SYSTEM AND METHOD FOR NAVIGATING WITHIN A TRACK NETWORK

Abstract

A system for navigating within a track network includes, as system components, a system central, a track maintenance machine and a communication device. The system central is set up for administering network data representing a model of the track network. The track maintenance machine is suited for treatment of track sections of the track network. The track maintenance machine includes a navigation device for processing navigation data derived from the network data. The communication device is provided for data exchange between the system central and the navigation device. The system includes at least one movable or stationary carrier platform with sensors for collecting raw data representing characteristic information of the track network. A big data framework is set up in the system central to evaluate the raw data and synchronize them with the network data. Automated updating of the network data can be carried out with the system.

Claims

1-15. (canceled)

16. A system for navigating within a track network, the system comprising: a system central for administering network data representing a model of the track network; a track maintenance machine for treating track sections of the track network, said track maintenance machine including a navigation device for processing navigation data derived from the network data; a communication device for data exchange between said system central and said navigation device; at least one movable or stationary carrier platform having sensors for collecting raw data representing characteristic information of the track network; and a big data framework disposed in said system central for evaluating the raw data and synchronizing the raw data with the network data.

17. The system according to claim 16, wherein said track maintenance machine is constructed as a carrier platform including a sensor system having said sensors collecting the raw data during travel on the track network.

18. The system according to claim 16, which further comprises a carrier platform constructed as a measuring vehicle or a track-bound vehicle equipped with sensors.

19. The system according to claim 16, which further comprises a flying carrier platform or a drone equipped with sensors.

20. The system according to claim 16, wherein the network data are stored as a graph with track objects as nodes having relationships between said track objects as edges.

21. The system according to claim 20, which further comprises characteristic data patterns specified for one of said track objects.

22. A method of operating a system for navigating within a track network, the method comprising the following steps: providing a system according to claim 16; using said sensors to collect the raw data; transmitting the raw data to said system central; generating object data from the raw data by using identification algorithms; and synchronizing the network data with the object data to update the network data.

23. The method according to claim 22, which further comprises, after performing an update of the network data, transferring all of the updated data or part of the updated data to said navigation device of said track maintenance machine.

24. The method according to claim 22, which further comprises assigning probability values or probability functions to the respective object data in dependence on at least one of said sensors or a carrier platform or identification algorithms.

25. The method according to claim 24, which further comprises carrying out an update of the network data by new object data in dependence on the assigned probability values or probability functions.

26. The method according to claim 22, which further comprises organizing the object data based on a detected motion pattern of a carrier platform, to supply track objects represented by the object data being strung together as an object chain for synchronization with the network data stored as a graph.

27. The method according to claim 26, which further comprises subdividing the object chain into segments, and synchronizing a segment with the graph based on distinctive track objects.

28. The method according to claim 27, which further comprises synchronizing a segment with a partial graph by specifying an extent of agreement, and replacing the partial graph with the segment when an extent of agreement exceeds a pre-set minimum extent.

29. The method according to claim 27, which further comprises synchronizing a segment with a partial graph, and maintaining a non-verifiable track object as a node of the partial graph until a pre-set number of failed verifications has been reached.

30. The method according to claim 22, which further comprises using the sensors disposed on the track maintenance machine to record surrounding track objects, and determining a current position of the track maintenance machine by synchronization the recorded track objects with the network data.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The invention will be described below by way of example with reference to the accompanying drawings. There is shown in a schematic manner in:

[0028] FIG. 1 track maintenance machine

[0029] FIG. 2 track network

[0030] FIG. 3 system layout

[0031] FIG. 4 structure of track sections

[0032] FIG. 5 method sequence

DESCRIPTION OF THE EMBODIMENTS

[0033] A track maintenance machine 1 shown in FIG. 1, which is to be navigated to a work assignment, is a component of the system according to the invention. This track maintenance machine 1 comprises working units 2 for treatment of a track section 3 of a track network 4. The track network 4 includes various objects such as, for example, tracks 5, switches 6, crossings 7, masts 8, tunnels 9, stations 10, underpasses 11, level crossings 12 or balises 13, as shown by example in FIG. 2.

[0034] The track maintenance machine 1 is further equipped with various sensors or sensor systems 14 in order to record the surroundings of the track 5 traveled upon, and the current position. These are, for example, a camera 15, a positioning system 16, a clearance gauge scanner 17 or a rail scanner 18. In this manner, the track maintenance machine 1 functions as a carrier platform for the sensors or sensor systems 14.

[0035] In order to get to a track section 3 to be worked on, the track maintenance machine 1 has a navigation device 19. This is configured as a computing- and controlling unit and serves for navigation within the track network 4 which is represented by network data. By means of the navigation device 19, navigation data derived from the network data are processed and synchronized with sensor data in order to determine the current position of the track maintenance machine 1.

[0036] One object of the present invention is to continuously update the network data in an automatized way. To that end, at raw data representing characteristic information of the track network 4 are first collected by means of the sensors or sensor systems 14. In further sequence, the raw data are evaluated and synchronized with the network data administered in a system central 20. From this data synchronization it is possible to draw conclusions with regard to the condition of the track network 4 or individual track sections 3. For example, a frequent change of collected position data allows conclusions as to an unstable track position.

[0037] Besides the track maintenance machine 1, other carrier platforms can be used for collecting the raw data, for example a measuring vehicle 21, another rail vehicle 22 or a flying carrier platform 23. Infrastructure facilities equipped with sensors 14 can be used as a stationary carrier platform 24. This might be, for example, a mast 8 with a camera 15 fastened to it which observes a track section 3. Also, fibre optic cables installed adjacent to the track 5 can be used as sensors 14. The basis for this is the so-called Distributed Acoustic Sensing (DAS) in which laser impulses are sent through fibre optic cables in order to thus register in real time along a track section 3 sound signals and activities which can be derived therefrom. By means of such stationary carrier platforms 24, raw data of an observed track section 3 are collected over time. Aside from detecting object changes, these raw data can also be used for position verification of moving carrier platforms 1, 21, 22, 23.

[0038] The measuring vehicle 21 is equipped, for example, with a GNSS receiver 25, a clearance gauge scanner 17 and a rail scanner 18. The other rail vehicle 22 includes a GNSS receiver 25, and the flying carrier platform 23 comprises a camera 15 or other devices for recording aerial views. By means of all these sensors or sensor systems 14, various raw data are collected and supplied for evaluation. In this, depending on data volume and available computing power, the raw data are either pre-processed on the carrier platform 1, 21, 22, 23, 24 or transmitted directly to the system central 20.

[0039] As can be seen in the illustrative system layout in FIG. 3, communication means 26 are provided as additional system components. These communication means 26 serve for data exchange between the system central 20, the sensor systems 14 and the navigation device 19. These means are, for example, devices for wireless communication via a mobile radio network. In a very simple embodiment of the system according to the invention, the system central 20 is accommodated in the track maintenance machine 1, so that there is a self-sufficient system for navigation and network data updating. In this, the communication means 26 may be elements of an installed bus system.

[0040] In the system central 20, a big data network 27 is installed. This also supports various machine learning algorithms besides conventional databases 28 and data analysis tools. Examples for this are noSQL or Hadoop. In this manner, the system central 20 serves for collecting, storing and processing the data.

[0041] From data points m.sub.r,s.sup.i(t), the sensor systems 14 generated a data tensor S.sub.k(t) with an arbitrary dimension k at a point in time t:

S.sub.k(t)=(M.sub.1(t), . . . ,M.sub.k(t)) [0042] wherein i=1, . . . , k are M.sub.irsare matrixes, therefore

[00001]$M_i\left(t\right)=\left(\begin{array}{ccc}{m}_{1,1}^i\left(t\right)& \mathrm{.Math.}& m_{1,s}^i\left(t\right)\\ \mathrm{.Math.}& & \mathrm{.Math.}\\ m_{r,1}^i\left(t\right)& \mathrm{.Math.}& m_{r,s}^i\left(t\right)\end{array}\right)$

[0043] Sensors 14 arranged on moving carrier platforms 1, 21, 22, 23 deliver in particular data points m.sub.r,s.sup.i(t) with spatial information. In stationary carrier platforms 24, however, the data points recorded by means of sensors 14 display especially temporal changes.

[0044] From characteristic features in the data points m.sub.r,s.sup.i(t) of the sensor systems 14, virtually indexed objects 29 (objekt.sub.i) are compiled as object data. These represent track objects 4-13 which are recognizable with stochastic reliability in the track environment and can serve as reference for navigation. Specifically, the objects 29 are characterized by significantly reproducible patterns. In an object register, the definitions (or algorithms) of the objects 29 are updated by new data.

[0045] During this, a probability is calculated for each object characteristic or a probability function P.sub.x is assigned:

[00002]${\mathrm{objekt}}_{v.Math.i}\left(t\right)=\left(\begin{array}{cc}\mathrm{objekt}.Math..Math.\mathrm{id}& P_{\mathrm{id}}\\ \mathrm{time}.Math..Math.t& P_t\\ \mathrm{coordinates}.Math..Math.x,y,z& P_{x,y,z}\\ \mathrm{offset}.Math..Math..Math..Math.x,.Math..Math.y,.Math..Math.z& P_{.Math..Math.x,.Math..Math.y,.Math..Math.z}\\ \mathrm{value}.Math..Math..Math..Math.t& P_{.Math..Math.t}\\ \mathrm{next}.Math..Math.\mathrm{objekt}.Math..Math.O_{i+1}& P_{O_{i+1}}\\ .Math..Math.t_{O_{i+1}}& P_{.Math..Math.t_{O_{i+1}}}\\ \mathrm{data}.Math..Math.\mathrm{points}.Math..Math.M_1& P_{M_1}\\ \mathrm{.Math.}& \mathrm{.Math.}\\ \mathrm{data}.Math..Math.\mathrm{points}.Math..Math.M_x& P_{M_x}\end{array}\right)$

[0046] Aside from the metadata resulting from the object verification, the data points in particular describe a current state of the virtually indexed object 29.

[0047] The respective probability function P is dependent on the type of sensor or sensor system 14, the type of carrier platform 1, 21, 22, 23, 24 and the algorithms in the object register. For example, a probability function with little scattering is prescribed for newer sensor systems 14. For older sensors 14 with less precision, however, a greater scattering is prescribed. A recognition probability derived from stored observation processes takes into account the number of objects 29 recognized so far. In this way, a degree of unambiguousness of an object 29 in relation to a data stock stored in the system central 20 is established. Accordingly, a classification of the objects 29 by means of the evaluated information content takes place with respect to the total track network and the information therein.

[0048] In addition, the evaluation method is continuously improved in that the recognizable objects 29 are continuously expanded and adapted by means of machine learning or manual learning. In this, the autonomous machine learning within the big data framework is based on those data which are newly collected in the system central 20.

[0049] As shown in FIG. 4, virtually indexed objects 29 are linked relative to one another according to a motion pattern of the carrier platform 1, 21, 22. Thus, in the case of track-bound carrier platforms 1, 21, 22 strung-together virtually indexed objects 29 are generated, wherein the resulting object chain 30 can be regarded as a semantic chain K.sub.t.sub.n:

K.sub.t.sub.n=(objekt.sub.vi(t.sub.n), . . . ,objekt.sub.vi(t.sub.n+m)),

wherein t.sub.n is a starting time and t.sub.n+m is an ending time of a run of the carrier platform 1, 21, 22.

[0050] Within this chain, distinctive objects 29 (for example, object data of a switch 6) are specified as discretization points for subdivision into segments 31. These can be determined dynamically, from the probability of recognition or from the relevancy in the network (degree of unambiguousness). In FIG. 4 for example, the strung-together objects 29 represent, from left to right, a switch 6, a balise 13, a track 5, a mast 8, a tunnel portal, a tunnel 9, a tunnel portal, a switch 6, a mast 8, a balise 13, two masts 8, and a switch 6.

[0051] New object chains 30 are synchronized with the network data, i.e. the model 31 of the track network 4. In this, the network data are stored as a graph N(t.sub.a), wherein t.sub.a is a time of updating of the particular object 29:

N(t.sub.a)=(objekt.sub.v(t),edge.sub.v(t))

N(t.sub.new)=objekt.sub.vi(t).fwdarw.N(t.sub.old)

[0052] Segments 32 of the object chain 30 are synchronized with the model 31 of the track network 4 (mapped) via distinctive objects 29. If a segment 29 and a partial graph coincide with a high probability, then the virtually indexed objects 29 contained in the segment 29 are transferred to the graph. In this manner, the characteristics of the particular virtually indexed object 29 are used for updating the characteristics of the model 31 (update of the network data) while taking into account the associated probability functions P. In this, with growing collection of raw data, the reliability and precision of the network data increases:

[00003]$N\left(t_a\right)=\left(\left(\begin{array}{cc}\mathrm{objekt}.Math..Math.\mathrm{id}& P_{\mathrm{id}}\\ \mathrm{update}.Math..Math.t_a& P_{t_a}\\ \mathrm{coordinates}.Math..Math.x,y,z& P_{x,y,z}\\ \mathrm{value}.Math..Math.d& P_d\\ \mathrm{data}.Math..Math.\mathrm{points}.Math..Math.M_1& P_{M_1}\\ \mathrm{.Math.}& \mathrm{.Math.}\\ \mathrm{data}.Math..Math.\mathrm{points}.Math..Math.M_x& P_{M_x}\end{array}\right).Math.\left(\begin{array}{cc}\mathrm{next}.Math..Math.\mathrm{objekt}.Math..Math.O_{v+1}& P_{O_{+1}}\\ \mathrm{distance}.Math..Math.\mathrm{next}.Math..Math.\mathrm{objekt}.Math..Math.x,y,z& P_{x,y,z}\\ \mathrm{.Math.}& \mathrm{.Math.}\end{array}\right)\right)$

[0053] Data points collected by means of stationary carrier platforms 24 are also synchronized in a corresponding manner with the network data, wherein here the information about temporal changes is paramount.

[0054] During an updating process, it may happen that, based on sensor faults or obstructions during object detection, individual objects 29 cannot be verified. Then it is practical if these remain existent in the network data untilduring a new collecting procedure 33a falsification takes place, or a verification fails to happen several times.

[0055] Shown FIG. 5 is the method sequence in an overview. In this, the method is divided roughly into the following method steps: [0056] input 34 of the various carrier platforms 1, 21, 22, 23, 24 and the sensor systems 14 [0057] signal analysis 35 for object indexing with statistical values [0058] repeated collecting procedures 33 (observations) for verification or falsification of the object indexing [0059] mapping 36 of various collection types or observation perspectives [0060] output 37 of the track network model 31 (updated network data) [0061] navigation procedure 38 of a track maintenance machine 1

[0062] In an initial phase, the system forms an initial model 31 for the track network 4 on the basis of the pre-set algorithms. To that end, for example, aerial views, raw data of a measuring vehicle 21, raw data of the track maintenance machine 1 and raw data of another vehicle 22 are evaluated. In FIG. 5, the respective signal analysis 35 initially results in different objects 29 or object classes which can be associated with individual track objects 5-13. In aerial views, these are, for example, tracks 5, switches 6, tunnels 9, masts 8, stations 10, etc. The raw data of the track-bound carrier platforms 1, 22, 23 can be associated with the track objects tracks 5, switches 6, tunnels 9, etc.

[0063] In the course of the repeated collecting procedures 33, the individual objects 29 are verified or falsified with regard to their relation to one another. In the case of the track-bound carrier platforms 1, 22, 23, object chains 30 are formed which depict a track section 3 traveled upon. A merging of these evaluation results by means of mapping 36 results at last in the model 31 of the recorded track network 4.

[0064] For reliably carrying out a navigation procedure 38, the network data are transmitted from the system central 20 to the navigation device 19 of the track maintenance machine 1 after a data update has taken place or at prescribed time intervals. In this, it is favourable if, via the desired confidence, it is parameterized which objects 29 contained in the network data are transmitted for navigation to the track maintenance machine 1.

[0065] During the navigation procedure 38 itself, objects or object data 29 are synchronized with those track objects 5-13 which are currently detected in the surroundings of the track maintenance machine 1 by means of the sensors or sensor systems 14. In this way, detected track objects 5-13 serve as reference for position determination. Additionally, results of a track gauge measurement can be used for more precise positioning on the track 5. In this, the detected progression of the track gauge forms further data points of those objects 29 which represent the corresponding track 5. In an extension, it is also possible to use detectable characteristic features of sleepers 39 or rails 40 (markings, material characteristics, etc.).

[0066] Additionally, the raw data collected during the navigation process 38 serve as new data input for updating the network data. With the present system, changes in the track network 4 due to reconstruction or malfunctions are automatically taken into account for subsequent navigation procedures 38.

[0067] Cognition takes place cause- or time-related automatically in dependence on the speed of migration of the track network 4.

[0068] Favourably, recognized and anticipated objects 29 in the current surroundings of the track maintenance machine 1 are displayed to a machine operator 41 for orientation. Work instructions can additionally be included in this display. Also, work parameters can be prescribed to the machine operator 41 in a location-dependent way or transmitted directly to a working unit 2. In this way, an automatized local adaptation of work parameters takes place, thus enabling an optimized operation of the track maintenance machine 1. In a tamping machine, these are, for example, lifting- and lining values as well as time stipulations for the tamping cycles. In other track maintenance machines 1, work parameters like ballast demand, spoil quantities of old ballast can be adjusted dependent on location.