JP7635640B2 - Traffic condition detection device, traffic condition detection method, and program - Google Patents

Description

The present invention relates to a traffic condition detection device, a traffic condition detection method, and a program.

In recent years, various methods have been known for detecting traffic conditions on roads. For example, a technology has been disclosed that detects traffic conditions based on images captured by a camera (see, for example, Patent Document 1). Also, a technology has been disclosed that detects traffic conditions using two free-flow antennas (see, for example, Patent Document 2).

Furthermore, a technology has been disclosed that detects traffic conditions at all points on a road based on traffic volume and vehicle speed measured by a traffic counter and vehicle position information and vehicle speed detected by a probe antenna (see, for example, Patent Document 3).

Patent No. 6060058 Patent No. 5017170 JP 2020-86647 A

However, it would be desirable to provide technology that would enable vehicle speed to be measured with high accuracy and precision while keeping costs down.

In order to solve the above problem, according to one aspect of the present invention, a traffic condition detection device is provided that includes: a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result; a traffic volume distribution determination unit that calculates a first frequency distribution for each predetermined time of a first traffic volume corresponding to the predetermined position, calculates a first parameter by approximating the first frequency distribution with a first statistical distribution, and calculates the probability that the traffic flow state on the road is a congested flow and the probability that the traffic flow state is a free flow based on the first parameter, data on the variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning; a traffic flow determination unit that obtains a determination result of the traffic flow state based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and a speed determination unit that determines the vehicle speed corresponding to the predetermined position based on the determination result of the traffic flow state, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed prepared in advance.

The traffic volume distribution determination unit may calculate a second frequency distribution of a second traffic volume corresponding to the specified position in a free flow, calculate a second parameter by approximating the second frequency distribution with the first statistical distribution, and calculate a reference parameter corresponding to the free flow based on the second parameter.

The traffic volume distribution determination unit may calculate a reference parameter corresponding to the free flow by approximating a frequency distribution of the correspondence between the second parameter and data related to the variance of the second frequency distribution with a second statistical distribution.

The traffic volume distribution determination unit may calculate a third frequency distribution of a third traffic volume corresponding to the specified position in the congestion flow, calculate a third parameter by approximating the third frequency distribution with the first statistical distribution, and calculate a reference parameter corresponding to the congestion flow based on the third parameter.

The traffic volume distribution determination unit may calculate a reference parameter corresponding to the congestion flow by approximating a frequency distribution of the correspondence between the third parameter and data related to the variance of the third frequency distribution with a third statistical distribution.

The traffic volume distribution determination unit may calculate a second frequency distribution of a second traffic volume corresponding to the specified position in a free flow, calculate a second parameter by approximating the second frequency distribution with the first statistical distribution, calculate a third frequency distribution of a third traffic volume corresponding to the specified position in a congested flow, and calculate a third parameter by approximating the third frequency distribution with the first statistical distribution, and calculate the reference parameter based on a combination of learning data including the second parameter and data related to a variation of the second frequency distribution and a teacher label indicating a free flow, and a combination of learning data including the third parameter and data related to a variation of the third frequency distribution and a teacher label indicating a congested flow.
The reference parameters in this case may be parameters of a learning model generated by a learning process based on a combination of learning data and teacher labels.

According to another aspect of the present invention, a traffic condition detection device is provided that includes: a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result; a delay time determination unit that calculates a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in an on-board device to when it is acquired by a vehicle speed acquisition unit, and calculates the probability that the traffic flow state on the road is a congested flow and the probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter previously calculated by machine learning; a traffic flow determination unit that obtains a determination result of the traffic flow state based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and a speed determination unit that determines the vehicle speed corresponding to the predetermined position based on the determination result of the traffic flow state, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed that has been prepared in advance.

The delay time determination unit may calculate a second frequency distribution of the second delay time in the free flow for each predetermined time, and calculate a reference parameter corresponding to the free flow by approximating the second frequency distribution with a first statistical distribution.

The delay time determination unit may calculate a third frequency distribution of a third delay time in the traffic jam flow for each predetermined time, and calculate a reference parameter corresponding to the traffic jam flow by approximating the third frequency distribution with a first statistical distribution.

The vehicle speed acquisition unit may include a probe antenna.

The vehicle detection unit may detect the vehicle in real time.

The vehicle detection unit may include a free-flow antenna.

According to another aspect of the present invention, there is provided a traffic condition detection method, the method comprising: a computer acquiring a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculating a traffic volume corresponding to the predetermined position based on the vehicle detection result; calculating a first frequency distribution for each predetermined time of a first traffic volume corresponding to the predetermined position, approximating the first frequency distribution with a first statistical distribution to calculate a first parameter, calculating a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first parameter, data related to a variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning; obtaining a traffic flow state determination result based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and determining a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed prepared in advance.

According to another aspect of the present invention, there is provided a traffic condition detection method, the method comprising: a computer acquiring a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculating a traffic volume corresponding to the predetermined position based on the vehicle detection result; calculating a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in an on-board device to when it is acquired by a vehicle speed acquisition unit, and calculating a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter calculated in advance by machine learning; obtaining a traffic flow state determination result based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and determining the vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed prepared in advance.

According to another aspect of the present invention, a program is provided that causes a computer to function as a traffic condition detection device that includes a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result; a traffic volume distribution determination unit that calculates a first frequency distribution for each predetermined time of a first traffic volume corresponding to the predetermined position, calculates a first parameter by approximating the first frequency distribution with a first statistical distribution, and calculates the probability that the traffic flow state on the road is a congested flow and the probability that the traffic flow state is a free flow based on the first parameter, data on the variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning; a traffic flow determination unit that obtains a determination result of the traffic flow state based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and a speed determination unit that determines the vehicle speed corresponding to the predetermined position based on the determination result of the traffic flow state, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed prepared in advance.

According to another aspect of the present invention, a program is provided that causes a computer to function as a traffic condition detection device that includes a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result; a delay time determination unit that calculates a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in the on-board device to when it is acquired by the vehicle speed acquisition unit and calculates the probability that the traffic flow state on the road is a congested flow and the probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter previously calculated by machine learning; a traffic flow determination unit that obtains a determination result of the traffic flow state based on the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow; and a speed determination unit that determines the vehicle speed corresponding to the predetermined position based on the determination result of the traffic flow state, the traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed that has been prepared in advance.

As described above, the present invention provides technology that makes it possible to measure vehicle speed with high accuracy and precision while keeping costs down.

FIG. 11 is a diagram showing the relationship between the interval at which vehicles are detected (vehicle detection interval) and the probability density when the traffic state is free flow. FIG. 13 is a diagram showing the relationship between the interval at which vehicles are detected (vehicle detection interval) and the probability density at a location where the traffic state is congestion. FIG. 13 is a diagram showing a comparison of the frequency distribution of parameters obtained by approximating the correspondence relationship between vehicle detection intervals and probability density to an exponential distribution when the traffic state is free flow and when the traffic state is congested flow. FIG. 11 is a diagram showing a comparison of the correspondence between the parameters obtained by approximating the correspondence between the vehicle detection interval and the probability density to an exponential distribution and the approximation error when the traffic state is free flow and when the traffic state is congested flow. 1 is a diagram illustrating an example of a functional configuration of a traffic condition detection device according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of free flow data stored by a free flow data storage unit. 4 is a diagram showing an example of probe data stored in a probe data storage unit; FIG. 13 is a flowchart showing an example of a traffic volume calculation process executed by a traffic volume calculation unit. FIG. 2 is a diagram showing an example of a QV diagram represented by QV data. 13 is a flowchart showing an example of a QV data creation process executed by a QV data creation unit. 13 is a flowchart illustrating an example of a process for generating a judgment criterion parameter executed by a speed calculation unit. FIG. 13 is a diagram showing the correspondence relationship between the traffic volume and the traffic volume difference at the end time of a time interval for each change in traffic flow state. FIG. 13 is a diagram showing the correspondence relationship between parameters relating to statistical distribution and standard deviation ratios. A figure showing an example of probe delay time (seconds) at each time in front of a free flow antenna. FIG. 13 is a diagram showing a frequency distribution of probe delay times for each predetermined time period. 11 is a flowchart illustrating an example of a velocity calculation process executed by a velocity calculation unit. 1 is a diagram illustrating a hardware configuration of an information processing device as an example of a traffic condition detection device according to an embodiment of the present invention.

The preferred embodiment of the present invention will be described in detail below with reference to the attached drawings. Note that in this specification and the drawings, components having substantially the same functional configuration are designated by the same reference numerals to avoid redundant description.

In addition, in this specification and drawings, multiple components having substantially the same functional configuration may be distinguished by adding different numbers after the same reference symbol. However, if there is no particular need to distinguish between multiple components having substantially the same functional configuration, only the same reference symbol will be used. In addition, similar components in different embodiments may be distinguished by adding different letters after the same reference symbol. However, if there is no particular need to distinguish between similar components in different embodiments, only the same reference symbol will be used.

(0. Overview)
First, an overview of the embodiment of the present invention will be described.

In recent years, various methods have been known for detecting traffic conditions on roads. For example, a first existing technology is disclosed that detects traffic conditions based on images captured by a camera (see, for example, Patent Document 1). However, since the first existing technology is difficult to apply to certain external environments (for example, rainy weather, nighttime, afternoon sun, etc.), it is difficult to obtain vehicle speed with high accuracy using the first existing technology.

A second existing technology has been disclosed that uses two free-flow antennas to detect traffic conditions (see, for example, Patent Document 2). However, since the second existing technology can only estimate the average vehicle speed between the two free-flow antennas, it is difficult to obtain vehicle speed with high precision using the second existing technology.

Furthermore, a third existing technology has been disclosed that detects traffic conditions at all points on a road based on traffic volume and vehicle speed measured by a traffic counter and vehicle position information and vehicle speed detected by a probe antenna (see, for example, Patent Document 3). However, because the third existing technology uses a traffic counter to measure traffic volume and vehicle speed, the costs of installing and maintaining the traffic counter are high.

In contrast, the technology according to the embodiment of the present invention detects traffic volume and vehicle speed using a free-flow antenna instead of a traffic counter. Therefore, the technology according to the embodiment of the present invention does not require the expense of installing and maintaining a traffic counter, and therefore the necessary costs can be reduced.

There is also a fourth existing technology that calculates vehicle speed using statistical information on the time intervals between vehicles detected by a free-flow antenna. Generally, a probe antenna can acquire vehicle speed from an on-board unit, but there may be points where it takes a significant amount of time (delay time) from when the vehicle speed is recorded in the on-board unit until the vehicle speed is acquired by the probe antenna. The fourth existing technology is a technology that estimates vehicle speed based on the results of vehicle detection by a free-flow antenna on roads (e.g., expressways) where both a probe antenna and a free-flow antenna are installed.

However, there are areas in the fourth existing technology that need to be improved. Below, we will explain the areas in the fourth existing technology that need to be improved with reference to Figures 1 to 4.

Figure 1 is a diagram showing the relationship between the interval at which vehicles are detected (vehicle detection interval) and the probability density when traffic conditions are free flow. Referring to Figure 1, the correspondence relationship between vehicle detection interval and probability density when traffic conditions are free flow is shown with the horizontal axis x representing the vehicle detection interval and the vertical axis y representing the probability density. The fourth existing technology approximates the correspondence relationship between vehicle detection interval x and probability density y when traffic conditions are free flow to an exponential distribution y = λ * exp (-λ × x) to calculate the parameter λ and the approximation error.

Figure 2 is a diagram showing the relationship between the interval at which vehicles are detected (vehicle detection interval) at a location where the traffic state is congested flow and the probability density. Referring to Figure 2, the correspondence relationship between the vehicle detection interval and the probability density at a location where the traffic state is congested flow is shown with the horizontal axis x representing the vehicle detection interval and the vertical axis y representing the probability density. The fourth existing technology approximates the correspondence relationship between the vehicle detection interval x and the probability density y at a location where the traffic state is congested flow to an exponential distribution y = λ * exp (-λ × x) to calculate the parameter λ and the approximation error.

Figure 3 shows a comparison of the frequency distribution of a parameter obtained by approximating the correspondence between vehicle detection intervals and probability density to an exponential distribution when the traffic state is free flow and when the traffic state is congested flow. Referring to Figure 3, there is an overlap in the frequency distribution of the parameter λ when the traffic state is free flow and when the traffic state is congested flow. Therefore, it may be difficult to determine whether the traffic state is free flow or congested flow from the frequency distribution of the parameter λ.

Figure 4 is a diagram showing a comparison of the correspondence between the parameters obtained by approximating the correspondence between the vehicle detection interval and the probability density to an exponential distribution and the approximation error when the traffic state is free flow and when the traffic state is congested flow. Referring to Figure 4, there is an overlap in the correspondence between the parameter λ and the approximation error when the traffic state is free flow and when the traffic state is congested flow. Therefore, it may be difficult to determine whether the traffic state is free flow or congested flow from the correspondence between the parameter λ and the approximation error.

In particular, when traffic volume is high and the resolution of vehicle detection time is low, it can be difficult to determine whether the traffic state is free flow or congested flow. And when it is difficult to determine whether the traffic state is free flow or congested flow, it is difficult to estimate vehicle speed with high accuracy.

As an example, in the learning stage, the technology according to the embodiment of the present invention generates a judgment criterion parameter by machine learning based on the average and variance of the frequency distribution of traffic volume for each given time period obtained from the accumulated free flow data. Then, in the inference stage, the technology according to the embodiment of the present invention judges whether the traffic state is free flow or congested flow based on the generated judgment criterion parameter and a judgment parameter based on the acquired free flow data.

As another example, in the learning stage, the technology according to the embodiment of the present invention generates a judgment criterion parameter by machine learning based on the delay time of the accumulated probe data. Then, in the inference stage, the technology according to the embodiment of the present invention determines whether the traffic state is free flow or congested flow based on the generated judgment criterion parameter and the delay time based on the acquired probe data.

The technology according to the embodiment of the present invention can estimate with higher accuracy whether the traffic state is free flow or congested flow. As a result, the technology according to the embodiment of the present invention can estimate the vehicle speed with higher accuracy.

The above describes an overview of an embodiment of the present invention.

1. DETAILED DESCRIPTION OF THE EMBODIMENTS
First, the details of the embodiment of the present invention will be described.

(1-1. Configuration of the Traffic Condition Detection Device)
First, a configuration example of the traffic condition detection device 1 according to the embodiment of the present invention will be described. Fig. 5 is a diagram showing an example of the functional configuration of the traffic condition detection device 1 according to the embodiment of the present invention.

Referring to FIG. 5, vehicles M1 to M3 are shown as examples of vehicles traveling on a road. Referring further to FIG. 5, vehicles M1 and M2 are traveling in the rear lane (vehicle M1 is traveling following vehicle M2), and vehicle M3 is traveling in the front lane in the opposite direction to vehicles M1 and M2 in the rear lane. Thus, in the embodiment of the present invention, it is mainly assumed that the road is made up of multiple lanes, but the road may also be made up of a single lane.

The traffic condition detection device 1 according to the embodiment of the present invention includes a probe data storage unit 121, a free flow data storage unit 122, a QV data storage unit 123, a parameter storage unit 124, a QV data creation unit 130, a traffic volume calculation unit 140, a speed calculation unit 150, and a traffic condition detection unit 170. A probe antenna 112 is connected to the probe data storage unit 121, and a free flow antenna 114 is connected to the free flow data storage unit 122.

The QV data creation unit 130, the traffic volume calculation unit 140, the speed calculation unit 150, and the traffic condition detection unit 170 each include a calculation device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and their functions can be realized by a program stored in a ROM (Read Only Memory) being expanded into a RAM by the calculation device and executed. At this time, a computer-readable recording medium on which the program is recorded can also be provided.

Alternatively, the QV data creation unit 130, the traffic volume calculation unit 140, the speed calculation unit 150, and the traffic condition detection unit 170 may be configured with dedicated hardware, or may be configured with a combination of multiple pieces of hardware. Data necessary for the calculation by the calculation device is appropriately stored in a storage unit (not shown).

The probe data storage unit 121, the free flow data storage unit 122, the QV data storage unit 123, and the parameter storage unit 124 are stored in a storage unit (not shown). Such a storage unit may be configured with a memory such as a RAM (Random Access Memory), a hard disk drive, or a flash memory.

(Free flow antenna 114)
The free flow antenna 114 functions as an example of a vehicle detection unit that detects a vehicle traveling at a predetermined position on a road. That is, in the embodiment of the present invention, the vehicle detection unit mainly assumes a case where the vehicle detection unit includes the free flow antenna 114. As a result, since an ETC free flow antenna of an already constructed ETC (Electronic Toll Collection) system can be used as a vehicle detection unit, there is no need to provide a new vehicle detection unit. However, instead of the free flow antenna 114, another vehicle detection unit (for example, an infrared sensor or an ultrasonic sensor) may be used.

More specifically, the free flow antenna 114 detects vehicles in real time by receiving vehicle identification information (vehicle ID) from the vehicle through communication with an onboard unit mounted on the vehicle. In the embodiment of the present invention, it is mainly assumed that an ETC onboard unit is used as an example of the onboard unit. It is assumed that the free flow antenna 114 is compatible with multiple versions of ETC onboard units, while the probe antenna 112 described later is compatible only with a specific version of ETC onboard units. In other words, the free flow antenna 114 can detect more vehicles than the probe antenna 112.

The "predetermined position" on the road is not particularly limited as long as it is the position of the vehicle that can be detected by the vehicle detection unit. In the following, the position of the vehicle that can be detected by the free flow antenna 114 (predetermined position) may be simply referred to as "in front of the free flow antenna." However, the position of the vehicle that can be detected by the free flow antenna 114 may also be a position that is shifted from in front of the free flow antenna.

When the free flow antenna 114 detects a vehicle in real time by receiving a vehicle ID through communication with an onboard device mounted on the vehicle, it outputs the vehicle detection result (hereinafter also referred to as "free flow data") to the free flow data storage unit 122 in real time. The vehicle detection result output in real time from the free flow antenna 114 to the free flow data storage unit 122 can then be used in real time by the traffic volume calculation unit 140 and the speed calculation unit 150.

Here, "real-time" can mean any short period of time between when a vehicle arrives in front of the free-flow antenna and before the traffic conditions on the road change. If a vehicle is detected at this time, some measure can be taken according to the traffic conditions at the time the vehicle is detected, before the traffic conditions on the road change.

Figure 6 is a diagram showing an example of free flow data stored by the free flow data storage unit 122. As shown in Figure 6, the free flow data is made up of a "vehicle ID" and a "passing time" associated with each other. The "vehicle ID" is vehicle identification information received from the vehicle through communication between the free flow antenna 114 and an onboard device mounted on the vehicle. The "passing time" is the time when the vehicle ID is received from the vehicle by the free flow antenna 114, and may correspond to the time when the vehicle passes in front of the free flow antenna (vehicle detection time).

(Probe antenna 112)
The probe antenna 112 functions as an example of a vehicle speed acquisition unit that acquires the vehicle speed. That is, in the embodiment of the present invention, a case is mainly assumed in which the vehicle speed acquisition unit includes the probe antenna 112. As a result, an ETC probe antenna of an already constructed ETC system can be used as the vehicle speed acquisition unit, so there is no need to provide a new vehicle speed acquisition unit. However, instead of the probe antenna 112, another vehicle speed acquisition unit may be used.

More specifically, when the probe antenna 112 acquires the vehicle speed from the vehicle through communication with an on-board device mounted on the vehicle, the probe antenna 112 outputs the acquired vehicle speed to the probe data storage unit 121. The vehicle speed output from the probe antenna 112 to the probe data storage unit 121 can then be used by the QV data creation unit 130.

Figure 7 is a diagram showing an example of probe data stored by the probe data storage unit 121. As shown in Figure 7, the probe data is made up of a correspondence between a "passing time", a "kilometer post", a "vehicle speed", and a "collection time". The "passing time" is the time when the vehicle passes a distance marker (kilometer post) from the start of the road. A "kilometer post" is a distance marker from the start of the road. The "vehicle speed" is the speed (e.g., average speed) of the vehicle passing in front of the distance marker (kilometer post) from the start of the road. The "collection time" is the time when the vehicle speed is collected by the probe antenna 112.

(Traffic volume calculation unit 140)
Returning to Fig. 5, the explanation will be continued. The traffic volume calculation unit 140 acquires the detection result (free flow data) of vehicles traveling in front of the free flow antenna from the free flow data storage unit 122 (Fig. 6). Then, based on the free flow data (vehicle IDs and passing times), the traffic volume calculation unit 140 calculates the number of vehicles that pass in front of the free flow antenna per unit time set in advance (the number of vehicle IDs detected per unit time by the free flow antenna 114) as the traffic volume in front of the free flow antenna. The calculation of the traffic volume by the traffic volume calculation unit 140 may be repeatedly performed for each unit time.

Note that not all vehicles that pass in front of the free flow antenna are equipped with an on-board device capable of communicating with the free flow antenna 114. In other words, not all vehicles that pass in front of the free flow antenna are necessarily detected by the free flow antenna 114. Therefore, if the ratio of vehicles equipped with on-board devices capable of communicating with the free flow antenna 114 to the total number of vehicles (including vehicles that do not have on-board devices capable of communicating with the free flow antenna 114) is preset as the "on-board device installation ratio," and the number of vehicles detected by the free flow antenna 114 is the "number of detected vehicles," it is desirable for the traffic volume calculation unit 140 to more accurately estimate the traffic volume in front of the free flow antenna using the following formula (1).

(Estimated traffic volume) = (Number of detected vehicles) ÷ (Proportion of vehicles equipped with on-board devices) ... (1)

However, in cases where vehicle-mounted devices capable of communicating with the free flow antenna 114 are widely available, such estimation may be omitted. The traffic volume estimated by the traffic volume calculation unit 140 is output to the speed calculation unit 150 each time it is estimated. In addition, the traffic volume estimated by the traffic volume calculation unit 140 and the unit time corresponding to that traffic volume are output to the QV data creation unit 130 for a predetermined period of time.

Typically, the specified period may be one day. In that case, the latest traffic volume for one day may be output to the QV data creation unit 130 only once a day. However, the specified period is not limited to one day. For example, the specified period may be one month. Furthermore, past traffic volume (e.g., traffic volume for one day one year ago) may be output to the QV data creation unit 130, not limited to the latest traffic volume.

Figure 8 is a flowchart showing an example of the traffic volume calculation process executed by the traffic volume calculation unit 140. First, the free flow antenna 114 detects vehicles in real time by receiving vehicle identification information (vehicle ID) from the vehicle through communication with an on-board device mounted on the vehicle. The vehicle detection result (free flow data) is output to the free flow data storage unit 122 in real time.

The traffic volume calculation unit 140 acquires the detection results (free flow data) of vehicles traveling in front of the free flow antenna from the free flow data storage unit 122 (Figure 6) (S11). Then, based on the free flow data (vehicle ID and passing time), the traffic volume calculation unit 140 calculates the number of vehicles that pass in front of the free flow antenna per unit time (number of detected vehicles) set in advance as the traffic volume in front of the free flow antenna (S12). The traffic volume calculation unit 140 estimates the traffic volume in front of the free flow antenna based on the number of detected vehicles and the ratio of vehicles equipped with on-board units (S13).

The traffic volume estimated by the traffic volume calculation unit 140 is output to the speed calculation unit 150 each time it is estimated. In addition, the traffic volume estimated by the traffic volume calculation unit 140 and the unit time corresponding to that traffic volume are output to the QV data creation unit 130 for a predetermined period of time.

(QV data creation unit 130)
When the QV data creation unit 130 acquires the traffic volume and unit time for a predetermined period in front of the free flow antenna output from the traffic volume calculation unit 140, it acquires the vehicle speed in front of the free flow antenna corresponding to the predetermined period (i.e., the vehicle speed corresponding to the "kilometer post" at the free flow antenna position) and the passing time from the probe data storage unit 121. The QV data creation unit 130 functions as an example of a correspondence relationship creation unit that creates a correspondence relationship between traffic volume and vehicle speed (in the following example, an approximation formula) by machine learning based on the traffic volume and unit time for a predetermined period in front of the free flow antenna, and the vehicle speed and passing time in front of the free flow antenna corresponding to the predetermined period.

More specifically, the QV data creation unit 130 associates the traffic volume corresponding to a unit time with the vehicle speed corresponding to the passing time that falls within that unit time. The QV data creation unit 130 creates a correspondence diagram (QV diagram) between traffic volume (Q) and vehicle speed (V) by associating the traffic volume with the vehicle speed for all unit times output from the traffic volume calculation unit 140. The QV data creation unit 130 saves the created QV diagram in the QV data storage unit 123.

Figure 9 is a diagram showing an example of a QV diagram shown by QV data. In the QV diagram shown in Figure 9, the horizontal axis indicates traffic volume (Q) and the vertical axis indicates vehicle speed (V). In such a QV diagram, each result in which traffic volume and vehicle speed are associated is plotted as each point. Here, in the QV diagram, when the traffic volume is less than a predetermined traffic volume, even if the traffic volume is set to one, two types of traffic flow states corresponding to that traffic volume can be assumed: a jammed flow (vehicle speed is lower than the boundary speed) and a free flow (vehicle speed is higher than the boundary speed).

The QV data creation unit 130 therefore creates an approximation equation corresponding to a congested flow and an approximation equation corresponding to a free flow as examples of the correspondence between traffic volume and vehicle speed based on the QV data. The QV data creation unit 130 stores the approximation equation corresponding to a congested flow and the approximation equation corresponding to a free flow in the QV data storage unit 123. For example, the creation of the approximation equation corresponding to a congested flow and the approximation equation corresponding to a free flow can be learned by machine learning.

The approximation equation corresponding to the congestion flow may be any equation, but as an example, it may be expressed by a quadratic function as shown in the following equation (2). In equation (2), V is the vehicle speed and Q is the traffic volume. In more detail, the QV data creation unit 130 obtains the approximation equation corresponding to the congestion flow by determining parameters a and b so that the function expressed in equation (2) approximates each point plotted on the QV diagram (however, all or some of the points in the area where the vehicle speed V is lower than the boundary speed (below the boundary line)).

V=aQ ² +b...(2)

On the other hand, the approximation equation corresponding to the free flow may be any equation, but as an example, it may be expressed by a linear function as shown in equation (3) below. In more detail, the QV data creation unit 130 obtains the approximation equation corresponding to the free flow by determining parameters c and d so that the function expressed in equation (3) approximates each point plotted on the QV diagram (however, all or some of the points in the area where the vehicle speed V is higher than the boundary speed (above the boundary line)).

V=cQ+d...(3)

The QV data creation unit 130 stores the QV data in the QV data storage unit 123, and also stores the parameters a and b of the approximation equation corresponding to the congested flow and the parameters c and d of the approximation equation corresponding to the free flow in the parameter storage unit 124.

The shape of the QV diagram may differ depending on the route, road shape, the number of lanes on which the vehicle travels, etc. Therefore, as described above, it is desirable to create separate QV data for each location (it is desirable to create separate QV data corresponding to each free flow antenna) rather than creating QV data common to all locations. Also, as shown in Figure 9, the vehicle speed when traffic volume is at its highest (i.e., the vehicle speed corresponding to "maximum traffic volume") is generally uniquely determined to be the boundary speed between free flow and congested flow.

Furthermore, it is not necessary to use all of the traffic volume for a specified period in front of the free flow antenna and the vehicle speed in front of the free flow antenna corresponding to that specified period to create a correspondence relationship between traffic volume and vehicle speed. For example, if there is no congestion throughout a certain specified period, it is expected that a useful correspondence relationship between traffic volume and vehicle speed will not be obtained.

Therefore, if the "vehicle speed" for a predetermined period does not include a vehicle speed below a predetermined speed (e.g., a vehicle speed below 40 km/h), or a vehicle speed below the predetermined speed that continues for more than a predetermined duration (e.g., several minutes), the QV data creation unit 130 may exclude the traffic volume and vehicle speed for the predetermined period from the data for creating the correspondence between traffic volume (Q) and vehicle speed (V). This may allow a more useful correspondence to be created.

As described above, the free flow antenna 114 is capable of detecting vehicles traveling in each of the multiple lanes at the same time. Therefore, when a specific event occurs in which vehicles can only travel in some of the multiple lanes (e.g., road construction or an accident), it is expected that the number of vehicles detected by the free flow antenna 114 will decrease. Therefore, it is desirable for the QV data creation unit 130 to create a correspondence relationship between traffic volume and vehicle speed for each number of lanes in which vehicles can travel. For example, when a specific event occurs in which vehicles can only travel in some of the multiple lanes, it is desirable for the QV data creation unit 130 to create a correspondence relationship for the occurrence of the event.

Figure 10 is a flowchart showing an example of the QV data creation process executed by the QV data creation unit 130. First, the QV data creation unit 130 acquires the traffic volume and unit time for a predetermined period in front of the free flow antenna output from the traffic volume calculation unit 140, and then acquires the vehicle speed and passing time in front of the free flow antenna corresponding to the predetermined period from the probe data storage unit 121 (S21).

The QV data creation unit 130 associates the traffic volume corresponding to a unit time with the vehicle speed corresponding to the passing time that falls within that unit time. The QV data creation unit 130 creates a correspondence relationship (QV data) between the traffic volume (Q) and the vehicle speed (V) by associating the traffic volume with the vehicle speed for all unit times output from the traffic volume calculation unit 140 (S22). Then, based on the created QV data, the QV data creation unit 130 creates an approximation equation corresponding to a congested flow and an approximation equation corresponding to a free flow by machine learning (S23).

The QV data creation unit 130 stores the QV data in the QV data storage unit 123, and also stores the parameters a and b of the approximation equation corresponding to the congested flow and the parameters c and d of the approximation equation corresponding to the free flow in the QV data storage unit 123 (S24).

<Learning stage (generation of judgment criteria parameters)>
Next, the traffic condition detection device 1 generates a determination criterion parameter that is a criterion for determining whether the traffic flow state is a congested flow or a free flow by executing a learning stage. The generation of the determination criterion parameter can be executed at a preset date and time.

Figure 11 is a flowchart showing an example of the process of generating the judgment criteria parameters executed by the speed calculation unit 150. First, the speed calculation unit 150 acquires QV data for a set period from the QV data storage unit 123 (S31). The set period may be one month, for example, but is not limited to this. Furthermore, the speed calculation unit 150 acquires free flow data for the set period from the free flow data storage unit 122 (S32).

Next, the statistical information analysis unit 160 generates judgment criteria parameters through a learning process based on the QV data and free flow data acquired by the speed calculation unit 150.

As an example, the statistical information analysis unit 160 includes a traffic volume time series change determination unit 151, a vehicle detection interval determination unit 152, a traffic volume distribution determination unit 153, and a probe delay time determination unit 154. The traffic volume time series change determination unit 151 generates a determination criterion parameter according to a determination based on a traffic volume time series change (S33), the vehicle detection interval determination unit 152 generates a determination criterion parameter according to a determination based on a vehicle detection interval (S34), the traffic volume distribution determination unit 153 generates a determination criterion parameter according to a determination based on a traffic volume distribution (S35), and the probe delay time determination unit 154 generates a determination criterion parameter according to a determination based on a probe delay time (S36).

However, each of the criterion parameters may not necessarily be generated by all of these four blocks. For example, each of the criterion parameters may be generated by some of these four blocks (e.g., one, two, or three blocks). Below, the generation of the criterion parameters by each of these four blocks will be described in detail with reference to Figures 12 to 15.

(Determination based on time-series changes in traffic volume S33)
The traffic volume time series change determination unit 151 extracts the traffic volume at each start time and end time of a preset time interval (e.g., 10 minutes) based on the free flow data for a predetermined period. Furthermore, the traffic volume time series change determination unit 151 determines whether the traffic flow state at each start time and end time is a jam flow (vehicle speed is lower than the boundary speed) or a free flow (vehicle speed is higher than the boundary speed) based on the QV data for the predetermined period.

Furthermore, the traffic volume time series change determination unit 151 calculates the difference between the traffic volume at the start time and the traffic volume at the end time as the traffic volume difference. The traffic volume time series change determination unit 151 determines whether the traffic flow state remains free flow, whether the traffic flow state remains congested flow, whether the traffic flow state switches from free flow to congested flow, or whether the traffic flow state switches from congested flow to free flow, from the start time to the end time.

Figure 12 shows the relationship between traffic volume and traffic volume difference at the end of a time interval for each change in traffic flow state. In the example shown in Figure 12, the horizontal axis shows the traffic volume at the end of a time interval, and the vertical axis shows the traffic volume difference. "○: open circle" indicates the relationship when the traffic flow state remains free flow. "●: filled circle" indicates the relationship when the traffic flow state remains congested flow. "▲: filled triangle" indicates the relationship when the traffic flow state switches from congested flow to free flow. "■: filled square" indicates the relationship when the traffic flow state switches from free flow to congested flow.

The traffic volume time series change determination unit 151 calculates a determination criterion parameter by machine learning based on the correspondence between the traffic volume (hereinafter also simply referred to as "traffic volume") at the end time of the time interval and the traffic volume difference for each of the following cases: when the traffic flow state remains free flow, when the traffic flow state remains congested flow, when the traffic flow state switches from congested flow to free flow, and when the traffic flow state switches from free flow to congested flow.

For example, the traffic volume time series change determination unit 151 may approximate the correspondence between the traffic volume and the traffic volume difference for each of these cases with a predetermined statistical distribution (e.g., a two-dimensional normal distribution, etc.). As a result, the traffic volume time series change determination unit 151 may calculate the parameters for each of these cases (e.g., the mean value μ, the standard deviation σ, etc.) as the determination criterion parameters.

Alternatively, the traffic volume time series change determination unit 151 may use labels indicating each of these cases as teacher labels, and the correspondence between the traffic volume and the traffic volume difference as learning data, and cause some kind of machine learning algorithm to execute a learning process based on a combination of the teacher labels and the learning data. For example, the machine learning algorithm may be a Support Vector Machine (SVM) or other machine learning algorithm. In this way, the traffic volume time series change determination unit 151 may calculate the parameters of the trained model as the determination criterion parameters.

The traffic volume time series change determination unit 151 stores the determination criterion parameters calculated in this manner in the parameter storage unit 124 as determination criterion parameters for determination based on traffic volume time series changes.

(Determination based on vehicle detection interval S34)
The vehicle detection interval determination unit 152 calculates the interval at which a vehicle is detected (vehicle detection interval) based on the free flow data for a predetermined period. The relationship between the vehicle detection interval and the probability density when the traffic state is free flow is as shown in Fig. 1. The relationship between the vehicle detection interval and the probability density when the traffic state is congested flow is as shown in Fig. 2.

The vehicle detection interval determination unit 152 calculates the frequency distribution of the vehicle detection interval for each predetermined time when the traffic state is free flow. The vehicle detection interval determination unit 152 approximates the frequency distribution of the vehicle detection interval for each predetermined time when the traffic state is free flow with a one-variable statistical distribution (e.g., exponential distribution) and calculates the parameter λ and the approximation error.

In the same manner, the vehicle detection interval determination unit 152 calculates the frequency distribution of the vehicle detection interval for each predetermined time when the traffic state is a congested flow. The vehicle detection interval determination unit 152 approximates the frequency distribution of the vehicle detection interval for each predetermined time when the traffic state is a congested flow with a one-variable statistical distribution (e.g., exponential distribution) to calculate the parameter λ and the approximation error.

The frequency distribution of the parameter λ calculated when the traffic state is free flow and when the traffic state is congested flow is shown in Figure 3. Furthermore, the correspondence between the parameter λ calculated when the traffic state is free flow and when the traffic state is congested flow and the approximation error is shown in Figure 4.

The vehicle detection interval determination unit 152 calculates the determination criterion parameter by machine learning based on the correspondence between the calculated parameter λ and the approximation error for each of the cases where the traffic flow state is free flow and where the traffic flow state is congested flow.

For example, the vehicle detection interval determination unit 152 may approximate the correspondence between the parameter λ and the approximation error for each of these cases using a predetermined statistical distribution (e.g., a two-dimensional normal distribution). As a result, the vehicle detection interval determination unit 152 may calculate the parameters (e.g., the mean value μ, the standard deviation σ, etc.) for each of these cases as the determination criterion parameters.

Alternatively, the vehicle detection interval determination unit 152 may use labels indicating each of these cases as teacher labels, and the correspondence between the parameter λ and the approximation error as learning data, and cause some kind of machine learning algorithm to execute a learning process based on a combination of the teacher labels and the learning data. For example, the machine learning algorithm may be SVM or another machine learning algorithm. In this way, the vehicle detection interval determination unit 152 may calculate the parameters of the trained model as the determination criterion parameters.

The vehicle detection interval determination unit 152 stores the determination criterion parameters calculated in this manner in the parameter storage unit 124 as the determination criterion parameters for determination based on the vehicle detection interval.

(Determination based on traffic volume distribution S35)
The traffic volume distribution determination unit 153 calculates a frequency distribution of traffic volume for each predetermined time period based on the free flow data for a predetermined period. Then, the traffic volume distribution determination unit 153 calculates a parameter λ by approximating the calculated frequency distribution with a predetermined statistical distribution (e.g., Poisson distribution, etc.). The traffic volume distribution determination unit 153 also calculates data related to the variance of the frequency distribution.

Here, we mainly assume that the ratio of the standard deviation of the frequency distribution to the standard deviation of the statistical distribution (standard deviation ratio) is used as an example of data related to the variance of the frequency distribution. However, the data related to the variance of the frequency distribution may be the standard deviation of the frequency distribution, the variance of the frequency distribution, the ratio of the variance of the frequency distribution to the variance of the statistical distribution, or other data indicating the variance of the frequency distribution.

Figure 13 is a diagram showing the correspondence between the parameter λ related to the statistical distribution and the standard deviation ratio. In the example shown in Figure 13, the horizontal axis indicates the parameter λ related to the statistical distribution (here, Poisson distribution), and the vertical axis indicates the standard deviation ratio. "○: open circle" indicates the correspondence when the traffic flow state is free flow. "●: filled circle" indicates the correspondence when the traffic flow state is congested flow. "▲: filled triangle" indicates the correspondence when the traffic flow state is congested flow. It can be seen that the case where the traffic flow state is free flow and the case where the traffic flow state is congested flow are roughly separated in the example shown in Figure 13.

The traffic volume distribution determination unit 153 calculates a criterion parameter by machine learning based on the parameter λ related to the statistical distribution for each of the cases where the traffic flow state is free flow and the traffic flow state is congested flow. More specifically, the traffic volume distribution determination unit 153 calculates a criterion parameter by machine learning based on the correspondence between the parameter λ related to the statistical distribution and the standard deviation ratio for each of the cases where the traffic flow state is free flow and the traffic flow state is congested flow.

For example, the traffic volume distribution determination unit 153 may approximate the correspondence between the parameter λ related to the statistical distribution and the standard deviation ratio with a predetermined statistical distribution (e.g., a two-dimensional normal distribution, etc.) for each of the cases where the traffic flow state is free flow and the traffic flow state is congested flow. In this way, the traffic volume distribution determination unit 153 may calculate the parameters (e.g., the mean value μ, the standard deviation σ, etc.) for each of these cases as the determination criterion parameters.

Alternatively, the traffic volume distribution determination unit 153 may use labels indicating each of these cases as teacher labels, and may use the correspondence between the parameter λ related to the statistical distribution and the standard deviation ratio as learning data, and may cause some kind of machine learning algorithm to execute a learning process based on a combination of the teacher labels and the learning data. For example, the machine learning algorithm may be SVM or another machine learning algorithm. In this way, the traffic volume distribution determination unit 153 may calculate the parameters of the trained model as the determination criterion parameters.

The traffic volume distribution determination unit 153 stores the determination criterion parameters calculated in this manner in the parameter storage unit 124 as determination criterion parameters for determination based on traffic volume distribution.

(Determination based on probe delay time S36)
The probe delay time determination unit 154 acquires probe data for a predetermined period from the probe data storage unit 121. As described above, a time (probe delay time) may occur between when the vehicle speed is recorded in the on-board device and when the vehicle speed is acquired by the probe antenna 112. Therefore, depending on the probe antenna 112, only the vehicle speed that passed in front of the free flow antenna several minutes to several tens of minutes ago may be acquired.

This delay may vary depending on the traffic conditions from the free-flow antenna to the next probe antenna (i.e., the downstream probe antenna). For example, when the traffic flow state is free flow, the probe delay time may be several minutes, and when the traffic flow state is congested flow, the probe delay time may be several tens of minutes. Based on the acquired probe data, the probe delay time determination unit 154 calculates the difference between the passing time and the collection time as the probe delay time.

Figure 14 is a diagram showing an example of the probe delay time (seconds) at each time in front of the free flow antenna. Referring to Figure 14, it can be seen that the probe delay time remains at approximately 150 to 200 seconds until around 14:00. After that, the probe delay time increases, and when the traffic flow state is congested, it can be seen that the probe delay time exceeds approximately 500 seconds. In addition, it can be seen that after 20:00, the traffic flow state switches from congested to free flow, and the probe delay time decreases, returning to approximately 150 to 200 seconds again.

Then, the probe delay time determination unit 154 calculates a frequency distribution of the probe delay time for each predetermined time period based on the calculated probe delay time.

Figure 15 shows the frequency distribution of probe delay times for each given time. Referring to Figure 15, the frequency distribution of probe delay times for each given time when the traffic flow state is free flow is shown by a histogram (light hatching). On the other hand, the frequency distribution of probe delay times for each given time when the traffic flow state is congested flow is shown by a histogram (dark hatching).

The probe delay time determination unit 154 calculates the determination criterion parameters by machine learning based on the calculated frequency distribution for each of the cases where the traffic flow state is free flow and where the traffic flow state is congested flow. More specifically, the probe delay time determination unit 154 calculates parameters (e.g., the mean value μ and standard deviation σ, or the shape parameter k and scale parameter θ, etc.) by approximating the frequency distribution with a predetermined statistical distribution (e.g., normal distribution or gamma distribution, etc.) for each of these cases.

The probe delay time determination unit 154 stores the determination criterion parameters calculated in this manner in the parameter storage unit 124 as determination criterion parameters for determination based on the probe delay time.

<Inference stage (speed calculation)>
Next, the traffic condition detection device 1 generates target data by executing an inference step, and calculates the vehicle speed based on the generated target data and the previously generated criterion parameters. The calculation of the vehicle speed can be performed in real time.

Figure 16 is a flowchart showing an example of speed calculation processing executed by the speed calculation unit 150. First, the traffic volume calculation unit 140 acquires free flow data for the most recent unit time from the free flow data storage unit 122, calculates traffic volume based on the free flow data for the most recent unit time, and outputs the calculated traffic volume to the speed calculation unit 150 (S41). Furthermore, the speed calculation unit 150 acquires traffic volume from the traffic volume calculation unit 140, and acquires judgment criterion parameters from the parameter storage unit 124 (S42).

Then, the statistical information analysis unit 160 calculates the probability that the traffic flow state is free flow and congested flow by inference processing based on the traffic volume and judgment criteria parameters acquired by the speed calculation unit 150.

As an example, the traffic volume time series change determination unit 151 calculates the probability for each case according to the determination based on the traffic volume time series change (S43), the vehicle detection interval determination unit 152 calculates the probability for each case according to the determination based on the vehicle detection interval (S44), the traffic volume distribution determination unit 153 calculates the probability for each case according to the determination based on the traffic volume distribution (S45), and the probe delay time determination unit 154 calculates the probability for each case according to the determination based on the probe delay time (S46).

However, each probability does not have to be calculated using all four blocks. For example, each probability may be calculated using some of the four blocks (e.g., one, two, or three blocks). Below, the calculation of the probability using each of the four blocks is described in detail.

(Determination based on time-series changes in traffic volume S43)
The traffic volume time series change determination unit 151 extracts the traffic volume at each start time and end time of a preset time interval (e.g., 10 minutes) based on the free flow data for the most recent predetermined period. Furthermore, the traffic volume time series change determination unit 151 calculates the difference between the traffic volume at the start time and the traffic volume at the end time as the traffic volume difference.

The traffic volume time series change determination unit 151 calculates the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow based on the calculated traffic volume and traffic volume difference and the determination criterion parameters.

More specifically, the traffic volume time series change determination unit 151 calculates the probability that the traffic flow state will remain in congested flow based on the calculated traffic volume and traffic volume difference, and the determination criterion parameters for when the traffic flow state remains in congested flow. Furthermore, the traffic volume time series change determination unit 151 calculates the probability that the traffic flow state will switch from free flow to congested flow based on the calculated traffic volume and traffic volume difference, and the determination criterion parameters for when the traffic flow state switches from free flow to congested flow.

The traffic volume time series change determination unit 151 can calculate the probability that the traffic flow state is congested flow by adding the probability that the traffic flow state remains congested flow and the probability that the traffic flow state switches from free flow to congested flow. For example, if the probability that the traffic flow state remains congested flow is 20% and the probability that the traffic flow state switches from free flow to congested flow is 50%, the probability that the traffic flow state is congested flow can be calculated to be 70%.

The traffic volume time series change determination unit 151 also calculates the probability that the traffic flow state will remain free flow based on the calculated traffic volume and traffic volume difference, and the determination criterion parameters for when the traffic flow state remains free flow. The traffic volume time series change determination unit 151 also calculates the probability that the traffic flow state will switch from congested flow to free flow based on the calculated traffic volume and traffic volume difference, and the determination criterion parameters for when the traffic flow state switches from congested flow to free flow.

The traffic volume time series change determination unit 151 can calculate the probability that the traffic flow state is free flow by adding the probability that the traffic flow state remains free flow and the probability that the traffic flow state switches from congested flow to free flow. For example, if the probability that the traffic flow state remains free flow is 30% and the probability that the traffic flow state switches from congested flow to free flow is 0%, the probability that the traffic flow state is free flow can be calculated to be 30%.

For example, when the criterion parameters for determining whether the traffic flow state remains congested are parameters related to a statistical distribution (e.g., a two-dimensional normal distribution), the traffic volume time series change determination unit 151 may calculate the probability that the traffic flow state remains congested by the integral value of a specified section of the probability density function of the statistical distribution determined by the criterion parameters. As an example, when the calculated traffic volume is x0 and the traffic volume difference is y0, the specified section may be a section whose distance from x=0 in the standard normal distribution determined by the criterion parameters is farther than the two-dimensional coordinates (x0, y0).

The probability that the traffic flow state changes from free flow to congested flow, the probability that the traffic flow state remains free flow, and the probability that the traffic flow state changes from congested flow to free flow may also be calculated in a similar manner.

Alternatively, the traffic volume time series change determination unit 151 may input the calculated traffic volume and traffic volume difference to a trained model, and based on the traffic volume and traffic volume difference and the parameters (determination criterion parameters) of the trained model, output from the trained model the probability that the traffic flow state will remain in congested flow, the probability that the traffic flow state will switch from free flow to congested flow, the probability that the traffic flow state will remain in free flow, and the probability that the traffic flow state will switch from congested flow to free flow.

(Determination based on vehicle detection interval S44)
The vehicle detection interval determination unit 152 calculates the interval at which a vehicle is detected (vehicle detection interval) based on the free flow data for the most recent predetermined period. The vehicle detection interval determination unit 152 then calculates the frequency distribution of the vehicle detection interval for each predetermined time. The vehicle detection interval determination unit 152 approximates the frequency distribution of the vehicle detection interval for each predetermined time with a one-variable statistical distribution (e.g., exponential distribution, etc.) to calculate the parameter λ and the approximation error.

The vehicle detection interval determination unit 152 calculates the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow based on the parameter λ, the approximation error, and the determination criterion parameters.

More specifically, the vehicle detection interval determination unit 152 calculates the probability that the traffic flow state is a congested flow based on the calculated parameter λ and approximation error, and the determination criterion parameters when the traffic flow state is a congested flow. For example, the probability that the traffic flow state is a congested flow may be calculated to be 50%.

The vehicle detection interval determination unit 152 also calculates the probability that the traffic flow state is free flow based on the calculated parameter λ and approximation error, and the determination criterion parameters when the traffic flow state is free flow. For example, the probability that the traffic flow state is free flow can be calculated to be 50%.

For example, similar to the judgment based on traffic volume time series change S43, when the judgment criterion parameter for when the traffic flow state is a congested flow is a parameter related to a statistical distribution (e.g., a two-dimensional normal distribution), the vehicle detection interval judgment unit 152 may calculate the probability that the traffic flow state is a congested flow by the integral value of a specified section of the probability density function of the statistical distribution determined by the judgment criterion parameter. Note that the probability that the traffic flow state is a free flow may be calculated in a similar manner.

Alternatively, the vehicle detection interval determination unit 152 may input the calculated parameter λ and approximation error to a trained model, and output the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow from the trained model based on the parameter λ and the approximation error and the parameters of the trained model (criterion parameters).

(Determination based on traffic volume distribution S45)
The traffic volume distribution determination unit 153 calculates a frequency distribution of traffic volume for each predetermined time period based on the free flow data for the most recent predetermined period. Then, the traffic volume distribution determination unit 153 calculates a parameter λ by approximating the calculated frequency distribution with a predetermined statistical distribution (e.g., Poisson distribution, etc.). The traffic volume distribution determination unit 153 also calculates data related to the variance of the frequency distribution.

Here, we mainly assume that the ratio of the standard deviation of the frequency distribution to the standard deviation of the statistical distribution (standard deviation ratio) is used as an example of data related to the variance of the frequency distribution. However, as described above, the data related to the variance of the frequency distribution may be the standard deviation of the frequency distribution, the variance of the frequency distribution, the ratio of the variance of the frequency distribution to the variance of the statistical distribution, or other data indicating the variance of the frequency distribution.

The traffic volume distribution determination unit 153 calculates the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow based on the calculated parameter λ and standard deviation ratio and the determination criterion parameters.

More specifically, the traffic volume distribution determination unit 153 calculates the probability that the traffic flow state is a congested flow based on the calculated parameter λ and standard deviation ratio, and the determination criterion parameters when the traffic flow state is a congested flow. For example, the probability that the traffic flow state is a congested flow can be calculated to be 60%.

The traffic volume distribution determination unit 153 also calculates the probability that the traffic flow state is free flow based on the calculated parameter λ and standard deviation ratio, and the determination criterion parameters when the traffic flow state is free flow. For example, the probability that the traffic flow state is free flow can be calculated to be 40%.

For example, similar to the judgment based on traffic volume time series change S43, when the judgment criterion parameter for when the traffic flow state is a congested flow is a parameter related to a statistical distribution (e.g., a two-dimensional normal distribution), the traffic volume distribution judgment unit 153 may calculate the probability that the traffic flow state is a congested flow by the integral value of a specified section of the probability density function of the statistical distribution determined by the judgment criterion parameter. Note that the probability that the traffic flow state is a free flow may be calculated in a similar manner.

Alternatively, the vehicle detection interval determination unit 152 may input the calculated parameter λ and standard deviation ratio to a trained model, and output the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow from the trained model based on the parameter λ and the standard deviation ratio and the parameters of the trained model (criterion parameters).

(Determination based on probe delay time S46)
The probe delay time determination unit 154 acquires probe data for a predetermined period from the probe data storage unit 121. Based on the acquired probe data, the probe delay time determination unit 154 calculates the difference between the passing time and the collection time as the probe delay time. Then, based on the calculated probe delay time, the probe delay time determination unit 154 calculates a frequency distribution of the probe delay time for each predetermined time period set in advance.

The probe delay time determination unit 154 calculates the probability that the traffic flow state is a congested flow and the probability that the traffic flow state is a free flow based on the frequency distribution of the probe delay time for each specified time period and the determination criterion parameters.

More specifically, the probe delay time determination unit 154 calculates the probability that the traffic flow state is a congested flow based on the frequency distribution of the calculated probe delay time for each predetermined time and the determination criterion parameters when the traffic flow state is a congested flow. For example, the probability that the traffic flow state is a congested flow can be calculated to be 80%.

The probe delay time determination unit 154 also calculates the probability that the traffic flow state is free flow based on the frequency distribution of the calculated probe delay time for each predetermined time and the determination criterion parameters when the traffic flow state is free flow. For example, the probability that the traffic flow state is free flow can be calculated to be 20%.

For example, similar to the judgment based on traffic volume time series change S43, when the judgment criterion parameter for when the traffic flow state is a congested flow is a parameter related to a statistical distribution (e.g., a two-dimensional normal distribution), the probe delay time judgment unit 154 may calculate the probability that the traffic flow state is a congested flow by the integral value of a specified section of the probability density function of the statistical distribution determined by the judgment criterion parameter. Note that the probability that the traffic flow state is a free flow may be calculated in a similar manner.

Alternatively, the probe delay time determination unit 154 may input the frequency distribution of the calculated probe delay time for each specified time into the trained model, and output the probability that the traffic flow state is congested flow and the probability that the traffic flow state is free flow from the trained model based on the frequency distribution of the probe delay time for each specified time and the parameters (criterion parameters) of the trained model.

(Determination of traffic flow state S47)
The traffic flow determination unit 155 comprehensively determines whether the traffic flow state is free flow or congested flow based on the results of the determination based on the traffic volume time series change S43, the determination based on the vehicle detection interval S44, the determination based on the traffic volume distribution S45, and the determination based on the probe delay time S46 (S47). For example, the traffic flow determination unit 155 may calculate the overall probability that the traffic flow state is free flow and the overall probability that the traffic flow state is congested flow, and may take the traffic flow state with the larger overall probability as the overall determination result.

The overall probability may be calculated in any manner. For example, the overall probability that the traffic flow state is congested flow may be calculated by averaging (e.g., arithmetic mean, weighted average, or geometric mean) the probability of simultaneous occurrence of the probabilities that the traffic flow state is congested flow in each determination result. Similarly, the overall probability that the traffic flow state is free flow may be calculated by averaging (e.g., arithmetic mean, weighted average, or geometric mean) the probability of simultaneous occurrence of the probabilities that the traffic flow state is free flow in each determination result.

For example, the overall probability that the traffic flow state is congested flow can be calculated as 64% by taking the geometric mean of the probabilities that the traffic flow state is congested flow in each judgment result. Also, the overall probability that the traffic flow state is free flow can be calculated as 33% by taking the geometric mean of the probabilities that the traffic flow state is free flow in each judgment result.

When there is only one probability that the traffic flow state is congested flow and one probability that the traffic flow state is free flow, there is no need to calculate the overall probability. That is, the traffic flow determination unit 155 may obtain a traffic flow state determination result based on the probability that the traffic flow state is congested flow and the probability that the traffic flow state is free flow. More specifically, the traffic flow determination unit 155 may determine as the determination result the traffic flow state with the greater probability between the probability that the traffic flow state is congested flow and the probability that the traffic flow state is free flow.

(Determination of Vehicle Speed S48)
The speed determination unit 156 determines the vehicle speed in front of the free flow antenna based on the traffic flow state determination result obtained by the traffic flow determination unit 155, the traffic volume in front of the free flow antenna calculated by the traffic volume calculation unit 140, and the QV parameters prepared in advance (S48). More specifically, the QV parameters may correspond to the parameters a and b of the approximation formula corresponding to the above-mentioned jammed flow, or the parameters c and d of the approximation formula corresponding to the free flow.

For example, when the speed determination unit 156 determines that the traffic flow state is a congested flow, it may substitute the traffic volume in front of the free-flow antenna calculated by the traffic volume calculation unit 140 into the above formula (2) to determine the vehicle speed in front of the free-flow antenna. On the other hand, when the speed determination unit 156 determines that the traffic flow state is a free flow, it may substitute the traffic volume in front of the free-flow antenna calculated by the traffic volume calculation unit 140 into the above formula (3) to determine the vehicle speed in front of the free-flow antenna.

(Detection of traffic conditions)
The traffic condition detection unit 170 detects the traffic condition at the free flow antenna installation point based on the traffic volume calculated by the traffic volume calculation unit 140 and the vehicle speed determined by the speed determination unit 156. In addition, the traffic condition detection unit 170 estimates the delayed portion of the probe data based on the traffic volume and vehicle speed at the free flow antenna installation point by the method disclosed in Patent Document 3, and is able to estimate the traffic condition not only at the free flow antenna installation point and the probe antenna installation point but also at all points on the road.

The above describes an example of the configuration of the traffic condition detection device 1 according to an embodiment of the present invention.

(1-2. Effects)
As described above, in the learning stage, the technology according to the embodiment of the present invention generates a judgment criterion parameter by machine learning based on the average and the variation of the frequency distribution of traffic volume for each predetermined time period obtained from the accumulated free flow data. Then, in the inference stage, the technology according to the embodiment of the present invention judges whether the traffic state is free flow or congested flow based on the generated judgment criterion parameter and a judgment parameter based on the acquired free flow data.

As another example, in the learning stage, the technology according to the embodiment of the present invention generates a judgment criterion parameter by machine learning based on the delay time of the accumulated probe data. Then, in the inference stage, the technology according to the embodiment of the present invention determines whether the traffic state is free flow or congested flow based on the generated judgment criterion parameter and the delay time based on the acquired probe data.

The technology according to the embodiment of the present invention can estimate with higher accuracy whether the traffic state is free flow or congested flow. As a result, the technology according to the embodiment of the present invention can estimate the vehicle speed with higher accuracy.

In addition, the technology according to the embodiment of the present invention makes it possible to estimate traffic conditions at many more points on the road, making it difficult to obtain vehicle speeds with high precision.

In addition, the technology according to the embodiment of the present invention can be applied to specific external environments (e.g., rainy weather, nighttime, afternoon sun, etc.), so the technology according to the embodiment of the present invention can estimate the vehicle speed with higher accuracy.

Furthermore, the technology according to the embodiment of the present invention detects traffic volume and vehicle speed using a free flow antenna 114 instead of a traffic counter. Therefore, the technology according to the embodiment of the present invention does not require the expense of installing and maintaining a traffic counter, and therefore the necessary costs can be reduced.

Furthermore, according to an embodiment of the present invention, the free flow antenna 114 can detect vehicles in real time. That is, according to an embodiment of the present invention, the vehicle speed is calculated based on the vehicle detection results obtained in real time by the free flow antenna 114, so that the vehicle speed can be calculated in real time. Note that the vehicle speed can also be obtained by the probe antenna 112, but the vehicle speed cannot be obtained in real time by the probe antenna 112.

Furthermore, since the free flow antenna 114 may be compatible with multiple versions of ETC on-board units, according to an embodiment of the present invention, more vehicle speeds can be calculated based on the vehicle detection results by the free flow antenna 114. On the other hand, since the probe antenna 112 may be compatible with only a specific version of the ETC on-board unit, less vehicle speeds may be obtained depending on the probe antenna 112.

The above describes the effects of the technology according to the embodiment of the present invention.

(2. Hardware Configuration Example)
Next, an example of the hardware configuration of the traffic condition detection device 1 according to the embodiment of the present invention will be described.

Below, an example of the hardware configuration of the information processing device 900 will be described as an example of the hardware configuration of the traffic condition detection device 1 according to an embodiment of the present invention. Note that the example of the hardware configuration of the information processing device 900 described below is merely one example of the hardware configuration of the traffic condition detection device 1. Therefore, the hardware configuration of the traffic condition detection device 1 may be such that unnecessary components are deleted from the hardware configuration of the information processing device 900 described below, or new components may be added.

Figure 17 is a diagram showing the hardware configuration of an information processing device 900 as an example of a traffic condition detection device 1 according to an embodiment of the present invention. The information processing device 900 includes a CPU (Central Processing Unit) 901, a ROM (Read Only Memory) 902, a RAM (Random Access Memory) 903, a host bus 904, a bridge 905, an external bus 906, an interface 907, an input device 908, an output device 909, a storage device 910, and a communication device 911.

The CPU 901 functions as an arithmetic processing device and control device, and controls the overall operation of the information processing device 900 in accordance with various programs. The CPU 901 may also be a microprocessor. The ROM 902 stores programs and arithmetic parameters used by the CPU 901. The RAM 903 temporarily stores programs used in the execution of the CPU 901 and parameters that change appropriately during the execution. These are interconnected by a host bus 904 consisting of a CPU bus, etc.

The host bus 904 is connected to an external bus 906, such as a PCI (Peripheral Component Interconnect/Interface) bus, via a bridge 905. Note that the host bus 904, bridge 905, and external bus 906 do not necessarily need to be configured separately, and these functions may be implemented on a single bus.

The input device 908 is composed of input means for the user to input information, such as a mouse, keyboard, touch panel, button, microphone, switch, and lever, and an input control circuit that generates an input signal based on the user's input and outputs it to the CPU 901. A user who operates the information processing device 900 can input various data to the information processing device 900 and instruct processing operations by operating this input device 908.

The output device 909 includes, for example, display devices such as a CRT (Cathode Ray Tube) display device, a Liquid Crystal Display (LCD) device, an OLED (Organic Light Emitting Diode) device, and a lamp, as well as audio output devices such as a speaker.

The storage device 910 is a device for storing data. The storage device 910 may include a storage medium, a recording device for recording data on the storage medium, a reading device for reading data from the storage medium, and a deleting device for deleting data recorded on the storage medium. The storage device 910 is configured, for example, with a HDD (Hard Disk Drive). This storage device 910 drives the hard disk and stores the programs executed by the CPU 901 and various data.

The communication device 911 is, for example, a communication interface configured with a communication device for connecting to a network. The communication device 911 may support either wireless communication or wired communication.

The above describes an example of the hardware configuration of the traffic condition detection device 1 according to an embodiment of the present invention.

(3. Summary)
Although the preferred embodiment of the present invention has been described in detail above with reference to the accompanying drawings, the present invention is not limited to such an example. It is clear that a person having ordinary knowledge in the technical field to which the present invention pertains can conceive of various modified or altered examples within the scope of the technical ideas described in the claims, and it is understood that these also naturally belong to the technical scope of the present invention.

REFERENCE SIGNS LIST 1 Traffic condition detection device 112 Probe antenna 114 Free flow antenna 121 Probe data storage unit 122 Free flow data storage unit 123 QV data storage unit 124 Parameter storage unit 130 QV data creation unit 140 Traffic volume calculation unit 150 Speed calculation unit 151 Traffic volume time series change determination unit 152 Vehicle detection interval determination unit 153 Traffic volume distribution determination unit 154 Probe delay time determination unit 155 Traffic flow determination unit 156 Speed determination unit 160 Statistical information analysis unit 170 Traffic condition detection unit

Claims (16)

a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from the vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result;
a traffic volume distribution determination unit that calculates a first frequency distribution for each predetermined time period of a first traffic volume corresponding to the predetermined location, calculates a first parameter by approximating the first frequency distribution with a first statistical distribution, and calculates a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on data related to the first parameter and the variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning;
a traffic flow determination unit that obtains a determination result of the traffic flow state based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
a speed determination unit that determines a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between the traffic volume and the vehicle speed that is prepared in advance;
A traffic condition detection device comprising: the traffic volume distribution determination unit calculates a second frequency distribution of a second traffic volume corresponding to the predetermined position in a free flow, calculates a second parameter by approximating the second frequency distribution with the first statistical distribution, and calculates a reference parameter corresponding to the free flow based on the second parameter;
2. The traffic condition detection device according to claim 1. the traffic volume distribution determination unit calculates a reference parameter corresponding to the free flow by approximating a frequency distribution of a correspondence relationship between the second parameter and data related to a variance of the second frequency distribution with a second statistical distribution;
3. A traffic condition detection device according to claim 2. the traffic volume distribution determination unit calculates a third frequency distribution of a third traffic volume corresponding to the predetermined position in the congestion flow, calculates a third parameter by approximating the third frequency distribution with the first statistical distribution, and calculates a reference parameter corresponding to the congestion flow based on the third parameter;
A traffic condition detection device according to any one of claims 1 to 3. the traffic volume distribution determination unit calculates a reference parameter corresponding to the congestion flow by approximating a frequency distribution of a correspondence relationship between the third parameter and data related to a variation of the third frequency distribution with a third statistical distribution;
5. A traffic condition detection device according to claim 4. The traffic volume distribution determination unit
calculating a second frequency distribution of a second traffic volume corresponding to the predetermined position in a free flow, and approximating the second frequency distribution with the first statistical distribution to calculate a second parameter;
calculating a third frequency distribution of a third traffic volume corresponding to the predetermined position in the congestion flow, and approximating the third frequency distribution with the first statistical distribution to calculate a third parameter;
calculating the reference parameters based on a combination of learning data including the second parameter and data related to the variance of the second frequency distribution and a teacher label indicating a free flow, and a combination of learning data including the third parameter and data related to the variance of the third frequency distribution and a teacher label indicating a congested flow;
2. The traffic condition detection device according to claim 1. a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from the vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result;
a delay time determination unit that calculates a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in the vehicle-mounted device to when the vehicle speed is acquired by the vehicle speed acquisition unit, and calculates a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter that has been calculated in advance by machine learning;
a traffic flow determination unit that obtains a determination result of the traffic flow state based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
a speed determination unit that determines a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between the traffic volume and the vehicle speed that is prepared in advance;
A traffic condition detection device comprising: the delay time determination unit calculates a second frequency distribution of the second delay time in the free flow for each predetermined time, and calculates a reference parameter corresponding to the free flow by approximating the second frequency distribution with a first statistical distribution;
8. A traffic condition detection device according to claim 7. the delay time determination unit calculates a third frequency distribution of a third delay time in the traffic jam flow for each predetermined time, and calculates a reference parameter corresponding to the traffic jam flow by approximating the third frequency distribution with a first statistical distribution;
A traffic condition detection device according to claim 7 or 8. The vehicle speed acquisition unit includes a probe antenna.
A traffic condition detection device according to any one of claims 7 to 9. The vehicle detection unit detects the vehicle in real time.
A traffic condition detection device according to any one of claims 1 to 10. The vehicle detection unit includes a free flow antenna.
12. A traffic condition detection device according to claim 11. The computer
acquiring a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculating a traffic volume corresponding to the predetermined position based on the vehicle detection result;
calculating a first frequency distribution for each predetermined time period of a first traffic volume corresponding to the predetermined location, approximating the first frequency distribution with a first statistical distribution to calculate a first parameter, and calculating a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first parameter, data related to the variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning;
obtaining a traffic flow state determination result based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
determining a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, a traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed that is prepared in advance;
A traffic state detection method comprising: The computer
acquiring a detection result of a vehicle traveling at a predetermined position on a road from a vehicle detection unit, and calculating a traffic volume corresponding to the predetermined position based on the vehicle detection result;
calculating a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in the vehicle-mounted device to when the vehicle speed is acquired by the vehicle speed acquisition unit, and calculating a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter calculated in advance by machine learning;
obtaining a traffic flow state determination result based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
determining a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, a traffic volume corresponding to the predetermined position, and a correspondence relationship between traffic volume and vehicle speed that is prepared in advance;
A traffic state detection method comprising: Computer,
a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from the vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result;
a traffic volume distribution determination unit that calculates a first frequency distribution for each predetermined time period of a first traffic volume corresponding to the predetermined location, calculates a first parameter by approximating the first frequency distribution with a first statistical distribution, and calculates a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on data related to the first parameter and the variance of the first frequency distribution, and a reference parameter calculated in advance by machine learning;
a traffic flow determination unit that obtains a determination result of the traffic flow state based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
a speed determination unit that determines a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between the traffic volume and the vehicle speed that is prepared in advance;
A program that causes the device to function as a traffic condition detection device. Computer,
a traffic volume calculation unit that acquires a detection result of a vehicle traveling at a predetermined position on a road from the vehicle detection unit, and calculates a traffic volume corresponding to the predetermined position based on the vehicle detection result;
a delay time determination unit that calculates a first frequency distribution for each predetermined time of a first delay time from when the vehicle speed is recorded in the vehicle-mounted device to when the vehicle speed is acquired by the vehicle speed acquisition unit, and calculates a probability that the traffic flow state on the road is a congested flow and a probability that the traffic flow state is a free flow based on the first frequency distribution and a reference parameter that has been calculated in advance by machine learning;
a traffic flow determination unit that obtains a determination result of the traffic flow state based on a probability that the traffic flow state is a congested flow and a probability that the traffic flow state is a free flow;
a speed determination unit that determines a vehicle speed corresponding to the predetermined position based on the traffic flow state determination result, the traffic volume corresponding to the predetermined position, and a correspondence relationship between the traffic volume and the vehicle speed that is prepared in advance;
A program that causes the device to function as a traffic condition detection device.

Images