SYSTEM AND MEHTOD FORPREDICTING THE PROGRESSOFGUIDEDVEHICLES
The present invention relates to a system and method for predicting the progress of guided vehicles, as well as software for this purpose.
In guided transport systems, such as for instance tram, underground railway and train systems, there is available an infrastructure comprising yards, route sections, points, signals etc., using which the vehicles, such as rail vehicles, can be applied for transporting people and/or goods. Use is made of planning for efficient use of the infrastructure by the vehicles. If the system comprising the infrastructure and the vehicles were able to run wholly in accordance with this planning, the transport process would be optimal. In practice however, there occur disruptions which interfere with the planning. This creates uncertainty in implementation relative to the planning. In order to limit such uncertainty a very detailed computer model is known wherein parameters are used. Such parameters include parameters relating to the physical characteristics of the infrastructure, such as bends, inclines or relative positions of signals; physical characteristics of the rolling stock used such as rail vehicles, such as power, length, weight, the composition thereof; and relating to the planning, such as travel routes, setting times or rolling stock composition. If a parameter is missing, it is necessary to make assumptions. It further requires a large amount of work to achieve the most complete possible implementation of the parameters in order to arrive at reliable predictions. Updating of the implementation of the parameters likewise requires a great effort.
In order to provide a system for which obtaining input information is relatively simple for the purpose of obtaining reliable predictions, the present invention provides a system for predicting the progress of guided vehicles, such as rail vehicles, on the basis of time measurement data relating to the progress of the vehicles in the past, comprising:
- at least one computer comprising a working memory, a processing unit and a storage memory, - input means for entering into the system measurement data relating to the progress of the vehicles on an infrastructure,
- processing means for processing the measurement data to statistical data by means of statistical methods,
- predicting means for determining a prediction,
- output means for outputting the prediction. In order to obtain input information for such a system use will be made of per se known measuring systems. An advantage of a system according to the present invention is that the system functions without the prerequisite of process parameters that are necessary in a prior art system. Further advantages are that the average deviations from the predictions are smaller, the distribution in the deviations is smaller and the stability of the predictions will be greater. Adjustment is hereby required less frequently. A system according to the present invention is therefore cost-effective, has low development costs and low maintenance costs. In transport systems with relatively great differences between the vehicles, predictions relating to each vehicle can be measured in practically simple manner.
A predicting system according to the present invention is self-learning in respect of the present infra-configuration (infrastructure and vehicles) . Using
information present in due course in the system, the system is self-learning in respect of changes in the timetable and changes in deployment of vehicles. The application of time measurements relating to the progress of the vehicles has the further positive advantage of taking into account exogenous factors such as the behaviour of engine drivers, the influence of the weather and/or rush hour activity.
In a first preferred embodiment, the system comprises means for managing and determining the statistical data. An advantage of such means is that the system can store the statistical data in a manner which is optimized for the prediction process. Data can herein for instance be stored in the form of measuring points in time or measurement data relating to the travelling time of a part-route (for instance from signal to signal) or statistical data relating to deviations therefrom.
In a further preferred embodiment, the system comprises monitoring means for evaluating the quality of the predictions. Using such a module the quality of the predictions relative to the actually realized measurements can be analysed after the prediction. On the basis of such analyses the prediction model can for instance be further refined.
An embodiment preferably comprises setting means for setting the system parameters. A practical development hereof is a graphic user interface. It is for instance up to a train service manager who is going to modify his timetable on the basis of predictions relating to deviations to himself determine in practice how accurate the data must be. Examples hereof are rounding off the travelling time resolution to a predetermined number of seconds. The train resolution for instance can further be adjusted, or a system
parameter such as the minimum deviation in the number of times a train passes (train passages) .
In addition to the processing means for processing the measurement data into statistical data by means of statistical method, the system preferably comprises analysing means for analysing measurement data. Using these analysing means the data are arranged in a data structure in efficient manner. This data structure has the purpose of having the data as stated in the foregoing available in an efficient form for processes relating to the prediction process. The data and the data structure are stored on a data storage medium. For the purpose of for instance monitoring the quality of the predictions, the system preferably comprises analysing means for analysing predictions. In addition to the graphic display of the quality of the predictions, a quantitative analysis of predictions and the underlying statistical processes can also be performed using this embodiment. The system further preferably comprises request means for requesting measurement data relating to the progress of the vehicles on an infrastructure. Using these means an operator of a system can request specific measurement data located in a data storage. It is also possible to hereby couple the system to a data source such as a measuring system close to for instance a signal .
A further aspect of the present invention relates to a system for planning the running of vehicles over an infrastructure such as a rail infrastructure, comprising a system for predicting the progress of the vehicles as specified in the foregoing. Such a system is for instance used by a train service manager for optionally ad hoc modification of the planning, such as for instance travel routes. If a train service manager
observes on the basis of a prediction that a part of the infrastructure, such as for instance a point, will be free on the basis of the prediction, as opposed to being available on the basis of the planning, the train service manager can then, on the basis of the prediction, assign a route for a train making use of this point during the time the point is free according to the prediction. The train which can now make earlier use of the point on the basis of the prediction, and therefore does not have to wait for the other delayed train, is hereby not delayed. The total rail traffic therefore becomes more efficient by making use of the predictions in allocation of the travel routes.
A further aspect of the present invention relates to a method for predicting the progress of guided vehicles, such as rail vehicles, wherein using statistical methods the predictions are calculated on the basis of time measurements at transition points of part-routes relating to the progress of the vehicles over an infrastructure in the past. Further preferred embodiments based on this method are further stated in the appended claims 11-15. The application of this method achieves, among others, the advantages as specified in the foregoing. A further aspect of the present invention relates to a computer program product which, when loaded into a computer, is suitable for carrying out a method as specified in the foregoing and/or which is suitable for operation in a system as according to one or more of claims 1-10.
Further advantages, features and details of the present invention will be further elucidated hereinbelow on the basis of an embodiment, wherein reference will be made to the accompanying figures, in which:
- figure 1 shows an embodiment of a system according to the present invention;
- figure 2 is a block diagram of an embodiment according to the present invention; - figure 3a is a flow chart of an embodiment according to the present invention;
- figure 3b is a flow chart of an embodiment according to the present invention;
- figure 4 is a flow chart of an embodiment according to the present invention;
- figure 5 is a flow chart of an embodiment according to the present invention;
- figure 6 is a flow chart of an embodiment according to the present invention; - figure 7 is a flow chart of an embodiment according to the present invention;
- figure 8 is a flow chart of an embodiment according to the present invention;
- figure 9 is a flow chart of an embodiment according to the present invention;
- figure 10 is a schematic representation of a screen layout for use in an embodiment according to the present invention; and
- figure 11 is a schematic representation of a screen layout for use in a further embodiment according to the present invention.
A train service manager preferably makes use of a graphic user interface to operate and control the rail traffic. Such a user interface functions on a computer system 1 which is also connected by means of one or more technical interfaces to underlying infrastructure operating systems and infrastructure safety systems.
An embodiment according to the present invention is a prediction module which preferably also functions in this computer system, whereby the train service manager
can use the prediction data in practical manner in operation and control of the rail traffic. In one embodiment the prediction system is for instance an integrated module in the system of the train service manager for operating and controlling the rail traffic. On screen 2 (figure 1) the planned rail traffic is shown graphically in per se known manner. The predicted rail traffic is also shown, wherein this prediction is made in a manner according to the embodiment according to the present invention. If differences occur between the planning and the reality due to for instance delays, the train service manager can enter in relatively simple manner a new planning adapted to the situation. There is also the possibility of allowing the system to automatically make new (sub-)planning schedules within limits, to the extent that such schedules comply with safety regulations.
The operation of the embodiment to be described hereinbelow is based on the main concept that the basis for formulating predictions is a large number of observations, such as measurements when for instance signals are passed (signal passages) . These measurements are then analysed and patterns are determined using statistical analyses. These patterns are then converted into information for the purpose of formulating predictions. These predictions, which are made on the basis of data and the systematics of previous measurements, are used to predict the near future on the basis of the known situation of the rail traffic on an infrastructure.
Figure 2 shows schematically an embodiment according to the present invention. The system comprises a module 22 for processing measurements. These measurements can be requested by means of an interface 10 for sending information request instructions. The
system obtains the measurement information via interface 11 for receiving measurement data relating to positions in time of trains on the tracks.
An alternative method of obtaining information is that the system automatically goes into a data receiving mode after initialization hereof, in which mode an external data supply system supplies current data continuously via interface 11 to the module for processing the measurements 22. After processing of the measurements in module 22, the data are supplied to control module 18 for further processing and for grouping in a data structure. In this control module the data are structured and stored in a manner suitable for the predictions. Further statistical operations are further carried out on the data, whereafter they are stored for use in prediction module 20.
Prediction module 20 makes use of the information stored in module 18 to create predictions. These predictions relate for instance to positions on the track where a train will be located for the next 30 minutes. On the basis of these predictions these positions can be displayed graphically on the screen of a train service manager. In module 16 the data relating to the predictions are processed for output. Data relating to specific requests entering by means of module 16 via interface 12 are also processed to predictions so that the resulting data are suitable for processing in module 20 for the purpose of determining the desired prediction. Interface 13 serves to output data relating to predictions. Such data are outputted, for instance in the form of records or objects, comprising fields such as <dynamic object identification>, <static object identification>, <date latest measurements <time latest measurements
0 [<static object identification>, <date and time prediction>] .
Interface 12 serves to input into the system data relating to specific requests relating to predictions. A train service manager can hereby for instance request a prediction relating to a specific train which is for instance delayed. Such information transfers take place in the form of for instance objects or records comprising fields such as <dynamic object identification>, <static object identification>.
In the request measurement interface 10 use is made in similar manner of objects which are identified by for instance the field <static object identification>. For the data input interface 11 use is made of, among others, the fields <dynamic object identification>, <static object identification> and <date and time measurement>.
A further module is module 14, with which the quality of the predictions is determined by comparison of the predictions with measurements relating to the actual times that trains pass determined points on the track. This is displayed graphically in display window 126 of the graphic user interface of figure 11.
Control and configuration module 24 is further provided for the purpose of setting various system parameters. An example of a graphic user interface with which parameters of the system can be set is shown in figure 10.
Figure 3 relates to the steps performed by the system during start-up thereof. In step 26 the system is started up, whereafter settings are first read from a register "initialization of system menus" in step 28. If measurement data are present in this register, these are read by the system. The first data processing is herein carried out in module 22. The second process of
analysing and grouping the data in the data structure is optionally also performed in module 18. The system is then ready to perform predictions.
When the system is closed down, the closedown instruction is given in step 30, whereafter successively in step 32 the system is signed out of external measuring systems providing the supply of measurement data; the measurement data which have already entered the system are stored; the settings are stored in the register "initialization of the system menus"; and the system is stopped.
This module provides on the one hand for the request for measurement data from external measuring systems and on the other for recording and processing of received measurements. At a predetermined minimum number of required measurements, the measurements are transmitted to module 18 (management statistical data) .
Such a minimum can be set by the user or by the administrator of the system. A database is maintained in module 24 (management configuration data) . Data are however first requested by means of module 22 (process measurements) . This takes place as follows. The list of static objects such as for instance measuring points and signals, from which are received the measurements of time and of the passage of dynamic objects such as trains, is determined. This list is compiled on the basis of configuration data or with reference to static objects such as occur in the requests for predictions formulated for dynamic objects such as trains. This takes place in step 36 of figure 4 after starting step 34.
In step 38 the request for the determined measurements is then transmitted to the external measuring system. This process then ends in step 40.
The process for receiving of measurements is described in figure 5. A measurement is received in step 42. In step 44 received measurements are recorded and stored in the internal data structure. If the number of measurements of the dynamic object received at the static object is greater than or equal to the predetermined number of measurements, the measurement is then processed further in the process "determine statistical data" . In step 46 the subsequent measurement is awaited, after receiving which the program returns to step 42.
In figure 6 the process of preparing and determining statistical data is started in step 48 by means of receiving a measurement . In step 50 a number of statistical data are then determined from the received measurement data, such as minimum travel time between two signals, average travel time between two signals, maximum travel time between two signals and/or standard deviation of the travel time between two signals. These data are then stored in an internal data structure. In step 52 extreme values of the data are filtered by means of per se known mathematical formulas. These mathematical formulas establish whether there is a case of an extreme measurement. These measurement data are then not used to determine statistical travel time data. They are used in situations in which the past of the train must be taken into account. A significant deviation is then determined in step 54. The method for determining a significant deviation is shown in figure 7. The determination of the significant deviation is started in step 56. In step 58 is determined whether data are in accordance with one of the three following situations:
- 1. There are no predictions as yet, but there is a measurement, so there is a significant deviation in the signal passage. An action associated herewith is to still determine the initial prediction. - 2. There are predictions and a measurement, but there is no significant deviation from the signal passage. This does not require any action to be taken.
- 3. There are predictions and a measurement, and there is a significant deviation from the signal passage. The associated action is to once again determine the prediction. The predictions are determined in step 60 on the basis of these stated situations.
A starting point in the prediction is that, on the basis of past measurements and statistics, predictions are formulated in the form of times at which static objects are passed. A prediction is therefore made of when (time) for instance each train will pass a signal. This process relates for instance to trains which are travelling outside the measurement region (the location of the train service manager) at the moment the prediction is performed, or it relates to trains which are travelling inside the measurement region (the location of the train service manager) . Depending on predetermined settings, a prediction in respect of trains is determined, wherein the exact route which the train must travel over the static objects is specified in the request.
A further possibility is that in the formulation of predictions use is made of measurements of routes along static objects already covered previously by this specific train. In making an initial prediction a starting point is for instance that the train will travel as quickly as possible, wherein in principle no disruptions will occur. On the basis hereof an initial data set with a high punctuality is created. The
predictions are therefore accurate. If at some point it becomes known that a train is delayed, the known time delay of the train is added to the predicted times. If a measurement is performed inside the operational region of the location of a train service manager after a train has entered this region, and a deviation is detected on the basis of this measurement, predictions further along the route of this train are then also adjusted on the basis of this measurement. Use is made of a so-called threshold value
"deviation signal passage" . Only if the measured signal passage deviates more than this threshold value is a new prediction made on the basis of this deviation above the threshold value. New predictions are created from the measuring point where the measurement with a deviation above the threshold value is measured. The train service manager can hereby see, for all measuring points in his graphic display, at which predicted moment the train will arrive there. In this method use is made of so-called travel times between two signals instead of a time at which a train passes the signal. This has the advantage for instance that, if a train leaves too late, it does not generally occur that the same train is present in the existing measurements. If use is made of the travel times, the average travel time per signal step can be added to the current measurement once a deviation in a time a signal is passed is detected. A new prediction can hereby be generated in simple and rapid manner. After receiving a measurement and determining that there has been a deviation in the time a signal is passed, the history of the train is taken into account on the basis of the database with previous measurements. Physical limitations of the rolling stock are also taken into account. Such a limitation is that a train cannot depart
before it has arrived at a station. If the requester of the prediction indicates that a rail vehicle, for instance train with number 5638 (departure) , uses the same rolling stock as the train with number 5637 (arrival) , this information is then used to thus determine a lower limit for the departure time of train with number 5638.
Determining of the predictions takes place as according to figure 8. This process is started in step 62. One or more new predictions relating to a measuring point are formulated in step 64. As stated in the foregoing, the travel time per signal step is applied for this purpose. From the most recent measurement (at measuring point X) a prediction is made for each subsequent measuring point (static object) as to what the time of passage will be.
In respect of the first measuring point A this is the time of the measurement X plus the travel time from measuring point X to measuring point A. The prediction of the time of a second measuring point B located further away will amount to measurement X + travel time measuring point X to A + travel time A to B. This is further elucidated below. Predictions downstream are partly determined here by a relative addition of the travel times. The module 14 for monitoring predictions can further be provided in continuous manner with the new statistical information so that the quality of the predictions can be monitored and updated in real-time. In step 66 the predictions are for instance transmitted for the storage thereof and for graphic display thereof for the benefit of the train service manager.
In the formulation of predictions the time of the formulation of the prediction is also stored. Using this information it is possible to determine what the average age of predictions is. The age of the prediction is one
of the prediction quality factors. An older, accurate prediction is in this case better than a newer, accurate prediction with the same deviation.
Figure 9 shows the processing of the requests for predictions. This process serves for processing of the request for predictions. A user, such as a train service manager, makes use of the system in order to obtain predictions relating to a train and receives the formulated predictions from the system. The requesting of the predictions is started in step 68. In step 70 is determined on the basis of the data given in the request for which static objects the passage times of for instance a train have to be determined. The request for the predictions is then transmitted to module 20 "determine predictions". Finally, module 16 sends the received predictions back to the requester, such as the graphic system which displays the predictions to the train service manager.
Figure 10 shows schematically a screen layout of a possible embodiment of an operating interface for the settings of the system. 72 is the menu bar of the prediction system. 74 is a header "travel time resolution" of the predictions. 76 to 82 are setting options hereof. If 76 is selected, there is rounding-off to even seconds. If 78 is selected, rounding-off takes place to five seconds. If 80 is selected, rounding-off takes place to ten seconds and if 82 is selected, predictions are not rounded off. 84 designates a block for selecting the resolution of the trains. The prediction results are shown with the exact train number if 86 is selected. If 88 is selected, results are shown per ten trains. If 90 is selected, results are shown per 100 trains. 92 relates to settings for a demonstration model of a simulator. The rail traffic of a determined date, which can be entered in 94, can herein be
simulated. The starting time of the simulation is entered in 96. Further system parameters can be set in block 98. 100 herein relates to an input module for a minimum deviation in the train passages. 102 relates to a minimum prediction time that can be entered. Finally, 104 relates to an additional period that can be entered. An input field for a minimum number of measurements per signal can also be provided here. Finally, 106 and 108 represent respective buttons for storing and for closing or cancelling this input field.
Figure 11 shows a screen display of module 14. Window 110 herein serves to display statistical data such as the number of predictions 112, the average deviation 114, the distribution in deviation 116, the average time margin 118 and possible further statistical data such as for instance the distribution in the time margin. Button 120 serves to cause the predicting process to proceed and button 122 serves to store the predictions. Window 126 shows a graph with the deviation between measured points in time of train passages and the predicted times for train passages. The height of the block with number 128 shows the average deviation. The positive deviations are deviations where a train is delayed. The negative deviations are deviations in which a train passes a measuring point too early. This can occur for instance in the case of goods trains which, although travelling according to a timetable, sometimes depart earlier if all cargo is available and coupled on, and if the traffic situation on the track permits. On the basis of data relating to 9 December 2002 a simulation was performed using an embodiment according to the present invention compared to a prior art system. The data are as follows:
Accuracy signal passages
Description Prior art Embodiment according to the present invention
Database : DAV091202CSBO.mdb : dav091202SpectrumNegeerEersteMeting.mdb
Signals : All signals : All signals
5 Trains : All trains : All trains
Predictions : Only current : Only current predictions
Predictions, no time margin cfo not ignore ignore
No. of measurements : 927 : 878 806 Average deviation : 39 s : 25 s 13 10 Standard deviation : 170 s : 140 s 56 Average time margin : 1744 s : 1396 s 1044
Stability of prediction signal passages
I S D Deessccrriippttiioonn P Prriioorr A Arrtt E Eml bodiment according to the present invention
Database : dav091202CSBOmetfilter.mdb : davO91202SpectrumNegeerEersteMeting.mdb
Signals : All signals : All signals
Trains : All trains : All trains
Predictions : All predictions, time margin n.a. : All predictions, time margin n.a. 0 H Hiissttoorriiccaall w wiitthh c cuurrrreenntt : : 33110000 : 1552
Historical : 2545 : 976
Only current (never adjusted) : 376 : 302
Total number of predictions : 2912 : 1278
Average number of adjustments : 6 : 3
25
This shows that in a system according to the present invention significant improvements occur compared to the prior art. The average deviation has for instance decreased to 25 (13) seconds compared to 39
30 seconds for a prior art system. The distribution of the deviation has decreased to 140 (56) seconds compared to 170 for the prior art.
The number of adjustments of the predictions has decreased to three, wherein a prior art system had six
35 adjustments of the predictions.
Use is made here of prediction measurements wherein the first measurement of a passage of a train is ignored. This is a practical method if use is made of predictions of the travelling distance between different 0 so-called signal passages.
The present invention is not limited to the above described preferred embodiment. The rights sought are defined by the appended claims.