f sub-segment, TT f (Tr k ) is the travel time spent by the complete trajectory on the f sub-segment.

[0045] Step 5: Get the average travel time of the road section through the travel time of the complete trajectory and the incomplete trajectory, which is:

[0046]

[0047] Where, TT ave is the average travel time of the road section, TT(Tr m ) is the travel time of the mth complete trajectory section, TT(Tr n) is the travel time of the nth incomplete trajectory segment, M and N represent the number of complete trajectories and incomplete trajectories in the segment, respectively.

[0048] The beneficial effects of the present invention are as follows: based on sparse floating vehicle trajectory data, the present invention makes full use of incomplete trajectory information in the road section, constructs a KNN model to increase the sample size that can be used for travel time estimation, and effectively improves the accuracy of the average travel time estimation of the road section. The present invention helps to enrich the processing methods and application fields of trajectory data, can add new research ideas and means for travel time estimation, and further meet the needs of the intelligent traffic control system for traffic information of the entire road network, thereby improving the efficiency of the transportation system and the travel experience of travelers. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the process of the present invention;

[0050] Figure 2 It is a schematic diagram of trajectory classification;

[0051] Figure 3 It is a schematic diagram of trajectory repair;

[0052] Figure 4 is the trajectory repair pseudocode;

[0053] Figure 5 It is the spatiotemporal information graph of the KNN model;

[0054] Figure 6 This is a schematic diagram of travel time estimation based on the travel decomposition method. DETAILED DESCRIPTION

[0055] 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.

[0056] Example 1: Figure 1 As shown in FIG. 1 , a method for estimating urban road travel time based on complete trajectories and incomplete trajectories is shown in FIG. 1 . The specific steps are as follows:

[0057] Step 1: Perform data preprocessing and map matching on the original FCD. The specific steps are:

[0058] Step 1.1: Clean the original FCD data to solve the problems of duplicate records, data anomalies, and direction angle anomalies;

[0059] Step 1.2: Use the topological weighted map matching algorithm on the cleaned trajectory data to map and associate the floating vehicle GPS points with the urban road network to form a driving path. The specific map matching algorithm is as follows:

[0060] (a) Input trajectory data, including timestamp (t), longitude (x), latitude (y), speed (spe), and direction angle (dir);

[0061] (b) Input road network data, including link ID number, start node (SN) and end node (EN) and their x, y coordinates;

[0062] (c) Calculate the proximity score and heading score of the candidate road segment, and calculate the matching score, and select the candidate road segment with the highest matching score as the correct road segment for GPS positioning. The matching score of each candidate road segment can be defined as:

[0063]

[0064] Where P is the vertical distance between a GPS point to be matched and a candidate road section, B is the angle between the vehicle driving direction provided by the GPS point and the direction of the candidate road section, and w p is the weight of the impact of matching factor p on the matching score, P err The maximum error value allowed for GPS positioning;

[0065] Step 2: Construct the floating vehicle trajectory, then classify the trajectory based on the spatial coverage relationship between the trajectory and the target road section, and repair the trajectory with missing GPS points. The specific steps are as follows:

[0066] Step 2.1: According to the length of the floating vehicle passing through the target section, the floating vehicle trajectory is subdivided into 4 types, namely:

[0067] (1) Complete trajectory, i.e., the trajectory of the floating vehicle that has passed the upstream and downstream intersections of the road section;

[0068] (2) Incomplete trajectory type 1, i.e., the floating vehicle trajectory that enters from the upstream segment but leaves the target segment before the downstream intersection;

[0069] (3) Incomplete trajectory type 2, i.e., the trajectory of a floating vehicle that enters after the upstream intersection and passes through the downstream intersection;

[0070] (4) Incomplete trajectory type 3, i.e., the trajectory of a floating vehicle that enters after the upstream intersection and leaves before the downstream intersection.

[0071] Step 2.2: Repair lost GPS points. The specific method is as follows:

[0072] (a) Input trajectory Calculate neighboring points and The sampling interval Δt seconds (1 <n≤N);

[0073] (b) The triggering conditions of the data repair procedure are defined as: 60 seconds ≤ Δt ≤ 300 seconds;

[0074] (c) If 60 seconds ≤ Δt ≤ 300 seconds, then divide the average sampling interval into H (if the remainder of Δt / 30 is 0, then H = Δt / 30; otherwise, H = H + 1) time slices, and there are The last time slice dt H Should meet dt H ≤30 seconds;

[0075] (d) Set the initial cumulative distance and timestamp: S = 0 meters, seconds, in the range of h = 1 to H-1 time slices, a set of GPS points is searched from the trajectory database, with a time range of [T, T+30] and a spatial range of the current trajectory chain. and

[0076] (e) Calculate the average speed (labeled as aveSpeed(dt)) using the instantaneous speed of the selected GPS point set. h )) and update the cumulative distance and timestamp:

[0077] S=S+avespeed(dt h )×30

[0078] T=T+30

[0079] (f) If Then add a new GPS point: If not, the calculation is stopped and the process returns to S24 to search for a new GPS point set until the H-1th time slice ends.

[0080] Step 3: Estimation of the travel time of a single complete trajectory data segment. The specific algorithm is as follows:

[0081] TT(Tr i )=Δt 2,3 +TT2(Tr i )+TT3(Tr i )

[0082] In the formula, TT(Tr i ) is the travel time of a bicycle section; TT2(Tr i ) is the travel time from the upstream intersection to the first GPS point of the road section; TT3(Tr i) is the travel time from the last GPS point of the road section to the downstream intersection; Δt 2,3 is the travel time difference between the first GPS point and the last GPS point of the road section, that is, Δt 2,3 =t3-t2.

[0083] Step 4: Construct a KNN model to estimate the travel time of incomplete trajectory data. The steps are as follows:

[0084] Step 4.1: Construct a complete spatiotemporal trajectory dataset and determine the spatiotemporal search range of the matching mechanism;

[0085] (a) Determine the time search range and select the time period with high correlation with the current travel time as the time range of the matching mechanism;

[0086] (b) Determine the spatial search range and select the current section covered by the incomplete trajectory and its spatially associated upstream and downstream sections as the spatial range of the matching mechanism;

[0087] Step 4.2: Measure the similarity between trajectories. Based on the currently input incomplete trajectory data, find the K nearest complete trajectories that best match the incomplete trajectory from the spatiotemporal complete trajectory dataset. The trajectory similarity measurement method is as follows:

[0088] The length of the spatial overlap between the incomplete trajectory and the complete trajectory is set to L meters, and divided into F equally spaced sub-segments. The similarity between the two trajectories is quantified by calculating the difference in travel time between the incomplete trajectory and the complete trajectory in multiple sub-segments, which can be expressed as:

[0089]

[0090] Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory dataset, TT f (Tr0) is the travel time spent by the incomplete trajectory on the f sub-segment, TT f (Tr k ) is the travel time spent by the complete trajectory on the f sub-segment.

[0091] Step 4.3: Weight the travel time of K complete trajectories, and finally obtain the travel time of a single section based on incomplete trajectory data, so as to increase the sample size of travel time. The specific calculation formula is as follows:

[0092]

[0093] Where Tr0 is the incomplete trajectory, Tr kis the kth complete trajectory in the trajectory data set, TT(Tr0) is the travel time of the incomplete trajectory, and TT(Tr k ) is the travel time of the kth complete trajectory, w k is the weight of the kth complete trajectory, and the calculation formula is as follows:

[0094]

[0095] Step 5: Take the arithmetic average of the travel time of the complete trajectory and the incomplete trajectory to get the average travel time of the road section. The specific calculation formula is as follows:

[0096]

[0097] Where, TT ave is the average travel time of the road section, TT(Tr m ) is the travel time of the mth complete trajectory section, TT(Tr n ) is the travel time of the nth incomplete trajectory segment, M and N represent the number of complete trajectories and incomplete trajectories in the segment, respectively.

[0098] Example 2: Figure 1 As shown, a method for estimating urban road travel time based on complete trajectories and incomplete trajectories includes the following steps:

[0099] Step 1: Perform data preprocessing and map matching on the original FCD. The specific steps are as follows:

[0100] Step 1.1: The present invention takes the GPS trajectory data of more than 7,700 taxis collected in Nanjing from November 1 to December 6, 2011 as an example for analysis. In view of the problems of repeated records, abnormal speed, abnormal direction angle, etc. in the original data, the original FCD is cleaned according to the following steps:

[0101] (1) Delete data outside the Nanjing area;

[0102] (2) Delete data that are duplicate or missing;

[0103] (3) Delete data with a speed less than 0 or greater than 120 km / h;

[0104] (4) Delete data with a direction angle greater than 360 degrees;

[0105] (5) Delete the data with a direction angle of 0 degrees and a speed of 0 to 3 km / h.

[0106] Regarding the data processing step 5, it should be noted that this part of abnormal data is mainly caused by the static drift of GPS positioning when the vehicle is driving at a low speed, which makes it impossible to accurately determine the direction of the vehicle. In addition, since the latitude and longitude coordinate system of the original FCD is a geographic coordinate system (in degrees), it needs to be converted into a projection coordinate system (in meters).

[0107] Step 1.2: Based on the direction angle information and location information provided by the GPS points, a topological weighted map matching algorithm is implemented to match the FCD with the road network data of the main urban areas of Nanjing. The specific algorithm steps are as follows:

[0108] (a) Input trajectory data, including timestamp (t), longitude (x), latitude (y), speed (spe), and direction angle (dir);

[0109] (b) Input road network data, including link ID number, start node (SN) and end node (EN) and their x, y coordinates;

[0110] (c) Calculate the proximity score and heading score of the candidate road segment, and calculate the matching score of each candidate road segment, and select the candidate road segment with the highest matching score as the final matching road segment. The matching score can be defined as:

[0111]

[0112] Where P is the vertical distance between a GPS point to be matched and a candidate road section, B is the angle between the vehicle driving direction provided by the GPS point and the direction of the candidate road section, and w p is the weight of the impact of matching factor p on the matching score, P err The maximum error allowed by GPS positioning is set to 30 meters in this study. By randomly selecting 100 original GPS data on December 1, 2011 as test samples, the average accuracy rate of successfully restoring its driving trajectory reached 91.8%.

[0113] Step 2: Construct the floating vehicle trajectory, then classify the trajectory based on the spatial coverage relationship between the trajectory and the target road section, and repair the trajectory with missing GPS points. The specific steps are as follows:

[0114] Step 2.1: Construct the floating car trajectory and record is the trajectory Tr generated by the i-th floating vehicle in the road network i ,in, It is the nth GPS point in the trajectory chain, and includes the vehicle number id, longitude (x), latitude (y), timestamp (t), heading (dir), speed (spe) and passenger status (sta), and n=1,…,N.

[0115] Step 2.2: According to the length of the target road section passed by the floating vehicle, the floating vehicle trajectory is subdivided into four types, such as Figure 2 As shown, that is:

[0116] (1) Complete trajectory, i.e., the trajectory of the floating vehicle that has passed the upstream and downstream intersections of the road section;

[0117] (2) Incomplete trajectory type 1, i.e., the floating vehicle trajectory that enters from the upstream segment but leaves the target segment before the downstream intersection;

[0118] (3) Incomplete trajectory type 2, i.e., the trajectory of a floating vehicle that enters after the upstream intersection and passes through the downstream intersection;

[0119] (4) Incomplete trajectory type 3, i.e., the trajectory of a floating vehicle that enters after the upstream intersection and leaves before the downstream intersection.

[0120] Step2.3: Repair lost GPS points, such as Figure 4 As shown, the specific method is as follows:

[0121] (a) Input trajectory Calculate neighboring points and The sampling interval Δt seconds (1 <n≤N);

[0122] (b) The triggering conditions of the data repair procedure are defined as: 60 seconds ≤ Δt ≤ 300 seconds;

[0123] (c) If 60 seconds ≤ Δt ≤ 300 seconds, then divide the average sampling interval into H (if the remainder of Δt / 30 is 0, then H = Δt / 30; otherwise, H = H + 1) time slices, and there are The last time slice dt H Should meet dt H ≤30 seconds;

[0124] (d) Set the initial cumulative distance and timestamp: S = 0 meters, seconds, in the range of h = 1 to H-1 time slices, a set of GPS points is searched from the trajectory database, with a time range of [T, T+30] and a spatial range of the current trajectory chain. and

[0125] (e) Calculate the average speed (labeled as aveSpeed(dt)) using the instantaneous speed of the selected GPS point set. h )) and update the cumulative distance and timestamp:

[0126] S=S+avespeed(dt h )×30

[0127] T=T+30

[0128] (f) If Then add a new GPS point: If not, the calculation is stopped and the process returns to S24 to search for a new GPS point set until the H-1th time slice ends.

[0129] Step 3: Estimation of the travel time of a single complete trajectory data segment. The specific algorithm is as follows:

[0130] TT(Tr i )=Δt 2,3 +TT2(Tr i )+TT3(Tr i )

[0131] In the formula, TT(Tr i ) is the travel time of a bicycle section; TT2(Tr i ) is the travel time from the upstream intersection to the first GPS point of the road section; TT3(Tr i ) is the travel time from the last GPS point of the road section to the downstream intersection; Δt 2,3 is the travel time difference between the first GPS point and the last GPS point of the road section, that is, Δt 2,3 = t3-t2;

[0132] Step 4: Calculate the travel time of the road segment for incomplete trajectory data

[0133] Construct a KNN model to generate additional travel time samples. According to the currently input incomplete trajectory data, find the nearest neighbor that matches the incomplete trajectory from the spatiotemporal trajectory dataset, and finally obtain the section travel time based on the incomplete trajectory data. The steps are:

[0134] Step 4.1: Construct a spatiotemporal trajectory dataset. Figure 5 Describes the spatiotemporal search scope of the KNN model. In terms of time search, it is assumed that the start time of the analysis is t i (Take 9:00 as an example), the selected time period set includes: current time period t i ~t i+1 (9:00~9:15), 1 period before the current period t i-1 ~t i (8:45~9:00), 2 periods before the current period t i-2 ~t i-1 (8:30~8:45), and historical period t i ~t i+1 (9:00~9:15), the previous period of historical period t i-1 ~ti (8:45~9:00), the first two periods of the historical period t i-2 ~t i-1 (8:30~8:45), 1 period after the historical period t i+1 ~t i+2 (9:15~9:30). In the spatial search level, assuming that the current section is Link(x j+1 →x j ), combined with the characteristics of the three incomplete trajectory types, the selected segment set includes: the current segment Link (x j+1 →x j ), upstream section Link(x j+2 →x j+1 ) and the downstream link Link(x j →x j-1 ).

[0135] Step 4.2: Similarity measurement between trajectories. By analyzing the similarity between the travel time of the incomplete trajectory and the complete trajectory on multiple sections, the similarity comparison between the two is completed. The difference in travel time between the incomplete trajectory and the complete trajectory on multiple sub-sections is calculated to quantify the similarity between the two trajectories, which can be expressed as:

[0136]

[0137] Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory dataset, TT f (Tr0) is the travel time spent by the incomplete trajectory on the f sub-segment, TT f (Tr k ) is the travel time of the complete trajectory on the f sub-segment. The travel time calculation method for the complete trajectory and the incomplete trajectory on the sub-segment is Figure 6 Describes the travel time decomposition method.

[0138] Step 4.3: The specific calculation formula for the travel time of the road section based on the incomplete trajectory is as follows:

[0139]

[0140] Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory data set, TT(Tr0) is the travel time of the incomplete trajectory, and TT(Tr k ) is the travel time of the kth complete trajectory, w k is the weight of the kth complete trajectory, and the calculation formula is as follows:

[0141]

[0142] Step 5: Take the arithmetic average of the travel time of the complete trajectory and the incomplete trajectory to get the average travel time of the road section. The specific calculation formula is as follows:

[0143]

[0144] Where, TT ave is the average travel time of the road section, TT(Tr m ) is the travel time of the mth complete trajectory section, TT(Tr n ) is the travel time of the nth incomplete trajectory segment, M and N represent the number of complete trajectories and incomplete trajectories in the segment, respectively.

[0145] The specific implementation modes of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above implementation modes, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A method for estimating urban road travel time based on complete trajectories and incomplete trajectories, characterized by: Step 1: Use data preprocessing methods to improve the quality of the original floating vehicle data; use a topology-weighted map matching algorithm to map and associate the floating vehicle's global positioning system points to the urban road network; Step 2: Based on the spatial coverage relationship between the trajectory and the target road section, the current trajectory is classified into a complete trajectory and an incomplete trajectory, and data repair is implemented when the GPS point in the floating vehicle trajectory is lost; Step 3: Calculate the travel time of a single complete trajectory data segment as follows: TT(Tr i )=Δt 2,3 +TT2(Tr i )+TT3(Tr i ) In the formula, TT(Tr i ) is the travel time of a bicycle section, TT2(Tr i ) is the travel time from the upstream intersection to the first GPS point on the road section, TT3(Tr i ) is the travel time from the last GPS point on the road section to the downstream intersection, Δt 2,3 is the travel time difference between the first GPS point and the last GPS point of the road section, that is, Δt 2,3 = t3-t2; Step 4: Construct a KNN model for the travel time of incomplete trajectory data to obtain the travel time TT(Tr0) of a single road section based on incomplete trajectory data; Step 5: Get the average travel time of the road section through the travel time of the complete trajectory and the incomplete trajectory, which is: Where, TT ave is the average travel time of the road section, TT(Tr m ) is the travel time of the mth complete trajectory section, TT(Tr n ) is the travel time of the nth incomplete trajectory segment, M and N represent the number of complete trajectories and incomplete trajectories in the segment, respectively; Step 2: Data recovery when GPS points are lost is as follows: Step 2.1: Input trajectory Calculate neighboring points and The sampling interval is Δt seconds, 1 <n≤N; Step 2.2: Define the triggering conditions of the data repair program as: 60 seconds ≤ Δt ≤ 300 seconds; Step 2.3: If 60 seconds ≤ Δt ≤ 300 seconds, divide the time into H time slices according to the average sampling interval, and Among them, the last time slice dt H ≤30 seconds; If the remainder of Δt / 30 is 0, then H = Δt / 30, otherwise, H = H + 1; Step 2.4: Set the initial cumulative distance and timestamp: S = 0 meters, seconds, in the time range of h = 1 to H-1, search for a set of GPS points from the trajectory database, with a time range of [T, T+30] and a spatial range of the current trajectory chain. and Step 2.5: Calculate the average speed using the instantaneous speed of the selected GPS point set, labeled as aveSpeed(dt h ), and update the cumulative distance and timestamp: S=S+avespeed(dt h )×30 T=T+30 Step 2.6: If Then add a new GPS point If not, stop the calculation and return to Step 2.4 to find a new GPS point set until the H-1th time slice ends; The Step 4 is specifically as follows: According to the currently input incomplete trajectory data, find the K nearest complete trajectories closest to the incomplete trajectory from the spatiotemporal complete trajectory dataset, weight the travel time of the K complete trajectories, and finally obtain the travel time of a single section based on the incomplete trajectory data, as follows: Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory data set, TT(Tr0) is the travel time of the incomplete trajectory, and TT(Tr k ) is the travel time of the kth complete trajectory, w k is the weight of the kth complete trajectory; The weight w of the kth complete trajectory k for:

2. The urban road travel time estimation method based on complete trajectory and incomplete trajectory according to claim 1 is characterized in that: The map matching algorithm described in Step 1 is as follows: Step 1.1: Input trajectory data, including timestamp t, longitude x, latitude y, speed spe, and direction angle dir; Step 1.2: Input the road network data, including the link ID number, the starting node SN and the ending node EN and their x, y coordinates; Step 1.3: Calculate the proximity score and heading score of the candidate road segments, calculate the matching score, and select the candidate road segment with the highest matching score as the correct road segment for GPS positioning.

3. The urban road travel time estimation method based on complete trajectory and incomplete trajectory according to claim 2 is characterized in that: The specific steps for calculating the matching score in Step 1.3 are: Where P is the vertical distance between a GPS point to be matched and a candidate road section, B is the angle between the vehicle driving direction provided by the GPS point and the direction of the candidate road section, and w p is the weight of the impact of matching factor p on the matching score, P err The maximum error allowed for GPS positioning.

4. The urban road travel time estimation method based on complete trajectory and incomplete trajectory according to claim 1 is characterized by: The method of finding the K nearest complete trajectories closest to the incomplete trajectory from the spatiotemporal complete trajectory dataset is implemented by quantifying the familiarity of the two trajectories, specifically: The length of the spatial overlap between the incomplete trajectory and the complete trajectory is set to L meters, and divided into F equally spaced sub-segments. The similarity between the two trajectories is quantified by calculating the difference in travel time between the incomplete trajectory and the complete trajectory in multiple sub-segments, which can be expressed as: Where Tr0 is the incomplete trajectory, Tr k is the kth complete trajectory in the trajectory dataset, TT f (Tr0) is the travel time spent by the incomplete trajectory on the f sub-segment, TT f (Tr k ) is the travel time spent by the complete trajectory on the f sub-segment.