WO2023276428A1

WO2023276428A1 - Operation arrangement system and method for generating operation arrangement plan

Info

Publication number: WO2023276428A1
Application number: PCT/JP2022/018759
Authority: WO
Inventors: 靖英森; 雄一小林; 祐子山下; 優美子石戸
Original assignee: 株式会社日立製作所
Priority date: 2021-07-01
Filing date: 2022-04-25
Publication date: 2023-01-05
Also published as: JP2023007077A

Abstract

Provided is an operation arrangement system (100) that generates an operation arrangement plan, comprising: an arithmetic device that executes a prescribed process; and a storage device that can be accessed by the arithmetic device. The arithmetic device has a diagram prediction unit (11) that generates a predicted diagram, a state determination unit (14) that determines whether a false convergence state is included in the predicted diagram, and a deadlock site extraction unit (15) that determines whether a deadlock state is generated. The false convergence state is a condition in which a deadlock site occurs in the predicted diagram output by the diagram prediction unit and a predicted diagram having a high rating cannot be generated, and the deadlock site serves as a main point in an operation arrangement that eliminates the false convergence state.

Description

Traffic rescheduling system and method for generating traffic rescheduling plan

Import by reference

This application claims the priority of Japanese Patent Application No. 2021-110092 filed on July 1, 2021, and incorporates the contents thereof into the present application by reference.

The present invention relates to a traffic rescheduling system that efficiently generates many traffic rescheduling proposals.

There is a traffic rescheduling system that outputs traffic rescheduling plans.

As background technologies in this technical field, there are the following prior arts. Patent Document 1 (Japanese Patent Application Laid-Open No. 2019-209797) discloses a traffic rescheduling proposal creation system, a learning data generation unit that transmits learning data to the traffic rescheduling proposal creation system, and a traffic rescheduling proposal creation system for traffic rescheduling. Equipped with an operation management unit that transmits timetable data and receives timetable rescheduling proposals, and a timetable action learning unit that receives learning timetable data and outputs the degree of suitability for timetable change actions, and a simulation execution unit that outputs the degree of suitability for timetable change actions, and determines search rules from the degree of suitability. A traffic schedule management system is described which includes a search rule update unit that creates a traffic schedule plan search unit that searches for traffic schedule plans according to the updated search rules and outputs the results to the traffic management unit.

The traffic rescheduling plan output from the traffic rescheduling system may contain errors or be insufficient, so learning using sufficient teacher data is required. However, although rescheduling due to small train delays occurs frequently, large train delays due to service disruptions occur infrequently, so sufficient training data (traffic rescheduling data) for learning cannot be collected. In the random generation method for randomly generating traffic rescheduling plans, it is difficult to generate good traffic rescheduling plans, and the generation efficiency of teacher data is poor. For this reason, it is desired to find out key points of traffic rescheduling, efficiently increase teacher data corresponding to the key points, and perform machine learning.

The purpose of the present invention is to realize a traffic rescheduling system that efficiently generates many traffic rescheduling proposals.

A representative example of the invention disclosed in the present application is as follows. That is, a traffic rescheduling system for generating a traffic rescheduling plan includes an arithmetic device that executes a predetermined process and a storage device that can be accessed by the arithmetic device, and the arithmetic device predicts a timetable for generating a predicted timetable. a state determination unit for determining whether the predicted timetable includes a false convergence state; and a stalemate extraction unit for determining whether a stalemate has occurred. , the false convergence state is a state in which a stalemate occurs in the predicted timetable output by the timetable prediction unit, and a highly evaluated predicted timetable cannot be generated; It is characterized by being a key point in traffic rescheduling.

According to one aspect of the present invention, many traffic rescheduling plans can be efficiently generated. Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

1 is a block diagram showing a logical configuration of a traffic rescheduling system according to an embodiment of the present invention; FIG. 1 is a block diagram showing a physical configuration of a traffic rescheduling system according to an embodiment; FIG. It is a figure which shows the structural example of the plan timetable data of a present Example. It is a figure which shows the structural example of the feature-value data of a present Example. It is a figure which shows the structural example of the traffic rescheduling data of a present Example. It is a flowchart of the process which the traffic rescheduling system of a present Example performs. It is a flow chart of the sticky point detection processing of the present embodiment. 8 is a flowchart of search weight determination processing according to the embodiment; It is a figure which shows the example of the agglutination part display screen of a present Example.

First, an outline of the traffic rescheduling system 100 of the embodiment of the present invention will be described. The traffic rescheduling system 100 of this embodiment detects a false convergence state and a stalemate in the process of generating a traffic rescheduling proposal and operating the diagram simulator (diagram evaluation/prediction unit 11). The user inputs the response to the stalemate, or refers to the past successful traffic rescheduling, and changes the weight for generating the perturbation so that the train schedule of the stalemate is likely to be changed.

For example, in a state where the delay is relatively large, a false convergence state occurs when the time schedule simulator repeatedly performs calculations but does not generate a train rescheduling plan that resolves the stalemate, or when only the trains around the stalemate increase in delay. In other words, the false convergence state is a state in which the predicted timetable output by the timetable simulator has a stalemate in a specific range (time or station), and a predicted timetable that further reduces the KPI cannot be generated. Further, the stalemate point is a point (bottleneck) for resolving a state where train traffic rescheduling is necessary at a certain station, that is, a false convergence state.

FIG. 1 is a block diagram showing the logical configuration of the traffic rescheduling system 100 of this embodiment.

The traffic rescheduling system 100 of the present embodiment includes a timetable evaluation/prediction unit 11, an inference unit 12, a perturbation generation unit 13, a state determination unit 14, a stalemate extraction unit 15, a display unit 16, a factor estimation unit 17, and a weight determination unit 18. and a data recording unit 19 .

FIG. 2 is a block diagram showing the physical configuration of the traffic rescheduling system 100 of this embodiment.

The traffic rescheduling system 100 of this embodiment is configured by a computer having a processing unit 110 , a storage device 120 , an input/output interface 130 and a communication interface 140 .

The processing unit 110 is an arithmetic device having a processor (CPU) and memory. The processor executes programs stored in memory. The processor executes various programs (for example, timetable evaluation/prediction program 111, traffic rescheduling learning/suggestion program 112, perturbation generation program 113, state determination program 114, stalemate extraction program 115, factor estimation program 117, weight determination program 118, etc.) By executing, each part of the traffic rescheduling system 100 (for example, the diagram evaluation/prediction unit 11, the inference unit 12, the perturbation generation unit 13, the state determination unit 14, the stalemate extraction unit 15, the display unit 16, the factor estimation unit 17 , weight determination unit 18, data recording unit 19, etc.) are realized. Note that part of the processing performed by the processor by executing the program may be performed by another arithmetic device (for example, hardware such as ASIC and FPGA).

The memory includes ROM, which is a non-volatile storage element, and RAM, which is a volatile storage element. The ROM stores immutable programs (eg, BIOS) and the like. RAM is a high-speed and volatile storage device such as DRAM (Dynamic Random Access Memory) that temporarily stores programs executed by a processor and data used during program execution.

The storage device 120 is, for example, a large-capacity, non-volatile storage device such as a magnetic storage device (HDD) or flash memory (SSD). In addition, the storage device 120 stores data used by the processor when executing the program (for example, planned timetable data 121, predicted timetable data 122, feature amount data 123, traffic rescheduling data 124, successful case data 125, etc.), and Stores programs that That is, the program implements each function of the traffic rescheduling system 100 by being read from the storage device 120, loaded into the memory, and executed by the processor.

The input/output interface 130 is connected to an input device 160 such as a keyboard and a mouse, and an output device 170 such as a display device and a printer (not shown). It is an interface that outputs with . A user terminal connected to the traffic rescheduling system 100 via a network may provide the input device 160 and the output device 170 . In this case, the traffic rescheduling system 100 may have a web server function, and the user terminal may access the traffic rescheduling system 100 using a predetermined protocol (for example, http).

The communication interface 140 is a network interface device that controls communication with other devices according to a predetermined protocol.

Programs executed by the processor are provided to the traffic rescheduling system 100 via removable media (CD-ROM, flash memory, etc.) or a network, and stored in the non-volatile storage device 120, which is a non-temporary storage medium. Therefore, the traffic rescheduling system 100 preferably has an interface for reading data from removable media.

The traffic rescheduling system 100 is a computer system configured on one physical computer or on a plurality of logically or physically configured computers, and is constructed on a plurality of physical computer resources. It may operate on a virtual machine. For example, the diagram evaluation/prediction unit 11, the inference unit 12, the perturbation generation unit 13, the state determination unit 14, the stalemate extraction unit 15, the display unit 16, the factor estimation unit 17, the weight determination unit 18, and the data recording unit 19 are each They may operate on separate physical or logical computers, or may operate on a single physical or logical computer in combination.

Next, the function and operation of each functional unit and program will be explained.

The timetable evaluation/prediction unit 11 is a timetable simulator that generates a predicted timetable by executing the timetable evaluation/prediction program 111 based on the planned timetable data 121, failure details, perturbations, and stalemate points. The generated predicted timetable is stored in predicted timetable data 122 . The diagram evaluation/prediction unit 11 also calculates a KPI, which is an evaluation index for the generated predicted diagram. KPI can use total delay time, total delay time of superior trains such as express trains, number of suspended trains, number of stops of suspended trains, total delay time of users, time until schedule recovery, etc. Low KPI prediction It can be said that diamonds are highly convenient for users.

By executing the traffic rescheduling learning/proposing program 112, the inference unit 12 calculates a traffic rescheduling plan using the initial value of the predicted timetable.

The perturbation generation unit 13 generates perturbations by executing the perturbation generation program 113, and generates a traffic rescheduling plan to which the generated perturbations are added.

The state determination unit 14 determines whether the predicted timetable includes a false convergence state by executing the state determination program 114 . As described above, the false convergence state means that the predicted timetable output by the timetable simulator (timetable evaluation/prediction unit 11) has a stalemate in a specific range (time or station), and the predicted timetable further reduces the KPI. cannot be generated.

The deadlock point extraction unit 15 determines whether deadlock points have occurred for each train and station by executing the deadlock point extraction program 115 . A stalemate point is, as described above, a state that requires rescheduling of trains at a certain station, that is, a point that becomes a key point (bottleneck) for resolving a false convergence state. delay is greater than the surrounding area (time or station), delay increases after a certain point (time or station) (become the top of the delay), or a point that is inconsistent and cannot be realized (e.g. siding (e.g. overtaking at a station where there is no station), high-class trains (e.g. limited express trains, paid express trains, etc.) If you follow the line, delays and troubles are increasing, and there is a big impact on subsequent trains), it is determined that a stalemate has occurred. Details of the processing by the sticky point extraction unit 15 will be described with reference to FIG.

The display unit 16 is an interface that displays a stalemate display screen 900 (FIG. 9) and receives input from the operator.

The factor estimation unit 17 estimates the factors of the detected stalemate by executing the factor estimation program 117 .

By executing the weight determination program 118, the weight determination unit 18 determines the search weight according to the estimated stalemate factor. Details of the processing by the weight determining unit 18 will be described with reference to FIG.

The data recording unit 19 stores traffic rescheduling data with a low KPI in the success case data 125.

The planned timetable data 121 is data of planned operation times of trains, and the details thereof will be described with reference to FIG. The predicted timetable data 122 is data of operation times predicted from actual operation times of trains. The feature quantity data 123 is data indicating the characteristics of the amount of correction of the train schedule, and is output by the schedule evaluation/prediction unit 11 . Details of the feature amount data 123 will be described with reference to FIG. The traffic rescheduling data 124 is data of traffic rescheduling candidates estimated by the inference unit 12, and details thereof will be described with reference to FIG. Success case data 125 is data in which KPI was good as a result of traffic rescheduling.

FIG. 3 is a diagram showing a configuration example of the planned timetable data 121. FIG.

The planned timetable data 121 includes a train ID, which is unique identification information for a train, the departure station of the train, the destination of the train, the type of train (local train or superior train), and the identification information of multiple stations where the train stops or passes. station ID, the arrival/departure track number of the station, the arrival time, and the departure time are recorded for each train and station. Note that the predicted timetable data 122 may also have the same format as the planned timetable data 121 .

FIG. 4 is a diagram showing a configuration example of the feature amount data 123. As shown in FIG.

The feature amount data 123 is data indicating characteristics of correction of a train schedule, and includes a train ID, a station ID, a track number that is identification information of the boarding point where the train arrives and departs, a destination, and a delay time that is the difference between the predicted timetable and the planned timetable. is the data recorded for each train and station. The feature amount data 123 may include the type of train in addition to the illustrated data. The feature amount data 123 may include trouble information that caused the delay in train operation. The trouble information may include data on the troubled train, the station (or section) where the trouble occurred, the traveling direction of the troubled train, the time of occurrence, and the time of restart.

FIG. 5 is a diagram showing a configuration example of the traffic rescheduling data 124. As shown in FIG.

The traffic rescheduling data 124 is data in which train IDs, station IDs, and traffic rescheduling details are recorded for each train. The operation rescheduling contents include changes in the order of arrivals and departures, changes in track numbers, as well as partial suspension of service and disconnection of service. The traffic rescheduling data 124 may include, in addition to the illustrated data, the time required for traffic rescheduling and the track number on which traffic rescheduling occurs.

FIG. 6 is a flowchart of processing executed by the traffic rescheduling system 100. FIG.

First, it receives fault information input from the input/output interface 130 or the communication interface 140 (1001).

Next, the timetable evaluation/prediction unit 11 applies the input failure information to the planned timetable data 121 to calculate the initial value of the predicted timetable (1002).

Next, the inference unit 12 calculates a traffic rescheduling plan using the initial value of the predicted timetable (1003).

Next, the perturbation generation unit 13 generates a perturbation and generates a traffic rescheduling plan to which the generated perturbation is added. For example, the content of traffic rescheduling and parameters (delay time, arrival/departure number, etc.) are changed at random.

Next, the timetable evaluation/prediction unit 11 applies the generated timetable replanning plan to the initial predicted timetable to calculate predicted timetable data 122 (1005).

Next, the state determination unit 14 determines whether the calculated predicted timetable data 122 includes a false convergence state (1006). For example, the amount of decrease in the KPI of the predicted timetable repeatedly calculated in the loop of the diagram evaluation/prediction unit 11 and the state determination unit 14 from the KPI of the previous predicted timetable is smaller than a predetermined threshold (that is, the KPI of the predicted timetable or the amount of decrease from the previous time is less than a predetermined threshold), and the predicted timetable converges, if the KPI is greater than a predetermined threshold, it is determined to be in a false convergence state. As a result, when it is determined that the predicted timetable data 122 includes a false convergence state, the stalemate point extraction unit 15 extracts the stalemate point from the predicted timetable data 122 (1007). Details of the sticky point extraction process will be described with reference to FIG.

Next, the display unit 16 outputs display data for displaying an agglutination point display screen (Fig. 9), prompting the operator to select a response.

Next, the factor estimating unit 17 estimates the detected cause of the stalemate depending on whether it matches a predetermined stalemate pattern (1009).

Next, the weight determining unit 18 determines the search weight according to the stalemate factor estimated by the factor estimating unit 17 (1010), returns to step 1003, and creates the next traffic schedule plan. Details of the search weight determination process will be described with reference to FIG. The determined search weights are used by the perturbation generator 13 to generate perturbations. By increasing the weight of the portion that causes the stalemate, many traffic rescheduling plans that eliminate the stalemate are created.

The estimation of the stalemate factor by the factor estimator 17 and the determination of the search weight by the weight determiner 18 include, for example, the following patterns.
(1) Since there are trains waiting to overtake the delayed train and the departure order cannot be changed, there is a pattern in which the trains waiting for their turn and the following trains are affected more. In this case, the departure order of the train is changed, and the change of the order of the following trains is effective, and the weight determination unit 18 decides the weight that causes many changes in the departure order.
(2) There is a pattern that there is a station that cannot be entered because the scheduled arrival/departure number track is not available, and the delay of the following train increases. In this case, the weight determination unit 18 determines a weight that makes it effective to change the arrival/departure number line at the station, and that causes many changes in the arrival/departure number line.
(3) There is a pattern in which there is no appropriate post-operation for the return train, although the section where the failure occurred should be partially suspended and the train should be turned back before it. In this case, the weight determining unit 18 determines the weight that gives more operation during the time after the occurrence of the partial suspension.
(4) There are patterns in which many factors are involved and it is difficult to determine the main factor. In this case, the weight determination unit 18 determines the weight for generating many possible rescheduling operations for that train and that time.
(5) When the stalemate cannot be specified, the weight determination unit 18 determines weights so that perturbation occurs overall without assigning weights to specific traffic rescheduling methods.

On the other hand, if it is determined in step 1006 that the predicted timetable data 122 does not include a false convergence state, the timetable evaluation/prediction unit 11 determines whether the KPI is decreasing in this predicted timetable data 122 (1011). As described above, the KPIs include the total delay time, the total delay time of superior trains such as express trains, the number of suspended trains, the number of stops of suspended trains, the total delay time of users, the time until the timetable is restored, and the like. As a result, if the KPI has decreased, the traffic rescheduling data of the traffic rescheduling plan is sent to the data recording unit 19 and recorded in the success case data 125 (1012). On the other hand, if the KPI has not decreased, the process returns to step 1003 to calculate another traffic rescheduling plan.

FIG. 7 is a flow chart of the stalemate detection process.

Specifically, for each train and station, the stalemate extraction unit 15 determines whether the delay is concentrated in a specific range near the train and station (1022), and whether the train and station are delayed. (1023), whether there is a contradiction in the train and station (1024), and whether an honors train is involved (1025). If none of the conditions are met, the train and station are determined not to be a stalemate (1026). On the other hand, if any one of the conditions is satisfied, the train and station are determined to be a stalemate (1027). After determining whether or not there is a stalemate for all combinations of trains and stations, the stalemate detection process is terminated, and the caller's process is returned to.

FIG. 8 is a flowchart of search weight determination processing.

First, the weight determination unit 18 determines whether or not there is a corresponding input on the agglutination point display screen 900 (FIG. 9) for each agglutination point (1032).

If there is a corresponding input, the weight determination unit 18 sets a large weight (for example, maximum) for the organized content that matches the input content so that many perturbations related to the input content occur (1033). It should be noted that the past similar states may be searched automatically in step 1034 without determining whether or not there is a corresponding input in step 1032 .

On the other hand, if there is no corresponding input (if "no operation" is selected on the stalemate display screen 900 in FIG. 9), the weight determining unit 18 searches the success case data 125 for past similar stalemate states. (1034), the curtailment procedures that match the curator content in the past similar stalemate are weighted heavily (eg, maximally) so that more perturbations with respect to the retrieved curative procedures occur (1035).

After that, the weight determining unit 18 determines the search weights for all of the stalemate points, ends the search weight determination process, and returns to the calling process.

FIG. 9 is a diagram showing an example of a stalemate display screen 900. FIG.

The stalemate display screen 900 includes a diamond display area 910 , a stalemate display area 920 , and a corresponding selection input area 930 .

The diagram display area 910 graphically displays the train diagram in the area where the horizontal axis is the time and the vertical axis is the station (distance from the starting station of the route), and the detected deadlock points are superimposed on the train diagram. The stalemate display area 920 displays information about the detected stalemate. A countermeasure selection input area 930 displays countermeasures against the detected deadlock, and any one of them can be selected. When a plurality of stuck points are detected, it is possible to select a countermeasure for each detected stuck point. If it is not determined in step 1032 of the search weight determination process whether there is a corresponding input, the corresponding selection input area 930 may not be displayed.

In the above, the embodiment of the present invention has been described with a railway operation diagram as an example, but it can also be applied to a simulator of a transportation schedule for passengers and freight.

It should be noted that the present invention is not limited to the above-described embodiments, and includes various modifications and equivalent configurations within the scope of the attached claims. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and the present invention is not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment may be replaced with the configuration of another embodiment. Moreover, the configuration of another embodiment may be added to the configuration of one embodiment. Further, additions, deletions, and replacements of other configurations may be made for a part of the configuration of each embodiment.

In addition, each configuration, function, processing unit, processing means, etc. described above may be realized by hardware, for example, by designing a part or all of them with an integrated circuit, and the processor realizes each function. It may be realized by software by interpreting and executing a program to execute.

Information such as programs, tables, and files that implement each function can be stored in storage devices such as memory, hard disks, SSDs (Solid State Drives), or recording media such as IC cards, SD cards, and DVDs.

In addition, the control lines and information lines indicate those that are considered necessary for explanation, and do not necessarily indicate all the control lines and information lines necessary for implementation. In practice, it can be considered that almost all configurations are interconnected.

Claims

A traffic rescheduling system that generates a traffic rescheduling plan,
A computing device that executes a predetermined process, and a storage device that can be accessed by the computing device,
A diagram prediction unit in which the arithmetic unit generates a predicted diagram;
a state determination unit in which the arithmetic unit determines whether the predicted timetable includes a false convergence state;
The computing device has a stalemate extraction unit that determines whether a stalemate has occurred,
The false convergence state is a state in which a stalemate occurs in the predicted timetable output by the timetable prediction unit, and a highly evaluated predicted timetable cannot be generated;
A traffic rescheduling system, wherein the stalemate point is a critical point in traffic rescheduling for resolving the false convergence state.
The traffic rescheduling system according to claim 1,
The traffic rescheduling system, wherein the state determination unit determines that a false convergence state occurs if the estimated timetable repeatedly generated by the timetable prediction unit converges and the evaluation is lower than a predetermined threshold.
The traffic rescheduling system according to claim 1,
The stalemate extracting unit determines whether delays are concentrated in a specific range of trains and stations, trains and stations are the heads of delays, contradictions occur in trains and stations, and superior trains are involved. A traffic rescheduling system characterized by determining that there is a stalemate if at least one of the conditions is satisfied.
The traffic rescheduling system according to claim 1,
a factor estimator for estimating the determined stalemate factor,
a perturbation generation unit in which the computing device generates a perturbation and generates a traffic rescheduling plan to which the generated perturbation is added;
The computing device has a weight determination unit that determines a search weight according to the estimated stalemate factor,
The traffic rescheduling system, wherein the perturbation generation unit generates the perturbation by changing a portion reflected in the timetable according to the determined weight.
The traffic rescheduling system according to claim 4,
The traffic rescheduling system, wherein the weight determination unit sets a large weight for a rescheduling procedure that matches a rescheduling procedure in a similar past stalemate.
The traffic rescheduling system according to claim 4,
The factor estimating unit estimates the cause of the stalemate depending on whether it matches a predetermined stalemate pattern,
The traffic rescheduling system, wherein the weight determination unit determines a weight that generates many perturbations at the estimated stalemate factor location.
The traffic rescheduling system according to claim 1,
A traffic rescheduling system, comprising: a display unit for outputting display data for superimposing and displaying a stuck point on the predicted timetable.
A method for generating a traffic rescheduling plan executed by a traffic rescheduling system,
The traffic rescheduling system has a computing device that executes a predetermined process, and a storage device that can be accessed by the computing device,
The method for generating the traffic rescheduling plan includes:
A timetable prediction procedure in which the arithmetic unit generates a predicted timetable;
a state determination procedure in which the arithmetic unit determines whether the predicted timetable includes a false convergence state;
The computing device has a stalemate extraction procedure for determining whether a stalemate has occurred,
The false convergence state is a state in which a stalemate occurs in the predicted timetable output in the timetable prediction procedure, and a highly evaluated predicted timetable cannot be generated;
A method for generating a traffic rescheduling plan, wherein the stalemate point is a key point in traffic rescheduling for resolving the false convergence state.