JP6988504B2 - Data creation device, data creation method and program - Google Patents

Description

The present invention relates to a data creation device, on the data How to Create Ho及beauty program.

With the development of sensor technology, it has become possible to easily collect road information such as traffic volume. However, it may be difficult to record the route of each individual traveling on the road due to privacy issues, sensor placement issues, and the like.

On the other hand, a method of estimating the number of people by route (hereinafter, also referred to as "number of people by route") from road information has been proposed. For example, various models for estimating the number of people by route (hereinafter, also referred to as "number estimation model by route") have been proposed. Further, for example, a method of estimating the number of people by route by solving an optimization problem from the correspondence between a route and a road has been proposed (see, for example, Non-Patent Document 1).

Hitoshi Shimizu, Takuma Otsuka, Tomoharu Iwata, Hiroshi Sawada, Osamu Ueda, Estimating Agent Parameters in Human Flow Simulation, 20th Information Theory Learning Theory Workshop (2017), D2-50, Internet <http: // ibisml .org / archive / ibis2017 / poster_preview_D2.pdf ＞

However, in the conventional model for estimating the number of people by route, the parameters are estimated without using data such as the number of people by route and road information. Therefore, the estimation accuracy of the number of people for each route estimated by the estimation model for each route with this parameter set may not be high.

In addition, when solving an optimization problem from the correspondence between a route and a road, for example, information such as the connection between roads (connection relationship) cannot be utilized, and the estimation accuracy of the number of people for each route is not high. there were.

The present invention has been made in view of the above points, and an object of the present invention is to estimate the number of people for each route with high accuracy by using road information.

Therefore, in the data creation device according to the embodiment of the present invention, when the road structure representing the road structure and the number of people by route representing the number of people by route for each road structure are created or acquired, the data creation device is created or acquired for each route. Create road route data that includes at least the road conditions that represent the road conditions for each number of people, the route observation correspondence matrix that associates the route with the observation points that observe the number of people for each route, and the number of people for each route. It is characterized by having a means for creating data.

Further, in the parameter estimation device according to the embodiment of the present invention, at least a route observation correspondence matrix in which a route and an observation point for observing the number of people for each route are associated with each other and a road condition representing the road condition are input. It is a parameter estimation device that estimates the parameters of the model that outputs the number of people for each route, which represents the number of people for each route, and is output by inputting the road conditions and the route observation correspondence matrix included in the learning data into the model. It is characterized by having a parameter estimation means for estimating the parameter so as to minimize an error between the first number of people for each route and the second number of people for each route included in the learning data.

Further, in the route-based number estimation device according to the embodiment of the present invention, at least a route observation correspondence matrix in which a route and an observation point for observing the number of people in each route are associated with each other and a road condition representing a road condition are input. Then, the number of people for each route is estimated by a model that outputs the number of people for each route, and the parameters estimated by the parameter estimation means in the parameter estimation device are set. It is characterized by having a route-specific number estimation means for estimating the route-specific number of people by inputting at least the route observation correspondence matrix and the road condition to the model.

The number of people by route can be estimated with high accuracy using road information.

It is a figure which shows an example of the functional structure of the number of people estimation apparatus for each route in embodiment of this invention. It is a flowchart which shows an example of the process performed by the route-based number estimation apparatus in embodiment of this invention. It is a figure which shows an example of the hardware configuration of the number of people estimation apparatus for each route in embodiment of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the embodiment of the present invention, the route-specific number estimation device 10 for estimating the route-specific number of people from road information will be described. The number of people by route is not limited to the number of people by route (for example, pedestrians, etc.), and may be the number or amount of arbitrary moving objects that move on any graph. The graph is not limited to a road, and may be, for example, an indoor passage, a communication network, or the like.

Specifically, the number of people for each route may be, for example, the number of cars for each route, the number of bicycles for each route, or the like, in addition to the number of people for each route. When the communication network is targeted as a graph, the number of people for each route may be, for example, the amount of information (data) for each route in the communication network.

<Functional configuration of the number of people estimation device 10 for each route>
First, the functional configuration of the route-based number estimation device 10 according to the embodiment of the present invention will be described with reference to FIG. FIG. 1 is a diagram showing an example of the functional configuration of the number of people estimation device 10 for each route according to the embodiment of the present invention.

As shown in FIG. 1, the route-based number estimation device 10 according to the embodiment of the present invention has a data creation unit 110, a parameter estimation unit 120, and a route-specific number estimation unit 130. Each of these functional units is realized by a process of causing a CPU (Central Processing Unit) or the like to execute one or more programs installed in the route-based number estimation device 10. It should be noted that each of these functional units may be realized by a process executed by one program, or may be realized by a process executed by a different program.

Further, the route-based number estimation device 10 according to the embodiment of the present invention has a storage unit 140. The storage unit 140 can be realized by using, for example, an auxiliary storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage unit 140 may be realized by using a storage device or the like connected to the route-specific number estimation device 10 via a network.

The data creation unit 110 creates data (this data is referred to as "road route data") regarding the number of people for each route and the road information at that time by simulation. The data creation unit 110 includes a road structure creation unit 111, a route-specific number of people creation unit 112, a road condition creation unit 113, and a road route data creation unit 114. Any traffic volume simulator can be used as the simulator used for the simulation.

The road structure creation unit 111 is a set of N road structures G _n .

を作成する。以降では、道路構造Ｇ_ｎの集合を「道路構造集合」と表す。

To create. Hereinafter, _{the set of road structures Gn} will be referred to as a “road structure set”.

The road structure _Gn is, for example, information representing the structure of the road such as the connection between the roads, the width, and the slope. The road structure _Gn may be information obtained by observing a road in the real world with a sensor or the like, or may be information generated by a simulator or the like. The road structure _Gn can be represented by a graph, for example, where a node represents a road and an edge represents a connection between roads.

The route-specific number creation unit 112 is a set of M _n route-specific number Y _{nm for each road structure G n.}

を作成する。ここで、ｙ_ｎｍｋはｋ番目の経路の人数、Ｋ_ｎｍは経路数を表す。以降では、経路別人数Ｙ_ｎｍの集合を「経路別人数集合」と表す。経路別人数Ｙ_ｎｍとは、経路別の人数を表す情報である。経路別人数Ｙ_ｎｍは、ランダムに作成された情報であっても良いし、実世界の道路をセンサ等で観測することで得られる観測値（或る経路における実際の人数）と同一又は近くなるように作成された情報であっても良い。

To create. Here, y _nmk represents the number of k-th routes and K _nm represents the number of routes. Hereinafter, the set of the number of people by route Y _nm is referred to as the “set of people by route”. The number _{of people by route Y nm} is information representing the number of people by route. The number of people Y _{nm for each} route may be randomly created information, and may be the same as or close to the observed value (actual number of people in a certain route) obtained by observing a road in the real world with a sensor or the like. The information may be created as follows.

The road condition creation unit 113 creates a set of road conditions X _nm corresponding to each route-specific number of people Y _nm (hereinafter referred to as “road condition set”) by simulation. The road condition X _nm represents information on a moving body such as the number of people included in each road in a certain time zone, the number of people passing through a certain place, and the average moving speed of each road.

For example, assuming that the road condition is X _nm , the number of people passing through each point (observation point) at each time.

が利用できる。ここで、ｘ_ｎｍｉｔはｎ番目の道路構造Ｇ_ｎにおけるｍ番目の道路状況Ｘ_ｎｍで時刻ｔにｉ番目の観測地点を通過する人数、Ｉ_ｎｍは観測地点数、Ｔは観測の時刻数を表す。なお、時刻ｔは、シミュレーションの或る時刻から或る時刻までを或る時間幅に分割した場合における各時間幅を表すインデックスである。

Is available. _{Here, x NMIT} the number of people passing through the n-th road structure _{G n} m-th i th at time t in road conditions _{X nm} of the observation point _{in, I nm} is observing location number, T is representative of the time the number of observations .. The time t is an index representing each time width when the simulation is divided into a certain time width from a certain time to a certain time.

The road route data creation unit 114 is a road route data summarizing _{each road structure Gn} , each route observation correspondence matrix A _nm , the number of people Y _{nm for} each route, and each road condition X _nm.

を作成する。ここで、経路観測対応行列Ａ_ｎｍは、Ｋ_ｎｍ個の経路と、Ｉ_ｎｍ個の観測地点とを対応付けるＫ_ｎｍ×Ｉ_ｎｍの行列であり、ｉ番目の観測地点がｋ番目の経路に含まれていればｉ行ｋ列目の要素が１、そうでなければ０となる。道路状況Ｘ_ｎｍとして、１時刻のみの道路状況を用いても良いし、或る一定時間の道路状況を用いても良い。

To create. Here, the route observation corresponding matrix _{A nm,} and _{K nm} pieces of _route, a matrix of _{K nm} × _{I nm} associating the _{I nm} pieces of observation points, i th observation point is included in the k-th path If so, the element in the i-row and k-th column is 1, otherwise it is 0. As the road condition X _nm , the road condition for only one time may be used, or the road condition for a certain period of time may be used.

Hereinafter, the information including the road structure _Gn , the route observation correspondence matrix A _nm, and the road condition X _nm is also referred to as “road information”. Therefore, it can be said that the road route data is data that summarizes the road information and the number of people Y _{nm for each route.}

However, the road information may be information _{that does not include the road structure Gn} but includes the route observation correspondence matrix A _nm and the road condition X _nm .

The road route data created by the road route data creating unit 114 is stored in the storage unit 140. This road route data is learning data used for parameter estimation of the number estimation model for each route.

The data creation unit 110 does not necessarily have to create at least one of each road structure _Gn and the number of people Y _nm for each route. For example, data obtained by observing the real world with a sensor or the like is input from the outside. May be done. That is, the road structure creating unit 111 may acquire _{each road structure Gn input from the outside.} Similarly, the route-specific number creation unit 112 may acquire the _{route-specific number Y nm input from the outside.}

The parameter estimation unit 120 estimates the parameters of the route-specific number estimation model using the road route data stored in the storage unit 140. The parameter estimation unit 120 estimates the parameters of the route-specific number estimation model so that the route-specific number of people can be estimated more accurately when the road information is given. As a model for estimating the number of people by route, for example, a graph convolutional neural network can be used. However, the model is not limited to this, and any model such as a neural network, a support vector machine, a Gaussian process, etc. can be used for each route as long as it is a model that can estimate and output the number of people for each route when road information is given. It can be used as a model for estimating the number of people.

In the case of a graph convolutional neural network, it is assumed that each path and each observation point has a hidden layer. Hidden layer of the lth layer of observation point i

は、以下の式１のように計算される。

Is calculated as in Equation 1 below.

ここで、０層目の隠れ層は道路状況、すなわち、

Here, the hidden layer of the 0th layer is the road condition, that is,

であるとする。

Suppose that

In addition, the hidden layer of the lth layer of the path k

は、以下の式２のように計算される。

Is calculated as in Equation 2 below.

ここで、上記の式１及び式２において、

Here, in the above equations 1 and 2,

はそれぞれｌ層目の線形変換行列であり、これらの線形変換行列が推定すべきパラメータである。

Are the linear transformation matrices of the first layer, respectively, and these linear transformation matrices are the parameters to be estimated.

The size of the final hidden layer of the path k (that is, the hidden layer of the Lth layer) is set to 1, and the value of the final hidden layer is set to 1.

を出力とすることができる。その出力と、道路経路データの経路別人数Ｙ_ｎｍに含まれるｙ_ｎｍｋの値とがなるべく同じになるようにパラメータを推定する。例えば、以下の式３に示す絶対値誤差Ｅを、最小化すべき目的関数として用いることができる。

Can be output. _{The parameters are estimated so that the output and the value of y nmk} included in the _{number of people Y nm} for each route in the road route data are as equal as possible. For example, the absolute value error E shown in Equation 3 below can be used as the objective function to be minimized.

ただし、絶対値誤差以外にも、例えば、二乗誤差等を用いても良い。

However, in addition to the absolute value error, for example, a square error or the like may be used.

The parameters estimated by the parameter estimation unit 120 are stored in the storage unit 140.

The model for estimating the number of people by route is not _limited to a model that uses three types of information: _{road structure G n} , route observation correspondence matrix A _nm, and road condition X nm, but for example, route observation without using _{road structure G n.} It may be a model using two pieces of information, the corresponding matrix A _nm and the road condition X _nm. For example, in the case of a graph convolutional neural network that uses two pieces of information, the path observation correspondence matrix A _nm and the road condition X _nm , the following equation 4 may be used instead of the above equation 1.

経路別人数推定部１３０は、記憶部１４０に格納されているパラメータを設定した経路別人数推定モデルと、道路情報とを用いて、経路別人数を推定する。すなわち、例えば、経路別人数推定部１３０は、道路情報に含まれる道路構造Ｇ_＊と経路観測対応行列Ａ_＊と道路状況Ｘ_＊とを、当該パラメータを設定した経路別人数推定モデルに入力することで、経路別人数を出力として得る。

The route-specific number estimation unit 130 estimates the route-specific number of people by using the route-specific number estimation model in which the parameters stored in the storage unit 140 are set and the road information. That is, for example, the route-specific number estimation unit 130 _{inputs the road structure G *} , the route observation correspondence matrix A _*, and the road condition X _* included in the road information into the route-specific number estimation model in which the parameters are set. Then, the number of people by route is obtained as an output.

When the parameters of the number estimation model for each route are estimated using the two information of the route observation correspondence matrix A _nm and the road condition X _nm without using the road structure _{Gn, the route observation correspondence included in the road information is estimated.} By inputting the matrix A _* and the road condition X _* into the route-specific number estimation model in which the parameters are set, the route-specific number of people may be obtained as an output.

The number of people by route estimated by the number of people estimation unit 130 by route is output to a predetermined output destination. The predetermined output destination may be a display device such as a display, a storage unit 140, an external recording medium, or the like, or another device or the like connected via a network. ..

In the example shown in FIG. 1, a case where one device (route-specific number estimation device 10) has a data creation unit 110, a parameter estimation unit 120, and a route-specific number estimation unit 130 has been described. Not limited to. The data creation unit 110, the parameter estimation unit 120, and the number of people estimation unit 130 for each route may be distributed to a plurality of devices. For example, an embodiment of the present invention is realized by, for example, a data creation device having a data creation unit 110, a parameter estimation device having a parameter estimation unit 120, and a route-specific number estimation device having a route-specific number estimation unit 130. May be done.

<Process executed by the number estimation device 10 for each route>
Next, the flow of processing executed by the route-based number estimation device 10 according to the embodiment of the present invention will be described with reference to FIG. FIG. 2 is a flowchart showing an example of processing executed by the route-based number estimation device 10 according to the embodiment of the present invention.

Step S101: The data creation unit 110 creates road route data. That is, the data creation unit 110 _{creates or acquires each road structure Gn by the road structure creation unit 111, creates or acquires the number Y nm} for each route by the route-specific number creation unit 112, and the road condition creation unit 113. Create each road condition X _nm. Then, the data creation unit creates the road route data by the road route data creation unit 114, and stores the created road route data in the storage unit 140.

Step S102: Next, the parameter estimation unit 120 estimates the parameters of the route-specific number estimation model using the road route data stored in the storage unit 140. That is, the parameter estimation unit 120 estimates a parameter that minimizes an error between the output of the route-specific number estimation model using the road information included in the road route data and the route-specific number of people included in the road route data. Then, the parameter estimation unit 120 stores the estimated parameters in the storage unit 140.

Step S103: Next, the route-specific number estimation unit 130 estimates the route-specific number of people by using the route-specific number estimation model in which the parameters stored in the storage unit 140 are set and the road information.

That is, for example, the route-specific number estimation unit 130 _{inputs the road structure G *} , the route observation correspondence matrix A _*, and the road condition X _* included in the road information into the route-specific number estimation model in which the parameters are set. Then, estimate the number of people by route. Alternatively, when the parameters of the route-specific number estimation model are estimated using the two information of the route observation correspondence matrix A _nm and the road condition X _{nm without using the road structure Gn, the route-specific number estimation unit 130 is used.} , The route observation correspondence matrix A _* and the road condition X _* included in the road information are input to the route-specific number estimation model in which the parameters are set, so that the route-specific number of people is estimated.

As described above, the route-specific number estimation device 10 according to the embodiment of the present invention creates data (road route data) relating to the route-specific number of people and the road information at that time by simulation. Then, the route-based number estimation device 10 according to the embodiment of the present invention estimates the parameters of the route-based number estimation model using the road route data. As a result, in the route-specific number estimation device 10 according to the embodiment of the present invention, the route-specific number of people can be estimated with high accuracy by inputting road information into the route-specific number estimation model in which this parameter is set. become.

<Comparative example of the present invention with the prior art>
Here, a comparative example between the present invention and the prior art will be described. In order to compare the present invention with the prior art, the following experiments were conducted. That is, first, 700 road structures were created based on the roads around Japanese railway stations. In each road structure, a total of 100,000 people, including pedestrians heading to the station from some random places around the station and pedestrians heading to some random places from the station, were simulated for one hour. Next, the parameters of the number-of-routes estimation model for each route were estimated using the route road data including one route observation matrix and road conditions in each of the 700 road structures as learning data.

Then, 70 road structure data different from the training data were used as verification data, and 70 road structure data different from the training data and the verification data were used as test data. As a model for estimating the number of people by route, a graph convolutional neural network with 3 hidden layers and 5 hidden layers is used, and learning is performed 5 times by changing the initial value, and the absolute value error is the lowest in the verification data. It was adopted. On the other hand, as the prior art, the method proposed in Non-Patent Document 1 was used.

Further, as the evaluation scale of the present invention and the prior art, the normalized absolute value error shown in the following equation 5 was used.

テストデータを用いて、本発明により経路別人数を推定した場合における平均二乗誤差は０．７４７であった。一方で、テストデータを用いて、従来技術により経路別人数を推定した場合における平均二乗誤差は０．７９９であった。これにより、本発明は、従来技術と比較して、高い精度で（すなわち、誤差が少ない）経路別人数を推定できていることがわかる。

The mean square error when the number of people by route was estimated by the present invention using the test data was 0.747. On the other hand, the mean square error was 0.799 when the number of people by route was estimated by the conventional technique using the test data. From this, it can be seen that the present invention can estimate the number of people for each route with high accuracy (that is, with less error) as compared with the prior art.

<Hardware configuration of the number of people estimation device 10 for each route>
Finally, the hardware configuration of the route-based number estimation device 10 according to the embodiment of the present invention will be described with reference to FIG. FIG. 3 is a diagram showing an example of the hardware configuration of the route-based number estimation device 10 according to the embodiment of the present invention.

The route-based number estimation device 10 shown in FIG. 3 communicates with an input device 201, a display device 202, an external I / F 203, a RAM (Random Access Memory) 204, a ROM (Read Only Memory) 205, and a CPU 206. It has an I / F 207 and an auxiliary storage device 208. Each of these hardware is connected so as to be communicable via the bus B.

The input device 201 is, for example, a keyboard, a mouse, a touch panel, or the like, and is used for a user to input various operations. The display device 202 is, for example, a display or the like, and displays the processing result of the number of people estimation device 10 for each route. The route-based number estimation device 10 does not have to have at least one of the input device 201 and the display device 202.

The external I / F 203 is an interface with an external device. The external device includes a recording medium 203a and the like. The route-specific number estimation device 10 can read or write the recording medium 203a or the like via the external I / F 203. The recording medium 203a may record a program or the like that realizes each function of the route-based number estimation device 10.

The recording medium 203a includes, for example, a flexible disk, a CD (Compact Disc), a DVD (Digital Versatile Disk), an SD memory card (Secure Digital memory card), a USB (Universal Serial Bus) memory card, and the like.

The RAM 204 is a volatile semiconductor memory that temporarily holds programs and data. The ROM 205 is a non-volatile semiconductor memory that can hold programs and data even when the power is turned off. The ROM 205 stores, for example, OS (Operating System) settings, network settings, and the like.

The CPU 206 is an arithmetic unit that reads programs and data from the ROM 205, the auxiliary storage device 208, and the like onto the RAM 204 and executes processing.

The communication I / F 207 is an interface for connecting the route-based number estimation device 10 to the communication network. The program that realizes each function of the route-based number estimation device 10 may be acquired (downloaded) from a predetermined server device or the like via the communication I / F 207.

The auxiliary storage device 208 is, for example, an HDD, an SSD, or the like, and is a non-volatile storage device that stores programs and data. The programs and data stored in the auxiliary storage device 208 include, for example, an OS, an application program that realizes various functions on the OS, a program that realizes each function of the route-based number estimation device 10.

The route-based number estimation device 10 according to the embodiment of the present invention can realize the above-mentioned various processes by having the hardware configuration shown in FIG.

The present invention is not limited to the above-described embodiment disclosed specifically, and various modifications and modifications can be made without departing from the scope of claims.

10 Number of people estimation device by route 110 Data creation unit 111 Road structure creation unit 112 Number of people creation unit by route 113 Road condition creation unit 114 Road route data creation unit 120 Parameter estimation unit 130 Number of people estimation unit by route 140 Storage unit

Claims (3)

At least N road structures G _n (n = 1, ..., N) expressing the connection relationship between roads in a graph structure and the number of people by route representing the number of people for each road structure G _n Y _{nm (m = 1, ···, M} n) when a is created or acquired, the road structure G _n and, road conditions X representing the number of passing people said each path-specific number Y _nm at each observation point on the road _nm and the a path observation corresponding matrix a _nm representing the whether or observation point is included in the path, the path-specific number Y _nm and n and combined data _{G n respect _m, {a _{nm, Y nm} , X _nm ; m = 1, ..., M _n }; n = 1, ..., N} as road route data.
A data creation device characterized by having. At least N road structures G _n (n = 1, ..., N) expressing the connection relationship between roads in a graph structure and the number of people by route representing the number of people for each road structure G _n Y _{nm (m = 1, ···, M} n) when a is created or acquired, the road structure G _n and, road conditions X representing the number of passing people said each path-specific number Y _nm at each observation point on the road _nm and the a path observation corresponding matrix a _nm representing the whether or observation point is included in the path, the path-specific number Y _nm and n and combined data _{G n respect _m, {a _{nm, Y nm} , X _nm ; m = 1, ..., M _n }; n = 1, ..., N} as road route data.
A data creation method characterized by a computer performing. The computer program for functioning as a data creation equipment according to claim 1.

Images