Patent Application
20240359716
The present disclosure relates to an automatic train operation assistance device, an automatic train operation assistance system, and an automatic train operation assistance method for assisting an automatic train operation system.
Trains normally arrive at and depart from stations according to specified schedules. However, due to factors such as crowding, the stop period of time may exceed a specified time period, which leads to departure delays. When one train is delayed, the subsequent train may rapidly approach the delayed train, and may suddenly decelerate or suddenly stop. The delay of one train may affect the operation of another train. Therefore, to efficiently operate trains, the future positions of the trains are predicted. For example, Patent Literature 1 discloses a technique to predict the stop period of time at a station or travel time of a train by statistical processing based on actual value data on past stop time.
In general, the states of crowding on the vehicles of a train are not uniform across the vehicles. Crowding is expected on vehicles close to stairs, a ticket gate, or the like at a station that is the place of departure or destination of passengers. However, the above conventional technique does not take into consideration differences in crowding between the vehicles. There is thus a problem in that when too many passengers get on and off a vehicle that stops near stairs, a ticket gate, or the like, the time taken to get on and off the vehicle may exceed a predicted stop time at the station.
The present disclosure has been made in view of the above, and an object thereof is to provide an automatic train operation assistance device capable of improving the accuracy of prediction of the future operating conditions of a train.
To solve the above problems and achieve an object, an automatic train operation assistance device according to the present disclosure, includes: a data extraction unit to acquire operation data including an operation of a train and load factors of vehicles of the train from a data collection device to collect equipment data on the train; a feature extraction unit to extract, using the operation data, operation data on an inter-station including a travel time taken by the train to travel from a first station to a second station and a stop time for which the train is stopped at the first station or the second station, as features; and an operation prediction unit to construct a prediction model using the features, and predict future operating conditions of the train including a future position of the train, using the prediction model and current features.
The automatic train operation assistance device of the present disclosure has the effect of being able to improve the accuracy of prediction of the future operating conditions of a train.
Hereinafter, an automatic train operation assistance device, an automatic train operation assistance system, and an automatic train operation assistance method according to embodiments of the present disclosure will be described in detail with reference to the drawings.
The data collection device 20 is an onboard device that collects equipment data on the train 10. Specifically, from onboard equipment (not illustrated) installed on the train 10, the data collection device 20 collects equipment data on the operating conditions of the train 10, the operating states of the onboard equipment, measurement values measured by the onboard equipment, etc. For example, the equipment data include, but are not limited to, the acceleration, the speed, door opening and closing information, travel position information, the load factors of the vehicles, etc. of the train 10. The data collection device 20 collects the equipment data in units of seconds, in units of some hundreds of milliseconds, or in units of some tens of milliseconds. The data collection device 20 includes an equipment data database 21 and stores the collected equipment data in the equipment data database 21.
The automatic train operation assistance device 30 acquires operation data including the operation of the train 10 and the load factors of the vehicles of the train 10 from the equipment data database 21 of the data collection device 20. The automatic train operation assistance device 30 extracts features from the acquired operation data, performs machine learning using the features, and predicts the future operating conditions of the train 10 including the future position of the train 10.
The automatic train operation system 40 performs the automatic operation of the train 10, based on the future operating conditions of the train 10 including the future position of the train 10 predicted by the automatic train operation assistance device 30. The automatic train operation system 40 is, for example, automatic train operation (ATO).
The configuration and operation of the automatic train operation assistance device 30 will be described in detail. As illustrated in
The data extraction unit 31 acquires the operation data including the operation of the train 10 and the load factors of the vehicles of the train 10 from the equipment data database 21 included in the data collection device 20 that collects the equipment data on the train 10 (step S1). The data extraction unit 31 holds, that is, stores the acquired operation data in the operation and load factor database 32 (step S2). Since the period of collection of the equipment data by the data collection device 20 is a short period, the data extraction unit 31 may acquire the operation data at intervals longer than the period of collection of the equipment data by the data collection device 20, with calculation load etc. in the automatic train operation assistance device 30 taken into consideration.
The operation and load factor database 32 is a database in which the operation data extracted by the data extraction unit 31 is stored.
The feature extraction unit 33 extracts, as features, operation data on each station, that is, on each inter-station, including a travel time taken by the train 10 to travel from a first station to a second station and a stop time for which the train 10 is stopped at the first station or the second station, using the operation data stored in the operation and load factor database 32 (step S3). The feature extraction unit 33 holds, that is, stores the extracted features in the feature database 34 (step S4).
The feature database 34 is a database in which the features extracted by the feature extraction unit 33 are stored.
A description is given of the relationships between operation data that is equipment data extracted by the data extraction unit 31 from the equipment data database 21 of the data collection device 20, and features extracted by the feature extraction unit 33.
In
In
The features for one station, that is, each inter-station includes a departure time, a day of the week, a stop time at the station, a station identifier (ID), destination information, load factor variations during getting on and off of the vehicles, load factor averages during travel of the vehicles, etc. An inter-station travel time may be included in the features because the travel time may be changed when the operation of the train 10 is delayed. When the load factors during travel are measured by the weights of the vehicles, measured values may exhibit fluctuations. Thus, the load factors during travel as features use average values. When there are no branch lines on the route on which the train 10 operates, the destination information may be replaced with an operation direction. The feature extraction unit 33 may narrow down these features, for example, by determining the degree of importance of each feature, instead of treating the features equally. The load factor variations during getting on and off are indexes representing the degrees of change in the load factors associated with the getting on and off of passengers when the doors of the train 10 are open, and are effective for the prediction of a delay time and an excess of stop time due to the intensity of getting on and off, the load factors, etc.
For example, as a load factor variation representing the degree of load factor variation, the magnitude of the value of a standard deviation illustrated in
In addition, as features, the difference between the load factor at the time of door opening and the minimum value of the load factor variations, the difference between the load factor at the time of door closing and the minimum value of the load factor variations, the time from the door opening to the minimum value of the load factor variations, which is the time taken to get off the vehicle, the time from the minimum value of the load factor variations until the load factor at the time of door closing is reached, which is the time taken to get on the vehicle, etc. may be used. Furthermore, an arrival time, a delay time, a date, a day of the week, the weather or information such as whether wipers are operated, a row of holidays, door opening and closing time points, door opening and closing time periods, the number of times of door opening and closing, a next station ID, an operation number, a train number, vehicle types, the presence or absence of an event, the scale of an event, etc. may be added to the features. A row of holidays is, for example, information on how many days after or how many days before a holiday, or how many consecutive holidays. Vehicle types are, for example, the presence or absence of a bathroom, a women-only vehicle, etc. Information on an event is an effective feature for a station that is near an event venue and is used less by passengers on days without events, for example. For a route on which trains 10 such as express trains other than local trains travel, information on train types indicating a local train, an express train, etc. may be added to the features. By using these pieces of information, the automatic train operation assistance device 30 can construct a more detailed prediction model. When the number of dimensions of the features is too large, the feature extraction unit 33 may use only features useful for prediction, for example, by determining the degrees of importance of the features.
As described above, the load factors of the vehicles of the train 10 included in the operation data include, for each vehicle of the train 10, the load factor when the train 10 is stopped at a station and passengers are getting on and off, and the load factor when the train 10 is traveling. The feature extraction unit 33 extracts features including at least one of a maximum load factor, a minimum load factor, an average load factor, or a standard deviation in load factor when passengers are getting on and off The maximum load factor, the minimum load factor, the average load factor, and the standard deviation indicate load factor variations during a passengers' getting on and off time when the train 10 is stopped at a station. The feature extraction unit 33 may extract features including a station ID, or may exclude a station ID and extract features for the same items for each station ID, that is, for each station.
The operation time-series data creation unit 35 creates operation time-series data by combining features on two or more different inter-stations stored in the feature database 34 in chronological order (step S5). That is, the operation time-series data creation unit 35 converts features extracted by the feature extraction unit 33 and stored in the feature database 34 into features with the time series taken into account.
In the example of
The operation prediction unit 36 performs machine learning using features to construct a prediction model. Specifically, the operation prediction unit 36 constructs a prediction model by machine learning with the operation time-series data created by the operation time-series data creation unit 35 as explanatory variables and the future operating conditions of the train 10 as objective variables (step S6). The operation prediction unit 36 predicts the future operating conditions of the train 10 including the future position of the train 10, using the constructed prediction model and current features (step S7).
In the present embodiment, the operation prediction unit 36 associates explanatory variables that are operation data on each inter-station with objective variables that are future operating conditions to perform learning, and constructs a prediction model by regression using supervised learning among machine learning algorithms. At the time of operation of the train 10, the operation prediction unit 36 inputs current data, that is, current features to the prediction model learned in advance, thereby predicting a delay time, a stop time at a station, the load factors, etc. as the operating conditions of the train 10 in the future, for example, several stations ahead. The operation prediction unit 36 may use the features on, of the vehicles included in the train 10, the vehicle with the longest time of getting on and off, the vehicle with the largest number of passengers who get off the vehicle, the vehicle with the largest number of passengers who get on the vehicle, or the vehicle with the greatest standard deviation, to predict the future operating conditions of the train 10 including that vehicle. The operation prediction unit 36 outputs the predicted future operating conditions of the train 10 to the automatic train operation system 40 installed on the train 10 (step S8).
Next, a hardware configuration of the automatic train operation assistance device 30 according to the first embodiment will be described. In the automatic train operation assistance device 30, the operation and load factor database 32 and the feature database 34 are memory. The data extraction unit 31, the feature extraction unit 33, the operation time-series data creation unit 35, and the operation prediction unit 36 are implemented by processing circuitry. The processing circuitry may be memory storing a program and a processor that executes the program stored in the memory, or may be dedicated hardware. The processing circuitry is also referred to as a control circuit.
The program can be said to be a program that causes the automatic train operation assistance device 30 to perform: a first step in which the data extraction unit 31 acquires operation data including the operation of the train 10 and the load factors of the vehicles of the train 10 from the data collection device 20 that collects the equipment data on the train 10; a second step in which the feature extraction unit 33 extracts, using the operation data, as features, operation data on an inter-station including a travel time taken by the train 10 to travel from a first station to a second station and a stop time for which the train 10 is stopped at the first station or the second station; a third step in which the operation prediction unit 36 performs machine learning using the features to construct a prediction model; and a fourth step in which the operation prediction unit 36 predicts, using the prediction model and current features, the future operating conditions of the train 10 including the future position of the train 10.
Here, the processor 91 is, for example, a central processing unit (CPU), a processing device, an arithmetic device, a microprocessor, a microcomputer, a digital signal processor (DSP), or the like. The memory 92 corresponds, for example, to nonvolatile or volatile semiconductor memory such as random-access memory (RAM), read-only memory (ROM), flash memory, an erasable programmable ROM (EPROM), or an electrically EPROM (EEPROM) (registered trademark), or a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, a digital versatile disc (DVD), or the like.
As described above, according to the present embodiment, the automatic train operation assistance device 30 extracts data on the operation of the train 10 and the load factors of the vehicles as operation data from the data collection device 20, which is an onboard device collecting equipment data on the train 10, and converts the operation data into time-series features on each inter-station, to predict the future operating conditions of the train 10 using machine learning. Consequently, the automatic train operation assistance device 30 can grasp a future delay, crowding, etc. of the train 10 in advance, and thus can assist the automatic train operation system 40 to improve the efficiency of the operation of the train 10.
The automatic train operation assistance device 30 extracts the intensities of getting on and off of the vehicles due to crowding, using the equipment data on the train 10, and thus can extract detailed passengers' conditions, and can improve prediction accuracy using machine learning, as compared with cases where getting on and off conditions of the vehicles are not taken into account. For example, when there are multiple trains 10 similar in the load factors of the vehicles but different in the degrees of getting on and off of passengers of the vehicles, that is, the standard deviations of the vehicles, the automatic train operation assistance devices 30 can predict different delay times as future delays. The automatic train operation assistance device 30 can improve the accuracy of predicting the future operating conditions of the train 10.
Further, the automatic train operation assistance device 30 predicts the future operating conditions of the train 10, so that, in order to prevent a departure delay of the train 10 due to an excess of stop time or the like at a station several stations ahead, the automatic train operation system 40 can use the predicted future operating conditions to generate a travel pattern that allows early arrival at the station, and guide passengers to vehicles with less getting on and off at the station, for example. Furthermore, the automatic train operation assistance device 30 predicts the future operating conditions of the train 10 for several stations ahead, so that if the automatic train operation assistance device 30 takes time before prediction, and the automatic train operation system 40 cannot deal with the latest delay, the automatic train operation system 40 can take measures to recover the delay within several stations ahead.
The first embodiment has described the case where the automatic train operation assistance device 30 is installed on the train 10. A second embodiment describes a case where an automatic train operation assistance device is installed in an operation control apparatus, as a case where an automatic train operation assistance device is installed outside a train. The same reference numerals are assigned to the same components as those of the first embodiment without detailed explanations.
On each train 10a, the communication unit 11 communicates with the communication unit 61 of the operation control apparatus 60a. Specifically, the communication unit 11 acquires the future operating conditions of the train 10a predicted by an operation prediction unit 36a of the automatic train operation assistance device 30a from the communication unit 61 of the operation control apparatus 60a, and outputs the acquired future operating conditions to the automatic train operation system 40. The communication unit 11 outputs equipment data acquired from onboard equipment (not illustrated) installed on the train 10a to the communication unit 61 of the operation control apparatus 60a. The communication unit 11 may directly acquire the equipment data from each piece of onboard equipment, or may acquire the equipment data of each piece of onboard equipment from an onboard device (not illustrated) that collects the equipment data from each piece of onboard equipment.
The operation control apparatus 60a performs wireless communications with a plurality of trains 10a and controls the operations of the plurality of trains 10a. Although there are two trains 10a in the example of
In the operation control apparatus 60a, the communication unit 61 communicates with the communication units 11 of the trains 10a. Specifically, the communication unit 61 acquires the equipment data from the onboard equipment (not illustrated) installed on the trains 10a from the communication units 11 of the trains 10a, and outputs the acquired equipment data to the data collection device 63. The communication unit 61 outputs the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a to the communication units 11 of the trains 10a. The communication unit 61 and the communication units 11 of the trains 10a communicate by wireless communication, but are not limited to a particular communication method for wireless communication.
The data collection device 63 collects the equipment data on the trains 10a via the communication unit 61. The data collection device 63 includes an equipment data database 64 and stores the collected equipment data in the equipment data database 64. Unlike the data collection device 20 of the first embodiment, the data collection device 63 collects the equipment data on the plurality of trains 10a. The items of the equipment data collected from each train 10a by the data collection device 63 are the same as the items of the equipment data collected from the onboard equipment of the train 10 by the data collection device 20 of the first embodiment.
The train order acquisition unit 62 acquires information on the operation order of the plurality of trains 10a stored in the equipment data database 64 of the data collection device 63. The operation control apparatus 60a typically controls operation schedules of the trains 10a. Thus, the train order acquisition unit 62 can acquire the information on the operation order of the plurality of trains 10a from a database (not illustrated) or the like of the operation control apparatus 60a.
The automatic train operation assistance device 30a is obtained by replacing the data extraction unit 31 and the operation prediction unit 36 with a data extraction unit 31a and the operation prediction unit 36a in the automatic train operation assistance device 30 of the first embodiment illustrated in
The data extraction unit 31a acquires the operation data on the plurality of trains 10a from the equipment data database 64 included in the data collection device 63 installed in the operation control apparatus 60a. The data extraction unit 31a acquires the information on the operation order of the plurality of trains 10a from the train order acquisition unit 62. The data extraction unit 31a holds, that is, stores the operation data and the information on the operation order of the plurality of trains 10a acquired in the operation and load factor database 32. When the automatic train operation assistance device 30a predicts the future operating conditions of the plurality of trains 10a without taking into account the trains 10a placed before and after such as the preceding train and the following train, the data extraction unit 31a may not acquire the information on the operation order of the plurality of trains 10a from the train order acquisition unit 62.
The operation prediction unit 36a constructs a prediction model by machine learning with the operation time-series data created by the operation time-series data creation unit 35 as explanatory variables and the future operating conditions of each train 10a as objective variables. The operation prediction unit 36a predicts the future operating conditions of each train 10a including the future position of the train 10a, using the constructed prediction model and the current features. The operation prediction unit 36a outputs the predicted future operating conditions of the trains 10a to the trains 10a via the communication unit 61. At this time, for the plurality of trains 10a, the operation prediction unit 36a predicts the future operating conditions of each train 10a. For example, when the operation prediction unit 36a has predicted for the preceding train of a certain train for which prediction is performed that departure from the next station will be delayed, the operation prediction unit 36a can predict for the train for which prediction is performed that arrival at or departure from the next station will also be delayed. Upon acquiring this future operating condition, the automatic train operation system 40 installed on the train 10a that is the train for which prediction has been performed can make, for travel to the next station, a travel pattern to reduce travel speed so that the arrival time is later than the scheduled time. Consequently, the automatic train operation system 40 of the train for which prediction has been performed can cause the train 10a to travel without wasteful energy consumption.
The feature extraction unit 33 and the operation time-series data creation unit 35 may perform the same operations as in the first embodiment for each of the plurality of trains 10a, or may add operation data on the preceding train of a train for which prediction is performed to the features described in the first embodiment since operation data on the plurality of trains 10a can be handled. Delays of the trains 10a that may occur include, in addition to a delay due to a passengers' getting on and off time, a delay of the following train caused by a delay of the preceding train. Therefore, information on the preceding train is useful for the prediction of delay. Thus, the feature extraction unit 33 may extract, as features, first features on a train for which prediction is performed and, in addition, second features on the preceding train of the train for which prediction is performed. The operation time-series data creation unit 35 may create operation time-series data using the features consisting of the first features on the train for which prediction is performed and the second features on the preceding train extracted by the feature extraction unit 33.
When the operation prediction unit 36a predicts, taking the plurality of trains 10a into consideration, the future operating conditions of each train 10a, the operation prediction unit 36a may predict the future operating conditions of a train for which prediction is performed, using, as features, features on the preceding train in addition to features on the train for which prediction is performed, or using the results of prediction of the future operating conditions of another train 10 such as the preceding train, or using both.
The operation of the automatic train operation assistance device 30a is different from that of the automatic train operation assistance device 30 of the first embodiment in that prediction is performed on the plurality of trains 10a, but the flow of the operation itself is the same as the flow of the operation of the automatic train operation assistance device 30 of the first embodiment illustrated in the flowchart of
Here, assume a case where delays occur in wireless communications between the communication unit 61 of the operation control apparatus 60a and the communication units 11 of the trains 10a, and wireless communications are not performed in real time. In this case, it may take time for the future operating conditions of the trains 10a, which are predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a installed in the operation control apparatus 60a, to reach the automatic train operation systems 40 of the trains 10a, and the automatic train operation systems 40 may not be able to effectively use the future operating conditions of the trains 10a. Therefore, when it is known in advance that wireless communications between the communication unit 61 of the operation control apparatus 60a and the communication units 11 of the trains 10a are not performed in real time, the process up to the construction of a prediction model may be performed in the operation control apparatus, and prediction may be performed on a train using the constructed prediction model.
The automatic train operation assistance device 30b is obtained by replacing the operation prediction unit 36a with a learning unit 37 in the automatic train operation assistance device 30a illustrated in
Each automatic train operation assistance device 30c is obtained by replacing the operation prediction unit 36 with the prediction unit 38 in the automatic train operation assistance device 30 illustrated in
In the following description, the automatic train operation assistance device 30b is sometimes referred to as a first automatic train operation assistance device. In the automatic train operation assistance device 30b, the data extraction unit 31a is sometimes referred to as a first data extraction unit, the operation and load factor database 32 as a first operation and load factor database, the feature extraction unit 33 as a first feature extraction unit, the feature database 34 as a first feature database, and the operation time-series data creation unit 35 as a first operation time-series data creation unit. Each automatic train operation assistance device 30c is sometimes referred to as a second automatic train operation assistance device. In each automatic train operation assistance device 30c, the data extraction unit 31 is sometimes referred to as a second data extraction unit, the operation and load factor database 32 as a second operation and load factor database, the feature extraction unit 33 as a second feature extraction unit, the feature database 34 as a second feature database, and the operation time-series data creation unit 35 as a second operation time-series data creation unit.
With the configuration as illustrated in
The automatic train operation assistance devices 30a, 30b, and 30c may not include the operation time-series data creation units 35, like the automatic train operation assistance device 30 of the first embodiment, and the operation prediction unit 36a, the learning unit 37, or the prediction units 38 may directly acquire features from the feature database 34 and perform the above-described operation. The same applies to the following embodiments.
A hardware configuration of the automatic train operation assistance device 30a will be described. In the automatic train operation assistance device 30a, the operation and load factor database 32 and the feature database 34 are memory. The data extraction unit 31a, the feature extraction unit 33, the operation time-series data creation unit 35, and the operation prediction unit 36a are implemented by processing circuitry. The processing circuitry may be memory storing a program and a processor that executes the program stored in the memory, or may be dedicated hardware.
A hardware configuration of the automatic train operation assistance device 30b will be described. In the automatic train operation assistance device 30b, the operation and load factor database 32 and the feature database 34 are memory. The data extraction unit 31a, the feature extraction unit 33, the operation time-series data creation unit 35, and the learning unit 37 are implemented by processing circuitry. The processing circuitry may be memory storing a program and a processor that executes the program stored in the memory, or may be dedicated hardware.
A hardware configuration of each automatic train operation assistance device 30c will be described. In each automatic train operation assistance device 30c, the operation and load factor database 32 and the feature database 34 are memory. The data extraction unit 31, the feature extraction unit 33, the operation time-series data creation unit 35, and the prediction unit 38 are implemented by processing circuitry. The processing circuitry may be memory storing a program and a processor that executes the program stored in the memory, or may be dedicated hardware.
As described above, according to the present embodiment, the automatic train operation assistance device 30a performs machine learning that handles a massive amount of data and prediction in the operation control apparatus 60a. Consequently, the trains 10a do not perform machine learning that handles a massive amount of data and prediction, as compared with the train 10 of the first embodiment, and thus do not require a high-spec device. Further, the automatic train operation assistance device 30a can improve prediction accuracy as compared with the automatic train operation assistance device 30 of the first embodiment, by adding data on the preceding train to features at the time of prediction on a train for which prediction is performed. Furthermore, since the automatic train operation assistance device 30a performs predictions on the plurality of trains 10a, when a delay, an excess of stop time at a station, or the like has been predicted on a certain preceding train, the automatic train operation system 40 of the following train that has acquired information on this prediction can appropriately control travel, for example, by generating a low-speed, energy-saving travel pattern so as not to be too close to the preceding train, to prevent sudden deceleration, a stop, or the like between stations.
Moreover, when real-time wireless communications cannot be performed between the operation control apparatus 60b and the trains 10c, the automatic train operation assistance device 30b performs machine learning that handles a massive amount of data in the operation control apparatus 60b, and the automatic train operation assistance devices 30c perform prediction on the trains 10c. Even in this case, the trains 10c do not perform machine learning that handles a massive amount of data as compared with the train 10 of the first embodiment, and thus do not require a high-spec device. Note that in the configuration as illustrated in
In the present embodiment, when delays occur in wireless communications between the communication unit 61 of the operation control apparatus 60a and the communication units 11 of the trains 10a, and wireless communications are not performed in real time, the automatic train operation assistance system 70 consisting of the automatic train operation assistance devices 30b and 30c is used, but the application of the automatic train operation assistance system 70 is not limited thereto. For example, iwhen transmission and reception of the equipment data etc. are not performed between the operation control apparatus 60b and the trains 10c during the operation of the trains 10c, the operation control apparatus 60b and the trains 10c may exchange, after the completion of the operation of the trains 10c, data such as the equipment data and the prediction models via storage media, or may exchange data such as the equipment data and the prediction models by wired communication. The same applies to the following embodiments.
The first embodiment has described the case where the automatic train operation assistance device 30 is installed on the train 10. A third embodiment describes a case where an automatic train operation assistance device is installed in a cloud, as a case where an automatic train operation assistance device is installed outside a train. The cloud is a use mode that allows a service to be provided on the Internet, and may be referred to as a cloud server or the like. In the present embodiment, the word “cloud” is used, and the same applies to the following embodiments. The same reference numerals are assigned to the same components as those of the first and second embodiments without detailed explanations.
On the trains 10a, the communication units 11 communicate with the communication unit 51 of the cloud 50a. Specifically, the communication units 11 acquire the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a from the communication unit 51 of the cloud 50a, and output the acquired future operating conditions to the automatic train operation systems 40. The communication units 11 output equipment data acquired from onboard equipment (not illustrated) installed on the trains 10a to the communication unit 51 of the cloud 50a.
The cloud 50a performs wireless communications with a plurality of trains 10a. In the example of
In the cloud 50a, the communication unit 51 communicates with the communication units 11 of the trains 10a. Specifically, the communication unit 51 acquires the equipment data from the onboard equipment (not illustrated) installed on the trains 10a from the communication units 11 of the trains 10a, and outputs the acquired equipment data to the data collection device 53. The communication unit 51 outputs the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a to the communication units 11 of the trains 10a. The communication unit 51 and the communication units 11 of the trains 10a communicate by wireless communication, but are not limited to a particular communication method for wireless communication.
The data collection device 53 collects the equipment data on the trains 10a via the communication unit 51. The data collection device 53 includes an equipment data database 54 and stores the collected equipment data in the equipment data database 54. The data collection device 53 has the same function as the data collection device 63 of the second embodiment.
The train order acquisition unit 52 acquires information on the operation order of the plurality of trains 10a stored in the equipment data database 54 of the data collection device 53. The train order acquisition unit 52 can acquire the information on the operation order of the plurality of trains 10a from, for example, an operation control apparatus (not illustrated) that controls operation schedules of the trains 10a.
The automatic train operation assistance device 30a has the same configuration as the automatic train operation assistance device 30a of the second embodiment illustrated in
Here, as in the second embodiment, assume a case where delays occur in wireless communications between the communication unit 51 of the cloud 50a and the communication units 11 of the trains 10a, and wireless communications are not performed in real time. In this case, it may take time for the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a installed in the cloud 50a to reach the automatic train operation systems 40 of the trains 10a, and the automatic train operation systems 40 may not be able to effectively use the future operating conditions of the trains 10a. Therefore, when it is known in advance that wireless communications between the communication unit 51 of the cloud 50a and the communication units 11 of the trains 10a are not performed in real time, the process up to the construction of a prediction model may be performed in the cloud, and prediction may be performed on a train using the constructed prediction model.
The automatic train operation assistance device 30b has the same configuration as the automatic train operation assistance device 30b of the second embodiment illustrated in
As described above, according to the present embodiment, the automatic train operation assistance device 30a performs machine learning that handles a massive amount of data and prediction in the cloud 50a. Consequently, the same effects as those in the second embodiment can be obtained.
The second embodiment has described the case where the automatic train operation assistance device 30a is installed in the operation control apparatus 60a. The third embodiment has described the case where the automatic train operation assistance device 30a is installed in the cloud 50a. A fourth embodiment describes a case where both an operation control apparatus and a cloud are used, and the automatic train operation assistance device 30a is installed in the operation control apparatus, and the data collection device 53 is installed in the cloud. The same reference numerals are assigned to the same components as those of the first to third embodiments without detailed explanations.
The trains 10a perform wireless communications with both the cloud 50d and the operation control apparatus 60d. Specifically, the communication units 11 of the trains 10a acquire the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a from the communication unit 61 of the operation control apparatus 60d, and output the acquired future operating conditions to the automatic train operation systems 40. The communication units 11 output equipment data acquired from onboard equipment (not illustrated) installed on the trains 10a to the communication unit 51 of the cloud 50d.
In the cloud 50d, the communication unit 51 communicates with the communication units 11 of the trains 10a. Specifically, the communication unit 51 acquires the equipment data from the onboard equipment (not illustrated) installed on the trains 10a from the communication units 11 of the trains 10a, and outputs the acquired equipment data to the data collection device 53. The communication unit 51 and the communication units 11 of the trains 10a communicate by wireless communication, but are not limited to a particular communication method for wireless communication. In the cloud 50d, the configuration and operation of the data collection device 53 are as described above.
In the operation control apparatus 60d, the communication unit 61 communicates with the communication units 11 of the trains 10a. Specifically, the communication unit 61 outputs the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a to the communication units 11 of the trains 10a. The communication unit 61 and the communication units 11 of the trains 10a communicate by wireless communication, but are not limited to a particular communication method for wireless communication. In the operation control apparatus 60d, the configurations and operations of the automatic train operation assistance device 30a and the train order acquisition unit 62 are as described above.
Here, as in the second and third embodiments, assume a case where delays occur in wireless communications between the communication unit 51 of the cloud 50d and the communication unit 61 of the operation control apparatus 60d, and the communication units 11 of the trains 10a, and wireless communications are not performed in real time. In this case, it may take time for the future operating conditions of the trains 10a predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a installed in the operation control apparatus 60d to reach the automatic train operation systems 40 of the trains 10a, and the automatic train operation systems 40 may not be able to effectively use the future operating conditions of the trains 10a. Therefore, when it is known in advance that wireless communications between the communication unit 51 of the cloud 50d and the communication unit 61 of the operation control apparatus 60d, and the communication units 11 of the trains 10a are not performed in real time, the process up to the construction of a prediction model may be performed in the operation control apparatus, and prediction may be performed on a train using the constructed prediction model.
The automatic train operation assistance device 30b has the same configuration as the automatic train operation assistance device 30b of the second embodiment illustrated in
As described above, according to the present embodiment, the automatic train operation assistance device 30a performs machine learning that handles a massive amount of data and prediction in the operation control apparatus 60e. Consequently, the same effects as those in the second embodiment can be obtained.
A fifth embodiment describes a case where the data collection device 63 installed in the operation control apparatus 60a in the second embodiment is installed on a train as the data collection device 20. The same reference numerals are assigned to the same components as those of the first to fourth embodiments without detailed explanations.
In the fifth embodiment, the communication units 11 of the trains 10f output, of equipment data acquired from onboard equipment (not illustrated) installed on the trains 10f, only equipment data to be used in the automatic train operation assistance device 30a, that is, to be used for learning and prediction in the automatic train operation assistance device 30a, to the operation control apparatus 60f. Consequently, as compared with the case in the second embodiment, the communication units 11 can reduce the amounts of data to be transmitted to the operation control apparatus 60f and can improve communication efficiency. The operations of the other components are as described above.
Here, as in the second embodiment, assume a case where delays occur in wireless communications between the communication unit 61 of the operation control apparatus 60f and the communication units 11 of the trains 10f, and wireless communications are not performed in real time. In this case, it may take time for the future operating conditions of the trains 10f predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a installed in the operation control apparatus 60f to reach the automatic train operation systems 40 of the trains 10f, and the automatic train operation systems 40 may not be able to effectively use the future operating conditions of the trains 10f. Therefore, when it is known in advance that wireless communications between the communication unit 61 of the operation control apparatus 60f and the communication units 11 of the trains 10f are not performed in real time, the process up to the construction of a prediction model may be performed in the operation control apparatus, and prediction may be performed on a train using the constructed prediction model.
The automatic train operation assistance device 30b has the same configuration as the automatic train operation assistance device 30b of the second embodiment illustrated in
As described above, according to the present embodiment, the automatic train operation assistance device 30a performs machine learning that handles a massive amount of data and prediction in the operation control apparatus 60f. Further, when real-time wireless communications cannot be performed between the operation control apparatus 60g and the trains 10g, the automatic train operation assistance device 30b performs machine learning that handles a massive amount of data in the operation control apparatus 60g, and the automatic train operation assistance devices 30c perform prediction on the trains 10g. Consequently, the same effects as those in the second embodiment can be obtained. Furthermore, in the present embodiment, the amounts of data to be transmitted from the trains 10f to the operation control apparatus 60f can be reduced, and the amounts of data to be transmitted from the trains 10g to the operation control apparatus 60g can be reduced. Consequently, the communication units 11 can improve communication efficiency as compared with the case in the second embodiment.
A sixth embodiment describes a case where the data collection device 53 installed in the cloud 50a in the third embodiment is installed on a train as the data collection device 20. The same reference numerals are assigned to the same components as those of the first to fifth embodiments without detailed explanations.
In the sixth embodiment, the communication units 11 of the trains 10f output, of equipment data acquired from onboard equipment (not illustrated) installed on the trains 10f, only equipment data to be used in the automatic train operation assistance device 30a, that is, to be used for learning and prediction in the automatic train operation assistance device 30a, to the cloud 50f. Consequently, as compared with the case in the third embodiment, the communication units 11 can reduce the amounts of data to be transmitted to the cloud 50f and can improve communication efficiency. The operations of the other components are as described above.
Here, as in the third embodiment, assume a case where delays occur in wireless communications between the communication unit 51 of the cloud 50f and the communication units 11 of the trains 10f, and wireless communications are not performed in real time. In this case, it may take time for the future operating conditions of the trains 10f predicted by the operation prediction unit 36a of the automatic train operation assistance device 30a installed in the cloud 50f to reach the automatic train operation systems 40 of the trains 10f, and the automatic train operation systems 40 may not be able to effectively use the future operating conditions of the trains 10f. Therefore, when it is known in advance that wireless communications between the communication unit 51 of the cloud 50f and the communication units 11 of the trains 10f are not performed in real time, the process up to the construction of a prediction model may be performed in the cloud, and prediction may be performed on a train using the constructed prediction model.
The automatic train operation assistance device 30b has the same configuration as the automatic train operation assistance device 30b of the third embodiment illustrated in
As described above, according to the present embodiment, the automatic train operation assistance device 30a performs machine learning that handles a massive amount of data and prediction in the cloud 50f. Further, when real-time wireless communications cannot be performed between the cloud 50g and the trains 10g, the automatic train operation assistance device 30b performs machine learning that handles a massive amount of data in the cloud 50g, and the automatic train operation assistance devices 30c perform prediction on the trains 10g. Consequently, the same effects as those in the third embodiment can be obtained. Furthermore, in the present embodiment, the amounts of data to be transmitted from the trains 10f to the cloud 50f can be reduced, and the amounts of data to be transmitted from the trains 10g to the cloud 50g can be reduced. Consequently, the communication units 11 can improve communication efficiency as compared with the case in the third embodiment.
The configurations described in the above embodiments illustrate an example, and can be combined with another known art. The embodiments can be combined with each other. The configurations can be partly omitted or changed without departing from the gist.
10, 10a, 10c, 10f, 10g train; 11, 51, 61 communication unit; 20, 53, 63 data collection device; 21, 54, 64 equipment data database; 30, 30a, 30b, 30c automatic train operation assistance device; 31, 31a data extraction unit; 32 operation and load factor database; 33 feature extraction unit; 34 feature database; 35 operation time-series data creation unit; 36, 36a operation prediction unit; 37 learning unit; 38 prediction unit; 40 automatic train operation system; 50a, 50b, 50d, 50f, 50g cloud; 52, 62 train order acquisition unit; 60a, 60b, 60d, 60e, 60f, 60g operation control apparatus; 70 automatic train operation assistance system.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2021/041572 | 11/11/2021 | WO |
