JP6528660B2 - Traffic congestion prediction method - Google Patents

Description

  The present invention relates to a congestion estimation method for predicting congestion in a traffic network using probe information.

  A probe information system in which a probe center receives probe information (vehicle position, traveling road, etc.) which is information related to traveling from a plurality of vehicles (probe cars), and the probe center estimates traffic information from the received probe information. It has been put to practical use. For example, Patent Document 1 discloses a technique for predicting traffic congestion based on probe information.

  In this probe information system, the probe center sequentially receives probe information from a plurality of vehicles traveling on various roads (hereinafter referred to as "links"), and stores it in a database as a history of probe information configured in time series for each link. accumulate. And estimation of traffic information, such as prediction of traffic congestion, is performed based on this probe information history.

JP, 2013-200809, A

  It is free and not regular that which link the vehicle travels at which time. Thus, for example, with regard to any link, the vehicle is traveling at a certain time, but may not be traveling at the next time.

  In the above probe information system, probe information is collected for links on which a vehicle is traveling, but probe information is not collected for links on which a vehicle is not traveling. For this reason, for example, in an arbitrary link as described above, a time zone in which probe information is not collected occurs. Therefore, in the history of probe information accumulated in the database, information loss occurs in time series. The loss of the probe information may deteriorate the accuracy of the traffic congestion prediction.

  The present invention has been made in view of the above problems, and there is a risk that the accuracy of traffic congestion prediction based on probe history information may be degraded even if a defect occurs in the probe history information configured in time series for each link. An object of the present invention is to provide a traffic congestion prediction method that can be suppressed.

In order to solve the above problems, the traffic congestion prediction method of the present invention receives probe information including at least travel time of a link that is a road segment from a vehicle traveling on the link, stores it in a database, and stores it in the database. A traffic congestion prediction method executed by a traffic congestion prediction apparatus based on probe history information in which probe information is configured in time series for each link, and a probe in which a time zone in which information is missing exists based on probe history information of a database the method comprising the steps of: detecting a link with the history information as the target link, for the detected target link, information is the day and the same day and the information is travel time based on the missing information of the target link of the time zone prior to the time zone missing subject phosphorus hours after the first curve and date information is missing and times the same day and information is missing is the change curve Is the change curve of travel time based on the information expected from the second curve, calculating the statistical processing lines and curves information representing a change in travel time of the missing time period, the calculated straight lines and curves Based on the step of estimating the information of the time zone in which the information of the target link is missing, and complementing the probe history information, and based on the probe history information in which the information of the missing time zone is complemented, And the step of

  In the traffic congestion prediction method of the present invention, a target link having probe history information in which a time zone in which information is missing exists is detected from among the probe history information configured in time series for each link. Then, if the target link is detected, the travel time of the time zone in which the information of the target link is missing is estimated and complemented based on the change of the travel time before and after the time zone. Specifically, the change in travel time in the information missing time zone predicted from the change in information in the time prior to the information missing time zone and the change in the information in the time zone later than the relevant time zone Find the forecast straight line or forecast curve to represent.

According to this processing, in the traffic congestion prediction method of the present invention, the prediction straight line representing the change in travel time in the information missing time zone predicted from the change in information in the time zones before and after the time zone where the information of the target link is missing Based on the prediction curve, it is possible to infer information of a time zone in which information in the target link is missing. Therefore, probe history information in the target link can be complemented by this inferable information.

  As described above, according to the traffic congestion prediction method of the present invention, even if there is a defect in the probe history information configured in time series for each link, the information of the missing time zone can be complemented. Therefore, it is possible to suppress the possibility that the accuracy of traffic congestion prediction based on the probe history information may be degraded.

A diagram showing a schematic configuration of a traffic congestion prediction system in which a traffic congestion prediction method according to an embodiment of the present invention is executed A flowchart illustrating the procedure of probe prediction processing executed by the probe prediction processing unit of the traffic congestion prediction device A flowchart illustrating the procedure of probe statistical processing performed by the probe statistical processing unit of the traffic congestion prediction apparatus

[Overview]
The present invention is a congestion estimation method for predicting congestion in a traffic network using probe information transmitted from a vehicle. In this congestion estimation method, a link (target link) in which a time zone in which information is missing exists in probe history information configured in time series for each link which is a road segment is detected, and information on the missing time zone is detected. It complements based on the information change before and behind the time zone concerned. Therefore, when performing traffic congestion prediction based on probe historical information, it is possible to suppress the possibility that the accuracy of traffic congestion prediction may deteriorate.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[Overall configuration of traffic congestion prediction system]
FIG. 1 is a diagram showing a schematic configuration of a traffic congestion prediction system 1 in which a traffic congestion prediction method according to an embodiment of the present invention is performed. The traffic congestion prediction system 1 comprises a plurality of vehicles (hereinafter referred to as "probe cars") 10 for transmitting probe information Pb, and a traffic congestion prediction apparatus 20 for collecting congestion information by collecting probe information Pb transmitted from the probe car 10. Composed of

  The probe information Pb is information on the traveling of the probe car 10, and includes, for example, an identifier for specifying the probe car 10 of the information transmission source, information on the traveling of the probe car 10, and the like. The probe information Pb is generated, for example, based on data obtained from a GPS (Global Positioning System) receiver mounted on the probe car 10 and various sensors (not shown). The probe car 10 transmits the generated probe information Pb to the traffic congestion prediction apparatus 20 via a wireless communication unit (not shown).

  The traffic congestion prediction apparatus 20 includes a probe information reception unit 21, a probe travel information database 22, a probe prediction processing unit 24, a probe statistical processing unit 25, a probe history database 26, and a probe prediction database 27. .

  The probe information receiving unit 21 sequentially receives probe information Pb transmitted respectively by the plurality of probe cars 10 via the wireless communication unit. Then, the probe information receiving unit 21 stores the sequentially received probe information Pb in the probe travel information database 22. In the probe travel information database 22, a plurality of pieces of probe information Pb received by the probe information receiving unit 21 are stored in time series for each link. The probe travel information includes link ID, link distance, route number, start / end coordinates, link entry / exit time, and the like.

  The probe prediction processing unit 24 obtains the travel time of each link (the time required for the vehicle to travel on the entire link) based on the plurality of probe information Pb accumulated in the probe travel information database 22, and from this travel time Calculate the congestion degree of each link (probe prediction function). The calculated congestion degree of each link is stored in the probe prediction database 27 and is used for congestion prediction. The link whose congestion degree is stored in the probe prediction database 27 can be, for example, a link of a secondary mesh area partitioned by a 10 km × 10 km rectangular area defined by JIS.

  The probe statistic processing unit 25 generates information of a missing time zone out of the information arranged in time series of each link in the information accumulated in the probe history database 26, and probe statistics having a time zone where the information is missing Create information (probe statistics function).

[Process Performed by Traffic Congestion Predictor]
Hereinafter, the probe prediction function and the probe statistical function realized by the traffic congestion prediction apparatus 20 will be described with further reference to FIGS. 2 and 3. FIG. 2 is a flow chart for explaining the procedure of probe prediction processing performed by the probe prediction processing unit 24 of the traffic congestion prediction device 20. FIG. 3 is a flow chart for explaining the procedure of probe statistical processing performed by the probe statistical processing unit 25 of the traffic congestion prediction apparatus 20.

(1) Probe prediction function This probe prediction function is based on the congestion degree obtained from the history of the past probe information Pb accumulated in the probe travel information database 22 and the congestion degree obtained from the current probe information Pb. It is a function to predict the degree of traffic congestion in the future. The probe prediction function is realized, for example, by the probe prediction processing unit 24 executing the processing described below at a predetermined timing.

  Please refer to FIG. When the processing is started, the probe prediction processing unit 24 first acquires, from the probe travel information database 22, probe information Pb related to all travel link strings for the last several minutes (for example, 5 minutes) from the time of processing start (step S201). .

  When the probe information Pb is acquired from the probe traveling information database 22, the probe prediction processing unit 24 determines that the traveling link row on the route traveled by the probe car 10 (user) is abnormal based on the vehicle speed information obtained from the probe information Pb. It is determined whether or not there is a link indicating a certain vehicle speed (step S202). Specifically, the probe prediction processing unit 24 determines whether there is a link having vehicle speed information exceeding a predetermined upper limit vehicle speed or a link having vehicle speed information falling below a predetermined lower limit vehicle speed (step S202). Then, when there is a link indicating an abnormal vehicle speed, the probe prediction processing unit 24 excludes the link from the target of the probe prediction process (step S203). This process is also called noise removal process, and is performed for each probe car 10 (user).

  When there is a link having vehicle speed information exceeding the upper limit vehicle speed and a link having vehicle speed information lower than the lower limit vehicle speed in the traveling link row on the route on which the probe car 10 travels, not only the link All links constituting the traveling link string including the links may be collectively excluded from the target of the probe prediction process.

  When the noise removal processing is completed for each probe car 10 (user), the probe prediction processing unit 24 determines from the departure point (Origin) corresponding to the processing target held in advance in the traveling link row after noise removal. It is determined whether an OD route exists up to a destination (Destination) (step S204). In this determination, when it is determined that the OD route exists, the probe prediction processing unit 24 further determines whether there is a link section (spatial missing link) not having the probe information Pb in the OD route. It judges (step S205). The determination in step S205 is performed, for example, as follows.

  In the OD route, if the distance difference between the end coordinates of the front link and the start coordinate of the rear link spatially continuous in the order in which the probe car 10 travels is larger than a predetermined distance, the front link and the rear It is determined that the link with the link is missing. Alternatively, in the OD route, if the time difference between the exit time from the spatially continuous front link and the entrance time to the back link in the order in which the probe cars 10 traveled is longer than a predetermined time, It is determined that the link between the link and the back link is missing.

  If it is determined that the link between the front link and the rear link is missing, the probe prediction processing unit 24 determines the vehicle speed of the missing link and the probe information Pb of the rear link based on the probe information Pb on the front link and the probe information Pb on the rear link. The travel time is calculated (step S206). The vehicle speed of the missing link can be calculated by calculating the average of the vehicle speed of the front link and the vehicle speed of the rear link. In addition, the travel time of the missing link can be obtained from the length of the missing link obtained from the coordinates and the vehicle speed of the missing link. If there are multiple missing links in the OD route, the vehicle speed and travel time are calculated for each missing link. This process complements the OD route where some links were missing. This process is also referred to as surface interpolation process, and is performed for each probe car 10 (user).

  When the surface interpolation processing is completed for each probe car 10 (user), the probe prediction processing unit 24 performs weighted moving average processing for the travel time of the same link by all the probe cars 10 (user) for each of all the links. The harmonic averaging process is executed to smooth the travel time of each link on a link basis (step S207). This yields a statistical travel time in link units (or mesh units).

  Then, when the travel time smoothed in link units is obtained, the probe prediction processing unit 24 calculates the congestion degree in link units from the travel time in link units smoothed (step S208). By this, the traffic congestion level in link units can be obtained. The congestion degree is an index for evaluating the degree of congestion, and is, for example, information indicating the degree of congestion in stages (“congestion”, “congestion”, “vacancy”, etc.). The travel time in the link unit and the calculated congestion degree in the link unit are stored in the probe history database 26 as the congestion degree in the corresponding date and time zone.

Finally, the probe prediction processing unit 24 acquires travel times and congestion degrees of all links in a predetermined corresponding date and time zone from the probe history database 26 (step S209). Then, the probe prediction processing unit 24 calculates an error average of the traffic congestion level while performing weighting in accordance with the number of samples in link units (step S210). Then, the probe prediction processing unit 24 ranks the minimum errors in the error averages of the calculated traffic congestion levels, obtains the average of the travel times for several pieces of information ranked higher, and calculates the average of the obtained travel times The predicted degree of congestion is calculated (step S210). The calculation result is stored in the probe prediction database 27.

  The congestion degree stored in the probe prediction database 27 is used as follows. When predicting the future traffic congestion level of a travel link string, for example, the traffic congestion degree at time t of the travel link string in the secondary mesh area and a predetermined time around time t closest to the traffic congestion status at time t Information that extracts the congestion level of the relevant traveling link string within x (for example, 30 minutes), and changes the congestion level from time t to time t + x or time t−x predicts the congestion degree in the future (after time t + x) It is used as

(2) Probe Statistics Function The probe statistics function is performed independently of the above-described probe prediction function. This probe statistical function generates and supplements information of a missing time zone among the information arranged in time series of each link in the information (travel time and congestion degree in link unit) accumulated in the probe history database 26 It is a function. The probe statistics function is performed periodically, for example, at predetermined time intervals.

  Please refer to FIG. When the process is started, the probe statistical processing unit 25 acquires, for example, information on travel times and congestion degrees of all links for the previous day from the probe history database 26 (step S301).

  When the information is acquired from the probe history database 26, the probe statistical processing unit 25 determines whether there is a link time zone (temporal missing link) having no information in the information arranged in time series in link units. It judges (step S302). At that time, if the information (travel time in link unit and congestion degree) is information already subjected to surface complementation by the above-described probe prediction function, the probe statistical processing unit 25 also receives the time zone having the information. Judge as link time zone where information is missing. The determination in step S302 is performed, for example, as follows.

  If, on a certain link, the time difference between the information acquisition date and time of the front probe information that is temporally continuous and the information acquisition date and time of the rear probe information is longer than a predetermined time, between the front probe information and the rear probe information It is determined that the probe information of the time zone is missing.

When it is determined that information of a predetermined time zone is missing in an arbitrary link, the probe statistical processing unit 25 uses the arbitrary link as a processing target link and is sufficient for calculating information of a time zone in which the information is missing. It is determined whether or not a predetermined number or more of samples are collected (step S303). The number of samples mentioned here means the number of pieces of information in the time zone before and after the time zone in which the information is missing. When the probe statistical processing unit 25 determines that the number of samples is equal to or more than a predetermined number, the probe statistical processing unit 25 calculates the travel time of the information missing time zone based on the information before and after the predetermined time zone (step S304). The calculation of the travel time of the information missing time zone is performed as follows.

  The probe statistic processing unit 25 generates a curve (pre-curve) based on travel time change in a time zone (for example, 15 minutes) before a time zone in which information is lost, and a time zone (for example, 15 minutes) Curves (post-curves) based on the travel time change of) are determined respectively. Next, the probe statistical processing unit 25 determines a straight line or a curve (a first-order line, a second-order curve, a third-order curve) of a travel time change predicted in a time zone where information is missing based on the front curve and the rear curve. , Or draw a logarithmic curve). Then, the probe statistical processing unit 25 estimates the travel time of an arbitrary point on the drawn straight line or curve as the travel time of the time zone in which the information is missing.

Averaging the travel line time that the estimated information missing time zone, by calculating the degree of congestion from travel times were averaged to create a probe statistics (step S305). In this process, if probe statistical information already exists, a new statistical value ts to be updated is determined using the following equation. Here, t0 is the statistical travel time stored in the probe history database 26, n0 is the number of statistical samples stored in the probe history database 26, and t1 is all corresponding to the factors existing in the probe history database 26. Travel time, n1 is the number of samples present in the probe history database 26, and ns is the number of new statistical samples to be updated.
ts = (t0 × n0 + Σt1) / (n0 + n1)
ns = n0 + n1

  The generated probe statistical information is stored in the probe prediction database 27 (step S306). Through this processing, time-series probe statistical information on the target link for which information on a part of time zones is missing is complemented. This process is also referred to as time series complementing process. The probe statistical information supplemented with the missing information is used for the subsequent traffic congestion prediction.

[Operation and Effect According to Embodiment]
A traffic congestion prediction method according to an embodiment of the present invention includes a target link having probe history information having a time zone in which information is missing from among probe history information in which probe information Pb is configured in time series for each link. To detect. Then, if the target link is detected, the information of the missing time zone of the target link is based on the change of the information of the time zone before the time zone and the change of the information of the time zone after the time zone. Complement.

According to this processing, in the traffic congestion prediction method of the present invention, a prediction straight line representing a change in travel time in a time zone in which information is lost predicted from a change in information in time zones before and after a time zone in which information in a target link is lost Based on the prediction curve , it is possible to infer information change in a time zone in which information in the target link is missing. Therefore, the information of the missing time zone in the target link can be complemented by this inferable information.

  The traffic congestion prediction method of the present invention is useful when it is desired to suppress the possibility that the accuracy of traffic congestion prediction based on probe history information may deteriorate even if a defect occurs in probe information configured in time series for each link.

1 Traffic jam prediction system 10 Probe car (vehicle)
Reference Signs List 20 traffic congestion prediction device 21 probe information receiving unit 22 probe traveling information database 24 probe prediction processing unit 25 probe statistical processing unit 26 probe history database 27 probe prediction database Pb probe information

Claims (1)

Probe history information including at least travel time of a link which is a road segment is received from a vehicle traveling on the link and accumulated in a database, and probe history information in which probe information is configured in time series for each link accumulated in the database Traffic congestion prediction method executed by the traffic congestion prediction device based on
Detecting, as a target link, a link having the probe history information in which a time zone in which information is missing exists, based on the probe history information of the database;
Wherein the detected target link, the first curve and the information a change curve of the information day is missing the same day and the information travel time based on the above target link time zone information prior to missing hours There are predicted from missing days and second curve, which is the change curve of travel time based on the information of the target link of the same day and time zone later than the information missing time zone, the information is the time zone missing Calculating a straight line and a curve representing a change in travel time by statistical processing;
Estimating information of a time zone in which the information of the target link is missing based on the calculated straight line and curve, and complementing the probe history information;
Predicting congestion at the target link based on the probe history information supplemented with the information of the missing time zone;
Traffic congestion forecasting method.

Images