JPH11306480A - Required traveling time predictive device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately predict time required for traveling in any arbitrary section. SOLUTION: This device is provided with a reference file preparing means 4 for preparing a standard required traveling time for each object unit section from a pass data of an expressway; means 8-11 for extracting similar reference required traveling time for several days in the past from this prepared standard required traveling time, estimating the parameters of a statistical model and predicting a required traveling time on the relevant day while using these estimated parameters; a relevant day prediction correcting means 13 for correcting the predicted required traveling time on the relevant day while using the standard required traveling time of respective unit sections provided from the latest nearby pass data at the current time point of the relevant day; an updating means 15 for updating a prepared time slice table while using this corrected required traveling time on the relevant day; and a required traveling time calculating means 17 for calculating a required traveling time for any arbitrary section among the object sections while using these successively updated contents of the table.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a travel time estimating apparatus used in a traffic control system for managing a motorway such as an expressway, and more particularly to a travel time estimating method for estimating a travel time in a specific target area. The present invention relates to a required time prediction device.

[0002]

2. Description of the Related Art On an expressway, traffic congestion occurs along with an increase in traffic volume, and the required travel time varies greatly. In such a road situation, for example, when traveling in a predetermined section such as between a specific interchange on an expressway or a specific city, the time required for traveling the section is provided to the driver of the vehicle every moment. It is important to improve the service of users and to operate the expressway smoothly.

Conventionally, several methods have been proposed for providing a predicted value of the required travel time from such a request.

One of the methods is to divide a target section into a plurality of unit sections, install a vehicle sensor in each unit section, and use a traveling vehicle speed measured by each vehicle sensor to determine a travel requirement of each unit section. This is a method of calculating the required travel time of the target section by calculating the time and summing the required travel times of these unit sections.

This method is a method of estimating a required time which is currently used on an expressway.

The other is to divide a target section into a plurality of unit sections, install a vehicle sensor in each unit section, and use the traveling vehicle speed measured by each vehicle sensor to travel in each unit section. The required time is calculated, and the calculated required time of each unit section is stored in time series in advance. Then, the required travel time of the target section is calculated by summing the required travel time of each unit section while considering the elapse of time due to travel. This required time prediction method is called a time slice total method.

Another one is an AVI system (Au
tomatic Vehicle Identification Sy-stem), and more specifically, TV on both sides of the target section.
This is a method of measuring the required travel time of each vehicle that has actually traveled in the target section by installing a camera and recognizing the number of the traveling vehicle by image recognition.

[0008]

However, the following problems have been pointed out in the above required time estimating method.

At present, the predicted value of the required travel time provided on the expressway is simply the sum of the required travel times of each unit section, and the traffic condition of each unit section does not change much with the passage of time. In this case, for example, when the vehicle is flowing smoothly over the entire target section, it is possible to accurately predict the required travel time.

However, in practice, the traffic situation in each unit section changes with the passage of time, so that the prediction error increases. In particular, when the occurrence and elimination of traffic congestion are repeated, such as during morning and evening rush hours, the prediction error becomes extremely large. Here, it is desired to provide an accurate predicted value of the required travel time even in the above-described road conditions.

In addition, the method of obtaining the actual travel time using an AVI system or the like requires that a new vehicle sensor or AVI device be installed at an appropriate place on the expressway. In addition, the traveling time is obtained when the traveling vehicle arrives at the end of the target section. In this case,
It is difficult for a driver who actually travels on the road to use this required time as a measure of how long it takes to reach the destination.

The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a travel time estimating apparatus for easily estimating the time required for traveling in a target section at low cost.

Another object of the present invention is to provide a travel time estimating apparatus for accurately estimating the travel time irrespective of traffic conditions.

[0014]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention relates to a traveling time estimating apparatus for estimating the time required for an automobile traveling on a predetermined section of an automobile road. A reference file creating means for calculating and storing a reference travel time of each unit section of the target section from the ticket data; and a unit of each unit section for a predetermined number of days in the past according to a selection condition from the stored reference travel time. data selection means for extracting and storing the reference travel time required every n (n is an integer) minutes; and a parameter of a prediction model of the reference travel time required every n minutes from the reference travel time selected by the data selection means. Using parameter estimating means to be estimated, parameters estimated by the estimating means, and reference traveling time required every n minutes for a predetermined number of past days in each unit section Means for predicting the required travel time for each unit section of the day; and means for predicting the required travel time for each unit section of the day obtained by the means for predicting data for the day. Means for creating a required travel time pattern table for each unit section by arranging the reference travel required time for each unit section created by the reference file creating means from the toll ticket data near the latest time of the day and the current day prediction data On the day, the travel time required for each unit section is corrected using the travel time required for each unit section on the day predicted by the calculating means, and the travel time pattern table for each unit section is updated as needed. Using the prediction correction means and the required travel time pattern for each unit section of the pattern table, the required travel time of an arbitrary section in the target section is predicted and calculated in advance. A drive elapsed time prediction apparatus provided with a that traveling required time calculating means.

By adopting such means, of the actual reference travel time required for each unit section obtained from the toll ticket data, the actual reference travel time required for each unit section for a predetermined number of similar days in the past is calculated. And the parameters of the prediction model for the required travel time for every n minutes are estimated, and the required travel time for each n minutes of each unit section of the day is predicted using the estimated parameters, , A pattern table of the required travel time of each unit section is created.

The above series of processing is performed in the middle of the day.

Thereafter, the required travel time per n minutes of each unit section of the day predicted based on the required reference travel time per n minutes of each unit section obtained from the latest and closest toll ticket data at the present day of the day. Is corrected and the pattern table is updated as needed, so that the required travel time for each n minutes of each unit section on the day is predicted, but is corrected according to the latest road condition which is the latest nearby pass data, and Since the required travel time pattern table is updated as needed, a highly accurate required travel time can be obtained from the required travel time pattern table.

Further, unlike the related art, it is not necessary to provide a vehicle sensor or an AVI device for each unit section, and the toll ticket data collected by, for example, a toll collection system installed in an interchange or the like is used. It can be realized inexpensively and easily.

[0019]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an overall configuration diagram showing one embodiment of a travel time estimation device according to the present invention.

The travel time estimation device is a unit section of a target section, for example, each IC (interchange).
This is a computer system that receives and processes toll ticket performance data sent from a toll collection system, etc., which is installed in, for example, a central traffic control center or a place that plays an equivalent role, The required travel time is displayed on a display panel at an appropriate place in the target section.

In FIG. 1, reference numeral 1 denotes pass ticket data.
Collected as vehicles pass through the exit toll booth of the IC.
This toll ticket data can be automatically collected by passing the toll ticket through the toll collection system. However, a system other than the toll collection system may be used as long as it is dedicated to reading and transmitting, for example, toll ticket data. The following data can be obtained from the general format of a toll ticket as data available from the toll ticket.

Vehicle type NO, entrance IC. NO, entrance month, entrance day, entrance time, entrance minute Exit IC. NO, exit month, exit day, exit time, exit minute The vehicle type NO1 is, for example, a large vehicle, and the vehicle type NO2 is a small vehicle.

Reference numeral 2 denotes entrance-specific file conversion means.
Based on the time at which the vehicle enters the entrance tollgate C, the time to exit the exit tollgate is predicted.
Is converted into a file for each entrance tollgate, for example.
At this time, it classified into entrance toll gate by file 3 ₁ to 3 _n of the unit sections for example segment 1 interval n, the file 3 ₁ -
The 3 _n store the actual duration data for each vehicle minus the inlet time from the outlet time on the ticket format.

Reference numeral 4 denotes a reference file creating means. In the case of traveling by car, there is a case where there is a break in a service area on the way, and there is a car which travels at a speed higher than the legal speed, so that the traveling time required for each car obtained at intervals of, for example, 5 minutes varies. (See FIGS. 4 and 5). Therefore, the reference file creating means 4 performs a filtering process to remove the variation, obtains a reference required time required for the majority of vehicles to travel in the unit section from the required time for each vehicle, and obtains the unit section reference value data. stored in a file 5 ₁ to 5 _n.

Reference numeral 6 denotes a calendar mode data file which has a calendar mode data which affects the required traveling time of the vehicle, for example, weekdays, holidays (Monday), holidays (Saturday), and holidays.

Numeral 7 is a data selection means. For example, in the case of mode selection, the data is classified according to the calendar mode of the day, for example, weekday, after holiday (Monday), before holiday (Saturday), or holiday, and section reference is made. It reads the file 5 ₁ to 5 _n 5 minutes each reference time required for one day in each unit interval from, and stored in the daily data file 8. Normally, when performing at midnight of the day, for example, around 1 am, 5 minutes for the past 5 days of each unit section
Retrieve and save the required time per minute.

Here, the past five days are the data of the previous day, the data of two days ago,..., When the calendar mode is a weekday.
In the case of a holiday, it means one week ago data, two weeks ago data,..., Five weeks ago data, etc. The number of days is not limited to five, and may be any number of days required.

The data selecting means 7 can select the data for each mode, date, and section, and save the data in the daily data file 8.

When the reference traveling time required for, for example, five days is stored in the daily data file 8 as described above, the processing by the parameter estimating means 9 is subsequently executed.
The parameter calculating means 9 calculates model parameters using a statistical model such as a Kalman filter using the reference traveling time required every five minutes for five days, and stores the model parameters in the parameter file 10.

FIG. 2 is a diagram for explaining an example of processing by the data selecting means 7 or the parameter estimating means 9.

That is, data reading conditions (mode, date, section) are set by the data selecting means 7 (S
1) reads the 5 days of the reference drive elapsed time data described above from the section reference file 5 ₁ to 5 _n, and stores it in the daily data file 8.

Thereafter, an initial value of the Kalman filter is set (S3), and the known time required two to five days ago and the parameter initial values a1 to a4 are used by the successive least squares method, and the previous day is calculated by the following equation. Travel time parameter Q
(The day before) is calculated (S4).

Q (previous day) = a1 · Q (2 days ago) + a2 · Q (3 days ago) + a3 · Q (4 days ago) + a4 · Q (5 days ago) (1) The parameters obtained in this way are parameters Write to file 10 (S5).

Then, using the parameters estimated by the parameter estimating means 9, the day-of-day forecast data calculating means 11
Estimates the travel time of each unit section on the day (Fig. 3),
Same day forecast data file 12 ₁ -12 for each unit section
Creating a _n (12 _{2 ~12 n).} That is, the day-of-day prediction data calculation means 11 fetches, for example, parameters from the previous day to four days ago from the parameter file 10 as shown in FIG. 3 (S11), calculates the day-of-day prediction data based on the following equation (S12), data file 12 ₁ ~12 _n
(S13).

Q (today) = a1 · Q (previous day) + a2 · Q (2 days ago) + a3 · Q (3 days ago) + a4 · Q (4 days ago) (2) Therefore, each estimated as described above If the required travel times of the unit sections are arranged in chronological order and the required travel times of the respective unit sections are totaled while taking into account the elapsed time due to travel, the required time of an arbitrary section is predicted in advance, for example, at about 1 am on the day. (Time slice method).

[0037] In addition, the section reference file 5 ₁ to seeking the latest standards required time of every moment on the day by the online
5 _n , the day-of-day prediction correction means 13 corrects the travel time required for each unit section, which is the predicted data of the day, in advance with the latest reference travel time required for the day, so that each unit section with higher accuracy can be obtained. Corrected day-of-day forecast data 14 ₁
14 _n (14 ₂ ~14 _n are not shown) can be obtained.

The time slice updating means 15 sequentially updates the time slice table 16 based on the current day prediction data without correction or the corrected current day prediction data. The required travel time calculation means 17 includes a time slice table 16
, The required travel time of each section when entering the tollgate of the vehicle can be summed up, and the required travel time of any section can be extracted.

The parameter estimating means 9 estimates the parameters of the prediction model of the required traveling time every n minutes using the statistical model of the Kalman filter. However, in addition to the statistical model, a parameter such as a neural network or a fuzzy model may be used according to the situation. Can be estimated using

In the neural network, conditions relating to the required travel time such as weekdays, before holidays, holidays, and the like are input in advance for each unit section, and a weighting factor for estimating the actual travel time of the car is estimated. Using the coefficients, the conditions of the day and the weighting coefficients are set the day before, and the travel time required for the day is predicted.

Also, a fuzzy model for predicting the required travel time of a unit section under various conditions is constructed, parameters are created based on past results and knowledge, and parameters for the current day are determined on the previous day based on fuzzy rules. It is also possible to input and predict the required travel time of the unit section.

Therefore, according to the above-described embodiment, every n minutes of each unit section of the day is predicted based on the reference traveling time required for each n minutes of each unit section obtained from the pass ticket data. The travel time is corrected based on the reference travel time for every n minutes of each unit section obtained from the latest nearest pass ticket data at the current time of the day, and the pattern table is updated as needed. The required travel time for every n minutes is predicted, but it is corrected according to the latest road condition which is the latest nearby toll ticket data, and the required travel time pattern table is updated as needed. In addition, it is possible to obtain a highly accurate travel time from the pattern table. Furthermore, it is possible to always predict the required travel time with high accuracy while responding to road conditions and aging.

Further, unlike the related art, it is not necessary to provide a vehicle sensor or an AVI device for each unit section, and the use of the toll ticket data makes it possible to reduce the predicted travel time in consideration of the time at low cost and in a simple manner. Can be calculated.

Further, since the parameters of the prediction model of the required travel time for every n minutes are estimated and the prediction data of the day are obtained from these parameters, the prediction accuracy can be improved.

(Other Embodiments) (1) FIG. 4 is an actual required time pattern diagram of all types of vehicles in a unit section of the expressway obtained from actual toll ticket data, and FIG. It is an actual required time pattern diagram of all the vehicle types of a section.

As is apparent from a comparison between FIG. 4 and FIG. 5, the normal vehicle occupies 90% or more of the whole, and the pattern of the normal vehicle may be determined as the whole pattern. .

Accordingly, even if only ordinary cars are targeted from the actual situation and the forecast data of the day is created by the apparatus shown in FIG. 1 using this, the reliability of the calculation of the reference required time can be improved.

Therefore, the use of ordinary vehicle pass ticket data can reduce the number of data to be handled by about 10%.
In addition, noise data that slips through traffic jams of vehicles such as motorcycles can be automatically removed, and data reliability can be further improved.

(2) The reference value file creating means 4 calculates the reference travel time required for each unit section from the actual travel time required for the vehicle after performing the filtering process. A frequency distribution (histogram) of k minutes (in the example, 5 minutes) of the travel time required for exiting the toll gate at 5 minute intervals is obtained, and the most frequent travel time at which the frequency is maximum is calculated as follows. It can be set as a reference required time of each unit section.

FIGS. 6 and 7 are diagrams showing an example of a histogram creation process.

That is, after a time-series sort file is created from the toll ticket data or the data of the entrance tollgate-specific files 3 _{1 to} 3 _n (S 21), the time of the required time at intervals of 5 minutes, for example, every 5 minutes is calculated from this file data. A histogram file representing the frequency value is created (S22). Then, the mode value file is created by extracting and writing the mode value at intervals of 5 minutes from the created histogram file (S23).

Histogram file creation step S22
After performing the total calculation for converting all the required times of each vehicle to "minutes" as shown in FIG. 7 (S22).
1), a histogram width, for example, a 5-minute interval width is determined (S
222). Thereafter, the required travel time data of each vehicle is sequentially extracted (S223), and it is determined which histogram time width the required travel time data falls into. After counting up the frequency of the width, the travel time data is stored (S224 to S226).

After that, the mode is taken out of the histogram at intervals of 5 minutes, set in, for example, an array of histograms at intervals of 5 minutes, and written in a histogram file (S227, S228).

Incidentally, among the patterns of the actual traveling time required of each vehicle in a certain section of the expressway shown in FIG. 8, for example, 5: 00-6: 05 and 12: 00-12: 05
FIGS. 9 and 10 show examples in which a minute-by-minute histogram is created. Further, FIG. 11 plots the mode of the same histogram for many times.

As a result of examining these histograms, it can be seen that the mode is almost along the lower boundary line of the pattern. That is, the vehicle often travels in a short travel time. This means that most cars are running as fast as possible.

Therefore, according to the above-described embodiment, the actual traveling time obtained from the toll ticket data of each vehicle varies even when the vehicle enters the entrance tollgate at the same time. From the required travel time, the reference required travel time of each unit section (the time required for the majority of vehicles to pass through the unit section) can be realized using a simple algorithm.

(3) Further, a description will be given of the processing of a time zone with a small amount of data and the filtering processing for determining a good reference traveling time required for each unit section excluding the influence of vehicles traveling in a group.

[0058] One of them is, from the converted entrance toll gate by file 3 ₁ ~3 _n by the entrance toll gate from the ticket data,
A frequency distribution is obtained at intervals of k minutes (for example, 5 minutes) at intervals of n minutes (for example, 5 minutes) of the travel time required for each vehicle to actually exit the exit tollgate, and the frequency in the frequency distribution for every n minutes is obtained. Based on the value, an average time is obtained for α% or more of the travel time frequency, and this is set as a reference travel time.

For example, in the histogram of FIG.
By averaging 以上 (α = 66.7) or more of the frequency of the mode, the reference travel time is (15 minutes × 14 + 20 minutes × 11) /25=17.2 minutes (3).

The other one is to select the mode as follows, for example.

B. From the ticket is converted from data by the entrance toll booth entrance toll gate by file 3 ₁ ~3 _n, each vehicle is actually frequency distribution of the k-minute increments n of each of the traveling time required to take the exit toll gate ( When the average time is obtained for α% or more of the mode frequency based on the mode for every n minutes, the average time is calculated excluding a specific mode.

B. However, for example, in the case of the histogram as shown in FIG. 13B, there is a case where there is no frequency at three intervals of 5 minutes, 5 minutes, and 5 minutes, and the mode appears next. In such a case, the mode value is within 2/3 or more, but based on a preset continuous coefficient = 3 or 3 or more, there is no frequency at three intervals in the time series direction of the histogram or the frequency is very small. At this time, the calculation of the average time is executed excluding the mode value (see S57 and S58 in FIG. 14).

C. In the time zone without the calculation result data, the previous value having the same condition is used (S61).

D. Further, when the mode value is 3 or less, the first frequency value = 2 of the histogram is used as shown in FIG. 13C (S62).

Although there are a plurality of modes, when the mode is separated by a predetermined amount or more from the mode required time, the average time may be calculated excluding the mode.

FIG. 14 shows an example of the above series of processing.

Therefore, according to the above-described embodiment, the reference traveling of good performance in each unit section is excluded except for the processing of the time zone with a small data amount and the influence of the vehicle traveling by group. The required time can be determined.

(4) Further, required travel time calculating means 1
7, the travel time required for each section is calculated by calculating the travel time required for each section at the time of entering the entrance tollgate from the travel time pattern for the unit section tabulated in the time slice table 16 to calculate and predict the travel time required for any section. As described above, the travel time required for an arbitrary section in the target section may be predicted using, for example, a time slice total method.

This prediction example will be described with reference to FIG. For example, the time when the vehicle that has entered the unit section 1 at 0: 5 arrives at the end of the unit section 5, that is, the time required for traveling in the target section is predicted.

First, the vehicle takes 8 minutes to travel in the unit section 1 and enters the unit section 2 at 0:13. At this time, the vehicle that has entered the unit section 2 requires 10 minutes to travel in the unit section 2. Similarly, considering the passage of time, a portion surrounded by a double line shown in FIG. 15 is added to the travel of each unit section.
As a result, the required travel time of the target section from the unit section 1 to the unit section 5 is 54 minutes in total, and the vehicle arrives at the destination at the end of the target section at 0:59.

FIG. 16 is a diagram showing an example in which the required travel time is estimated using the time slice method. The solid line in the figure is the estimated travel time, and the dotted line is the actual time.
Although there is a certain error between the estimated travel time and the actual time, the estimated pattern matches the actual pattern well, and follows the start of traffic congestion and the time of traffic jam elimination without delay. .

The error (estimated−actual) is on the + side as shown in FIG. 17, but this is added to the waiting time at the exit tollgate in each section in the case of the time slice method estimation. Because it is. In addition, it can be seen that subtracting a constant value from the error graph of FIG. 17 reduces the error on average.

Therefore, a certain value may be subtracted from the required travel time in the target section estimated by the above-described time slice totaling method, and the resulting value may be used as the predicted travel time.

In addition, there is some variation in the results of the required travel time shown in FIG.

Therefore, assuming that the on-line calculation for calculating the reference travel time in advance is performed in the middle of the night when all the actual data are collected, for example, an average of about five pieces is calculated around the time corresponding to the calculation. Reduce variability using the average technique. The moving average without delay means, for example, that the data of the previous day has already been obtained from the calculation in the middle of the night as well as the data of the past time. Therefore, averaging is performed with substantially no delay.

Calculation method of the moving average without delay The running time required at time t (unit: minute) is defined as Tr (t),
The average time without delay Trm (t) at the calculation corresponding time t is obtained based on the following equation. There are five samples, with two data at the past time and two data at the previous time centered on the time of calculation.

[0077]

(Equation 1)

However, in the above equation, Δt is 5 minutes.

FIG. 18 shows the result of applying the average without delay (number: 5) to the data of FIG. As is clear from this figure, there is no fluctuation such as noise shown in FIG. FIG. 19 shows FIG.
8 shows the error (estimated-actual) of the graph of FIG. The average value of this error is about 9.25 minutes. FIG. 20 shows a result obtained by subtracting a fixed value (error average value) from the estimated value in FIG. 18 as the improved estimated value. Therefore, as is clear from FIG. 20, the fluctuation and the error are considerably reduced, and an estimated value larger than that in FIG.

Therefore, according to the above-described embodiment, when estimating the required travel time of an arbitrary section of the target section, it is necessary to use not only the time slice total method but also the travel time of the unit section of the table 16. Since the difference between the estimated value and the actual result is subtracted from the required time pattern to obtain the required travel time for obtaining the prediction, errors and fluctuations can be significantly reduced.

(5) The congestion start time of the day is determined from the pattern of the reference required time obtained using the latest nearby toll ticket data group obtained at the present time of the day, and the travel time pattern of the unit section predicted in advance is determined. If the correction is made, it is possible to predict the required traveling time pattern of the unit section with higher accuracy.

FIG. 21 is a diagram for explaining an example of such a processing procedure.

First, the actual data of the latest nearest toll ticket data group obtained at the present day of the day is read (S 31), and the above-mentioned is performed based on the actual data obtained at the exit tollgate by the file conversion means 2 for each entrance. After converting the data for each entrance tollgate in the same manner as described above (S32), the reference file creating means 4
Similarly, in a histogram above, estimates the latest reference travel time required from the most frequently used data, and stores the interval reference file 5 ₁ ~5 _n (S33).

Thereafter, the day prediction correcting means 13 determines whether or not a movement completion crag has been raised for the latest reference travel time of the day (S34). If the movement has not been completed yet, the latest day of the day is determined. The determination of the rise (congestion start) of the reference travel required time pattern of the reference travel required time, it is determined whether the reference travel required time at time n and the time (nk) after k steps both exceeds a predetermined congestion start determination required time. A determination is made (S35). If it exceeds, the time n is set to the latest traffic congestion start time ncs of the day (S36).

Further, the traffic congestion start time ncs is compared with the traffic congestion start time tcs obtained from the prediction data of the same day for each unit section predicted in advance (S37). If the time ncs is early, the current prediction data pattern is calculated by tcs-ncs. Is moved forward (S38), and a movement completion flag “1” is set on the day (S39). In other words, if the congestion start time of the forecast data of the day is advanced by tcs-ncs, the pattern becomes closer to the latest reference travel time pattern of the day.

On the other hand, if the time ncs is not earlier than the judgment result in step S37, the process shifts to step S39, where the time n is compared with the traffic congestion start time tcs (S40). Data pattern is n
It is moved backward by -tcs (S41).

Thereafter, a time error Δter between the actual required travel time and the moved reference required travel time is calculated (S4).
2) The time slice table 1 is corrected by adding a time error Δter to the moved reference travel required time pattern used in the time slice method and updating the time slice at any time (S43).
6 is a required travel time pattern of the unit section.

Therefore, according to this embodiment, the traffic congestion start time of the nearest neighborhood at the current time of the day is compared with the traffic congestion start time obtained from the prediction data of each unit section predicted in advance on the day. Since the current day prediction data pattern is moved to the current day prediction data in accordance with the deviation, the current day's road congestion state can be appropriately reflected.

Further, after moving the prediction data pattern of the day in accordance with the difference between the traffic congestion start times, the required travel time pattern of the unit section in the time slice table 16 is updated. The required travel time of any section of the section can be obtained,
A highly accurate travel time can be provided to the driver.

(6) In the above embodiment, when estimating the required travel time of each unit section on the day, an autoregressive model coefficient is estimated using a Kalman filter, or a neural network or a fuzzy model is used. Although the parameters of the required travel time prediction model are estimated every n minutes, the parameters of the prediction model may be estimated using, for example, a chaos model or a genetic algorithm.

[0091]

As described above, according to the present invention, the travel time required for each unit section of the target section can be reduced at low cost.
Can be easily predicted. In addition, the travel time required for the day predicted in advance is corrected with the latest road condition and the contents of the time slice table are updated as needed, so that the travel time required for an arbitrary section can be accurately predicted.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram showing an embodiment of a travel required time prediction device according to the present invention.

FIG. 2 is a view for explaining an operation procedure of a parameter estimating unit shown in FIG. 1;

FIG. 3 is a view for explaining the operation procedure of the day-of-day forecast data calculation means shown in FIG.

FIG. 4 is an actual required time pattern diagram of all vehicle types in a unit section.

FIG. 5 is an actual required time pattern diagram of only ordinary cars in a unit section.

FIG. 6 is a diagram illustrating a schematic operation of creating a histogram.

FIG. 7 is a diagram showing an example of a histogram creation process.

FIG. 8 is a pattern diagram of the actual required traveling time of each vehicle in a section of the expressway.

9 is a histogram diagram of the pattern diagram of FIG. 8 at intervals of 5 minutes from 6:00 to 6:05.

FIG. 10 shows the pattern diagram of FIG.
FIG. 5 is a histogram diagram at a 5-minute interval of FIG.

FIG. 11 is a diagram in which mode values of a histogram at many times are plotted.

FIG. 12 is a view for explaining an example of calculating a reference required time from a histogram.

FIG. 13 is a view for explaining another example of calculating a reference required time from a histogram.

FIG. 14 is a view for explaining a processing example when calculating a reference required time from a histogram.

FIG. 15 is a time slice table in which required traveling times of each unit section are arranged in time series and for each unit section.

FIG. 16 is a graph showing an example of a required travel time of an estimated value and an actual result.

FIG. 17 is a graph showing an error, which is a difference between a required travel time between an estimated value and an actual result with respect to FIG. 16;

FIG. 18 is a graph in which the required travel times of the estimated value and the actual result are averaged for each five, and are graphed.

FIG. 19 is a graph in which an error, which is a difference between a required travel time between an estimated value and an actual result with respect to FIG. 18, is graphed.

FIG. 20 is a diagram showing a graph of a required travel time between an estimated value obtained by subtracting an error from the estimated required travel time in FIG. 18 and an actual result;

FIG. 21 is a flowchart illustrating a current-day prediction correction process in consideration of a traffic jam start time.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Toll ticket data 4 ... Reference file creation means 7 ... Data selection means 9 ... Parameter estimation means 11 ... Same day forecast data calculation means 13 ... Same day forecast correction means 15 ... Time slice table updating means 16 ... Time slice table 17 ... Travel requirement Time calculation means

Claims (13)

[Claims] 1. A travel time prediction device for predicting a time required for a vehicle traveling on a preset target section of a motorway, comprising: calculating a reference travel time of each unit section of the target section from pass ticket data; A reference file creating means for calculating and storing, and extracting a reference traveling time required for each of n (n is an integer) minutes of each unit section for a predetermined number of days in the past according to a selection condition from the stored reference traveling time required. Data selecting means to be stored; parameter estimating means for estimating parameters of a prediction model of the reference traveling time required every n minutes from the reference traveling time selected by the data selecting means; and parameters estimated by the estimating means. The day-of-the-day forecast for estimating the travel time required for each unit section on the day using the reference travel time required every n minutes for the past predetermined number of days in each unit section. Data calculating means, and means for arranging the required traveling time of each unit section of the day obtained by the current day prediction data computing means in a time-series manner and for each unit section to create a required traveling time pattern table of each unit section. And the reference travel time required for each unit section created by the reference file creation means from the toll ticket data near the latest time of the day and the travel time required for each unit section of the day predicted by the day forecast data calculation means. Using, and the same day prediction correction updating means for correcting the predicted travel time required for each unit section and updating the travel time pattern table for each unit section as needed, and for each unit section of the pattern table Travel time calculation means for predicting and calculating the travel time of an arbitrary section in the target section in advance using the travel time pattern, Traveling required time prediction apparatus and butterflies. 2. The method according to claim 1, wherein said parameter estimating means estimates a parameter of a prediction model of a reference traveling time required every n minutes using one of a statistical model, a neural network, and a fuzzy model. Item 1. A travel time estimation device according to item 1. 3. The travel time estimating device according to claim 1, wherein the toll ticket data of the car is only ordinary cars. 4. The reference file creating means, based on the toll ticket data classified for each toll gate, determines a predetermined amount of time required for each vehicle to travel from each toll gate to an actual exit toll gate every n minutes. The travel time prediction device according to claim 1, wherein a frequency distribution at intervals is obtained, and a mode value in the frequency distribution is set as a reference travel time required for each unit section. 5. The reference file creating means, based on the toll ticket data classified for each toll gate, determines a predetermined amount of time required for each vehicle to travel from each toll gate to an actual exit toll gate every n minutes. The frequency distribution in steps is obtained, an average value of a predetermined value% or more of the mode frequency in these frequency distributions is taken, and this average value is used as a reference traveling time required for each unit section. The travel time estimation device according to claim 1. 6. The reference file creating means, based on the toll ticket data classified for each toll gate, determines a predetermined amount of time required for each vehicle to travel from each entrance tollgate to the actual exit tollgate every n minutes. When a frequency distribution at intervals is obtained and an average value of a predetermined value% or more of the mode frequency in these frequency distributions is taken, the average value is a predetermined value% or more of the mode frequency, but in an increasing direction of the time series of the frequency distribution. The travel time estimation device according to claim 1, wherein an average value is calculated excluding a travel time interval that is longer than a predetermined time interval, and the average value is used as a reference travel time required for each of the unit sections. . 7. The reference file creating means, based on the toll ticket data classified for each toll gate, determines a predetermined amount of time required for each vehicle to travel from each toll gate to the actual exit toll gate every n minutes. When the frequency distribution at intervals is determined and the average value of the mode frequency in these frequency distributions is equal to or more than a predetermined value%, except for the required time separated by a predetermined time or more from the mode travel time, The travel time prediction device according to claim 1, wherein an average value is obtained, and the average value is used as a reference travel time required for each of the unit sections. 8. The reference file creating means, based on the toll ticket data classified for each toll gate, determines a predetermined amount of time required for each vehicle to travel from each entrance tollgate to the actual exit tollgate every n minutes. When a frequency distribution at intervals is obtained, and when an average value of a predetermined value% or more of the mode frequency in these frequency distributions is taken, an average value is obtained by excluding the predetermined value% at both ends of the frequency distribution, and this average value is calculated as The travel time estimation device according to claim 1, wherein the travel time is a reference travel time of a unit section. 9. The required travel time prediction unit according to claim 1, wherein the required travel time calculation means obtains a required travel time predicted value of an arbitrary section in the target section using a time slice method. apparatus. 10. The required travel time calculating means obtains a required travel time of an arbitrary section in the target section using a time slice method, and calculates a time obtained by subtracting a predetermined time from the required travel time in the target section. The travel required time prediction device according to claim 1, wherein the travel required time prediction value is an estimated travel time value of an arbitrary section within the range. 11. The travel time calculation means calculates a running average of the travel time of an arbitrary section in the target section without delay using a time slice method, and subtracts a predetermined time from the moving average time. Is a travel time prediction value of an arbitrary section in the target section.
Traveling time prediction device according to 1. 12. The day-of-the-week prediction correction / updating means includes a reference travel time required for each unit section between the latest time of the day created from the toll ticket data group and a time after a predetermined time has elapsed, and a predetermined congestion start determination. Means for comparing the required time and the latest traffic congestion start time; and the latest traffic congestion start time and the predicted traffic congestion start time of the traveling required time prediction pattern for each unit section of the day obtained by the same day prediction data calculation means. A first pattern moving means for moving forward the traveling time prediction pattern for each unit section on the day when the latest traffic congestion start time is earlier, and 9. The traveling time according to claim 1, further comprising: a second pattern moving means for moving a traveling time prediction pattern for each unit section backward. Inter prediction device. 13. The day-of-the-week prediction correction updating unit includes: an error calculating unit configured to calculate a time error from each of the constituent elements described in claim 12; an actual traveling required time and the traveled traveling required time prediction pattern; 12. A travel time prediction device according to claim 1, further comprising means for updating the travel time pattern table of each unit section as needed using the time error. .