WO2022190827A1

WO2022190827A1 - Operation state estimation system and operation state estimation method

Info

Publication number: WO2022190827A1
Application number: PCT/JP2022/006797
Authority: WO
Inventors: 慶行但馬; 由泰高橋; 雄一小林
Original assignee: 株式会社日立製作所
Priority date: 2021-03-10
Filing date: 2022-02-18
Publication date: 2022-09-15
Also published as: JP7371040B2; JP2022138355A

Abstract

This operation state estimation system comprises a processor. The processor (1) executes a passage time estimation model construction unit and uses at least path information, a timetable, and location information about a moving body based on the traveling history, or uses at least the location of a stop place and timetable information at the stop place of the moving body, and information that can be acquired by linearly complementing the timetable information, and constructs, in a unit of commonality in operation patterns, an elapsed time estimation model which estimates an elapsed time from the start of the moving body in response to an input of information about the current location of the moving body. The processor (2) executes an inference unit and obtains the current delay time of the moving body from the elapsed time. The processor (3) executes a disclosure unit and presents a user with an operation state of the moving body.

Description

Operation state estimation system and operation state estimation method

The present invention relates to technology for estimating the delay of mobile objects such as buses, railways, trams, ships, and taxis.

　Moving vehicles such as buses, trains, trams, ships, and some taxis operate according to timetables. Although it is desirable that these moving bodies operate according to the timetable, they may be delayed for various reasons such as accidents, construction work, weather, and traffic jams. Here, in the case of constant delays, it may be possible to respond by changing the timetable, but it is not possible to depart earlier than the normal timetable (for example, it is regulated in Japan). , it is not possible to properly determine the timetable based on the average running time. Therefore, for this reason, even if the timetable is updated, there are situations where constant delays continue to occur. Further, in consideration of user-friendliness, a timetable is often set such that flights on the same route run at the same time (for example, with a difference of one hour) even if the departure times are different.

And now, due to the development of information and communication technology, it has become possible to use location information of mobile objects that changes from moment to moment. In addition to this, studies on data formats that can be commonly used among business operators, such as GTFS (General Transit Feed Specification) and GTFS-RT, are underway both in Japan and overseas. Against this background, there is a demand for a common delay prediction technique for improving user convenience.

Here, Patent Document 1 discloses that "a calculation unit that calculates a plurality of explanatory variables using at least one of the arrival time and departure time of each train at each station in a predetermined section of the railway; a delay time at a predetermined station and a predetermined train in a section is used as an objective variable, an explanatory variable having a predetermined contribution to the objective variable is selected from the plurality of explanatory variables, and a regression equation for the objective variable and a display control unit that causes a display unit to display information about the delay time based on the regression equation."

JP 2019-123479 A

Here, in recent years, GPS devices for specifying positions are often attached to mobile bodies. Therefore, there is a problem that the delay cannot be estimated by reflecting real-time mobile information acquired by GPS (Global Positioning System) or the like.

Furthermore, on buses and trams, etc., there may be no means of obtaining accurate arrival and departure times at stops (the intended place to stop, for example, a bus stop). In addition, when there are no users, the train may pass as it is, and it may be difficult to obtain the arrival and departure times. Therefore, the information corresponding to the arrival time and departure time for each station, which is the basic data of the technology described in Patent Document 1, cannot be obtained, and the technology described in Patent Document 1 cannot be used.

Therefore, an object of the present invention is to provide an operation state estimation system and an operation state estimation method that can estimate the current delay time with finer granularity.

According to the first aspect of the present invention, the following operating state estimation system is provided. That is, the operating state estimation system includes a processor. The processor (1) executes the passage time estimation model construction unit, and uses at least the route information, the timetable, and the position information of the mobile object based on the travel history, or the position of the stop of the mobile object and timetable information at stops, and information that can be obtained by linearly interpolating the information, and inputting information on the current position of the moving object in a unit common to the operation pattern Construct a passing time estimation model that estimates the elapsed time from the departure of the moving object according to The processor (2) executes the inference section to obtain the current delay time of the moving object from the elapsed time output from the passage time estimation model. The processor (3) executes the disclosure part to present to the user the operating state of the moving body obtained by executing the inference part.

According to the second aspect of the present invention, the following operation state estimation method is provided. That is, the operation state estimation method uses (1) at least the route information, the timetable, and the position information of the moving body based on the travel history, or the position of the stop of the moving body and the time at the stop By using at least table information and information that can be obtained by linearly interpolating the information, the movement of the moving object in accordance with the input of the current position information of the moving object in a unit common to the operation pattern constructing a passage time estimation model for estimating the elapsed time from the departure of the body, (2) obtaining the current delay time of the moving object from the passage time output from the passage time estimation model, and (3) obtaining the movement It presents the user with the running state of the body.

According to the present invention, it is possible to estimate the current delay time with fine granularity by using a passage time estimation model that estimates the elapsed time from departure for information specifying the position of an arbitrary mobile object.

Furthermore, if information for specifying the position of a moving vehicle in operation can be obtained, it is possible to predict the delay time of the moving vehicle even when accurate arrival and departure times cannot be obtained. . For example, even if it is difficult to obtain accurate arrival and departure times at stops, such as buses and trams, which may pass without passengers, the delay time of the moving object is can be predicted.

FIG. 2 is a diagram for explaining the system configuration and functional configuration in this embodiment; FIG. 2 is a diagram for explaining the hardware configuration of the system according to the embodiment; FIG. The figure which shows an example of flight data. The figure which shows an example of timetable data. The figure which shows an example of route data. The figure which shows an example of real-time position data. The figure which shows an example of passage time estimation model data. FIG. 4 is a diagram showing an example of estimated passage time and delay time data; The figure which shows an example of delay time data for every route number. The figure which shows an example of delay time prediction model training data. The figure which shows an example of delay time prediction model data. The figure which shows an example of the predicted delay time data classified by path|route order. The figure which shows an example of reliability data. 4 is a flowchart for explaining model construction processing; 4 is a flowchart for explaining the details of learning processing of a passage time estimation model; 4 is a flowchart for explaining standardization of estimated delay time sequences; FIG. 4 is a diagram for explaining a delay time prediction model; 4 is a flowchart for explaining inference processing; The figure which shows an example of an operation status display screen.

Hereinafter, representative modes for carrying out the present invention will be described with appropriate reference to the drawings. In this embodiment, an example in which a bus is used as a moving body will be described. However, the same applies to other moving bodies such as railways, streetcars, ships, and taxis. First, an outline is given.

The processing of the operation state estimation system in this embodiment is divided into a construction phase and an inference phase. The construction phase consists of static data such as flights, timetables, and routes, dynamic data such as locations, departure and arrival information, congestion conditions inside the vehicle, etc. obtained from the vehicle and environmental equipment during actual operation, and weather information. and surrounding information such as traffic information, etc., and constructing a passing time estimation model and a delay time prediction model used for estimating the operation state. In the inference phase, the above-mentioned static data, dynamic data of the vehicle in motion, and surrounding information are used to calculate an estimate of the current or future delay time and information on the reliability of the estimation result. and present the results to the user.

In the construction phase, the aforementioned static and dynamic data are first collected. Dynamic data is collected, for example, for one month at the time of initial construction. However, this period can be freely selected. Next, using static data and dynamic data, a passage time estimation model is constructed to estimate the elapsed time (sometimes referred to as passage time) from the time of the first train to a certain position on the route as 0. be. Here, when the same location is passed twice or more on the operation route, information that can specify the number of visits is taken into consideration. For the sake of simplicity of explanation, it is assumed that the aforementioned information is taken into account even when the position is simply written hereinafter.

The passage time estimation model is intuitively like a timetable for any position, and by using this model and real-time position information, it is possible to estimate the state with finer granularity.

Although it is possible to calculate the estimated delay time at each time using dynamic data and the transit time estimation model (that is, it is possible to calculate the estimated delay time at the time when the position information is acquired), Assuming that the estimated delay time for each flight at each time is called an estimated delay time string, the length of the estimated delay time string differs for each flight because the manner in which each flight is delayed differs depending on the day. Therefore, standardization (fixed length) is performed using a representative position on the route (predetermined position). Then, using information such as the standardized estimated delay time sequence, date and time, and the above-mentioned peripheral information, a delay time (called a predicted delay time) is calculated from the partially given estimated delay time sequence and other information. A delay time prediction model is constructed.

In the inference phase, dynamic data about the vehicle in motion is first collected. Then, the estimated delay time for the running vehicle is calculated using the collected dynamic data, static data, and passage time estimation model. Next, a predicted delay time is calculated using the estimated delay time, the peripheral information, and the delay time prediction model. In addition, a confidence measure is calculated based on the reconstruction error and the adjustment (mask) of the inputs used for the prediction. These pieces of information are then presented to the user as the operation status.

Next, the configuration of the operating state estimation system according to this embodiment will be described with reference to FIG. The operation state estimation system includes an operation state estimation server 10 that estimates the operation state, a vehicle 11 that transports passengers, an environmental facility 12 that measures the state of the vehicle with a sensor such as a camera, a traffic information service 13, and weather information. It comprises a service 14 and a user terminal 15 such as a smart phone or a personal computer (PC). A user 16 who operates the user terminal 15 may be a passenger of the vehicle 11 or may not be a passenger of the vehicle 11 .

The components of the operation state estimation system are interconnected via the WWW (World Wide Web). In this embodiment, the connection is made via WWW, but other communication means may be used. Moreover, the above components are examples, and the number of components may be increased or decreased. For example, the operation state estimation server 10 may be composed of two or more servers. Also, it is sufficient that the dynamic data described above can be collected, and the dynamic data may be obtained from either the vehicle 11 or the environmental facility 12, or from both. Moreover, when acquiring dynamic data from the vehicle 11, the environmental equipment 12 may be omitted.

Next, referring to Figures 1 and 2, the correspondence between functions and hardware will be described. A CPU (Central Processing Unit) 1H101 loads a program stored in a ROM (Read Only Memory) 1H102 or an external storage device 1H104 into a RAM (Random Access Memory) 1H103, and uses a communication I/F (Interface) 1H105, mouse, keyboard, etc. By controlling the external input device 1H106 represented by, the external output device 1H107 represented by the display etc., the collection unit 101 provided in the operation state estimation server 10 of the operation state estimation system, the passing time estimation model construction unit 102, the delay The temporal prediction model construction unit 103, the inference unit 104, the disclosure unit 105, and the data management unit 106 are implemented (function).

Next, the structure of each data will be explained with reference to the diagram. First, the flight data 1D1 collected by the operation state estimation server 10 from the traffic information service 13 and managed by the data management unit 106 will be described with reference to FIG. The flight data 1D1 is data representing basic information of the operated flight, and includes a flight ID (1D101), a flight name (1D102), a timetable ID (1D103), a group ID (1D104), and a route ID ( 1D105). A flight ID (1D101) is an identifier for specifying a flight. The flight name (1D102) is a name corresponding to the flight ID (1D101). Timetable ID (1D103) is an identifier for specifying a timetable. The group ID (1D104) is an identifier for ensuring that two flights have the same route and the same elapsed time in the timetable (elapsed time when the first departure time is time 0). be. In other words, it is an identifier for ensuring that there is commonality in operation patterns of moving bodies. A route ID (1D105) is an identifier for specifying a route.

Supplement the specific explanation of the data shown in FIG. "XY-LINE-OUT-W600" represents an outbound flight that departs at 6:00 on weekdays from station X to station Y, and operates according to the route specified by R0 and the timetable specified by ST10000. . "XY-LINE-OUT-H610" represents a holiday flight corresponding to "XY-LINE-OUT-W600". In this example, the route on the holiday is the same, but the departure time is different, so the group ID (1D104) is different. On the other hand, "XY-LINE-OUT-W700" represents an outbound flight from X station to Y station that departs one hour after "XY-LINE-OUT-W600". Since this flight has the same elapsed time as the route of "XY-LINE-OUT-W600" and the timetable, G10000, which has the same group ID (1D104) as "XY-LINE-OUT-W600", is registered. there is

Next, the timetable data 1D2 collected by the operation state estimation server 10 from the traffic information service 13 and managed by the data management unit 106 will be described with reference to FIG. The timetable data 1D2 is data representing a timetable, and includes a timetable ID (1D201), a timetable sequence (1D202), an arrival time (1D203), a departure time (1D204), and a stop ID (1D205). , provided. The timetable ID (1D201) corresponds to the timetable ID (1D103) of the flight data 1D1, and the timetable data is specified in combination with the timetable sequence (1D202). The timetable order (1D202) represents the order of stops of a flight, and represents the order of arrival from the smallest numerical value to the largest numerical value. Note that the numbers do not have to be consecutive. Arrival Time (1D203) and Departure Time (1D204) represent the arrival and departure times of the stop. In addition, when there is almost no stop time, the same time is registered for these two. Buses often stop at the same time, but trains often stop for one to several minutes. A stop ID (1D205) represents an identifier for specifying a stop on a certain flight.

Supplement the specific explanation of the data shown in FIG. This example includes the timetable of "XY-LINE-OUT-W600" described above, and indicates that the stop IDs are "S1, S2, S3, . . . ". It also indicates that the arrival and departure times are "6:00, 6:01, 6:04, ...".

Next, the route data 1D3 collected by the operation state estimation server 10 from the traffic information service 13 and managed by the data management unit 106 will be described with reference to FIG. The route data 1D3 represents a route that the vehicle travels, and includes a route ID (1D301), a route number (1D302), a route latitude (1D303), a route longitude (1D304), a stop ID (1D305), and a stop name (1D306). Note that the route data 1D3 has finer granularity than the timetable data 1D2. The route ID (1D301) corresponds to the route ID (1D105) of the flight data 1D1, and is an identifier for specifying the route data 1D3 together with the route number (1D302). The route number (1D302) represents the running order of each position on the route, and serial numbers such as 0, 1, 2, . . . are registered. Route Latitude (1D303) and Route Longitude (1D304) represent positions in a geographic coordinate system. A stop ID (1D305) and a stop name (1D306) represent an identifier and a name for specifying it when the position is a stop. N/A is registered when it is not a stop position (for example, when it is not a bus stop).

Supplement the specific explanation of the data shown in FIG. In this example, a route with a route ID (1D105) of R0 is shown, and position information and stop information are registered for each consecutive route number (ID302). In particular,

route numbers

0, 2, and 4 are positions of stops, and stop IDs (1D305) and stop names (1D306) are registered. On the other hand,

route numbers

1 and 3 are not stops, and N/A is registered in the stop ID (1D305) and stop name (1D306).

Next, the real-time position data 1D4 collected from the vehicle 11 or the environmental equipment 12 by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. The real-time position data 1D4 is data relating to the position acquired in real time, and is highly accurate data with respect to the measurement time. The real-time position data 1D4 includes a flight ID (1D401), date and time (1D402), vehicle latitude (1D403), vehicle longitude (1D404), and passed route number (1D405). Flight ID (1D401) corresponds to flight ID (1D101) of flight data 1D1, and real-time position data 1D4 is specified by flight ID (1D401) and date and time (1D402). The date and time (1D402) represents the measurement time of the vehicle position. Vehicle latitude (1D403) and vehicle longitude (1D404) represent positions in a geographic coordinate system. In addition to this, the passed route number (1D405) is registered to distinguish cases where the same position is passed multiple times as described above.

Supplement the specific explanation of the data shown in FIG. This example includes location data collected for flight "XY-LINE-OUT-W600", with four data values noted from 6:00 to 6:01:31. Also, the passed route number (1D405) is 0 at the initial stage and becomes 1 at the stage of the third data. This indicates that the first stop was passed before the third data.

Next, the passage time estimation model data 1D5 generated by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. The passage time estimation model data 1D5 includes a group ID (1D501), construction date (1D502), model parameters (1D503), and offset (1D504). The group ID (1D501) corresponds to the group ID (1D104) of the flight data 1D1. Two flights have the same route, and the elapsed time in the timetable (when the first departure time is set to time 0) Elapsed time) is an identifier for ensuring the same. Then, in this embodiment, a passage time estimation model is constructed for each group ID (1D501). Here, the construction date (1D502) is the date when the passage time estimation model was constructed. The model parameter (1D503) is a parameter for expressing the passage time estimation model. The offset (1D504) is a value representing the average difference between the built passage time estimation model and the timetable. In this embodiment, the unit is minutes. Note that the offset (1D504) is not an essential requirement and may be omitted.

Supplement the specific explanation of the data shown in FIG. In this example, the passage time estimation model is constructed on February 1, 2021 for the group ID "G10000". Also, the offset (1D504) is 0.3 minutes.

Next, the estimated passage time and delay time data 1D6 generated by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. Estimated passage time and delay time data 1D6 includes flight ID (1D601), date and time (1D602), vehicle latitude (1D603), vehicle longitude (1D604), passed route number (1D605), and estimated passage time ( 1D606) and a delay time (1D607). Flight ID (1D601) corresponds to flight ID (1D101) of flight data 1D1, and flight ID (1D601) and date and time (1D602) specify estimated transit time and delay time data 1D6. The date and time (1D602) represents the measurement time of the vehicle position. Vehicle Longitude (1D603) and Vehicle Longitude (1D604) represent positions in a geographic coordinate system. The passed route number (1D605) is registered to distinguish cases where the same position is passed multiple times. The estimated passage time (1D606) is an estimated value of the passage time calculated using the passage time estimation model described above, and is the elapsed time with the time of departure from the starting place set to 0. In this embodiment, the unit is minutes. The delay time (1D607) is an estimated delay time obtained from the difference between the date and time (1D602) and the estimated passage time (1D606). The delay time (1D607) can be considered to represent the estimated delay time at each time in the above outline.

Supplement the specific explanation of the data shown in FIG. In this example, the estimated transit time and delay time data 1D6 of the flight "XY-LINE-OUT-W600" is registered, and at the fourth position (latitude: 35.399571, longitude: 139.539084), The estimated passage time is 1.4 minutes when the time of departure from the place is 0, and the delay time is 0.13 minutes.

Next, the delay time data 1D7 for each route number generated by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. The route number delay time data 1D7 includes a flight ID (1D701), first departure date and time (1D702), route number (1D703), and delay time (1D704). The flight ID (1D701) corresponds to the flight ID (1D101) of the flight data 1D1, and the first flight date and time (1D702) and the route number (1D703) specify the route number delay time data 1D7. The first departure date and time (1D702) is the first departure date and time of the flight. The route number (1D703) corresponds to the route number (1D302) of the route data 1D3, and represents the running order of each position on the route. be done. Delay Time (1D704) is an estimate of the delay time at each location. Delay (1D704) can be considered to represent the normalized estimated delay in the above schematic.

Supplement the specific explanation of the data shown in FIG. In this example, delay time data 1D7 for each route number of flight "XY-LINE-OUT-W600" is registered, and the fourth position (route ID: R0, route number: 3, latitude: 35.399711, Longitude: 139.539513), the delay time is 0.12 minutes.

Next, with reference to FIG. 10, the delay information generated by combining the information collected from the traffic information service 13 and the weather information service 14 and the information generated by the operation state estimation server 10 and managed by the data management unit 106 is The temporal prediction model training data 1D8 will be explained. The delay prediction model training data 1D8 includes a group ID (1D801), a flight ID (1D802), the first departure date and time (1D803), weather (1D804), traffic volume (1D805), and a delay time sequence (1D806). , provided. The group ID (1D801) corresponds to the group ID (1D104) of the flight data 1D1, and specifies the delay prediction model training data 1D8 together with the flight ID (1D802) and first departure date and time (1D803). Flight ID (1D802) corresponds to flight ID (1D101) of flight data 1D1. The first departure date and time (1D803) is the first departure date and time of the flight. Weather (1D804) is collected from the weather information service 14 and in this embodiment registers one of four values associated with the flight: clear, cloudy, rain and snow. Note that the types of weather are not limited to four. Further, more detailed information such as precipitation probability, temperature, humidity, and wind speed may be registered, or may be registered with finer granularity. Traffic volume (1D805) is collected from the traffic information service 13 and represents road congestion. In this embodiment, an integer value from 0 to 100 is registered for the congestion status related to flights, and the congestion status is indicated by the magnitude of the integer value. In addition, you may register by finer granularity. The delay time sequence (1D806) is data obtained by grouping the route number delay time data 1D7 with group ID (1D801), flight ID (1D802), and first departure time (1D803).

Supplement the specific explanation of the data shown in FIG. In this example, the group ID is G10000, the flight ID is XY-LINE-OUT-W600, and the first departure date and time is 2021-03-01T06:00:00. ([0.00, 0.04, 0.10, 0.12, . . . ]) is registered.

Next, the delay time prediction model data 1D9 generated by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. The delay time prediction model data 1D9 includes a group ID (1D901), construction date (1D902), and model parameters (1D903). The group ID (1D901) corresponds to the group ID (1D104) of the flight data 1D1. Two flights have the same route, and the elapsed time in the timetable (when the first departure time is time 0) Elapsed time) is an identifier for ensuring the same. In this embodiment, a delay time prediction model is constructed for each group ID (1D901). The construction date (1D902) is the date when the delay time prediction model was constructed. A model parameter (1D903) is a parameter for expressing a delay time prediction model.

Supplement the specific explanation of the data shown in FIG. In this example, the delay time prediction model is constructed on February 1, 2021 for the group ID "G10000".

Next, with reference to FIG. 12, the predicted delay time data by route sequence 1D10 generated by the operation state estimation server 10 and managed by the data management unit 106 will be described. The route sequence predictive delay time data 1D10 includes a flight ID (1D1001), first departure date and time (1D1002), route number (1D1003), and predictive delay time (1D1004). The flight ID (1D1001) corresponds to the flight ID (1D101) of the flight data 1D1, and the delay time data 1D7 for each route number is specified by the first departure date (1D1002) and the route number (1D1003). The first departure date and time (1D1002) is the first departure date and time of the flight. The route number (1D1003) corresponds to the route number (1D302) of the route data 1D3, and represents the running order of each position on the route. be done. The predicted delay time (1D1004) is the estimated value of the delay time at each position calculated using the delay time prediction model. Unlike the delay time (1D704) of the delay time data 1D7 for each path number, it contains the predicted value of the future delay time.

Supplement the specific explanation of the data shown in FIG. In this example, route-ordered predicted delay time data 1D10 for flight "XY-LINE-OUT-W600" is registered, and the fourth position (route ID: R0, route number: 3, latitude: 35.399711 , longitude: 139.539513), the predicted delay time is 0.11 minutes.

Next, the reliability data 1D11 generated by the operation state estimation server 10 and managed by the data management unit 106 will be described with reference to FIG. The reliability data 1D11 includes a flight ID (1D1101), first departure date and time (1D1102), restoration error base reliability (1D1103), and mask base reliability (1D1104). Flight ID (1D1101) corresponds to flight ID (1D101) of flight data 1D1, and is an identifier for specifying reliability data 1D11. The restoration error base reliability (1D1103) is a reliability calculated from the delay time (1D704) of the delay time data for each route number 1D7 and the predicted delay time (1D1004) of the predicted delay time data for each route sequence 1D10. Mask-based reliability (1D1104) is reliability based on multiple predictions by the delay time prediction model.

Supplement the specific explanation of the data shown in FIG. In this example, the flight "XY-LINE-OUT-W600" that departed at 6:00 on March 1, 2021 has a reconstruction error base reliability (1D1103) of 88 and a mask base reliability (1D1104) of 89.

Next, the processing flow of the construction phase in this embodiment will be described using FIG.

First, the collection unit 101 of the operation state estimation server 10 collects flight data 1D1, timetable data 1D2, and route data 1D3 from the traffic information service 13, and stores them in the data management unit 106 (step 1F101). Note that flight data 1D1, timetable data 1D2, and route data 1D3 may be generated from data collected from the traffic information service 13 and stored.

Next, the collection unit 101 of the operation state estimation server 10 collects real-time position data 1D4 of the vehicle 11 from the vehicle 11, the environmental equipment 12, the traffic information service 13, etc. over a predetermined period. Estimated passage time and delay time data 1D6 is generated and stored in data management unit 106. FIG. Also, traffic volume data from the traffic information service 13 and weather data from the weather information service 14 are collected, and these data are stored in the data management unit 106 (step 1F102).

Next, the passage time estimation model construction unit 102 of the operation state estimation server 10 constructs a passage time estimation model using the flight data 1D1, the timetable data 1D2, and the real-time position data 1D4 for each group ID, and estimates the passage time. Model data 1D5 is generated and stored in the data management unit 106 (step 1F103). Note that this processing will be described later in detail.

Next, the delay time prediction model building unit 103 of the operation state estimation server 10 uses the built passage time estimation model data 1D5 and the position information (vehicle latitude, vehicle longitude, route number that has been passed) of the real-time position data 1D4, Calculate the estimated transit time (1D606) for each flight and each date and time. Then, the delay time prediction model construction unit 103 of the operation state estimation server 10 calculates the delay time (1D607) obtained as the difference from the date and time, generates the estimated passage time and delay time data 1D6, and sends it to the data management unit 106 Store (step 1F104).

Next, the delay time prediction model construction unit 103 of the operation state estimation server 10 uses the estimated passage time and delay time data 1D6 and the route data 1D3 to generate a fixed length delay time sequence for each group ID, Delay time data 1D7 for each path number is generated and stored in the data management unit 106 (step 1F105). Note that this processing will be described later in detail.

Next, the delay time prediction model construction unit 103 of the operation state estimation server 10 combines the route number delay time data 1D7 with the traffic volume, weather, and delay time sequence to generate delay time prediction model training data 1D8, Store in the data management unit 106 (step 1F106).

Finally, the delay time prediction model building unit 103 of the operation state estimation server 10 learns (builds) the delay time prediction model, generates delay time prediction model data 1D9, and stores it in the data management unit 106 (step 1F107). . The delay time prediction model will be explained later in detail.

Next, the processing of the passage time estimation model construction unit 102 of the operation state estimation server 10 in step 1F103 will be described in detail with reference to FIG. The passage time estimation model construction unit 102 implements a passage time estimation model construction function that constructs a passage time estimation model using machine learning or probability statistics.

First, flight data 1D1 and real-time position data 1D4 are combined by flight ID (step 1F201).

Next, based on the time of the first train in the timetable data 1D2, the transit time (the elapsed time from the time of the first train to 0) is calculated (step 1F202). The position for which the passage time is calculated can be any position on the route. Here, the passing time of the stop position may be calculated by referring to the departure time (1D204) of the timetable data 1D2, for example. Also, the passage time at a position different from the stop on the route may be calculated by, for example, linearly interpolating data on the position of the stop.

Next, for each group ID, a passing time estimation model is learned by quantile regression of passing times with respect to vehicle latitude, vehicle longitude, and passed route numbers (step 1F203). In this embodiment, the quantile for regression is 10 percent, but it can be changed to an appropriate value as appropriate. Quantile regression finds the elapsed time less affected by delay by using quantiles that are closer to the lower bound. As an example, by estimating the elapsed time using quantile regression on relatively small percentile values (e.g., the 10th percentile), it is expected that the elapsed time at the stop will not differ greatly from that calculated from the timetable. be done. In this embodiment, an example using quantile regression has been described, but with regard to the cumulative distribution function that performs distribution estimation with a generative model and integrates from the larger to the smaller, the point where 90% has passed It may be calculated as time.

Finally, the average difference between the passage time obtained from the passage time estimation model at each stop and the passage time calculated from the actual timetable is calculated as an offset. Then, the passage time estimation model data 1D5 is generated together with the above information and stored in the data management unit 106. FIG. By using this offset (1D504) as a correction value, as described above, it is possible to deal with constant delays caused by circumstances such as being unable to operate earlier than the timetable. In this embodiment, the same (single) offset is used in all sections, but for example, the offset is calculated for each stop, and the closer offset is used in the section between them, or two offsets are used. may be used.

The delay time prediction model construction unit 103 implements a delay time prediction model construction function that constructs a delay time prediction model that predicts the future delay time of mobile objects. Next, with reference to FIG. 16, the processing of the delay time prediction model construction unit 103 of the operation state estimation server 10 in step 1F105 will be described in detail. It should be noted that in this explanation, the processing related to a certain flight ID will be explained.

First, the flight data 1D1 and the route data 1D3 are combined by the route ID, and the route number, route latitude, and route longitude associated with the flight ID of interest are obtained (step 1F301). Note that these are fixed lengths for each flight ID in the processing flow of FIG.

Next, from the estimated passage time and delay time data 1D6, five pieces of position data are acquired for each route number described above, starting with the shortest route number difference within 2 (step 1F302). In addition, in this embodiment, the difference between route numbers is set to within 2, but other values may be used. Similarly, although the number of pieces of position data to be acquired is five, other values may be used. In FIG. 16, there are four cases.

　Finally, the average value of the five obtained data is calculated and taken as the delay time for the corresponding route number. By this procedure, delay time data 1D7 for each route number is generated and stored in the data management unit 106 (step 1F303).

With the above processing, the estimated delay time is converted to a fixed length for the flight ID, or more precisely for the flight running on the same route (that is, the estimated delay time at the time is the estimated delay time at a predetermined position). ), resulting in a fixed-length sequence of delay times. This facilitates modeling by statistics and machine learning. More specifically, a structure is realized that accepts vectors of the same length for input and output of a delay time prediction model, which will be described later. In addition, it can be used to calculate reliability, which will be described later. In addition, using the above method (step 1F301 to step 1F303 method), a fixed length delay time for the position on the route may be obtained, or using a known data search method such as the k-neighborhood method, A fixed length delay time for a position on the path may be obtained.

Next, the delay time prediction model will be described in detail with reference to FIG. In this embodiment, a delay time prediction model is constructed for each group ID using a transformer, which is a type of neural network.

First, I will explain the overall input and output. The input is the partially masked delay sequence and weather and traffic, and the output is the (unmasked) delay sequence. By using a model trained on tasks with such inputs and outputs, it becomes possible to predict future delay times given known delay times and peripheral information during operation.

Next, I will explain the mask. When learning, a different mask is randomly generated each time for each mini-batch, and training is performed. At that time, a route number on the route is selected and the values after that route number are set to 0.

Next, the embedding process will be explained. First, a convolutional layer is applied to the vector representing the delay time sequence. It is then concat with the weather and traffic vectors to obtain a vector with both delay time and perimeter information.

Next, the Position Encoding process will be explained. In Position Encoding, in order to cope with the loss of position information about vector elements in the self-attention layer, a vector created with a different sine wave (appropriate sine wave) is added to the input vector. Common. In this embodiment, in addition to that, by adding a predetermined vector as an offset for each month, day of the week, weekday or holiday, and time period information, month, day of the week, weekday or holiday, and time period information are embedded. be

Next, the processing of Self-attention will be explained. Basically, a general multi-head self-attention layer and a layer in which Point Feed Forward is stacked three times is applied. However, since the input is masked, the value (the product of the query and the key) immediately before passing through the softmax function is masked correspondingly. Although the number of stacks is 3, it may be more or less. In addition, weather and traffic volume information may be appropriately added to queries, keys, and Point Feed Forward residual blocks.

Next, the processing of Linear Transform will be explained. Here the fully bonded layer is applied twice. Although the total number is 2, it may be larger or smaller. Also, if the delay time sequence is very long and difficult to restore, a deconvolution layer or the like may be used.

Finally, the loss function calculation process will be explained. Here, the restored estimated delay time and the MSE (Mean Squared Error) of the original delay time sequence are calculated to estimate the loss of the model. Learning of the model progresses by updating the parameters to minimize this loss using the back propagation method or the like. Although MSE is used in this embodiment, MAE (Mean Absolute Error), hinge loss, etc. may be used.

Next, the processing flow of the inference phase in this embodiment will be described with reference to FIG. It is assumed that the construction phase described in FIG. 14 has been executed prior to this process (that is, the passage time estimation model and the delay time prediction model have been constructed).

In the inference phase, first, the collection unit 101 of the running state estimation server 10 collects real-time position data 1D4 from the vehicle 11, the environmental equipment 12, the traffic information service 13, etc. for the running vehicle 11. Also, traffic volume data from the traffic information service 13 and weather data from the weather information service 14 are collected, and these data are stored in the data management unit 106 (step 1F401).

Next, the inference unit 104 of the operation state estimation server 10 uses the transit time estimation model built in the construction phase to estimate the transit time (1D606) and the delay time (1D607) from departure to the present for each flight. is calculated (step 1F402). Thus, the inference unit 104 realizes an inference function of calculating the current estimated passage time and delay time.

Next, the inference unit 104 of the operation state estimation server 10 uses the estimated passage time and delay time data 1D6 and the route data 1D3 to generate a delay time sequence for each group ID (step 1F403). Here, handling of data is facilitated by making the length fixed even in the case of prediction. This procedure is basically the same as described above with reference to FIG. However, if there is a location where the condition that the route number is within 2 is not satisfied, the length is not fixed.

Next, the inference unit 104 of the operation state estimation server 10 uses the delay time prediction model constructed for each group ID in the construction phase to determine the weather and traffic volume collected in step 1F401 and the delay obtained in step 1F403. From the time series, the predicted delay time for each flight ID, first departure date and time, and route number is calculated to generate route sequence predicted delay time data 1D10, which is stored in the data management unit 106 (step 1F404). As described above, the delay time sequence for each group ID may not have a fixed length. In this case, the value is set to Complement and fix the length.

Next, the inference unit 104 of the operation state estimation server 10 generates a delay time sequence for each group ID of a flight on a certain date and time, and a delay time sequence of the predicted delay time data 1D10 for each route order that the group and date and time are covered (that is, , the delay time sequence of the predicted delay times (1D1004)), and the average value of the absolute values of the differences (that is, the average value of the restoration error) is calculated. Then, with respect to a predetermined constant D, the result of 100×(D−(the average value))/D is used as the reconstruction error base reliability. Since this reconstruction error base reliability uses the delay time sequence, it becomes the average value of the absolute value of the vector difference in a short period at the beginning of the run, but as the run progresses, the absolute value of the vector difference in a longer period. average value. However, if the reconstruction error base reliability is less than zero, the reconstruction error base reliability is replaced with zero. Here, in this embodiment, the constant D is assumed to be the 90th percentile value of the average value of restoration errors during learning for the same delay time prediction model, but other values may be used (step 1F405).

Next, the inference unit 104 of the operation state estimation server 10 creates a delay time sequence by masking the delay time sequence for each group ID by 5 minutes, 10 minutes, and 15 minutes from the current time. That is, by using masks of different sizes, a delay time sequence is created by masking 5 minutes, 10 minutes, and 15 minutes before the current time. Then, the inference unit 104 of the operation state estimation server 10 predicts the predicted delay time by route order of four patterns including no mask (no mask, 5 minute mask, 10 minute mask, 15 minute mask), Calculate the standard deviation of the predicted delay time by route order for each pattern. Then, the value of 100×(S−(the standard deviation))/S becomes the mask base reliability. However, if the mask-based reliability is less than 0, it is replaced with 0, and if the mask-based reliability exceeds 100, it is replaced with 100. Here, in the present embodiment, the 90th percentile value of the standard deviation at the time of learning for the same delay time prediction model is used as the constant S, but other values may be used (step 1F406).

Next, the inference unit 104 of the operation state estimation server 10 generates reliability data 1D11 based on the obtained restoration error-based reliability and mask-based reliability, and the predicted delay by route order generated up to the above step It is stored in the data management unit 106 along with time. Then, the disclosure unit 105 of the operation state estimation server 10 notifies the user terminal 15 of the update information. The disclosure unit 105 implements a presentation function of presenting the operation status of the mobile object to the user. Finally, the user terminal 15 reads the latest predicted delay time by route order and reliability data 1D11 based on the notification, and presents these data to the user 16 as the operation status (step 1F407).

Next, the user interface will be explained with reference to the diagram. FIG. 19 is an example of an operation status display screen 1G1 for a certain flight presented to the user 16 by the user terminal 15. As shown in FIG. In addition, the user 16 can operate the user terminal 15 to specify the flight desired to be displayed.

The operation status display screen 1G1 includes a flight name label (1G101), an arrival time table (1G102), a reliability table (1G103), and a delay status graph (1G104). In the flight number label (1G101), the flight name of the flight specified by the user 16 is displayed. The arrival time table (1G102) displays the arrival times and delay times obtained by using the timetable information and the delay time prediction model for the flights specified by the user 16 described above. In the example of FIG. 19, it is displayed that the train was scheduled to arrive at stop G at 6:30 if the timetable had been followed, but was delayed by 4 minutes at 6:34. It also indicates that the train is scheduled to arrive at Stop F at 6:40 if the timetable is followed, but is scheduled to be delayed by 5 minutes at 6:45. The reliability table (1G103) displays reconstruction error-based reliability and mask-based reliability obtained using the delay time prediction model. In this embodiment, both reliability levels are displayed, but it is also possible to display either one or none. The delay status graph (1G104) displays the predicted delay times by route order obtained using the delay time prediction model. Here, the horizontal axis is information including the position of the stop on the route, and the vertical axis is information indicating the predicted delay time. The position of the stop can be indicated as appropriate, and as shown in FIG. 19, for example, a symbol may be displayed at the position of the stop. In addition, stop names corresponding to route numbers (corresponding to stop A, stop G, etc. in FIG. 19) may be displayed at the positions of the stops. Also, the current position is visualized as a dashed line (1G104a). In addition, the delay time (delay time obtained without using the delay time prediction model) obtained from the traveled position data is visualized as a solid line (1G104b). Also, the predicted (restored) delay time for the traveled area is visualized as a narrow dashed line (1G104c). On the other hand, the predicted delay time for the untraveled region is visualized as a wide dashed line (1G104d). The reconstruction error base reliability is a value calculated based on the average of the absolute values of the differences between the solid line (1G104b) and the narrow dashed line (1G104c). Also, the delay status graph (1G104) may be generated by an appropriate method. The delay status graph (1G104) may be generated, for example, by a regression curve based on the delay time series.

As described above, according to the operation state estimation system of the present embodiment, by using a passage time estimation model that estimates the elapsed time from departure for information specifying the position of an arbitrary moving body, It is possible to estimate the current delay time and predict the future delay time using the results of the model.

It should be noted that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the gist of the present invention. Also, various inventions can be formed by appropriate combinations of the plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiments.

The subject of processing of the operation state estimation server 10 is a processor, and an example of a processor is a CPU, but other semiconductor devices (eg, GPU) may be used as long as the subject executes predetermined processing.

The transit time estimation model and delay time prediction model can be generated by methods based on probability statistics and machine learning using acquired or generated data.

In the above description, when generating the delay time prediction model, the delay time prediction model construction unit 103 calculates the estimated passage time and the delay time. A delay time may be used.

In this specification, a "stop" is a place where a moving object stops, for example, a bus stop for a bus, or a station for a train.

10 Operation state estimation server 11 Vehicle 12 Environmental equipment 13 Traffic information service 14 Weather information service 15 User terminal 16 User 101 Collection unit 102 Passing time estimation model construction unit 103 Delay time prediction model construction unit 104 Inference unit 105 Disclosure unit 106 Data management unit

Claims

with a processor
The processor
(1) Execute the passage time estimation model construction part,
Using at least the route information, the timetable, and the location information of the moving object based on the travel history, or linearly interpolating the information with information on the location of the stop of the moving object and the timetable at the stop and at least using information that can be obtained from the above, in units that are common to operation patterns, in response to input of information on the current position of the moving object, estimating the elapsed time from the departure of the moving object. Build a time estimation model,
(2) Execute the reasoning part,
Obtaining the current delay time of the moving object from the elapsed time output from the passage time estimation model;
(3) Execute the public part,
presenting to the user the operating state of the moving object obtained by executing the inference unit;
An operation state estimation system characterized by:
The operating state estimation system according to claim 1,
The processor
Execute the delay time prediction model construction part,
Using the delay time information obtained from the elapsed time output from the passing time estimation model, the moving object is determined in accordance with the input of the delay time information of the moving object in a unit common to the operation pattern. build a delay time prediction model that predicts the future delay time of
executing the inference unit,
Obtaining the future delay time of the mobile from the information of the delay time of the mobile,
An operation state estimation system characterized by:
The operating state estimation system according to claim 1,
The processor
By executing the passage time estimation model construction unit,
Combining vehicles that have the same route and different starting times but the elapsed time from the starting time to each stop is the same, data is collected in common units. to build a transit time estimation model,
An operation state estimation system characterized by:
The operating state estimation system according to claim 2,
The processor
By executing the delay time prediction model construction unit,
Building the delay time prediction model further using at least one or more of weather information, traffic information, month information, day of the week information, weekday or holiday information, and time zone information,
executing the inference unit,
Determining the future delay time of the mobile from the information on the delay time of the mobile and the at least one or more information used to construct the delay time prediction model,
An operation state estimation system characterized by:
The operating state estimation system according to claim 1,
The processor
By executing the passage time estimation model construction unit,
constructing the passage time estimation model using quantile regression of the elapsed time with respect to the position of the moving object;
An operation state estimation system characterized by:
The operating state estimation system according to claim 2,
The processor
By executing the delay time prediction model construction unit,
The delay time information obtained from the elapsed time output from the passage time estimation model is converted into a delay time sequence that is fixed-length delay time information, and the delay time prediction model is constructed.
An operation state estimation system characterized by:
The operating state estimation system according to claim 6,
The processor
By executing the delay time prediction model construction unit,
Constructing the delay time prediction model by including data obtained by masking a portion of the fixed-length delay time sequence,
executing the inference unit,
A delay time sequence having the same length as the fixed-length delay time sequence is generated by masking the delay time from the current time to the end point with respect to the delay time sequence from the first train to the current time. Using the delay time sequence, restore the delay time from the first train to the present and predict the delay time from the present to the end point,
An operation state estimation system characterized by:
The operating state estimation system according to claim 6,
The processor
executing the inference unit,
Predicting the delay time at a predetermined position on the route based on the fixed-length delay time sequence,
executing the publishing part,
presenting to the user the predicted delay time of a predetermined position on the route;
An operation state estimation system characterized by:
The operating state estimation system according to claim 7,
The processor
executing the inference unit,
Restoration error-based reliability based on the restoration error of the delay time sequence of the moving body from the first train to the present and the delay time sequence of the same date and time restored from the delay time prediction model, and / or the advance from the present to the past calculating a mask-based reliability based on the delay time predicted by the delay time prediction model using the delay time sequence in which the defined range is masked;
executing the publishing part,
Presenting the calculated reconstruction error-based reliability and/or the calculated mask-based reliability to a user together with a future delay time;
An operation state estimation system characterized by:
using a computer,
(1) Using at least route information, a timetable, and location information of a moving object based on travel history, or information on a location of a stop of a moving object and a timetable at the stop, and the above information Elapsed time from the departure of the moving body according to the input of the information on the current position of the moving body, in units common to the operation patterns, at least using information that can be obtained by linear interpolation. Build a transit time estimation model to estimate,
(2) obtaining the current delay time of the moving body from the elapsed time output from the passage time estimation model;
(3) presenting the user with the requested operation state of the moving object;
An operation state estimation method characterized by: