JP5802823B2 - Traffic prediction system - Google Patents

Description

  The present invention relates to a system for predicting traffic volume on a road, and more particularly to a system for predicting traffic volume using information on a vehicle traveling on a road.

  In Japan, a general traffic survey of road traffic census, which measures the traffic volume of the road manually every five years, is known as an effort to measure the traffic volume. Similarly overseas, traffic surveys are conducted manually. Such traffic surveys require a lot of workers and time, and are therefore very expensive. For this reason, it is difficult to increase the frequency of traffic volume surveys.

  In view of this, Japanese Patent Laid-Open No. 2009-140007 (Patent Document 1) discloses an invention for obtaining the traffic volume of a road section based on the travel history of a vehicle collected using a probe car. In order to accurately obtain the traffic volume of the road section from the travel history of the probe car, it is necessary to create in advance a relationship between the travel speed of the vehicle and the traffic volume for each road. In order to create a relationship between the vehicle travel speed and the traffic volume, the actual traffic volume and the vehicle travel speed must be measured simultaneously.

  However, since it is very expensive to measure the actual traffic volume and the vehicle speed at the same time for all roads, in Patent Document 1, the relationship between the speed of the road section and the traffic volume is expressed as a high speed. It is described that the actual traffic volume and the traveling speed of the vehicle are approximated for each road type by creating for each road type such as a road and a national road. Since the traffic volume and the vehicle speed need only be measured at the same time for each road type, the cost of traffic volume survey can be reduced.

JP 2009-140007 A

  However, it is difficult to define the relationship between speed and traffic volume only by road type. This is because the road environment is different even on roads of the same road type, such as the number of lanes, the number of traffic lights, the presence or absence of a bus-dedicated lane, and the number of parking on the road. For this reason, when the traffic volume is predicted by obtaining the actual traffic volume and the traveling speed of the vehicle for each road type, the accuracy is limited.

  The traffic volume prediction system of the present invention estimates road characteristics such as free speed, critical speed, saturation density, and critical density from probe data, classifies the road characteristics into clusters corresponding to the probe data, and configures each cluster. A travel pattern for each cluster is created based on the speed and acceleration of the probe data. Next, when the probe data is received and the road of the probe data does not belong to any cluster, the travel pattern generated from the probe data is compared with the travel pattern for each cluster, The road characteristics of a high cluster are acquired, and the traffic volume of the road where the probe data is collected is predicted using the acquired road characteristics.

  According to the present invention, the travel pattern of the probe data is obtained, and the traffic volume is predicted from the road characteristics of a similar group among the road characteristics classified in advance. The amount can be predicted.

It is a block diagram which shows the whole structure of the traffic prediction system which concerns on embodiment of this invention. It is a figure which shows the format of the traffic volume data transmitted from the transmitter 200. It is a figure which shows the data format of a probe data storage device. It is a figure which shows the processing flow of a road characteristic estimation apparatus. It is a figure which shows the relationship between average speed and traffic density. It is a figure which shows the conceptual diagram of clustering of a road characteristic. It is a figure which shows the example of the running pattern of a wide road. It is a figure which shows the example of the running pattern of a narrow road. It is a figure which shows the processing flow of a travel pattern production apparatus. It is a figure which shows the table of a running pattern. It is a figure which shows the data format of a memory | storage device. It is a figure which shows the processing flow of a road characteristic search apparatus. It is a block diagram which shows the whole structure of the traffic prediction system which concerns on the deformation | transformation of embodiment of this invention. It is a figure which shows the data format of the speed data received with a receiver. It is a figure which shows the data format of a road information storage device. It is a figure which shows the processing flow of a road characteristic search apparatus.

  Embodiments of a traffic volume prediction system according to the present invention will be described with reference to the drawings.

  FIG. 1 shows an overall configuration of a traffic volume prediction system according to an embodiment of the present invention. The traffic volume prediction system 100 includes a traffic volume storage device 110, a probe data storage device 120, a road characteristic estimation device 130, a classification device 140, a travel pattern creation device 150, a storage device 160, a reception device 170, a road characteristic search device 180, a traffic An amount prediction device 190 and a transmission device 200 are provided.

  The traffic volume storage device 110 is a storage device such as a hard disk or a flash memory, and stores traffic volume information. The traffic volume information is used for estimation of road characteristics by the road characteristic estimation device 130. The probe data storage device 120 is a storage device such as a hard disk or a flash memory, and stores probe data. The probe data includes a vehicle identification ID (hereinafter referred to as vehicle ID), a road identification ID (hereinafter referred to as road ID), position information, time information, speed information, and the like. The vehicle uplinks the vehicle ID, position information, time information, and speed information to a probe center (not shown) via a communication device such as a mobile phone or a radio. Such a vehicle is called a probe car.

  The probe center specifies the road on which each vehicle has traveled using the position information, time information, and map information collected from each vehicle. The probe data stored in the probe data storage device 120 and the probe data received by the receiving device 170 are data processed by this probe center. The probe data storage device 120 is used for estimation of road characteristics by the road characteristic estimation device 130 and travel pattern creation by the travel pattern creation device 150.

  The road characteristic estimation device 130 estimates road characteristic data, which is a parameter of a relational expression of speed and traffic density, for each road section using the traffic volume information of the traffic volume storage device 110 and the speed information of the probe data storage device 120. To do. This road characteristic data is used when the traffic volume is predicted. The road characteristic data is provided to the classification device 140.

  The classification device 140 clusters the road characteristic data from the road characteristic estimation device 130, and creates representative road characteristic data of the generated cluster and road ID information constituting the cluster. The representative road characteristic data is road characteristic data which is a representative value of each cluster obtained from a plurality of clustered road characteristic data. This representative road characteristic data is stored in the storage device 160. Further, the road ID information associated with the road characteristic data constituting the cluster is stored in the storage device 160 and provided to the travel pattern creation device 150.

  The travel pattern creation device 150 generates travel pattern data for each cluster using the road ID information constituting the clusters of the classification device 140 and the probe data of the probe data storage device 120. The generated traveling pattern data is written in the storage device 160.

  The storage device 160 is a storage device such as a hard disk or a flash memory, and stores road ID information constituting the cluster created by the classification device 140, representative road characteristic data of the cluster, and travel pattern data created by the travel pattern creation device 150. ing.

  The receiving device 170 is a device that receives probe data from the probe center. The format of probe data to be received is the same as the format of probe data stored in the probe data storage device 120. The received probe data is stored in the probe data storage device 120 and provided to the road characteristic search device 180 and the traffic volume prediction device 190.

  The road characteristic search device 180 creates travel pattern data based on the probe data from the receiving device 170, and searches the storage device 160 for clusters having travel pattern data having high similarity to the travel pattern data. Then, the representative road characteristic data of the corresponding cluster is provided to the traffic prediction device 190.

  The traffic volume prediction device 190 predicts the traffic volume using the representative road characteristic data provided from the road characteristic search device 180 and the travel speed data of the probe data received by the reception device 170. The predicted traffic volume is provided to the transmission device 200. The transmission device 200 transmits the traffic volume data of the traffic volume prediction device 190 to the in-vehicle terminal or other center system.

  The road characteristic estimation device 130, the classification device 140, the travel pattern creation device 150, the road characteristic search device 180, and the traffic volume prediction device 190 are all microprocessors (not shown) mounted in the traffic volume prediction system 100, This is realized by software executed by a RAM, a ROM, or the like.

  FIG. 2 shows an example of the format of traffic volume data transmitted from the transmission apparatus 200. The traffic volume data includes a road ID 21, a start time 22 and an end time 23 of a time zone in which the traffic volume is measured, and a traffic volume 24 between the start time and the end time. This traffic volume is the number of vehicles passing the road in a specific time zone.

  Next, details of the traffic volume storage device 110 will be described. The traffic volume storage device 110 stores road traffic volume data. However, not all road traffic data is stored, but only information on roads whose traffic volume has been measured, such as road traffic census, is stored. This traffic volume measurement includes a method in which an observer counts the number of passing vehicles, and a method in which a road traffic infrastructure sensor mechanically counts the number of passing vehicles. The data format of the traffic volume storage device 110 is the same as that in FIG.

  Next, details of the probe data storage device 120 will be described. The probe data storage device 120 stores probe data for a certain period. FIG. 3 is an example of the data format of the probe data storage device 120.

  The probe data stored in the probe data storage device 120 includes a road ID 31 for identifying a road from which probe data has been collected, a vehicle ID 32 for identifying a vehicle from which probe data has been collected, and an inflow that is the time when the vehicle has entered the road. It consists of a time 33 and an outflow time 34 which is the time when the vehicle has flowed out of the road, an average speed 35 when the vehicle has passed the road, and detailed data 36.

  Detailed data 36 in the probe data storage device 120 includes data recording start time 37 that starts at regular time intervals, vehicle traveling direction speed 38, vehicle traveling direction acceleration 39, vehicle lateral speed 40, vehicle It consists of each data of the lateral acceleration 41. The traveling direction speed 38 is information obtained by differentiating the vehicle speed pulse information of the probe car and the position information with respect to time. The traveling direction acceleration 39 is information obtained by further differentiating the information of the acceleration sensor of the probe car and the speed data with respect to time. The lateral speed 40 is information on the moving speed of the vehicle in the lateral direction obtained by differentiating the probe car sensor and position information. The lateral acceleration 41 is acceleration information at the time of the moving speed in the lateral direction of the vehicle, which is obtained by differentiating information from the acceleration sensor of the probe car and speed data. Here, the lateral direction is a direction parallel to the ground and perpendicular to the traveling direction of the vehicle.

  Next, details of the road characteristic estimation device 130 will be described. The traffic volume Q (unit: vehicle / h) of the road is the traffic density k (unit: vehicle / km), which is the number of vehicles existing within a certain distance at a certain moment, and the average speed v of the vehicle passing through the certain distance. It is calculated by (Equation 1) from (unit: km / h).

Q = k × v (Formula 1)
Since this traffic density k is difficult to measure, it is determined from the average speed v from the traffic model. Various models have been proposed for the relationship between the average speed v and the traffic density k. For example, the Underwood model formula is expressed by the following (Formula 2).

v = vf × exp (−k / k0) (Formula 2)
exp () represents an exponential function, vf represents a free speed that is a speed when the traffic density is 0, and k0 represents a critical density that is a traffic density when the traffic volume is maximum. Greenberg's model formula is expressed by the following (formula 3).

v = c × log (kj / k) (Formula 3)
Here, log () is a logarithm, c is a constant, and kj is a saturation density that is a traffic density when the speed is zero.

  Further, when the relationship between the speed v and the traffic density k is expressed by an exponential function, for example, it is expressed as in (Expression 4) where a and b are constants.

v = a × exp (b × k) (Formula 4)
Thus, the relationship between the average speed v and the traffic density k can be expressed by the model and its parameters. In the Underwood model of (Equation 2), these parameters are the free velocity vf and the critical density k0, in the Greenberg model of (Equation 3), the constant c, the saturation density kj, and this is expressed as an exponential function (Equation 4 ) Are constants a and b. These parameters are called road characteristics. The road characteristic estimation device 130 estimates these parameters. In the following description, the underwood model formula is used. However, the relational expression between the traffic density k and the average speed v is not limited to the Underwood model expression.

  A processing flow of the road characteristic estimation apparatus 130 is shown in FIG. The road characteristic estimation device 130 processes a plurality of probe data stored in the probe data storage device 120 at a regular cycle such as one month or one year. Each step will be described in detail below.

  First, in step S10, a loop for repeating the processes from step S20 to S40 is started for all roads stored in the probe data storage device 120. In step S20, it is determined whether the traffic volume storage device 110 has the traffic volume data of the road ID to be processed. If there is traffic data (S20: Yes), the process proceeds to step S30. If there is no traffic data (S20: No), the process proceeds to step S40.

  In step S30, using the traffic volume data and probe data read from the traffic volume storage device 110, the road characteristic of the road ID to be processed is estimated using the traffic volume data as teaching data. In this estimation process, the parameter of (Expression 2), which is a road characteristic, is set so that the difference between the estimated traffic estimated using (Expression 1) and (Expression 2) and the traffic as teaching data is minimized. To decide. Therefore, the probe data collected from the start time 22 to the end time 23 of the traffic data with the matching road ID 21 is acquired from the probe data storage device 120. FIG. 5 is a graph in which the horizontal axis represents the average speed of probe data and the vertical axis represents traffic density. The traffic density, which is the vertical axis of this graph, can be calculated by dividing the value of the traffic volume read from the traffic volume storage device 110 by the speed of the probe data according to (Equation 1). One sample point 51 in the graph of FIG. 5 is a point where the speed of the probe data and the traffic density at the time of the speed (start time of the probe data storage device 120) are plotted. One point can be plotted on the graph of FIG. 5 for one probe data.

  Next, the underwood model formula (Formula 2) is applied to the plotted points so that the error of the free speed vf and the critical density k0 is minimized by using the data of a plurality of sample points 51. Find the curve set to. This is the curve 52 of FIG. In this way, the free speed vf and the critical density k0, which are parameters of the model formula, are obtained for the road ID to be processed. This parameter becomes the road characteristic of this road.

  In step S40 of FIG. 4, since there is no traffic volume data as teaching data for the target road in the traffic volume storage device 110, the road characteristics are estimated using only the probe data. For example, in this case, the free speed vf is set to the maximum value or the 90th percentile value among the speeds of the probe data that has traveled on the processing target road ID. However, since it is difficult to estimate the critical density k0 using only probe data, a predetermined value is used. For example, in the case of Japan, it is empirically known that the critical density is 40 to 60 vehicles / km per lane. Therefore, the median value of 50 vehicles / km is used.

  Further, even when the relationship between the traffic density and the average speed is an exponential function such as (Equation 4), the constants a and b representing the road characteristics cannot be obtained from the probe data alone. For this reason, when the relationship between the traffic density and the average speed is assumed by a function such as (Equation 4), the process proceeds to step S50 without creating road characteristics in step S40, and the process of step S40 is terminated. To do. In this case, since the road characteristic cannot be estimated from the probe data, only the value estimated in step S30 is used as the road characteristic.

  Step S50 in FIG. 4 determines whether or not the processing has been completed for all the roads. If the processing is completed, the processing flow ends. If the processing has not been completed for all roads, the process returns to step S20.

  Next, details of the classification device 140 will be described. The classification device 140 creates a cluster by clustering the road characteristic data of all roads estimated by the road characteristic estimation device 130. FIG. 6 shows a conceptual diagram of the cluster. FIG. 6A shows the result of mapping a road having these two parameters as road characteristics on a biaxial graph of the free speed vf and the critical density k0. One sample point 61 represents data of the free velocity vf and the critical density k0, and is a parameter obtained from the model equation of the curve 52 in FIG. That is, one road characteristic data can be calculated for one road, and this road characteristic data is plotted as one point on FIG. FIG. 6B shows a result of clustering these points into the cluster 62 by an existing method. Clustering is calculated so that roads with similar values are classified into the same cluster among road characteristic data of all roads, and roads with different values are classified into different clusters. This calculation ends when the number of clusters reaches a certain number. Each cluster calculates the barycentric position (for example, average value) of the road characteristic to which it belongs. This center-of-gravity position is used as the representative road characteristic of the cluster.

  The classification device 140 is not limited to two road characteristics (free speed vf and critical density k0). In the Underwood model formula, there are two road characteristics, the free speed vf and the critical density k0. Therefore, when the road characteristics are clustered, the road characteristics are plotted in a two-dimensional graph as shown in FIG. If a model formula that handles three road characteristics is used, the road characteristics are represented by a three-dimensional graph.

  Information about the cluster calculated by the classification device 140 is stored in the storage device 160. Each cluster is given an identification ID (hereinafter referred to as cluster ID). Further, the road ID of the road from which the probe data corresponding to the road characteristics belonging to the cluster is collected is stored in the storage device 160. Further, the vehicle ID of the probe car that has collected the probe data corresponding to the road characteristics belonging to the cluster is stored in a temporary storage area for use by the travel pattern creation device 150.

  Next, details of the travel pattern creation device 150 will be described. 7 and 8 are conceptual diagrams of speed patterns. The traveling pattern is defined by a distribution of acceleration in the traveling direction with respect to a speed in the traveling direction and a distribution of acceleration in the lateral direction with respect to the speed in the traveling direction. Hereinafter, the distribution of acceleration in the traveling direction with respect to the speed in the traveling direction is defined as a first traveling pattern, and the distribution of acceleration in the lateral direction with respect to the speed in the traveling direction is defined as a second traveling pattern. Here, the lateral direction is a direction parallel to the ground and perpendicular to the traveling direction of the vehicle as described above.

  FIG. 7A shows a wide road 71 (hereinafter referred to as a wide road) in which a probe car 72 runs while avoiding other vehicles such as a bicycle, a motorcycle, or a vehicle running at a lower speed than itself. Represents the situation. FIG. 7B shows the first traveling pattern on the wide road, FIG. 7C shows the second traveling pattern, the horizontal axis represents the speed in the traveling direction of the vehicle, and the vertical axis represents the vehicle traveling direction. The acceleration in the traveling direction is shown. In FIG. 7B, in order to make the explanation of the traveling pattern easy to understand, the acceleration distribution with respect to the speed in the traveling direction is divided into a high frequency region 73 (high frequency region) and a low frequency region 74 (low frequency region). It is shown separately.

  In the case of the wide road shown in FIG. 7, the vehicle can move freely in the lateral direction and can travel while avoiding other vehicles. For this reason, as shown in FIG. 7B, the acceleration in the traveling direction is widely distributed in the high speed region of the traveling direction, but is congested in the low speed region of the traveling direction speed. Narrow distribution. On the other hand, since the vehicle moves relatively in the lateral direction to avoid other vehicles, the lateral acceleration is widely distributed regardless of the speed range of the traveling direction as shown in FIG. 7C. .

  FIG. 8A shows a situation in which a probe car 82 cannot avoid other vehicles and travels continuously with a preceding vehicle on a narrow road 81 (hereinafter, narrow road). FIG. 8B shows the first traveling pattern on this narrow road, and FIG. 8C shows the second traveling pattern. Like FIG. 7, the distribution of acceleration with respect to the speed in the traveling direction. Are divided into a high frequency region 83 (high frequency region) and a low frequency region 84 (low frequency region). On a narrow road as shown in FIG. 8, since the vehicle cannot move freely in the lateral direction, the acceleration in the traveling direction and the acceleration in the lateral direction are narrowly distributed regardless of the speed range of the traveling direction.

  As described above, the travel pattern differs depending on the road, and the travel pattern creation device 150 creates a first travel pattern and a second travel pattern for each cluster classified by the classification device 140.

  FIG. 9 shows a processing flow of the travel pattern creation device 150. The travel pattern creation device 150 processes all clusters clustered by the classification device 140 in the loop processing (steps S110 to S140) from step S100 to step S150. And by the loop process from step S110 to S130, all the probe data which comprise the cluster of a process target are processed by step S120, and a travel pattern is produced. As described above, the vehicle ID of the probe car that has collected the probe data corresponding to the road characteristics belonging to the same cluster when the cluster constituting the processing target is generated is created by the classification device 140 as described above, and is stored in the temporary storage area. Is saved. Information on road IDs of roads for which probe data corresponding to road characteristics belonging to the cluster to be processed is collected is stored in the storage device 160. Probe data corresponding to these vehicle IDs and road IDs are extracted from the probe data storage device 120. And a driving | running pattern is produced | generated by repeating the process of step S120 with respect to the extracted probe data.

In step S120, the detailed data 36 of the probe data to be processed is acquired, and the first travel pattern and the second travel pattern of each probe data are created. Each traveling pattern is data representing the relationship of the traveling direction acceleration / lateral direction acceleration to the traveling direction speed, and is created using a table as shown in FIG. The table of FIG. 10 includes a column 101 divided for each velocity band and a row 102 divided for each acceleration region. In column 101 of this table, for example, a column of “0-10” represents a speed range of 0-10 km / h. In the row 102, for example, the “−3 to 3” row represents an acceleration region of −3 to 3 m / s ² . It is assumed that “0” is initially input to the cell 103 of the table as an initial value. In step S120, from the detailed data 36 of each probe data, a cell in the velocity region and acceleration region corresponding to the value of the traveling direction velocity 38, the traveling direction acceleration 39 and the lateral acceleration 41 is found, and the value of the corresponding cell is found. Repeat the process of adding 1 to. As a result, the larger the cell value, the more frequently the speed region / acceleration region is represented.

  In step S130, it is determined whether or not all the probe data constituting the cluster to be processed has been processed. When the processing is completed for all probe data in one cluster, the process proceeds to step S140. If the process has not been completed, step S120 is repeated.

  Next, in step S140, the running pattern of the cluster to be processed created in step S120 is normalized. Since the number of probe data constituting a cluster differs depending on the cluster, normalization is performed so that all clusters can be compared. The value of the cell in the i-th row and j-th column in the travel pattern table shown in FIG. 10 is Sij. When the normalized value of this cell is Rij, Rij is expressed as (Equation 5).

Rij = Sij / ΣΣSij (Formula 5)
The denominator ΣΣSij in (Equation 5) is the sum of the values of all cells. In this way, the cell values are normalized so that the sum of all the cells becomes 1. The normalized first travel pattern and second travel pattern are stored in the storage device 160 in association with the corresponding cluster ID.

  Finally, in step S150, it is determined whether all clusters clustered by the classification device 140 have been processed. When the process is completed for all clusters, the process of the travel pattern creation device 150 is terminated. However, if there are still clusters that have not been processed, the process returns to step S100.

  Next, the configuration of the storage device 160 will be described. FIG. 11 is a diagram showing a data format of the storage device 160. Data in the storage device 160 includes a cluster ID 111 classified by the classification device 140, a representative road characteristic 112 of the cluster, a road ID 113 constituting the cluster, a first travel pattern 114, and a second travel pattern 115. The representative road characteristic 112 is composed of a free speed vf and a critical density k0 when the underwood model formula is used in the road characteristic estimation device 130, but a different road is used when another model formula is used. The characteristics are stored. The first travel pattern 114 and the second travel pattern 115 are recorded in the same data format as that in FIG. 10 and the first travel pattern and the second travel pattern created by the travel pattern creation device 150 are stored. The

  Next, the road characteristic search device 180 will be described. The road characteristic search device 180 uses the probe data received by the reception device 170 to search the storage device 160 for clusters close to the road characteristic of the road from which the probe data was collected.

  The processing flow of this road characteristic search device 180 is shown in FIG. When there are a plurality of probe data received by the receiving apparatus 170 within a certain period, the processing of FIG. 12 is executed for the number of probe data. In step S200 of FIG. 12, there is data in which the road ID that matches the road ID of the traveled road included in the probe data from the receiving device 170 is included in the road ID list of the road ID 113 from the storage device 160. It is determined whether or not. If there is a match, the information of the representative road characteristic 112 of the cluster whose matching road ID is included in the road ID list of the road ID 113 is acquired, and the process is terminated. If they do not match, the process proceeds to step S210 to search for clusters from the travel pattern of the probe data.

  In step S210, a first travel pattern and a second travel pattern of the probe data from the receiving device 170 are created. The specific processing content is the same as step S120 in FIG. Next, in step S220, the first travel pattern and the second travel pattern of the probe data created in step S210 are normalized. The specific processing content is the same as step S140 in FIG. By normalizing, it can be compared with the running pattern of each cluster stored in the storage device 160.

  In step S230, a travel pattern having high similarity is extracted from the travel pattern created from the probe data created in step S220 and the travel pattern 114 of each cluster in the storage device 160, and the road characteristics of the cluster are acquired. In the first traveling pattern, the similarity is evaluated by an expression shown in (Expression 6). The cell value of the i-th row and j-th column in the first travel pattern normalized by the probe data is P1ij, and the cell value of the i-th row and j-th column is R1kij in the first travel pattern of the cluster ID “k”. To do. At this time, the evaluation value of the first travel pattern of the cluster ID “k” is set to W1k.

W1k = ΣΣ | P1ij−R1kij | ² (Formula 6)
In the second travel pattern, the similarity is evaluated by an expression shown in (Expression 7). The cell value of the i-th row and j-th column in the second travel pattern normalized by the probe data is P2ij, and the cell value of the i-th row and j-th column in the second travel pattern of the cluster ID “k” is R2kij. To do. At this time, the evaluation value of the second running pattern of the cluster ID “k” is set to W2k.

W2k = ΣΣ | P2ij−R2kij | ² (Expression 7)
At this time, the cluster having the smallest sum of evaluation values (W1k + W2k) is taken as the cluster to which the road on which the probe data traveled belongs, and information on the representative road characteristic 112 is acquired.

  Next, the traffic volume prediction device 190 will be described. The traffic volume prediction device 190 predicts the traffic volume on a road basis using the average speed 35 of the probe data of the reception device 170 and the information of the representative road characteristics 112 acquired from the road characteristic search device 180. At this time, the model set by the road characteristic estimation device 130 is used. For example, when the Underwood model formula is used, the traffic density k is calculated from (Formula 2) using the average speed v of the probe data and the road characteristic data (free speed vf, critical density k0) of the representative road characteristics 112. Is estimated. Next, the traffic volume is predicted from (Equation 1) using the estimated traffic density k and the average speed v of the probe data. The predicted traffic volume data and road ID information are provided to the transmission device 200.

  According to the embodiment described above, the traffic volume prediction system 100 can use the road characteristics of similar groups by using the driving pattern of the probe data, so that the traffic can be accurately detected without incurring a large cost. The amount can be predicted.

  Next, a traffic volume prediction system according to a modification of the embodiment of the present invention will be described. In a variation of this embodiment, road-based speed data collected from traffic infrastructure sensors or the like is received, road characteristics are searched using this speed data, and traffic volume is predicted.

  FIG. 13 shows the configuration of a traffic volume prediction system according to a modification of this embodiment. The traffic volume prediction system 300 in FIG. 13 includes a traffic volume storage device 110, a probe data storage device 120, a road characteristic estimation device 130, a classification device 140, a travel pattern creation device 150, a storage device 160, a reception device 210, and a road information storage device. 220, a road characteristic search device 230, a traffic volume prediction device 190, and a transmission device 200.

The receiving device 210 is a device that receives speed data for each road in addition to probe data from an external server. An example of the format of the received speed data is shown in FIG. The speed data includes the road ID 141 of the road from which the data was collected, the time 142 at which the data was collected, and the collected speed 143 data. The received speed data is provided to the road characteristic search device 230 and the traffic volume prediction device 190.
The road information storage device 220 is a storage device such as a hard disk or a flash memory, stores road attribute information, and is referred to by the road characteristic search device 230. An example of the data format of the road information storage device 220 is shown in FIG. The road information storage device 220 stores the number of lanes 152 and the road width 153 of the road corresponding to the road ID 151 as road attribute information. The road attribute information is detailed information about the road such as the number of lanes and the road width, but is not limited to the number of lanes and the road width.

  The road characteristic search device 230 is realized by software executed by a microprocessor (not shown) mounted on the traffic volume prediction system 300, RAM, ROM, or the like. The road characteristic search device 230 uses the road ID that has collected the speed data from the reception device 210, the road ID that constitutes each cluster stored in the storage device 160, and the road information in the road information storage device 220. The cluster similar to the road ID of the speed data is searched.

  The processing flow of the road characteristic search device 230 is shown in FIG. First, in step S300, whether or not the storage device 160 has data included in the road ID list in which the road ID matching the road ID included in the speed data from the receiving device 210 is recorded in the road ID 113 is determined. Determine whether. If there is data having a road ID list including the matching road ID (S300: Yes), the information of the cluster ID 111 and the representative road characteristic 112 of the data is acquired, and the process is terminated. If there is no data having a road ID list including the matching road ID (S300: No), the process proceeds to step S310 to search for clusters corresponding to similar roads from the road attribute information. In step S <b> 310, road attribute information of a road ID that matches the road ID of the speed data from the receiving device 210 is acquired from the road information storage device 220.

  In step S320, a cluster having high similarity is extracted from all the clusters in the storage device 160, and road characteristics of the cluster are acquired. Specifically, for all the clusters in the storage device 160, the road attribute information of the matching road ID is obtained from the road information storage device 220 using the road IDs included in the road ID list of the road ID 113 of each cluster. get. Next, among the roads included in the cluster, a road having a high similarity to the road attribute of the road of the speed data from the receiving device 210 is searched to find a cluster having the highest similarity. Finally, the road characteristics of the cluster determined to have high similarity are acquired from the storage device 160.

  A cluster similarity evaluation method for road attributes will be described. Here, the road attribute information of the road of the speed data from the receiving device 210 is the number of lanes and the road width as shown in FIG. Cluster evaluation value when the number of lanes of the i-th road constituting the cluster is A (i), the road width is B (i), the number of lanes of the road attribute information to be evaluated is C, and the road width is D E is calculated by the following (Equation 8).

E = Σ (| A (i) −C | ² + | B (i) −D | ² ) (Expression 8)
In (Expression 8), the number of lanes and the road width of the evaluation target road are compared for all roads constituting the cluster. Since the difference between the number of lanes and the road width of the road to be evaluated is added to the number of roads constituting the cluster, the similarity is higher as the evaluation value E is smaller. If the road attribute includes information that cannot be quantified, such as the presence or absence of a bus lane, the calculation is made as 1 if the road attribute is the same, and 0 if the road attribute is different.

  As a result, the traffic volume prediction system 300 can use the attribute information of the road on which the probe car has traveled to use the road characteristics of a similar group, so the traffic volume can be predicted with high accuracy without incurring a large expense. It becomes like.

100, 300 Traffic prediction system 110 Traffic storage device 120 Probe data storage device 130 Road characteristic estimation device 140 Classification device 150 Travel pattern creation device 160 Storage device 170, 210 Reception device 180, 230 Road characteristic search device 190 Traffic prediction device 200 Transmitter 220 Road Information Storage Device

Claims (4)

A traffic prediction system that predicts traffic using probe data,
A road characteristic estimation device that estimates road characteristics including any one of free speed, critical speed, saturation density, and critical density used in a model expression representing traffic volume from probe data;
A classification device for generating a cluster in which the estimated road characteristics are classified for each similar road characteristic;
A travel pattern creation device that creates a travel pattern of a vehicle on a road belonging to the cluster with respect to each cluster based on the speed of the probe data used for estimating the road characteristics constituting the cluster and the amount of change in the speed;
The road characteristics of a cluster having a high similarity in the travel pattern of the vehicle created from the probe data of the road that does not belong to the cluster and the travel pattern in the cluster created by the travel pattern creation device are compared. A road characteristic acquisition device for acquiring
A traffic volume prediction device that predicts the traffic volume of the road from the model formula based on the road characteristics acquired by the road characteristic acquisition device;
A traffic volume prediction system comprising:   2. The traffic volume prediction system according to claim 1, wherein the road characteristic estimation device estimates the road characteristic of the model equation using a vehicle speed obtained by statistically processing the accumulated probe data. Traffic volume prediction system.   3. The traffic volume prediction system according to claim 2, wherein the travel pattern creation device includes a relationship between a vehicle speed and a speed change amount in a traveling direction and a speed change amount in a lateral direction with respect to the vehicle speed and the travel direction as a travel pattern. A traffic volume prediction system, wherein the frequency of the probe data is calculated for the relationship.   The road characteristic acquisition device according to claim 2, in the case of a road that does not belong to the cluster and has no probe data, a road in which the number of lanes and the road width match among the attribute information of the road that belongs to the cluster. A traffic volume prediction system characterized in that road characteristics of a cluster having a large number are acquired as road characteristics of the road.

Images