Method And Apparatus for Real-Time Scheduling for Resolving Delay of Train Operation

Abstract

The present disclosure provides a real-time scheduling method and apparatus for resolving train operation delays. According to one aspect of the present disclosure, a real-time scheduling method for resolving train operation delays comprises: a process of collecting passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of said train; a process of using said passenger information and said operation information to calculate a number of passengers boarding and alighting from at least one preceding train that has stopped first at a scheduled dwell station of said train, and generating a first data set including said number of passengers boarding and alighting; a process of estimating a number of passengers boarding and alighting from said train at the scheduled dwell station using said first data set; and a process of estimating a required dwell duration of said train at the scheduled dwell station using said second data set including said number of passengers boarding and alighting from said train using said first data set; The present invention provides a real-time scheduling method characterized by including a schedule generation process that updates the train operation schedule by comparing the required stopping time with a preset dwell duration.

Description

Technology Field

[0001] The present disclosure relates to a real-time scheduling technology for estimating the number of passengers boarding and alighting a train and resolving operational delays. Background Technology

[0002] The following description merely provides background information related to the embodiments of the present disclosure and does not constitute prior art.

[0003] In high-density urban railway lines in downtown areas, the increase in passengers during peak hours leads to congestion on platforms and inside trains, and there are problems with delays in boarding and alighting at stations. Delays in boarding and alighting affect not only the delay of the train itself but also the operation of following trains, becoming a fundamental cause of chain delays.

[0004] If a child train is unable to depart from a station due to a departure delay, the following train becomes unable to enter the station. If the following train operates according to its existing schedule without prior knowledge of the child train's delay, it faces the problem of having to wait while stopped in the inter-station section. Consequently, the following train repeatedly accelerates and decelerates in this section, leading to a deterioration in train service quality and a decrease in the train's scheduled speed.

[0005] To resolve cascading delays of the following and preceding trains, technology is required to detect delays in the preceding train in advance and secure additional stopping time at the departure station. The problem to be solved

[0006] According to one aspect of the present disclosure, the main objective is to provide a technology for managing the schedule of a subsequent train in real time by predicting the passenger demand and required stopping time of the train at the scheduled arrival station in advance using a machine learning model. means of solving the problem

[0007] According to one embodiment of the present disclosure, a real-time scheduling method for resolving train operation delays comprises: a process of collecting passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of said train; a process of using said passenger information and said operation information to calculate a number of passengers boarding and alighting from at least one preceding train that has stopped first at said train's scheduled dwell station, and generating a first data set including said number of passengers boarding and alighting; a process of estimating a number of passengers boarding and alighting from said train at said station using said first data set; and a process of estimating a required dwell duration of said train at said station using said second data set including said number of passengers boarding and alighting from said train at said station using said first data set; The present invention provides a real-time scheduling method characterized by including a schedule generation process that updates the train operation schedule by comparing the required stopping time with a preset dwell duration.

[0008] According to another embodiment of the present disclosure, a method for learning a dwell duration estimation model for calculating a required dwell time, which is the time a train must stop, performed by a server, is provided, comprising: a process of calculating the passenger occupancy of the train, the boarding and alighting quantity, which is the number of passengers entering and exiting between the platform of the planned station and the train, and the record dwell duration, which is the actual time the train stopped at the platform of the planned station, based on the planned station where the train intends to enter; a process of generating a learning data set based on all or part of the passenger occupancy, the boarding and alighting quantity, the record dwell duration, and previously stored station information; and a process of training the dwell duration estimation model based on the learning data set.

[0009] According to another embodiment of the present disclosure, a real-time scheduling device for resolving train operation delays comprises: a data collection unit that collects passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of said train; a dataset generation unit that generates a first dataset including the preceding passenger number, which is the number of passengers boarding and alighting from at least one preceding train that has stopped first at the scheduled dwell station of said train, using the passenger information and the operation information; a passenger number estimation unit that estimates the passenger number, which is the number of passengers boarding and alighting from said train at the scheduled dwell station, using the first dataset; and a dwell time estimation unit that estimates the required dwell duration of said train at the scheduled dwell station using the second dataset including the passenger number. The present invention provides a real-time scheduling device characterized by including a schedule generation unit that updates the train operation schedule by comparing the required stopping time with a preset dwell duration.

[0010] According to another embodiment of the present disclosure, a server for training a dwell duration estimation model for calculating a required dwell time, which is the time a train must stop, comprises: a data calculation unit that calculates the passenger occupancy of the train, the boarding and alighting quantity, which is the number of passengers entering and exiting between the platform of the scheduled station and the train, and the record dwell duration, which is the actual time the train stopped at the platform of the scheduled station, based on the scheduled station where the train intends to enter; a dataset generation unit that generates a learning dataset based on all or part of the passenger occupancy, the boarding and alighting quantity, the record dwell duration, and previously stored station information; and a learning unit that trains the dwell duration estimation model based on the learning dataset.

[0011] According to another embodiment of the present disclosure, a computer program stored on a computer-readable recording medium is provided to execute each process included in a real-time scheduling method. Effects of the invention

[0012] According to one embodiment of the present disclosure, the schedule of a subsequent train is adjusted in real time based on the required stopping time estimated using a machine learning model, thereby minimizing the cascading delay of train operations.

[0013] According to another embodiment of the present disclosure, additional stopping time is secured at the stopping station to reduce congestion at the stopping station and reduce delays of subsequent trains. Brief explanation of the drawing

[0014] FIG. 1 is a conceptual diagram illustrating a real-time scheduling system for resolving train operation delays according to one embodiment of the present disclosure. FIG. 2 is a block diagram illustrating each configuration included in a real-time scheduling device for resolving train operation delays according to one embodiment of the present disclosure. FIG. 3 is an exemplary diagram illustrating the number of passengers boarding the train and the number of passengers alighting from the train at each station in a real-time scheduling method according to one embodiment of the present disclosure. FIG. 4 is an example diagram illustrating the boarding and alighting of a preceding train and a child train when entering a station scheduled to stop in a real-time scheduling method according to one embodiment of the present disclosure. FIG. 5 is a block diagram for explaining each configuration included in a server according to one embodiment of the present disclosure. FIG. 6 is a flowchart for explaining each process included in a real-time scheduling method according to one embodiment of the present disclosure. FIG. 7 is a flowchart for explaining each process included in the first dataset generation process in a real-time scheduling method according to one embodiment of the present disclosure. FIG. 8 is a flowchart for explaining each process included in the self-boarding / alighting vehicle estimation process in a real-time scheduling method according to one embodiment of the present disclosure. FIG. 9 is a flowchart for explaining each process included in the learning method of a stopping time estimation model according to one embodiment of the present disclosure. Specific details for implementing the invention

[0015] Hereinafter, some embodiments of the present invention will be described in detail with reference to the exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the embodiments of the present invention, if it is determined that a detailed description of related known components or functions could obscure the essence of the present invention, such detailed description is omitted.

[0016] In addition, terms such as first, second, A, B, (a), (b), etc. may be used when describing the components of the present invention. These terms are intended merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by these terms. Throughout the specification, when a part is described as 'comprising' or 'equipped' with a certain component, unless specifically stated otherwise, this means that it does not exclude other components but may include additional components. Furthermore, terms such as '…part' or 'module' described in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0017] The present disclosure provides a real-time scheduling technology for resolving train operation delays. Specifically, the real-time scheduling device calculates the number of passengers boarding and alighting of a preceding train and estimates the number of passengers boarding and alighting of a child train and the required dwell duration based on machine learning. The real-time scheduling device estimates the number of passengers boarding and alighting and the required dwell duration in advance before entering a scheduled dwell station. Accordingly, the real-time scheduling technology according to the present disclosure can determine the platform congestion of the scheduled dwell station in advance and update the operation schedule of the following train. The real-time scheduling device enters the scheduled dwell station and stops for the estimated required dwell duration, allowing passengers on the platform to board and alight. Accordingly, since the number of passengers boarding and alighting of the following train is reduced, the real-time scheduling technology according to the present disclosure can reduce the platform congestion of the scheduled dwell station.

[0018] In the present disclosure, a station scheduled for stopping refers to the station where the child train is scheduled to enter and stop first at the current time among a plurality of stations where the child train is scheduled to stop on the operation schedule. In the present disclosure, an initial station refers to the first station on the train line where the child train begins operation. In the present disclosure, a prior station refers to the station where the child train stopped immediately before the station scheduled for stopping.

[0019] In the present disclosure, a preceding train refers to at least one train that precedes a child train. In the present disclosure, a trailing train refers to at least one train that trails a child train.

[0020] In the present disclosure, the preceding passenger volume refers to the number of passengers boarding or alighting from the preceding train at the station scheduled for stopping. In the present disclosure, the self-boarding / alighting volume refers to the number of passengers predicted to board or alight from the self-train at the station scheduled for stopping. In the present disclosure, the required stopping time refers to the time the self-train must stop at the platform of the station scheduled for stopping in order to resolve train delays.

[0021] The detailed description set forth below, together with the accompanying drawings, is intended to describe exemplary embodiments of the present disclosure and is not intended to represent the only embodiment in which the present disclosure may be practiced.

[0022] FIG. 1 is a conceptual diagram illustrating a real-time scheduling system for resolving train operation delays according to one embodiment of the present disclosure.

[0023] A real-time scheduling system (10) according to one embodiment of the present disclosure includes a train (100), a first preceding train (102), a second preceding train (104), a following train (106), and a server (108).

[0024] In the present disclosure, the child train (100) refers to a train that serves as a reference for the real-time scheduling system (10). A real-time scheduling device included in the child train (100) calculates the number of passengers boarding and alighting ahead of at least one preceding train that precedes the child train (100). The real-time scheduling device estimates the number of passengers boarding and alighting and the required stopping time using a machine learning model. Here, the machine learning model includes a pre-learned passenger count calculation model and a stopping time estimation model. Since the real-time scheduling device estimates the number of passengers boarding and alighting and the required stopping time of the child train (100) based on the preceding passenger, it can determine in real-time whether congestion is occurring in a specific section of the route.

[0025] In one embodiment, the preceding train includes a first preceding train (102) and a second preceding train (104). Here, the first preceding train (102) refers to a preceding train that departs from the station scheduled for stopping after stopping at the station scheduled for stopping just before the train enters the station scheduled for stopping. The second preceding train (104) refers to a preceding train that departs from the station scheduled for stopping after stopping at the station scheduled for stopping just before the first preceding train (102) enters the station scheduled for stopping.

[0026] A trailing train (106) refers to at least one train that follows the child train (100). A real-time scheduling device updates the operation schedule of the child train and / or at least one trailing train (106) based on the required stopping time. The updated operation schedule is transmitted to each trailing train (106) using wireless communication. A real-time scheduling device included in the trailing train (106) estimates the number of passengers boarding and alighting at the scheduled stop station and the required stopping time before entering the scheduled stop station. In other words, since the real-time scheduling device included in the trailing train (106) can perform the same operation as the real-time scheduling device included in the child train (100), it can determine whether there is a train delay in real time and resolve the train delay.

[0027] The server (108) trains a machine learning model used by a real-time scheduling device using training data. The server (108) retrains a stopping time estimation model using an accumulated training dataset. The server (108) transmits the previously trained machine learning model to at least one train running on the operating route. Since the server (108) periodically retrains the machine learning model using an accumulated training dataset, the accuracy of the real-time scheduling device included in the train is improved.

[0028] FIG. 2 is a block diagram illustrating each configuration included in a real-time scheduling device for resolving train operation delays according to one embodiment of the present disclosure.

[0029] A real-time scheduling device (20) according to one embodiment of the present disclosure includes, in whole or in part, a data collection unit (200), a data set generation unit (202), a boarding and alighting quantity estimation unit (204), a dwell duration estimation unit (206), and a schedule generation unit (208). The real-time scheduling device (20) illustrated in FIG. 2 is according to one embodiment of the present disclosure, and not all blocks illustrated in FIG. 2 are essential components, and some blocks included in the real-time scheduling device (20) in other embodiments may be added, changed, or deleted.

[0030] Hereinafter, each component included in the real-time scheduling device (20) will be described with reference to FIG. 2.

[0031] The data collection unit (200) collects passenger information and train operation information. Here, passenger information refers to information regarding passengers entering or exiting at least one station. In one embodiment, passenger information includes transportation card tag information of passengers tagging in or out of at least one station. In another embodiment, passenger information may refer to the number of passengers waiting on the platform of the station. Here, the number of passengers on the platform is 5G communication (5) between the terminal carried by the passenger and the train th It can be calculated based on generation communications. The number of passengers at the platform may also be calculated based on Ultra Wide-Band (UWB) communications between the tag device carried by the passenger and the train. The method of calculating the number of passengers waiting at the platform is not limited to this embodiment. Operation information includes all or part of the position of at least one train, the arrival time of the train for at least one station, and the departure time of the train. In one embodiment, the data collection unit (200) may collect data by performing wireless communication with at least one of the Automatic Train Supervision (ATS), the station server, and other trains. Passenger information and / or operation information may be transmitted using Vehicle-to-Vehicle (V2V) communication and / or Vehicle-to-Infrastructure (V2I) communication. Specific methods for performing communication between trains or between a train and a ground device are known in the art, so further detailed description is omitted.

[0032] The dataset generation unit (202) calculates the number of passengers boarding and alighting from at least one preceding train using passenger information and operation information, and generates a first dataset including the number of passengers boarding and alighting. In one embodiment, the first dataset includes the number of passengers boarding and alighting from the preceding train, the number of passengers boarding and alighting from the train. Here, the number of passengers boarding and alighting from the preceding train refers to the number of passengers (passenger occupancy) who board the preceding train when the preceding train enters the station scheduled to stop. The number of passengers boarding and alighting from the train ( ) refers to the number of passengers who have boarded the train (100) when the train (100) enters the station scheduled for stopping. In this embodiment, it is assumed that the preceding trains considered by the dataset generation unit (202) to generate the first dataset include the first preceding train (102) and the second preceding train (104). Here, the first dataset is the number of passengers who have boarded the first preceding train (102) ), the number of passengers on the preceding train (104) ), the number of passengers on the train (100) ), the preceding boarding / alighting vehicle of the first preceding train (102) ( ) and the preceding boarding / alighting vehicle of the second preceding train (104) ...includes ). In another embodiment, the dataset generation unit (202) may consider information regarding other preceding trains other than the first preceding train (102) and the second preceding train (104) defined in this embodiment.

[0033] The dataset generation unit (202) estimates the number of passengers boarding per station from the number of passengers entering each station and estimates the number of passengers alighting per station from the number of passengers exiting each station in order to calculate the number of passengers boarding before or after. Here, the number of passengers entering the station or the number of passengers exiting the station can be calculated using passenger information. In one embodiment, the number of passengers entering or exiting the station can be calculated using transportation card tag information collected from the ticket gate of the station. A specific method for the dataset generation unit (202) to estimate the number of passengers boarding per station or the number of passengers alighting per station from the number of passengers entering or exiting the station is described later in FIG. 5. In another embodiment, the server (108) may estimate the number of passengers boarding per station or the number of passengers alighting per station from the number of passengers entering the station. The dataset generation unit (202) calculates the total boarding quantity by accumulating the number of passengers boarding from the departure station to the previous station, and calculates the total alighting quantity by accumulating the number of passengers alighting from the departure station to the previous station. The dataset generation unit (202) calculates the value obtained by subtracting the total alighting quantity from the total boarding quantity as the number of passengers alighting before or after.

[0034] FIG. 3 is an exemplary diagram illustrating the number of passengers boarding the train and the number of passengers alighting from the train at each station in a real-time scheduling method according to one embodiment of the present disclosure.

[0035] Referring to Fig. 3a, at the platform of the departure station (STA1), the train (TR x The number of passengers boarding ( ) and the number of passengers alighting from the train ( ) is illustrated. Referring to Fig. 3b, the number of passengers boarding the train at the platform of the second stop (STA2) ( ) and the number of passengers alighting from the train ( ) is illustrated. Referring to Fig. 3c, the previous station (STA j-1 The number of passengers boarding the train at platform ) ) and the number of passengers alighting from the train ( ) is illustrated. The dataset generation unit (202) is the number of passengers at the departure station ( ) or the previous station's ridership ( By accumulating ) the total number of passengers ( Calculate ) and the number of passengers alighting at the departure station ( ) or the number of passengers alighting at the previous station ( By accumulating ) the total number of vehicles alighting ( ) calculates. The dataset generation unit (202) calculates the value obtained by subtracting the total number of passengers alighting from the total number of passengers alighting when entering the scheduled stop station ( It is calculated as ). Using the method illustrated in FIG. 3, the dataset generation unit (202) calculates the number of passengers in the preceding train of the first preceding train ( ), number of passengers on the preceding train of the second preceding train ( ) and the number of passengers on the train ( Produces ).

[0036] The dataset generation unit (202) estimates the number of passengers boarding from the number of passengers entering the station scheduled for stopping and estimates the number of passengers alighting from the number of passengers exiting the station scheduled for stopping in order to calculate the number of passengers boarding or alighting in advance. A specific method for estimating the number of passengers boarding or alighting in advance from the number of passengers entering or exiting the station scheduled for stopping is described later in FIG. 5. The dataset generation unit (202) calculates the value obtained by adding the number of passengers boarding and the number of passengers alighting in advance as the number of passengers boarding ( and Calculate as ).

[0037] FIG. 4 is an example diagram illustrating the boarding and alighting of a preceding train and a child train when entering a station scheduled to stop in a real-time scheduling method according to one embodiment of the present disclosure.

[0038] Referring to Fig. 4a, the scheduled station (STA) jPassengers board and disembark from the second preceding train (104) at the platform of ). Here, the preceding passenger capacity of the passengers boarding the second preceding train (104) ( ) and the preceding disembarking vehicle (104) for passengers disembarking from the second preceding train (104) Summing ) the preceding number of passengers boarding and alighting for the second preceding train ( ) can be calculated. Referring to Fig. 4b, the scheduled station (STA) j Passengers board and disembark from the first preceding train (102) at the platform of ). Here, the preceding passenger capacity of the passengers boarding the first preceding train (102) ( ) and the preceding disembarking vehicle (102) for passengers disembarking from the first preceding train (102) Summing ) the preceding number of passengers boarding and alighting of the first preceding train ( ) can be produced.

[0039] Referring to FIG. 4c, the train (100) is at the previous station (STA j-1 Departing from ) and scheduled stop station (STA j It is driving in the ) direction. Here, using a passenger count calculation model, the self-boarding / alighting vehicle ( ) can be estimated. Hereinafter, the boarding / alighting vehicle estimation unit (204) illustrated in FIG. 2 uses the first dataset to estimate the self-boarding / alighting vehicle ( Explains the method for estimating ).

[0040] The passenger count estimation unit (204) uses a pre-trained passenger count calculation model to estimate the number of passengers boarding and alighting from the train at the station. Here, the passenger count calculation model is a machine-learned model based on a record data set in which at least one train operation record and the passenger count of at least one preceding train are accumulated and stored. Here, the record data set includes the actual number of passengers boarding and alighting from the first preceding train (102), the actual number of passengers boarding and alighting from the second preceding train (104), the actual number of passengers boarding and alighting from the child train (100), the actual number of passengers boarding and alighting from the first preceding train (102), and the actual number of passengers boarding and alighting from the second preceding train (104). In one embodiment, the machine learning method for training the passenger count calculation model may include the K-NN (K-Nearest Neighbors) method. Hereinafter, the passenger count estimation unit (204) uses a K-NN method passenger count calculation model to estimate the passenger count of the child train (100). Explains the method for estimating ).

[0041] The boarding and alighting vehicle estimation unit (204) extracts at least one data corresponding to the first dataset and time period from the actual data set as a candidate data set. Here, the candidate data set is the candidate preceding passenger count of the first preceding train (102). ), candidate advance re-passenger of the second advance train (104) ( ), candidate self-re-train personnel of the self-train (100) ), candidate preceding boarding / alighting vehicle of the first preceding train (102) ( ) and candidate preceding boarding / alighting vehicle of the second preceding train (104) ...includes ). The boarding / alighting vehicle estimation unit (204) calculates the Euclidean distance between the data of the first dataset and the data of the candidate dataset, respectively. Here, Euclidean distance is a method used to evaluate the similarity between data. Since the method of calculating Euclidean distance is common in the relevant technical field, a detailed explanation is omitted. The boarding / alighting vehicle estimation unit (204) determines the data of the candidate dataset corresponding to the top k (where k is a natural number greater than or equal to 1) with short Euclidean distances as the data of the analogous data set. Here, the value of k, which determines the number of analogous data sets, is set to an appropriate value to improve the classification accuracy of the model. For example, if the value of k is excessively small, over-fitting may occur. If the value of k is excessively large, under-fitting may occur. Therefore, the value of k can be varied according to the embodiments of the present disclosure. The passenger boarding / alighting vehicle estimation unit (204) calculates the weighted average of the data included in the similar dataset, thereby determining the passenger boarding / alighting vehicle ( Estimates )

[0042] The stopping time estimation unit (206) is the estimated self-boarding / alighting vehicle ( The required stopping time is estimated using a second data set including ). The stopping time estimation unit (206) uses a pre-trained required stopping time estimation model to estimate the required stopping time. The stopping time estimation unit (206) uses the number of passengers ( ), estimated self-boarding / alighting vehicle( ) and previously stored historical information (station information, Using all or part of ) as input parameter values ​​for the stop time estimation model, the required stop time ( ...estimates the required stopping time. During the required stopping time estimated by the stopping time estimation unit (206), passengers board and alight from the train (100) at the scheduled stopping station. Therefore, when traffic volume increases rapidly during peak hours, sufficient time is secured for passengers to board and alight during the extended required stopping time. Meanwhile, in estimating the required stopping time, there is a problem in that it is difficult to distinguish the specific factor itself that causes the delay. To solve this, the weight of the factor with a greater influence on the occurrence of the delay can be determined using an importance measure analysis method based on Random Forest (RF). Subsequently, during the process of rescheduling the schedule of the subsequent train, the departure times of multiple subsequent trains from their respective stopping stations can be changed by reflecting the factor with a greater influence.

[0043] The schedule generation unit (208) requires a stop time ( ) and pre-set dwell duration, By comparing ), the train operation schedule is updated. For example, the schedule generation unit (208) updates the required stopping time ( If ) is greater than the scheduled dwell duration on the operation schedule, the operation schedule is updated based on the dwell duration difference between the required dwell time and the scheduled dwell time. Here, the operation schedule of the train updated by the schedule generation unit (208) includes the operation schedule of the child train (100) and the operation schedule of at least one subsequent train following the child train (100). In one embodiment, the schedule generation unit (208) may update the schedule of the subsequent train (106) to delay the departure time of the subsequent train (106) from the stop station by 5 to 10 seconds. Here, the subsequent train (106) subject to the schedule update is the departure station (STA1) to the previous station (STA j-1It means at least one trailing train located at at least one stop station among them.

[0044] FIG. 5 is a block diagram for explaining each configuration included in a server according to one embodiment of the present disclosure.

[0045] A server (108) according to one embodiment of the present disclosure includes, in whole or in part, a calculation unit (500), a learning data generation unit (502), and a learning unit (504). The server (108) illustrated in FIG. 5 is according to one embodiment of the present disclosure, and not all blocks illustrated in FIG. 1 are essential components, and some blocks included in the server (108) in other embodiments may be added, changed, or deleted.

[0046] Hereinafter, each component included in the server (108) will be described with reference to FIG. 5.

[0047] The calculation unit (500) is the current station (STA). c Based on the train route, the number of passengers on board, the number of passengers entering and exiting between the platform of the current station and the train, and the actual dwell duration, which is the time the train actually stopped at the platform of the current station, are calculated. The calculation unit (500) calculates the number of passengers on board, the number of passengers entering and exiting, and the actual dwell duration, respectively, using passenger information and operation information of at least one train. In one embodiment, the passenger information may include transportation card tag information of passengers entering and exiting the station.

[0048] The calculation unit (500) estimates the number of passengers boarding and alighting at each station by matching passengers entering and exiting each station with trains stopping at the station based on transportation card tag information. The calculation unit (500) individually matches passengers and trains to calculate the number of passengers boarding and alighting at each station. The calculation unit (500) calculates the tag-in station, tag-in time, tag-out station, and tag-out time from the transportation card tag information of the first passenger, who is at least one passenger entering and exiting the current station. The calculation unit (500) determines whether the tag-out station matches the current station, and if it matches, matches the most recent train that arrived at the platform of the current station before the tag-out time as the train from which the first passenger alighted. The calculation unit (500) individually matches the first train that arrived at the platform of the current station after the entry time as the train that the first passenger boarded when the entry station is the current station. Using the individual matching described above, the calculation unit (500) estimates the number of passengers boarding and alighting by matching the passengers who entered and exited the current station with the trains that the passengers boarded and alighted.

[0049] The calculation unit (500) calculates the total number of passengers by accumulating the number of passengers boarding at each station from the departure station to the previous station, which is the station where the train stopped immediately before the current station, and calculates the total number of passengers alighting by accumulating the number of passengers alighting at each station from the departure station to the previous station. The calculation unit (500) calculates the number of passengers boarding by subtracting the total number of passengers alighting from the total number of passengers boarding. The calculation unit (500) calculates the number of passengers boarding and alighting by adding the number of passengers boarding and alighting at the current station. Based on the operation information, the calculation unit (500) extracts the actual stopping time by calculating the difference between the departure time, which is the time the train departs from the current station, and the arrival time, which is the time the train arrives at the current station.

[0050] The number of passengers, the number of passengers boarding and alighting, and the actual stopping time calculated by the calculation unit (500) can be accumulated and stored as a record data set. Specifically, at least one train operation record and the number of passengers boarding and alighting of a train preceding the train can be accumulated and stored as a record data set. The record data set includes the number of passengers boarding a train of the same type, the number of passengers boarding a first preceding train, the number of passengers boarding a second preceding train, the number of passengers boarding and alighting of passengers who boarded and alighted on the first preceding train, and the number of passengers boarding and alighting of passengers who boarded and alighted on the second preceding train. A train of the same type refers to at least one other train that entered the current station at a time corresponding to the time when the train enters the current station. The first preceding train refers to a preceding train that entered the current station immediately before the train of the same type entered the current station. The second preceding train refers to a preceding train that entered the current station immediately before the first preceding train entered the current station. As mentioned above, the performance dataset uses a pre-trained passenger count calculation model to determine the scheduled stops (STA j It is used to estimate the number of passengers boarding and alighting at the station. The server (108) transmits the performance dataset accumulated and stored by the calculation unit to at least one train running on the train line.

[0051] The learning data generation unit (502) generates a learning data set based on all or part of the number of passengers, the number of passengers boarding and alighting, the actual stopping time, and the previously stored station information.

[0052] The learning unit (504) trains a stopping time estimation model based on a learning dataset. The learning unit (504) can retrain the stopping time estimation model using an accumulated learning dataset. The accumulated learning dataset refers to a learning dataset accumulated in the learning dataset based on new passenger numbers, new passenger boarding and alighting numbers, and new actual stopping times. New passenger numbers, new passenger boarding and alighting numbers, and new actual stopping times can be calculated from newly collected new passenger information and new operation information. The method by which the learning unit trains the estimation model using the learning data is common in the relevant technical field, so a detailed explanation is omitted. The server (108) can periodically transmit the retrained stopping time estimation model to at least one train.

[0053] FIG. 6 is a flowchart for explaining each process included in a real-time scheduling method according to one embodiment of the present disclosure.

[0054] Hereinafter, each process included in the real-time scheduling method is described with reference to FIG. 6. Content that overlaps with the descriptions in FIG. 1 to 5 is omitted.

[0055] The data collection unit (200) collects passenger information and train operation information (S600). Specific data included in the passenger information and train operation information, respectively, has been explained in FIG. 2, so further explanation is omitted.

[0056] The dataset generation unit (202) calculates the number of passengers boarding and alighting for at least one preceding train that has stopped first at the train's scheduled station using passenger information and operation information, and generates a first dataset including the number of passengers boarding and alighting (S602). A specific method for calculating the information included in the first dataset is described later in FIG. 7.

[0057] The passenger boarding and alighting estimation unit (204) estimates the number of passengers boarding and alighting for a train at a scheduled stop station using the first dataset and a previously learned passenger count calculation model (S604). A specific method for estimating the number of passengers boarding and alighting is described later in FIG. 8.

[0058] The stopping time estimation unit (206) estimates the necessary stopping time for a train at a scheduled stopping station using a second dataset including the estimated number of self-boarding and alighting trains and a previously learned necessary stopping time estimation model (S606). As the specific input parameters used by the stopping time estimation unit (206) have been described in FIG. 2, further description is omitted.

[0059] The schedule generation unit (208) updates the train's operation schedule by comparing the required stopping time with the pre-set stopping time. A specific embodiment of how the schedule generation unit (208) updates the operation schedule is described in FIG. 2, so further description is omitted.

[0060] FIG. 7 is a flowchart for explaining each process included in the first dataset generation process in a real-time scheduling method according to one embodiment of the present disclosure.

[0061] The dataset generation unit (202) calculates the number of passengers for the preceding train when at least one preceding train enters the station scheduled to stop (S700). The specific method by which the dataset generation unit (202) calculates the number of passengers for the preceding train is described in detail in FIG. 2, so further explanation is omitted.

[0062] The dataset generation unit (202) calculates the number of passengers for the train when the train enters the station where it is scheduled to stop (S702). The specific method by which the dataset generation unit (202) calculates the number of passengers for the train is described in detail in FIG. 3, so further explanation is omitted.

[0063] The dataset generation unit (202) estimates the number of passengers who boarded and alighted in advance from the number of passengers who entered and exited the station scheduled for stopping, and calculates the number of passengers who boarded and alighted in advance (S704). The specific method by which the dataset generation unit (202) calculates the number of passengers who boarded and alighted in advance is described in detail in FIG. 4, so further explanation is omitted.

[0064] FIG. 8 is a flowchart for explaining each process included in the self-boarding / alighting vehicle estimation process in a real-time scheduling method according to one embodiment of the present disclosure.

[0065] The boarding and alighting vehicle estimation unit (204) extracts data corresponding to the first dataset and time period from a performance dataset in which the boarding and alighting vehicles of at least one train for each station are accumulated and stored, as a candidate dataset (S800). The specific data included in the candidate dataset is as described above, so further explanation is omitted.

[0066] The boarding / alighting vehicle estimation unit (204) calculates the Euclidean distance between the data of the first dataset and the data of the candidate dataset, respectively (S802). Since the method of calculating the Euclidean distance between data based on data values ​​is common in the relevant technical field, further explanation is omitted.

[0067] The vehicle boarding / alighting estimation unit (204) determines the data of the candidate dataset corresponding to the top k with short Euclidean distances as the data of the similar dataset (S804). The value of k, which serves as the criterion for selecting the similar dataset, can be varied according to the embodiments of the present disclosure.

[0068] The boarding / alighting vehicle estimation unit (204) estimates the number of passengers boarding / alighting by calculating the weighted average value of the data in a similar dataset (S806). As the boarding / alighting vehicle estimation unit (204) estimates the number of passengers boarding / alighting in advance, it has the effect of being able to predict the congestion occurring at the scheduled stop station in advance.

[0069] FIG. 9 is a flowchart for explaining each process included in the learning method of a stopping time estimation model according to one embodiment of the present disclosure.

[0070] Hereinafter, each process included in the learning method is explained with reference to FIG. 9. Content that overlaps with the descriptions in FIG. 1 to 8 is omitted.

[0071] The calculation unit (500) calculates the number of passengers on the train, the number of passengers boarding and alighting at the train at the train station, and the actual stopping time of the train at the train station based on the current station (S900).

[0072] The training data generation unit (502) generates a training dataset based on all or part of the number of passengers, the number of people boarding and alighting, the actual stopping time, and the previously stored station information (S902).

[0073] The learning unit (504) trains a stopping time estimation model based on a learning dataset (S904).

[0074] Although FIGS. 6 to 9 describe the respective processes as being executed sequentially, this is merely an illustrative explanation of the technical concept of some embodiments of the present invention. In other words, a person skilled in the art to which some embodiments of the present invention belong may modify and adapt the processes described in FIGS. 6 to 9 in various ways, such as by changing and executing the processes or by executing one or more of the processes in parallel, without departing from the essential characteristics of some embodiments of the present invention; therefore, FIGS. 6 to 9 are not limited to a chronological order.

[0075] Various embodiments of the apparatus and method described herein may be realized as digital electronic circuits, integrated circuits, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include being implemented as one or more computer programs executable on a programmable system. A programmable system comprises a storage system, at least one input device, and at least one programmable processor (which may be a special-purpose processor or a general-purpose processor) coupled to receive data and instructions from at least one output device and to transmit data and instructions to them. Computer programs (which are also known as programs, software, software applications, or code) include instructions for the programmable processor and are stored on a "computer-readable recording medium."

[0076] Computer-readable recording media include all types of recording devices in which data that can be read by a computer system is stored. Such computer-readable recording media may further include non-volatile or non-transitory media such as ROM, CD-ROM, magnetic tape, floppy disk, memory card, hard disk, magneto-optical disk, and storage device, or transitory media such as data transmission media. Additionally, computer-readable recording media may be distributed across networked computer systems, and computer-readable code may be stored and executed in a distributed manner.

[0077] Various embodiments of the apparatus and method described herein may be implemented by a programmable computer. Herein, the computer comprises a programmable processor, a data storage system (including volatile memory, non-volatile memory, or other types of storage systems, or a combination thereof), and at least one communication interface. For example, the programmable computer may be one of a server, a network device, a set-top box, an embedded device, a computer expansion module, a personal computer, a laptop, a PDA (Personal Data Assistant), a cloud computing system, or a mobile device.

[0078] The foregoing description is merely an illustrative explanation of the technical concept of the present invention. The embodiments of the present invention are intended to explain, not limit, the technical concept of the present embodiment, and the scope of the technical concept of the present embodiment is not limited by these embodiments. The scope of protection of the present embodiment shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present embodiment. Explanation of the symbols

[0079] 10: Real-time scheduling system 100: Self-train 102: 1st preceding train 104: 2nd preceding train 106: Trailing train 108: Server 20: Real-time scheduling device 200: Data collection unit 202: Dataset Generation Unit 204: Boarding and Alighting Vehicle Estimation Unit 206: Stopping time estimation unit 208: Schedule generation unit 500: Output Unit 502: Training Data Generation Unit 504: Learning Department

Claims

Claim 1 A real-time scheduling method for resolving train operation delays comprises: a process of collecting passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of said train; a process of using said passenger information and said operation information to calculate a number of passengers boarding and alighting from at least one preceding train that has stopped first at said train's scheduled dwell station, and generating a first data set including said number of passengers boarding and alighting; a process of estimating a number of passengers boarding and alighting from said train at said station using said first data set; and a process of estimating a required dwell duration of said train at said station using said second data set including said number of passengers boarding and alighting from said train at said station using said first data set; The method includes a schedule generation process for updating the train operation schedule by comparing the required stopping time and a pre-set dwell duration, wherein the process of estimating the number of passengers boarding and alighting utilizes a passenger count calculation model that has been trained to estimate the number of passengers boarding and alighting at the station, wherein the passenger count calculation model is machine-learned based on a record data set in which at least one train operation record and the number of passengers boarding and alighting of a train preceding at least one train are accumulated and stored, and the machine learning includes a K-NN (K-Nearest Neighbors) method, and the process of estimating the number of passengers boarding and alighting includes a process of extracting data corresponding to the first data set and time period from the record data set as a candidate data set;A real-time scheduling method characterized by: a process of calculating the Euclidean distance between the data of the first dataset and the data of the candidate dataset, respectively; a process of determining the data of the candidate dataset corresponding to the top k (where k is a natural number greater than or equal to 1) with the shortest Euclidean distances as the data of the similar dataset; and estimating the self-boarding vehicle by calculating the weighted average value of the data of the similar dataset. Claim 2 A real-time scheduling method according to claim 1, wherein the passenger information includes transportation card tag information of a passenger tagging in or tagging out of the at least one station, and the operation information includes the location of at least one train, the arrival time of the at least one train at the at least one station, and all or part of the departure time of the at least one train at the at least one station. Claim 3 A real-time scheduling method according to claim 1, wherein the process of generating the first dataset comprises: a process of calculating the preceding passenger occupancy, which is the passenger occupancy of the at least one preceding train when the at least one preceding train enters the station scheduled for stopping; a process of calculating the self-occupancy, which is the passenger occupancy of the train when the train enters the station scheduled for stopping; and a process of calculating the preceding number of passengers boarding and alighting. Claim 4 A real-time scheduling method according to paragraph 3, wherein at least one preceding train comprises a first preceding train that stopped at the station scheduled to stop immediately before the train, and a second preceding train that stopped at the station scheduled to stop immediately before the first preceding train. Claim 5 In paragraph 3, the process of calculating the preceding passenger or the self-passenger is characterized by estimating the boarding quantity for each station from the number of passengers entering each station and estimating the alighting quantity for each station from the number of passengers exiting each station, calculating the total boarding quantity by accumulating the boarding quantity from the initial station (where the train started operating) to the prior station (where the train stopped immediately before the scheduled station), calculating the total alighting quantity by accumulating the alighting quantity from the initial station to the prior station, and calculating the value obtained by subtracting the total alighting quantity from the total boarding quantity as the preceding passenger or the self-passenger. Claim 6 A real-time scheduling method according to paragraph 3, wherein the process of calculating the preceding boarding and alighting number is characterized by estimating the preceding boarding number from the number of passengers who have entered the station scheduled for stopping, estimating the preceding alighting number from the number of passengers who have exited the station scheduled for stopping, and calculating the value obtained by adding the preceding boarding number and the preceding alighting number as the preceding boarding and alighting number. Claim 7 delete Claim 8 delete Claim 9 delete Claim 10 A real-time scheduling method according to paragraph 3, wherein the process of estimating the required stopping time utilizes a required stopping time estimation model that has been trained to estimate the required stopping time, and estimates the required stopping time by using all or part of the self-boarding / alighting vehicles, the self-returning passengers, and previously stored station information as input parameter values ​​of the stopping time estimation model. Claim 11 A real-time scheduling method according to claim 1, wherein the schedule generation process further updates the schedule of at least one subsequent train of the train based on the dwell duration difference between the required dwell time and the scheduled dwell duration when the required dwell time is greater than the scheduled dwell duration on the operation schedule. Claim 12 A real-time scheduling method according to claim 11, wherein the schedule generation process is located at at least one stop station among the initial station, which is the station where the train started operation, and the prior station, which is the station where the train stopped immediately before the scheduled stop station, and further updates the schedule of the subsequent train so as to delay the departure time of at least one subsequent train following the train by 5 to 10 seconds. Claim 13 A method for learning a stopping time estimation model for calculating a required stopping time, which is the time a train must stop, performed by a server, comprising: a process of calculating the passenger occupancy of the train, the boarding and alighting quantity (the number of passengers entering and exiting between the platform of the station scheduled for stopping and the train), and the record dwell duration (the actual time the train stopped at the platform of the current station), based on the current station which is the station the train intends to enter; and a process of generating a learning data set based on all or part of the passenger occupancy, the boarding and alighting quantity, the record dwell duration, and previously stored station information. The method includes a process of training the dwell duration estimation model based on the above-mentioned training dataset, wherein the calculation process utilizes a previously trained passenger count calculation model to estimate the number of passengers boarding and alighting from the train at the station, and the training method comprises the process of accumulating and storing at least one train operation record and the number of passengers boarding and alighting from the train preceding the at least one train as a record data set;The learning method further comprises the process of transmitting the above performance dataset to at least one train, wherein the passenger count calculation model is a machine learning model based on the K-NN (K-Nearest Neighbors) method, and the above performance dataset includes the number of passengers on board a homogeneous train, which is at least one other train that entered the current station at a time corresponding to the time when the train enters the current station, and further includes the number of passengers on board a first preceding train, the number of passengers on board a second preceding train, the number of passengers boarding and alighting passengers on board a first preceding train, and the number of passengers boarding and alighting passengers on board a second preceding train, among a first preceding train and a second preceding train that entered the current station in sequence immediately before the homogeneous train entered the current station. Claim 14 A learning method according to claim 13, wherein the calculation process is characterized by calculating the number of passengers, the number of passengers boarding and alighting, and the actual stopping time, respectively, using passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of at least one train. Claim 15 In paragraph 14, the passenger information includes transportation card tag information of a passenger entering or exiting at least one station, and the calculation process comprises: a process of estimating the boarding quantity, which is the number of passengers boarding at each station, and the alighting quantity, which is the number of passengers alighting from the train at each station, by matching the passengers entering or exiting each station based on the transportation card tag information with the train stopping at each station; a process of calculating the total boarding quantity by accumulating the boarding quantity at each station from the initial station, which is the station where the train started operation, to the prior station, which is the station where the train stopped immediately before the scheduled station, and calculating the total alighting quantity by accumulating the alighting quantity at each station from the initial station to the prior station, and calculating the number of passengers alighting by subtracting the total alighting quantity from the total boarding quantity; and a process of calculating the boarding and alighting quantity by adding the boarding quantity and the alighting quantity at the current station. A learning method characterized by including a process of extracting the actual stopping time by calculating the difference between the departure time, which is the time when the train departs from the current stopping station, and the arrival time, which is the time when the train arrives at the current stopping station, based on the above-mentioned operation information. Claim 16 In claim 15, the estimating process comprises an individual matching process for determining whether the tag-in station, tag-in time, tag-out station, and tag-out time of the first passenger from the transportation card tag information of at least one passenger, and determining whether the tag-out station corresponds to the current stop station; if it corresponds, matching the most recent train that arrived at the platform of the current stop station before the tag-out time as the train from which the first passenger disembarked, and if the tag-in station corresponds to the current stop station, matching the first train that arrived at the platform of the current stop station after the tag-out time as the train from which the first passenger boarded; wherein the learning method is characterized by estimating the number of passengers boarding and the number of passengers alighting by matching the passengers who entered and exited the current stop station with the trains from which the passengers who entered and exited the current stop station boarded and alighted, respectively, using the individual matching process. Claim 17 A learning method according to claim 14, further comprising: a process of collecting new passenger information and new operation information to extract new passenger counts, new boarding and alighting vehicles, and new actual stopping times, and accumulating them in the learning dataset to obtain an accumulated learning dataset; a process of retraining the stopping time estimation model based on the accumulated learning dataset based on the new passenger information and the new operation information; and a process of transmitting the retrained stopping time estimation model to at least one train. Claim 18 delete Claim 19 delete Claim 20 delete Claim 21 A real-time scheduling device for resolving train operation delays comprises: a data collection unit that collects passenger information, which is information regarding passengers entering and exiting at least one station, and operation information of said train; a dataset generation unit that calculates a number of passengers boarding and alighting from at least one preceding train that has stopped first at the scheduled dwell station of said train using said passenger information and said operation information, and generates a first dataset including said number of passengers boarding and alighting; a boarding and alighting estimation unit that estimates a number of passengers boarding and alighting from said train at the scheduled dwell station using said first dataset; and a dwell time estimation unit that estimates a required dwell duration of said train at the scheduled dwell station using said second dataset including said number of passengers boarding and alighting from said train.The system includes a schedule generation unit that updates the train operation schedule by comparing the required stopping time with a preset dwell duration, wherein the boarding / alighting estimation unit for estimating the number of passengers boarding and alighting uses a passenger count calculation model that has been trained to estimate the number of passengers boarding and alighting the train at a stopping station, wherein the passenger count calculation model is machine-learned based on a record data set in which at least one train operation record and the number of passengers boarding and alighting of a train preceding the at least one train are accumulated and stored, and the machine learning includes a K-NN (K-Nearest Neighbors) method, wherein the boarding / alighting estimation unit for estimating the number of passengers boarding and alighting extracts data corresponding to the first data set and the time period from the record data set as a candidate data set, calculates the Euclidean distance between the data of the first data set and the data of the candidate data set, and selects the top k candidates (where k is a natural number greater than or equal to 1) with the shortest Euclidean distances. A real-time scheduling device characterized by determining data of a dataset as data of a similar dataset and calculating a weighted average value of the data of the similar dataset to estimate the self-boarding vehicle. Claim 22 A server for training a stopping time estimation model to calculate a required stopping time, which is the time a train must stop, comprising: a data calculation unit that calculates the passenger occupancy of the train, the boarding and alighting quantity, which is the number of passengers entering and exiting between the platform of the current station and the train, and the record dwell duration, which is the actual time the train stopped at the platform of the current station, based on the current station, which is the station the train intends to enter; and a dataset generation unit that generates a learning data set based on all or part of the passenger occupancy, the boarding and alighting quantity, the record dwell duration, and previously stored station information. The method includes a learning unit that trains the dwell duration estimation model based on the above-mentioned training dataset, wherein the data output unit utilizes a previously trained passenger count calculation model to estimate the number of passengers boarding and alighting from the train at the station, and the training server has a process of accumulating and storing at least one train operation record and the number of passengers boarding and alighting from the train preceding the at least one train as a record data set;A server characterized by further including the process of transmitting the above performance dataset to at least one train, wherein the passenger count calculation model is a machine learning model based on the K-NN (K-Nearest Neighbors) method, and the above performance dataset includes the number of passengers on board a homogeneous train, which is at least one other train that entered the current station at a time corresponding to the time when the train enters the current station, and further includes the number of passengers on board a first preceding train, the number of passengers on board a second preceding train, the number of passengers boarding and alighting passengers on board a first preceding train, and the number of passengers boarding and alighting passengers on board a second preceding train, among the first preceding train and the second preceding train that entered the current station in sequence immediately before the homogeneous train entered the current station. Claim 23 A computer program stored on a computer-readable recording medium to execute each process included in a real-time scheduling method according to any one of paragraphs 1 through 6 and paragraphs 10 through 12.

Images