JP2016091453A - Signal cycle length-of-intersection estimation system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a problem associated with obtaining plural clock times at which a vehicle traveling on a road has started after waiting for a traffic signal and estimating a signal cycle length of a traffic signal installed at an intersection on the basis of a difference in the aforementioned start clock times, the problem where as a traffic volume at an intersection decreases, the case of no vehicle waiting for a signal when the traffic light changes to red appearing increases, causing the chances of obtaining start clock times and making it impossible to estimate a signal cycle length accurately.SOLUTION: In a case where a traffic volume is small at an intersection, a relationship between "start clock time" and "the number of start clock times obtained" is defined as a function (a start interval function), an auto-correlation function is applied to the start interval function to obtain a cycle of the start interval function, and the cycle is estimated as a signal cycle length.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a traffic signal cycle length estimation system for estimating a signal cycle length, which is a cycle in which a display of a traffic signal installed at an intersection makes a round.

  Patent Document 1 discloses a traffic parameter calculation system (hereinafter also referred to as “conventional system”) that estimates a traffic parameter (for example, timing at which a traffic light display is switched) based on information transmitted from a vehicle traveling on a road. Have been described. The conventional system includes a “probe vehicle that is a vehicle that travels on the road and transmits information on the traveling state” and a “probe center that receives data from each probe vehicle”.

  Each of the probe vehicles transmits “probe data relating to the stop time and start time when entering the intersection, the vehicle position, etc.” to the probe center at a predetermined timing. The probe center estimates traffic parameters for each intersection based on probe data received from a plurality of probe vehicles. The probe center transmits traffic parameters regarding an intersection in the traveling direction of the vehicle to the probe vehicle. The probe vehicle provides information to the driver based on the traffic parameters received from the probe center.

  For example, the probe vehicle determines whether or not the vehicle should enter the intersection based on the timing when the traffic signal at the intersection in the traveling direction switches to red and the distance between the vehicle and the stop line at the intersection. The driver is notified of the determination result. Alternatively, when the probe vehicle is stopped before the intersection, the vehicle estimated based on “the timing when the traffic light display next switches to blue and the distance between the vehicle and the intersection stop line” is When the time until the next start is longer than a predetermined value, the driver is prompted to stop the engine. That is, since the engine stop time is longer than the predetermined value, the probe vehicle prompts the driver to perform a so-called idling stop when an effect such as improvement in fuel consumption and reduction in exhaust gas emission can be expected.

JP 2009-75709 A

  By the way, when the number of probe vehicles entering the intersection per unit time is small (that is, when the traffic volume at the intersection is small), the number of probe data acquired by the probe center is small. As a result, for example, the probe center cannot accurately estimate the time (blue signal start time) when the traffic light display changes to blue. Therefore, the probe center obtains the green light start time as the time obtained by adding the “time obtained by multiplying the signal cycle length by an integer” to the “green light start time obtained last time” when the traffic volume at the intersection is small. To do.

  However, Patent Document 1 does not disclose a method for accurately estimating the signal cycle length of a traffic light in which a vehicle waiting for a signal may not appear even if the traffic light display turns red due to a small amount of traffic. . Therefore, it is difficult to accurately estimate the green signal start time of an intersection with a small traffic volume by the above method using the signal cycle length.

  The present invention has been made to cope with this problem, and an object of the present invention is to provide an intersection signal cycle length estimation method capable of accurately estimating the signal cycle length of an intersection with a small traffic volume.

  An intersection signal cycle length estimation device for achieving the above object (hereinafter also referred to as “the device of the present invention”) includes a start time acquisition unit, a start interval acquisition unit, and an estimation unit.

The departure time acquisition unit
A plurality of leading vehicle start times, which are times when the leading vehicle in the train of vehicles stopped according to the red display of the traffic light installed at the intersection, starts according to the blue display of the traffic light are acquired.

The start interval acquisition unit
A start interval which is a time difference between the start vehicle start times adjacent to each other among the plurality of start vehicle start times and the number of times each of the start intervals has occurred are acquired.

The estimation unit includes
Based on the start interval, a signal cycle length that is a cycle in which the display of the traffic light makes a round is estimated.

More specifically, the estimation unit is
The autocorrelation function is applied to the start interval function x (n) whose input value is the start interval n and whose output value is the number x of occurrence of the start interval n as the input value.
Where the autocorrelation function is
The input value is the movement amount k of the start interval n, and the output value is the start interval function x (n) and “start interval function x (n + k) obtained by moving the start interval function x (n) by the movement amount k”. A function that is a value corresponding to the product of.

Furthermore, the estimation unit includes
In the range where the movement amount k is larger than 0, the minimum movement amount k among the movement amounts k at which the output value of the autocorrelation function becomes maximum is estimated as the signal cycle length.

  If the traffic volume at the intersection is large, that is, if the number of vehicles entering the intersection is large, a large number of “start vehicle start times” can be obtained from these vehicles. Therefore, a large number of “start intervals” can be acquired.

  More specifically, when there is a large amount of traffic at the intersection, every time the traffic light is displayed and the traffic light turns red, there is a vehicle waiting for the traffic light before the traffic light. Therefore, the start time of the first vehicle in the vehicle queue waiting for a signal (that is, the start time of the first vehicle), and then the start of the first vehicle in the vehicle queue waiting for the signal that appears when the traffic light display turns red again. The time difference from the time (that is, the start interval) is equal to the signal cycle length.

  However, when the traffic volume at the intersection is small, a vehicle that waits for a signal does not always appear every time the traffic light display turns red. That is, in this case, the start interval is not equal to the signal cycle length, and is a time that is an integral multiple of the signal cycle length.

  For example, when the actual signal cycle length is 120 seconds, the start interval acquired if the traffic volume is large is often a value in the vicinity of 120 seconds. On the other hand, if the traffic volume is small, the case where the acquired start interval becomes a value near 120 seconds decreases, and instead the case where the start interval becomes a multiple of 120 seconds, such as 240 seconds and 360 seconds, increases.

  That is, the distribution of start intervals changes according to the amount of traffic, but “start intervals equal to 120 seconds or multiples of 120 seconds” still occurs. In other words, when the input value (starting interval n) of the start interval function x (n) is equal to the signal cycle length and when the input value is equal to a multiple of the signal cycle length, the output value of the function becomes a maximum. That is, it can be said that the start interval function x (n) has periodicity, and the period is equal to the signal cycle length.

  For this reason, the period of the start interval function x (n) is obtained by applying the autocorrelation function (see equation (2) below) used to calculate the period of the function to the start interval function x (n). The period can be estimated as the signal cycle length. In the equation (2), the input value is the movement amount k of the start interval n, and the output value is the correlation amount between the start interval function x (n) and the start interval function x (n + k) moved by the movement amount k. It is an example of the autocorrelation function which is a value according to.

  For example, when the actual signal cycle length is 120 seconds, the period of the start interval function x (n) is 120 seconds. In this case, the output value of the autocorrelation function is as shown in the graph shown in FIG. In the example shown in FIG. 8, the output value of the autocorrelation function becomes maximum when the input value (movement amount k) is “0”, and the movement amount k is “the period of the start interval function x (n)”. It becomes a maximum when it is equal to a multiple of “120” and “120” (“240”, “360”, and “480”). Therefore, the estimation unit has a minimum movement amount k (120 seconds in this example) among the input values (movement amount k, where movement amount k> 0) of the autocorrelation function at which the output value of the autocorrelation function is maximized. ) As the period of the start interval function x (n).

  According to the system of the present invention, it is possible to accurately estimate the signal cycle length of a traffic light installed at the intersection even at an intersection with a small amount of traffic.

  It should be noted that the present invention extends to a method for estimating the signal cycle length of an intersection used by the system of the present invention.

It is the schematic of the vehicle (this vehicle) and information processing center (this center) with which the signal cycle length estimation system which concerns on embodiment of this invention is provided. It is an example of the map information memorize | stored in the database of this center. It is an example of the start interval calculated based on the start vehicle start time acquired from this vehicle and the start vehicle start time. It is a histogram showing distribution of start intervals. It is the table | surface showing the process of calculating the average value of start space | interval and estimating a signal cycle length. It is a histogram showing distribution of the estimated signal cycle length. It is a histogram showing distribution of the start interval when the signal cycle length becomes an outlier. It is a graph of the autocorrelation function applied with respect to a start space | interval function. It is a flowchart showing the start time accumulation | storage process routine which the central processing part of this center performs. It is a flowchart showing the signal cycle length estimation processing routine which the central processing part of this center performs.

(Constitution)
Hereinafter, an intersection signal cycle length estimation system (hereinafter also referred to as “present estimation system”) according to an embodiment of the present invention will be described with reference to the drawings. The estimation system includes a vehicle 10 and an information processing center 20 shown in FIG. The present invention is applied to the information processing center 20.

  The vehicle 10 includes a calculation unit 11, an in-vehicle communication unit 12, a GPS reception unit 13, a vehicle speed sensor 14, and a display device 15.

  The arithmetic unit 11 is an electronic circuit including a known microcomputer, and includes a CPU, a ROM, a RAM, an interface, and the like. The ROM stores a program executed by the CPU.

  The in-vehicle communication unit 12 performs data communication with the information processing center 20 via the network 31. The network 31 is a known wide area communication network including a mobile phone network (including a wireless data communication network) and an Internet network.

The GPS receiving unit 13 acquires the current position of the vehicle 10 based on a signal (radio wave) from the GPS satellite 32 and outputs a signal representing the current position to the calculation unit 11.
The vehicle speed sensor 14 detects the rotational speed of the axle of the vehicle 10 and outputs a signal representing the traveling speed (vehicle speed) Vs of the vehicle 10 to the computing unit 11.

  The display device 15 is disposed on a center console (not shown) provided in the passenger compartment of the vehicle 10. The display device 15 includes a display and a speaker. The display device 15 controls the display and the speaker according to the signal received from the calculation unit 11, and provides information to the driver through video and audio.

  The calculation unit 11 transmits operation information of the vehicle 10 to the information processing center 20 via the in-vehicle communication unit 12 every time a predetermined time elapses or the vehicle 10 travels a predetermined distance. The operation information includes information on a place where the vehicle 10 has traveled after the operation information was transmitted last time. Furthermore, if a vehicle stop has occurred after the previous transmission of operation information (that is, if there is a time zone in which the vehicle speed Vs is “0”), the operation information includes the stop position, stop time, and start time. Contains.

  The information processing center 20 includes a central processing unit 21, a central communication unit 22, and a database 23.

  The central processing unit 21 is an electronic circuit including a known general-purpose computer, and includes a CPU, a disk device (HDD), a RAM, an interface, and the like. The disk device stores a program executed by the CPU.

  The central communication unit 22 performs data communication with each of the vehicles 10 via the network 31. When the central communication unit 22 receives operation information from the vehicle 10, the central communication unit 22 stores the content of the operation information in the database 23.

  The database 23 is a known database device. The database 23 stores map information, operation information received from each of the vehicles 10, signal cycle lengths of intersections estimated by signal cycle length estimation processing described later, and the like. The map information includes each intersection, a traffic signal installed at the intersection, a “link that is a route (road) through which the vehicle flows into or out of the intersection”, and the like.

  FIG. 2 shows an example of map information. FIG. 2 shows an intersection S1 and an “intersection S2 adjacent to the intersection S1”. At the intersection S1, traffic lights T1a to T1d are installed. At the intersection S2, traffic lights T2a to T2d are installed.

  The links flowing into the intersection S1 are links L1a to L4a. The link L3a is also a link that flows out from the intersection S2. On the other hand, the links flowing out from the intersection S1 are links L1b to L4b. The link L3b is also a link that flows into the intersection S2.

  The links flowing into the intersection S2 include links L5a to L7a in addition to the link L3b. The links flowing out from the intersection S2 include links L5b to L7b in addition to the link L3a.

  For example, each of the link L1a and the link L1b represents an opposite lane on the same road. In many cases, two links are set because a vehicle passes in both directions on one road, but only one link is set on a one-way road.

  The traffic light T1a displays one of the signal colors of blue, yellow and red for the vehicle flowing into the intersection S1 from the link L1a. Similarly, the traffic light T1b displays a signal color for the vehicle flowing in from the link L2a. The traffic light T1c displays a signal color for the vehicle flowing in from the link L3a. The traffic light T1d displays a signal color for the vehicle flowing in from the link L4a.

  The traffic light T2a displays a signal color for the vehicle flowing into the intersection S1 from the link L3b. Similarly, the traffic light T2b displays a signal color for the vehicle flowing in from the link L5a. The traffic light T2c displays a signal color for the vehicle flowing in from the link L6a. The traffic light T2d displays a signal color for the vehicle flowing in from the link L7a.

  When the operation information stored in the database 23 includes a stop position, a start time, and the like, the central processing unit 21 determines whether the vehicle 10 corresponding to the operation information has been stopped for waiting for a signal. judge.

  More specifically, if the distance between the stop position and the intersection exceeds a predetermined value, the central processing unit 21 determines that the vehicle 10 has stopped for a reason other than waiting for a signal. To do. On the other hand, if the distance between the stop position and the intersection is less than or equal to the predetermined value, the central processing unit 21 determines that the vehicle 10 has stopped for signal waiting.

  When the vehicle 10 is waiting for a signal, the central processing unit 21 checks on which link the vehicle 10 is collated with the map information in the database 23 by comparing the position where the vehicle 10 was traveling before the vehicle 10 stopped. To determine if it has stopped.

  The central processing unit 21 causes the database 23 to store the link on which the vehicle is waiting for a signal and the time when the vehicle has started as signal waiting information. That is, the central processing unit 21 extracts signal waiting information from the operation information received from each of the vehicles 10 and stores the signal waiting information in the database 23.

(Signal cycle length estimation processing)
The central processing unit 21 executes signal cycle length estimation processing for estimating the signal cycle length for each intersection included in the map information of the database 23. Hereinafter, specific processing steps of the signal cycle length estimation process will be described using the intersection S1 shown in FIG. 2 as an example.

  The central processing unit 21 extracts links (link L1a to link L4a in this example) flowing into the intersection S1 from the map information in the database 23. Furthermore, the central processing unit 21 extracts from the database 23 signal waiting information (that is, start time) regarding the link L1a that is one of the links flowing into the intersection S1.

  By the way, traffic lights installed at intersections can be classified into multi-stage traffic lights whose signal cycle length changes according to the time zone during a day and single-stage traffic lights whose signal cycle length does not change throughout the day. As will be described later, the traffic signal installed at the intersection S1 is a multistage traffic signal. However, the central processing unit 21 cannot know in advance whether the traffic signal installed at each intersection (including the intersection S1) is a multistage signal or a single signal.

  Therefore, the central processing unit 21 estimates the signal cycle length for each time slot divided into 15-minute units in the signal cycle length estimation process. The estimated signal cycle length every 15 minutes is hereinafter also referred to as “estimated signal cycle length Ce”. If a single stage traffic light is installed at a certain intersection, the estimated signal cycle length Ce of the traffic light becomes equal to each other in all time zones. On the other hand, when a multistage traffic signal is installed at the intersection, the estimated signal cycle length Ce changes according to the time zone.

  The table shown in FIG. 3A represents the start times of the vehicles 10 that started after waiting for a signal on the link L1a between 11:00 and 11:15 on a certain day. . The “difference time” corresponding to each start time shown in this table represents a time difference between the start times adjacent to each other.

  This will be explained with an example. Each of the vehicles 10 corresponding to the start times indicated by “number 1” to “number 3” in FIG. 3A (also referred to as “vehicle 10a” to “vehicle 10c” for convenience) waits for a signal. This is shown in FIG. When the display of the traffic light T1a changes from red to blue, these vehicles 10 start in order from the leading vehicle (vehicle 10a in this example). After that, when the display of the traffic light T1a changes to red again after a round, the vehicle 10 corresponding to the start time indicated by “number 4” in FIG. 3A is also referred to as “vehicle 10d” for convenience. FIG. 3 [Bb] shows a state of waiting for a signal.

  For example, “difference time” (1.8 seconds) represented in the row of “number 2” in FIG. 3A is a time difference between the start time of the vehicle 10a and the start time of the vehicle 10b. . Similarly, “difference time” (1.2 seconds) shown in the row of “number 3” in FIG. 3A is a time difference between the start time of the vehicle 10b and the start time of the vehicle 10c. is there.

  When the display of the traffic light T1a changes from red to blue, the vehicle 10a, which is the leading vehicle, can start without waiting for the start of the preceding vehicle. However, in practice, the driver of the vehicle 10a recognizes the change of the display of the traffic light T1a to blue and then starts the vehicle 10a after confirming the safety in the vehicle traveling direction, so the display of the traffic light T1a is blue. There is a time difference (start delay time) between the time when the vehicle changes to 10 and the start time of the vehicle 10a.

  Since the vehicle 10d is also the leading vehicle, a start delay time occurs between the time when the display of the traffic light T1a changes to blue and the start time of the vehicle 10d.

  However, the start delay time included in the start time of the vehicle 10a is substantially equal to the start delay time included in the start time of the vehicle 10d. Therefore, when the time difference between the start time of the vehicle 10a and the start time of the vehicle 10d is calculated, the start delay time included in each of these two start times is offset. Therefore, the time difference between the start time of the vehicle 10a and the start time of the vehicle 10d is equal to the signal cycle length of the traffic light T1a.

  On the other hand, the start delay time of a vehicle (following vehicle) stopped at a position behind the leading vehicle is the preceding vehicle (including the leading vehicle, a vehicle waiting for a signal at a position before that vehicle) ) And the acceleration at the start of the preceding vehicle. In the example of FIG. 3 [Ba], the vehicle 10b, which is a subsequent vehicle, starts after the vehicle 10a, and the “vehicle 10c, which is a subsequent vehicle as well as the vehicle 10b” starts after the vehicle 10b.

  Therefore, it is difficult to estimate the signal cycle length based on the start times of the following vehicles 10b and 10c (for example, based on the time difference between the start time of the vehicle 10c and the start time of the vehicle 10d). is there. Therefore, when estimating the estimated signal cycle length Ce, the central processing unit 21 excludes the start time of the following vehicle from the set of start times accumulated in the database 23, and starts the remaining start time of the leading vehicle (that is, the leading vehicle). Refer only to (Start time).

  More specifically, since the vehicle 10b and the vehicle 10c, which are the following vehicles, can start following the start of the preceding vehicle, the difference time corresponding to the vehicle 10b and the difference time corresponding to the vehicle 10c are compared. Short.

  On the other hand, the vehicle 10d, which is the leading vehicle, starts after the vehicle 10c starts, and further, the display of the traffic light T1a changes from yellow to red, and then the display of the traffic light T1a changes to blue. Therefore, the difference time (116.8 seconds) corresponding to the vehicle 10d is longer than the difference time corresponding to the vehicle 10b and the difference time corresponding to the vehicle 10c.

  Therefore, when the difference time is smaller than a predetermined first time threshold value (5 seconds in this example), the central processing unit 21 determines that the vehicle 10 corresponding to the difference time is a subsequent vehicle. On the other hand, when the difference time is equal to or greater than the first time threshold, the central processing unit 21 determines that the vehicle 10 corresponding to the difference time is the leading vehicle. When the start time of the following vehicle is eliminated, a series of leading vehicle start times as shown in the table of FIG. 3 [C] are obtained.

  Further, the central processing unit 21 calculates a start interval that is a time difference between the start vehicle start times adjacent to each other. The central processing unit 21 generates a histogram for the start interval shown in FIG. The generated histogram is shown in FIG.

  Similarly, the central processing unit 21 calculates start intervals for the links L2a to L4a flowing into the intersection S1, and generates a histogram. The generated histograms are shown in FIGS.

  Furthermore, the central processing unit 21 generates a histogram as shown in FIG. 4 [E] by adding the histograms shown in FIGS. 4 [A] to [D]. FIG. 4E shows the distribution of start intervals at all links where the vehicle flows into the intersection S1.

  Among the traffic lights T1a to T1d installed at the intersection S1, when the traffic lights T1a and T1c are displayed in blue, the traffic lights T1b and T1d are displayed in red. In other words, the time zone in which the vehicle flows into the intersection S1 from the link L1a and the link L3a is different from the time zone in which the vehicle flows into the intersection S1 from the links L2a and L4a. However, the signal cycle lengths of these traffic lights are equal to each other. Accordingly, all of the start intervals appearing in the histogram shown in FIG. 4E are the start vehicle start times of the vehicles 10 waiting for the signals before each of the traffic lights operating with the same signal cycle length. Based on.

  The central processing unit 21 estimates the estimated signal cycle length Ce based on the start interval appearing in the histogram shown in FIG. That is, the estimated signal cycle length Ce is estimated based on the start time acquired from the vehicle 10 that has been waiting for the signal at all the links flowing into the target intersection. As a result, the number of information samples used for estimating the estimated signal cycle length Ce (in this case, the number of start intervals to be acquired) increases, so that the estimation accuracy of the estimated signal cycle length Ce can be increased. Become.

  The central processing unit 21 groups each start interval appearing in the histogram shown in FIG. 4E in order to estimate the estimated signal cycle length Ce. More specifically, when the time difference between two adjacent start intervals is smaller than a predetermined second time threshold value (10 seconds in this example), the central processing unit 21 has the same start interval. Is determined to belong to the group. On the other hand, when the time difference is equal to or greater than the second time threshold, the central processing unit 21 determines that these start intervals belong to different groups.

  As a result of the grouping, the start interval shown in FIG. 4E is divided into four groups, group Gr1 to group Gr4. That is, these start intervals are group Gr1, which is a set of start intervals in the vicinity of 120 seconds, group Gr2, which is a set of start intervals in the vicinity of 240 seconds, and groups Gr3, which are sets of start intervals in the vicinity of 360 seconds and near 480 seconds. Are included in one of the groups Gr4, which is a set of start intervals.

  In the example shown in FIG. 4E, there are 12 start intervals included in the group Gr1, 5 start intervals included in the group Gr2, 2 start intervals included in the group Gr3, and , There is one start interval included in the group Gr4. That is, the number of start intervals appearing in the group Gr1 is the largest compared to all other groups.

  The central processing unit 21 estimates that the average value of the start intervals included in the group including the most start intervals (representative group, group Gr1 in this example) is the estimated signal cycle length Ce.

A method of calculating the average value of the start intervals will be described using a simple example. For example, if the representative group contains 2 data for the start interval “119 seconds”, 4 data for the start interval “120 seconds”, and 3 data for the start interval “121 seconds”, the average value Av is calculated as in the following formula (1).

Av = (119 × 2 + 120 × 4 + 121 × 3) / (2 + 4 + 3) (1)

  FIG. 5 shows a process of calculating an average value for the start intervals included in the group Gr1 shown in FIG. In the table shown in FIG. 5, “start interval” corresponding to each of “numbers 1 to 8” is a start interval included in the group Gr1. In this table, “number of occurrences” is the number of occurrences of each start interval.

  The total value of “the value obtained by calculating the product of the start interval and the number of occurrences for each start interval” (24 minutes 01.5 seconds in this example) is the total number of occurrence times (in this example, 12 The value obtained by dividing by the number is the average value (in this example, 2 minutes and 10.125 seconds). This average value is the estimated signal cycle length Ce corresponding to the time zone from 11:00 to 11:15 of the intersection S1.

  The central processing unit 21 estimates a total of 96 (24 hours / 15 minutes) estimated signal cycle lengths Ce by performing the same processing for other time zones. Further, the central processing unit 21 generates a histogram representing the relationship between the estimated signal cycle length Ce and the number of occurrences. The generated histogram is shown in FIG.

  The central processing unit 21 groups the estimated signal cycle lengths Ce that appear in the histogram shown in FIG. More specifically, if the time difference between two adjacent estimated signal cycle lengths Ce is smaller than a predetermined third time threshold value (10 seconds in this example), the central processing unit 21 estimates these values. It is determined that the signal cycle length Ce belongs to the same group. On the other hand, if the time difference is equal to or greater than the third time threshold, the central processing unit 21 determines that these estimated signal cycle lengths Ce belong to different groups.

  As a result of the grouping, the estimated signal cycle length Ce shown in FIG. 6A is divided into five groups of groups Gp1 to Gp5. That is, these estimated signal cycle lengths Ce are a group Gp1 that is a set of estimated signal cycle lengths Ce near 120 seconds, a group Gp2 that is a set of estimated signal cycle lengths Ce near 140 seconds, and an estimated signal cycle near 240 seconds. Group Gp3, which is a set of long Ce, group Gp4, which is a set of estimated signal cycle length Ce in the vicinity of 280 seconds, and group Gp5, which is a set of estimated signal cycle length Ce in the vicinity of 360 seconds.

  In the example shown in FIG. 6, there are 58 estimated signal cycle lengths Ce included in the group Gp1, 34 estimated signal cycle lengths Ce included in the group Gp2, and an estimated signal cycle length Ce included in the group Gp3. There are two, the estimated signal cycle length Ce included in the group Gp4 is one, and the estimated signal cycle length Ce included in the group Gp5 is one.

  As described above, since the traffic signal installed at the intersection S1 is a multistage traffic signal, the estimated signal cycle length Ce is divided into a plurality of groups. In a multi-stage traffic light, the signal cycle length in a predetermined time zone during one day is different from the signal cycle length in other time zones. There may also be intersections where the signal cycle length switches more than three times during the day.

  However, usually, the predetermined period (that is, the time during which one of the signal cycle lengths continues) is set to a time longer than 90 minutes. Therefore, the central processing unit 21 determines that if the number of estimated signal cycle lengths Ce included in the group is equal to or less than a predetermined value (in this example, “6” corresponding to 90 minutes / 15 minutes), It is determined that the estimated signal cycle length Ce included in the group is an “outlier”.

  In this example, the estimated signal cycle length Ce included in the group Gp3, the group Gp4, and the group Gp5 corresponds to an outlier. The central processing unit 21 determines that the estimated signal cycle length Ce corresponding to the outlier deviates from the actual signal cycle length. On the other hand, the central processing unit 21 determines that each of the estimated signal cycle lengths Ce included in the group Gp1 and the group Gp2 correctly reflects the actual signal cycle length.

  On the other hand, the estimated signal cycle length Ce corresponding to the outlier has a value different from those even though the actual signal cycle length is a value in the vicinity of 120 seconds or a value in the vicinity of approximately 140 seconds. I understand that. This point will be described with reference to FIG. 7A showing a distribution of start intervals different from the distribution of start intervals appearing in the histogram shown in FIG. That is, the start interval appearing in FIG. 7A is calculated from the start time of the vehicle 10 waiting for a signal on the link L1a in a time zone different from 11:00 to 11:15.

  When the start intervals appearing in the histogram shown in FIG. 7A are grouped in the same manner as the start intervals appearing in FIG. 4E, the groups are divided into four groups Gr1a to Gr4a. That is, these start intervals are a group Gr1a which is a set of start intervals in the vicinity of 120 seconds, a group Gr2a which is a set of start intervals in the vicinity of 240 seconds, a group Gr3a which is a set of start intervals in the vicinity of 360 seconds and the vicinity of 480 seconds Are included in one of the groups Gr4a, which is a set of start intervals.

  In the histogram shown in FIG. 4E, the group Gr1 constituted by the start intervals near 120 seconds is a representative group, whereas in the histogram shown in FIG. 7A, the start intervals around 240 seconds are formed. The group Gr2a is a representative group.

  This difference is due to the following points. That is, in the example shown in FIG. 4E, there are many cases where the vehicle 10 waiting for a signal appears every time the display of the traffic light changes to red. On the other hand, in the example shown in FIG. 7A, when the traffic light display changes to red, the vehicle 10 waiting for the signal does not appear. Instead, when the traffic light display changes to red next time. In many cases, a vehicle 10 waiting for a signal appeared. That is, the frequency at which the start interval becomes a value in the vicinity of 240 seconds was high.

  As a result, the estimated signal cycle length Ce obtained by calculating the average value of the start intervals included in the representative group Gr2a in the histogram shown in FIG. 7A is a value in the vicinity of 240 seconds. Therefore, the estimated signal cycle length Ce is an outlier included in the group Gp3 shown in FIG. In other words, since the traffic volume at the intersection is small compared to the example shown in FIG. 4E, the estimated signal estimated based on the start interval included in the histogram shown in FIG. 7A. The cycle length Ce was approximately twice the actual signal cycle length (that is, a value in the vicinity of 240 seconds). For example, if the traffic volume at the intersection is still smaller, the estimated signal cycle length Ce is approximately three times the actual signal cycle length, or an integer multiple of the actual cycle length, and may be a larger value. That is, when the traffic volume at the intersection is small, the estimated signal cycle length Ce can be an outlier.

  The central processing unit 21 re-estimates the signal cycle length by applying the autocorrelation function to the start interval distribution corresponding to each of the estimated signal cycle lengths Ce determined to be outliers.

  More specifically, the central processing unit 21 calculates the relationship between various start interval values (input value n) and the number of times the start interval is acquired (output value x (n)) as a function. (Starting interval function) and an autocorrelation function is applied to the function x (n).

  For example, in the histogram shown in FIG. 7A, the horizontal axis represents the start interval n, and the vertical axis represents the number of times x (n) each start interval has occurred. It can be considered as a graph of x (n).

The autocorrelation function is specifically expressed by the following equation (2). In equation (2), N is the maximum value of the acquired start interval n (see FIG. 7A), and k is the amount of movement of the start interval function x (n). The movement amount k is an input value (variable) of the autocorrelation function R (k) expressed by the equation (2). The autocorrelation function R (k) is a graph of the start interval function x (n) and a graph obtained by moving the graph in the direction in which the value on the horizontal axis decreases by the movement amount k (ie, x (n + k)). And has a positive correlation with the area of the overlapping region.

  FIG. 8 is a graph of the autocorrelation function R (k) obtained by applying the start interval function x (n) shown in FIG. 7A to the equation (2). As understood from FIG. 8, the autocorrelation function R (k) has a maximum value Rmax when the movement amount k = 0, a maximum value RLm1 near the movement amount k = 120, and a maximum value near the movement amount k = 240. The value is RLm2. Further, the autocorrelation function R (k) has a maximum value RLm3 in the vicinity of the movement amount k = 360, and has a maximum value RLm4 in the vicinity of the movement amount k = 480.

  FIG. 7B is a graph obtained by moving the start interval function x (n) shown in FIG. 7A to the left in the horizontal axis direction for 120 seconds. In other words, FIG. 7B represents a graph of the start interval function x (n + k) when the movement amount k = 120. As can be understood from FIGS. 7A and 7B, when the movement amount k = 120, the area of the region where these two graphs overlap is the movement amount k is another value (for example, k = 60). It becomes larger compared with the case of.

  The area of the region where the graphs overlap increases once when the movement amount k further increases and the movement amount k = 240. Therefore, the graph of the autocorrelation function R (k) shown in FIG. 8 has a maximum value when the movement amount k = 120 and when the movement amount k = 240. Similarly, the autocorrelation function R (k) is also maximized when the movement amount k = 360 and when the movement amount k = 480.

  In other words, when the moving amount k is “120” and when the moving amount k is a multiple of “120”, the autocorrelation function R (k) is maximized. This means that the start interval function x (n) has periodicity and the period is “120”. That is, the minimum movement amount k (where k> 0) of the movement amounts k at which the autocorrelation function R (k) has a maximum value is the period of the start interval function x (n).

  The start interval function x (n) has periodicity when the start interval n which is an input value of this function is equal to the signal cycle length and when the start interval n is equal to a multiple of the signal cycle length. (I.e., the output value of this function increases). More specifically, as described above, when the traffic volume at the intersection is large, the vehicle 10 that waits for a signal appears every time the traffic light display changes to red, so the start interval n becomes equal to the signal cycle length. When the traffic volume at the intersection is small, there may occur a case where the vehicle 10 waiting for the signal does not appear even if the traffic light is displayed in red, so the start interval n is an integral multiple of the signal cycle length. As a result, the start interval function x (n) has an output value that increases when the start interval n is equal to the signal cycle length and when the start interval n is equal to a multiple of the signal cycle length, and thus a period equal to the signal cycle length. Have Therefore, the central processing unit 21 can estimate that the estimated signal cycle length Ce is a value equal to the period of the start interval function x (n).

  As described above, the central processing unit 21 corresponds to each of the estimated signal cycle lengths Ce determined as outliers (estimated signal cycle lengths Ce included in the group Gp3 to the group Gp5 in the example of FIG. 6A). The autocorrelation function is applied to the distribution of start intervals. As a result, the estimated signal cycle length Ce distributed as shown in FIG. 6 [A] is distributed as shown in FIG. 6 [B].

  The central processing unit 21 groups the estimated signal cycle lengths Ce that appear in the histogram shown in FIG. 6B. As a result of the grouping, the estimated signal cycle length Ce shown in FIG. 6B is divided into two groups, a group Gp1a and a group Gp2a. That is, these estimated signal cycle lengths Ce are included in either group Gp1a, which is a set of estimated signal cycle lengths Ce in the vicinity of 120 seconds, or group Gp2a, which is a set of estimated signal cycle lengths Ce in the vicinity of 140 seconds.

  As understood from FIG. 6B, the estimated signal cycle length Ce determined as an outlier based on the histogram shown in FIG. 6A is the result of re-estimation by the autocorrelation function R (k). , Included in any of group Gp1a and group Gp2a. In the example shown in FIG. 6, there are 60 estimated signal cycle lengths Ce included in the group Gp1a, and 36 estimated signal cycle lengths Ce included in the group Gp2a.

  Further, the central processing unit 21 calculates the average value of the estimated signal cycle length Ce constituting the group Gp1a and the average value of the estimated signal cycle length Ce constituting the group Gp2a. In this example, the average value of the estimated signal cycle length Ce constituting the group Gp1a is 120 seconds, and the average value of the estimated signal cycle length Ce constituting the group Gp2a is 140 seconds.

  That is, it can be seen that the traffic light installed at the intersection S1 is a multistage traffic light whose signal cycle length is switched between 120 seconds and 140 seconds. As described above, the start times used for estimating the estimated signal cycle length Ce are tabulated in units of 15 minutes. Therefore, since the estimated signal cycle length Ce constituting the group Gp1a is 60, the period in which the signal cycle length is 120 seconds in one day is 15 hours (15 minutes × 60). Similarly, since the estimated signal cycle length Ce constituting the group Gp1a is 36, the period in which the signal cycle length is 140 seconds in one day is 9 hours (15 minutes × 36).

  The central processing unit 21 sequentially estimates the signal cycle length for intersections other than the intersection S1 included in the map information in the database 23.

  The central processing unit 21 distributes the estimated signal cycle length to the vehicle 10 via the central communication unit 22. More specifically, the central processing unit 21 determines the signal cycle length of the intersection (that is, the intersection in the traveling direction of the vehicle 10) into which the link on which each vehicle 10 is traveling and the link acquired last time. Is transmitted to the vehicle 10 via the central communication unit 22. This start vehicle start time is the time when the start delay time is added to the time when the traffic light in front of the vehicle 10 turns blue in the past.

  The calculation unit 11 of the vehicle 10 that has received the information adds the “time obtained by multiplying the signal cycle length by an integer” to the start vehicle start time, and further reduces the “predetermined value corresponding to the start delay time”. Estimate the time when the front traffic light display next turns blue.

  If the vehicle 10 remains stopped at the estimated time, the calculation unit 11 determines that the driver has overlooked the display of the traffic light, and the display and speaker included in the display device 15 for the driver. Warn through. As a result, the driver does not notice that the display of the traffic light has changed to blue, and therefore, it is possible to avoid the vehicle 10 from continuing the stop state.

(Specific operation)
Next, a specific operation of the central processing unit 21 will be described. The CPU of the central processing unit 21 (hereinafter also simply referred to as “CPU”) executes the “start time accumulation processing routine” represented by the flowchart in FIG. 9 every time a predetermined time elapses.

  That is, when the appropriate timing is reached, the CPU starts the process from step 900 of FIG. 9 and proceeds to step 905. After the routine is executed last time, the central communication unit 22 newly receives the operation information received from the vehicle 10. It is determined whether or not exists.

  If new operation information has been acquired, the CPU makes a “Yes” determination at step 905 to proceed to step 910 to acquire operation information from the database 23.

  Next, the CPU proceeds to step 915 and extracts the link and start time at which the vehicle 10 was waiting for a signal from the location where the vehicle 10 traveled and the stop position included in the operation information. More specifically, the CPU travels before the vehicle 10 stops and the position where the vehicle 10 is waiting for a signal by comparing the location with the map information in the database 23 before the vehicle 10 stops. Extract the corresponding link. Furthermore, after the vehicle 10 waits for a signal at the link, the CPU extracts the time at which the vehicle 10 started (that is, the start time) from the operation information.

  Next, the CPU proceeds to step 920 and causes the database 23 to store the extracted link and start time combination. That is, the CPU generates signal waiting information.

  Next, the CPU proceeds to step 925 to determine whether or not links and start times have been extracted from all the operation information newly received by the central communication unit 22. If there is operation information that has not yet been processed, the CPU makes a “No” determination at step 925 to proceed to step 915 to repeatedly execute link and start time extraction.

  On the other hand, if the link and the start time are extracted from all the operation information, the CPU proceeds to step 995 and completes this routine once. If the central communication unit 22 has not received new operation information, the CPU makes a “No” determination at step 905 to proceed directly to step 995.

  On the other hand, the CPU executes the “signal cycle length estimation processing routine” represented by the flowchart in FIG. 10 every time a predetermined time elapses. That is, when the appropriate timing is reached, the CPU starts processing from step 1000, proceeds to step 1005, and selects an intersection for estimating the signal cycle length from the intersections included in the map information of the database 23.

  Next, the CPU proceeds to step 1010 and extracts a link (target link) into which the vehicle 10 flows into the selected intersection. Next, the CPU proceeds to step 1015 to acquire from the database 23 the start time of the vehicle 10 that has been waiting for the signal on the target link, and further extracts the start time (that is, the start vehicle start time) related to the top vehicle. To do. Next, the CPU proceeds to step 1020 to acquire a start interval based on the time difference between the start vehicle start times.

  Next, the CPU proceeds to step 1025 to estimate the estimated signal cycle length Ce for each time zone of 15 minutes. More specifically, the CPU groups the start intervals according to the start interval distribution, and selects a representative group that is a group including the most start intervals. Further, the CPU estimates that the average value of the start intervals included in the representative group is the estimated signal cycle length Ce.

  Next, the CPU proceeds to step 1030 to determine whether or not the estimated signal cycle length Ce has been estimated for all time zones. If there is a time zone in which the estimated signal cycle length Ce that has not yet been estimated exists, the CPU makes a “No” determination at step 1030 to proceed to step 1025 to repeat estimation of the estimated signal cycle length Ce.

  On the other hand, if the estimated signal cycle length Ce is estimated for all the time zones, the CPU makes a “Yes” determination at step 1030 to proceed to step 1035 to group the estimated signal cycle length Ce. Next, the CPU proceeds to step 1040 to determine whether or not an outlier exists in the estimated signal cycle length Ce.

  If an outlier exists, the CPU makes a “Yes” determination at step 1040 to proceed to step 1045, where the estimated signal cycle length Ce using the autocorrelation function is calculated for the estimated signal cycle length Ce that has become an outlier. Is estimated again.

  Next, the CPU proceeds to step 1050 to determine whether or not the estimated signal cycle length Ce is estimated using the autocorrelation function for all outliers. If there is an outlier whose estimated signal cycle length Ce has not been estimated yet by the autocorrelation function, the CPU makes a “No” determination at step 1050 to proceed to step 1045, where the estimated signal cycle using the autocorrelation function is determined. Repeat the estimation of the length Ce.

  On the other hand, if the estimated signal cycle length Ce is estimated by the autocorrelation function for all outliers, the CPU makes a “Yes” determination at step 1050 to proceed to step 1055, where the estimation included in each group An average value of the signal cycle length Ce is calculated. Next, the CPU proceeds to step 1095 to end the present routine tentatively.

  If no outlier is included in the estimated signal cycle length Ce, the CPU makes a “No” determination at step 1040 to proceed directly to step 1055.

  By the way, it is also possible to estimate the estimated signal cycle length Ce corresponding to all time zones by using an autocorrelation function. However, in this case, the calculation amount of the central processing unit 21 increases drastically, and as a result, it becomes difficult to estimate the estimated signal cycle length Ce for all intersections. Therefore, the central processing unit 21 first estimates the estimated signal cycle length Ce by calculating the average value of the start intervals included in the representative group, that is, by a relatively simple method (step 1025 in FIG. 10). . Further, if the estimated signal cycle length Ce includes an outlier, the central processing unit 21 estimates the autocorrelation function by applying an autocorrelation function to the start interval distribution that caused the outlier. The signal cycle length Ce is reestimated (step 1045 in FIG. 10). By doing so, it is possible to accurately estimate the estimated signal cycle length Ce while suppressing the calculation amount of the central processing unit 21.

As described above, this estimation system is
A plurality of leading vehicle start times, which are times when the leading vehicle in the row of vehicles (10) stopped according to the red display of the traffic lights (traffic signals T1a to T1d) installed at the intersection (intersection S1) started according to the blue display of the traffic light. A starting time acquisition unit to acquire (the flowchart shown in FIG. 9 and step 1015 in FIG. 10);
A start interval acquisition unit (FIG. 10) that acquires a start interval that is a time difference between the start vehicle start times adjacent to each other among the plurality of start vehicle start times and the number of times each of the start vehicle start times has occurred. Step 1020),
An estimation unit (step 1025 and step 1030 in FIG. 10) for estimating a signal cycle length, which is a cycle in which the display of the traffic light makes a round based on the start interval;
An intersection signal cycle length estimation system comprising:
The estimation unit includes
For the start interval function x (n) where the input value is the start interval n and the output value is the number x of occurrence of the start interval n as the input value, the input value is the movement amount k of the start interval n. , The output value is an autocorrelation function (R (R ()) which is a value corresponding to the product of the start interval function x (n) and the start interval function x (n + k) obtained by moving the start interval function x (n) by the movement amount k. k))
In the range where the movement amount k is larger than 0, the minimum movement amount k of the movement amount k at which the output value of the autocorrelation function becomes a maximum is estimated as the signal cycle length (steps of FIGS. 8 and 10). 1045),
It is configured as follows.

  According to this estimation method, the signal cycle length can be accurately estimated even when the traffic volume is small.

  The embodiment of the signal cycle length estimation method for an intersection according to the present invention has been described above. However, the present invention is not limited to the above embodiment, and various modifications can be made without departing from the object of the present invention. is there. For example, in the present embodiment, the central processing unit 21 reestimates the signal cycle length using the autocorrelation function only for the estimated signal cycle length Ce corresponding to the outlier. However, the central processing unit 21 may estimate all signal cycle lengths using an autocorrelation function.

  In addition, in the present embodiment, when the estimated signal cycle length Ce is grouped, the central processing unit 21 determines whether or not the estimated signal cycle length Ce included in the group is equal to or less than a predetermined value. Whether or not is an outlier. However, the central processing unit 21 may determine whether it is an outlier using another method. For example, the central processing unit 21 has a ratio of “the number of estimated signal cycle lengths Ce included in a specific group” to “the number of estimated signal cycle lengths Ce included in the group having the largest number of estimated signal cycle lengths Ce”. When the value is smaller than the predetermined value, the estimated signal cycle length Ce included in the specific group may be determined to be an outlier.

  In addition, in the present embodiment, the central processing unit 21 considers a link where the vehicle 10 flows into an intersection (target intersection) where the signal cycle length is estimated as a target link. However, if an intersection controlled by a signal cycle length equal to that of the target intersection is known (so-called system-controlled intersection), a link into which a vehicle flows into the system-controlled intersection may be treated as the target link. Good. That is, the central processing unit 21 may treat a plurality of system-controlled intersections as one intersection in the estimation of the signal cycle length.

  In addition, in the present embodiment, the central processing unit 21 does not consider the possibility that the signal cycle length changes depending on the day even at the same intersection. However, the central processing unit 21 may estimate the signal cycle length for each of weekdays, Saturdays, Sundays, and holidays.

  In addition, in this embodiment, a signal from the GPS satellite 32 is used to acquire the current position of the vehicle 10. However, in order to acquire the current position of the vehicle 10, a short-range communication system (for example, DSRC (Dedicated Short Range Communication)) may be used in addition to the signal from the GPS satellite 32.

  In addition, all or a part of the calculation unit 11, the in-vehicle communication unit 12, the GPS reception unit 13, and the display device 15 included in the vehicle 10 may be realized by a navigation system mounted on the vehicle 10.

  Vehicle 10, calculation unit 11, in-vehicle communication unit 12, GPS reception unit 13, vehicle speed sensor 14, display device 15, information processing center 20, central processing unit 21, central communication unit 22, database ... 23.

Claims (1)

A start time acquisition unit for acquiring a plurality of start vehicle start times, which are times when the start vehicle in the train of vehicles stopped according to the red display of the traffic lights installed at the intersection starts according to the blue display of the traffic lights;
A start interval acquisition unit that acquires a start interval that is a time difference between the start vehicle start times adjacent to each other among the plurality of start vehicle start times, and the number of times each of the start intervals has occurred,
An estimation unit that estimates a signal cycle length, which is a cycle in which the display of the traffic light makes a round based on the start interval;
An intersection signal cycle length estimation system comprising:
The estimation unit includes
For the start interval function x (n) where the input value is the start interval n and the output value is the number x of occurrence of the start interval n as the input value, the input value is the movement amount k of the start interval n. Then, an autocorrelation function whose output value is a value corresponding to the product of the start interval function x (n) and the start interval function x (n + k) obtained by moving the start interval function x (n) by the movement amount k is applied. ,
In the range where the movement amount k is larger than 0, the minimum movement amount k among the movement amounts k at which the output value of the autocorrelation function becomes maximum is estimated as the signal cycle length.
An intersection signal cycle length estimation system configured as described above.

Images