COMPUTER-IMPLEMENTED METHOD FOR DETERMINING RAILWAY VEHICLE MOVEMENT PROFILE TYPE AND TRACK CIRCUIT SYSTEM CONTROLLER

Abstract

This disclosure relates to a computer-implemented method for determining the rail vehicle motion profile type from a rail vehicle motion profile, wherein the rail vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver over time, comprising: - obtaining a rail vehicle motion profile; - normalizing the rail vehicle motion profile; - extracting one or more features from the normalized rail vehicle motion profile; - determining the distance of the extracted features from each centroid of a determined rail vehicle motion profile type in a classification process; and - assigning the rail vehicle motion profile to the rail vehicle motion profile type with the nearest centroid.

Description

COMPUTER-IMPLEMENTED METHOD FOR DETERMINING RAILWAY VEHICLE MOVEMENT PROFILE TYPE AND TRACK CIRCUIT SYSTEM CONTROLLER Description of the invention The present invention relates to a computer-implemented method for determining the type of railway vehicle motion profile from a railway vehicle motion profile as a function of a plurality of transmitted currents and measurements from a track circuit transceiver. According to another aspect, the present disclosure relates to a track circuit system controller that includes a transceiver that is connected to a pair of rails of a railway track and a controller that receives from the transceiver the transmitted and measured currents, wherein the controller is adapted to obtain the railway vehicle movement profile. Furthermore, this disclosure relates to a computer-readable non-transient storage medium comprising instructions. Document KR 2019000028 refers to a train position detection device that uses an audio frequency track circuit. The distance is determined based on the impedance. US patent 9,026,283B2 describes dynamic time warping (DTW) methods for comparing magnetic sensor data to determine the degree of agreement with an expected waveform for train motion. Track circuits could be used to locate a train's position within a signaled block for the purpose of enabling virtual signaling within advanced train control systems. These virtual block track circuits use the amount of current measured and transmitted to the front axles of an approaching train, or the rear axles of a reversing train, to determine the location of the nearest axle signaled block. When track circuit data is used to determine a train's position within a signaled block, errors in the determined location can occur, which are offset by the train's movement profile. For example, if the train were to accelerate at a point along the track, this could introduce a deviation into the computer's calculated relationship between track circuit data and the train's position. Generally speaking, it is true that the greater the acceleration, the greater the potential deviation. The same observation applies to the train's deceleration.Therefore, it is important to understand when a rail vehicle might be accelerating or decelerating significantly, so that it can be handled appropriately to optimize the accuracy of the computer-determined relationship between track circuit data and train position. According to one aspect, the invention relates to a computer-implemented method for determining the type of railway vehicle motion profile from a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of currents transmitted and measured by a track circuit transceiver with respect to time, comprising ML / a / ZUZZ / UU l ózo - obtain a rail vehicle movement profile; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. According to an additional aspect, a non-transient, computer-readable storage medium is provided comprising instructions, which, when executed by a computer, cause the computer to perform the following steps: - obtaining a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver with respect to time; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. According to another aspect, a track circuit system controller is provided, the track circuit system including a transceiver that is connected to a pair of rails of a railway track and the controller receives the transmitted and measured currents from the transceiver, where the controller is adapted to: - obtaining a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver with respect to time; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. The method steps could be carried out by means of hardware components, firmware, software, a computer programmed by means of the appropriate software, by means of any combination thereof, or in any other way. Therefore, the manner in which the aforementioned characteristics of the present invention can be understood in detail, the more particular description of the invention, summarized with ML / a / ZUZZ / UU l ózo brevity above, could be read with reference to the embodiments. However, it is noted that the accompanying drawings illustrate only the typical embodiments of this invention and therefore shall not be considered as limiting its scope. The accompanying drawings relate to embodiments of the invention and are described below: Figure 1 shows, schematically, the track circuit system according to the invention; Figure 2 shows, schematically, the calculation of a location using a DTW method; Figure 3 shows a comparison between the actual location and the calculation using a DTW method of an accelerating rail vehicle; Figure 4 shows a flowchart of a method according to one modality of this disclosure; Figure 5 shows an example curve of transmitted current measurements from a railway vehicle; Figure 6 shows the curve from Figure 5, which has been filtered with an exponential filter; Figure 7 shows the curve from Figure 5, which has been filtered with a Savitzky-Golay filter; Figure 8 shows the curve from Figure 5, which has been filtered with a Hodrick-Prescott filter; Figure 9 shows a curve from a calculation of the samples of the first derivative of the transmitted (filtered) and normalized current values; Figure 10 shows a curve from a calculation of the samples of the second derivative of the transmitted (filtered) and normalized current values; Figure 11 shows an area under the curve (AUC) of the transmitted (filtered) and normalized current values; Figure 12 shows an area under the line (AUL) of the transmitted (filtered) and normalized current values; Figure 13 shows an area between curves (ABC) of the transmitted (filtered) and normalized current values; Figure 14 shows a reference curve that establishes a relationship between distance and transmitted and measured current; Figure 15 shows sequences of the normalized (filtered) transmitted and measured currents for different simulated railway vehicles moving through the track; Figure 16 shows the first and second derivatives of the filtered and normalized currents that are transmitted and measured from Figure 15; Figure 17 shows the variation of the first derivative with respect to the ABC value; Figure 18 shows the negative percentage of the second derivative with respect to the ABC value; Figure 19 shows the variation of the first derivative with respect to the negative percentage of the second derivative; ML / a / ZUZZ / UU l OZO Figure 19A shows a three-dimensional graph of the train motion profile feature and the respective centroids; Figure 20 shows a plurality of transmitted and measured currents, which are filtered and normalized; Figure 21 shows the variation of the first derivative with respect to the ABC value; Figure 22 shows the negative percentage of the second derivative with respect to the ABC value; Figure 23 shows the variation of the first derivative with respect to the negative percentage of the second derivative; Figure 23A shows a three-dimensional graph of the train motion profile characteristic and the respective centroids; Figures 24, 25, and 26 show, respectively, the transmitted and time-normalized current of a railway vehicle classified as decelerating (Figure 24), classified as driving at a constant speed (Figure 25), and classified as accelerating (Figure 26); and Figure 27 shows a flow chart of a method according to a modality. Figure 1 schematically shows the track circuit system 1 according to the invention. The track circuit system 1 includes a railway track with two rails 3 and a railway vehicle 5 that can move on the rails 3 of the track. A railway vehicle could be a locomotive, a train, a multiple unit, or the like. The track circuit system 1 also includes a transceiver 6, which is coupled to the rails 3. A controller 7 is adapted to control the transceiver 6, to transmit signals to the rails 3, and / or to receive the measurement results from the transceiver 6. Transceiver 6 can transmit a signal on one or both of rails 3. The signals are fed to rails 3 at the feed point 9. In some configurations, the track circuit is a DC track circuit. In other configurations, the track circuit could be an AC track circuit. In other words, the track circuit signal is transmitted by transceiver 6. Furthermore, transceiver 6 is adapted to measure the transmitted current of the track circuit signal. The transmitted and measured current from transceiver 6 is provided by the transceiver to controller 7. In some configurations, the transceiver is adapted to measure the transmitted currents at regular time intervals. Controller 7 is adapted to collect a plurality of transmitted and measured currents and to store them as a sequence of transmitted and measured currents over time. These sequences of transmitted and measured currents over time could form a rail vehicle movement profile, particularly if a rail vehicle passes along the track during the measurement period. In different configurations, the railway track is divided into one or more blocks. In the case of multiple blocks, these are placed sequentially along the railway track. In the configuration shown in Figure 1, only a portion of a single block is shown. The wheels and axle of the rail vehicle create a short circuit between the rails 3 of the same track. The current transmitted and measured by transceiver 6 depends on the position of the rail vehicle. MA / a / ZUZZ / UU l OZO transmitted and measured current is reduced in case the railway vehicle 5 moves out of the power supply at points 9. According to the modalities, the transmitted and measured current is used to determine the distance between the nearest axle 11 of the railway vehicle 5 with respect to the feed at points 9. The nearest axle could be the rear axle 11 of a reversing railway vehicle 5 or a front axle 11 of an approaching railway vehicle 5. The feed at points 9 are the points where the transceiver 6 is electrically connected to the rails 3. In other words, the track circuits could be used to locate the position of the nearest axle 11 of the rail vehicle 5 within a block toward the feed at points 9, in particular, for the purpose of enabling virtual signaling. For example, this system could be used within advanced train control systems. In some embodiments, taps 13 are selectively provided at one or more predetermined positions on the track between the two rails 3 for the purpose of calibrating the track circuit system 1. For example, a tap 13 could selectively and electrically connect the two rails at one or more predetermined known positions so that the track circuit could be calibrated. The tap 13 then simulates the axle and wheels of the rail vehicle 5 and a specific position. As described above, the track circuits use the amount of measured current transmitted to the nearest axle 11 of rail vehicle 5 to determine where the nearest axle 11 is located in the signaled block. Typically, the relationship between the transmitted current and the location of the nearest axle 11 of the railway vehicle 5, which could be determined based on the distance determined between the feed at points 9 and the nearest axle 11 of the railway vehicle 5, cannot be presented analytically, for example, using linear models. For example, the relationship between the transmitted current and the distance between the power supply at points 9 and the nearest axle 11 could also be different between geographical locations, due to different factors such as track circuit length, rail resistance, ballast resistance, train axle resistance, weather conditions, etc. Therefore, according to one modality, which could be combined with other modalities described herein, an excited data method is used to automatically determine the relationship between the transmitted current and the location of the rail vehicle using data collected from many passing rail vehicles. The ratio will be understood to be the transmitted current against the location (or the distance between the nearest axis 11, the rail vehicle 5 and the feed at points 9). According to the invention, a dynamic time-warping (DTW) method is used. A DTW method could be used to determine the similarity between two sequences. For example, the sequences, here of the transmitted currents, could vary with respect to the vehicle speed. ML / a / ZUZZ / UU l Railway OZO. DTW is a method that can calculate a match between two given sequences. For this purpose, each time index of a first sequence is compared with one or more indices of another sequence, or vice versa. The mapping of the indices of the first and second sequences must increase monotonically and vice versa. According to modalities, the DTW method is used to translate or transform the time domain into a relative time domain, for example, the distance d between the nearest axis 11 of the railway vehicle 5 and the feed at points 9. Figure 2 shows an example of a DTW method. On the left side, the small curves 15 show the transmitted current from several rail vehicles passing along the track including rails 3, collected against time to calculate the transmitted current versus rail vehicle distance d (the dashed curve 17a in the right graph) using the DTW process. For this purpose, each of the curves 15 in the transmitted current versus the time domain, which is also called the transmitted current with respect to time in this disclosure, is transformed into the transmitted current versus the distance domain using the dynamic time-warp (DTW) method. Then, from the plurality of curves, a reference curve 17a is calculated, for example, by calculating the average value. In the right-hand diagram of Figure 2, a comparison of the DTW calculation (the dashed curve 17a) with the actual relationship (the solid curve 19) shows a very good match. The solid curve 19 is generated by using taps 13 on the heles at specific locations. For these locations, the transmitted current is measured. Then, an interpolation is performed to generate the solid curve 19 in the right-hand diagram. The movement of rail vehicles 5 is randomly selected and includes rail vehicles at constant speed, rail vehicles with increasing speed, and rail vehicles with decreasing speed. The thick curve 17b in the left diagram corresponds to a projection or transformation of curve 17a onto the transmitted current versus time diagram. In one mode, the maximum value of the distance d in the right-hand diagram corresponds to the lengths of a block of a railway track. In the previous stages, the DTW method is used in particular to avoid the periodic manual calibration of data, for example, when using the derivations. Once this relationship is established, track circuit data can be used to provide the train's location in real time, which better supports the resolution of rail vehicle detection and allows for greater track capacity. While the dynamic time warping (DTW) method used can translate the transmitted current being captured from an absolute time domain to a common relative time domain (corresponding in the present case to a distance d, in particular, the distance from the nearest axis 11 of the rail vehicle 5 to the feed at points 9), there are errors that can occur depending on ML / a / ZUZZ / UU l ózo of the acceleration or deceleration of passing railway vehicles. As an example, if all the passing railway vehicles used to calibrate the circuit system of track 1 accelerate at the same location, there will be a deviation or error in the relationship between the transmitted current and the location or distance of the railway vehicle using the above calibration method comprising a DTW method. Figure 3 shows a comparison between the true relationship between transmitted current and distance (shown as the thin line 20), which corresponds to the solid curve 19 in the right-hand diagram of Figure 2, and the calculation or calibration comprising a DTW method (shown as the DTW reference curve 22) if only accelerating rail vehicles are taken into account. For a given transmitted current level of 3400 mA, the DTW reference curve (curve 22) would show the rail vehicle further out of transceiver 6 or the feed at points 9 when compared to the true relationship (curve 20), which for the following rail vehicles is in the unsafe direction. The following method steps, which could be used to avoid deviation errors, are presented below. These method steps could be used in a method for calibrating a track circuit. For example, method steps are provided to identify the acceleration of the rail vehicle and to remove those measurements from the measurement result set, which are used to calibrate the system according to the method described with respect to Figure 2. In other embodiments, the method steps could be used in a method for calculating a distance d of a railway vehicle 5, in particular, the distance between the nearest axle 11 of a railway vehicle 5 to the feed at points and the feed at points 9. Figure 4 shows a flowchart for a method according to a modality of the present disclosure. The flowchart in Figure 4 shows three main stages, which are used with simulated data (on the left) and with real data (on the right). The simulated and real data correspond to the transmitted current, which is measured at constant time intervals. In other words, the data includes rail vehicle movement profiles comprising a sequence of the transmitted currents over time. The simulated rail vehicle motion profiles, such as curve 15 in Figure 2, are calculated using different types of rail vehicle motion profiles, as will be explained in further detail with reference to Figure 14. For example, some rail vehicles are moving at a constant speed, while others are accelerating or decelerating. The simulated transmitted current depends on other fixed parameters, which can be set for the track, such as ballast resistance, rail material, etc. The simulated motion profiles could also be considered as reference rail vehicle motion profiles, which are particularly useful for learning about System 1. The different stages of the flowchart in Figure 4 are explained below. ML / a / ZUZZ / UU l ózo In an optional first stage 30, the data is filtered. This will be further explained with reference to Figures 5-8. In a second stage 32, features are extracted, which is explained with reference to Figures 9-13, and in a third stage 34, the rail vehicle movement profiles are classified, which is explained with reference to Figures 14-19A. The classification can then be used to classify the actual rail vehicle movement profiles, as will be explained with reference to Figures 20-26, which corresponds to portion 36 of the flowchart in Figure 4. Next, with regard to Figures 5-8, the filtering of the moving profiles of railway vehicles is explained. As explained previously, a railway vehicle movement profile represents the transmitted current of a track circuit over time, specifically within a block of track. The transmitted current is measured at regular time intervals. Figure 5 shows an example curve of the rail vehicle motion profile. As can be seen in Figure 5, the transmitted currents of the rail vehicle motion profile include some interspersed samples. To better visually represent the rail vehicle motion profile, a straight diagonal line is also included in the drawings. Each measurement point is shown as a circle. Depending on the modality, the rail vehicle movement profile can be filtered using different methods. In one scenario, an exponential filter could be used to filter the rail vehicle movement profile. Figure 6 shows the rail vehicle motion profile from Figure 5, which has been filtered using an exponential filter. The filtered rail vehicle motion profile is shown as the continuous curve 44. The circles correspond to the measurement points, which have already been shown in Figure 5. The index on the horizontal axis corresponds to the time domain. Each index point represents a specific time at which the transmitted time measurement is taken. An exponential filter is used to remove outliers and to search for trends by exponentially aging past samples with a certain value, commonly referred to as the aging parameter. In the example in Figure 6, an exponential filter is used with an aging parameter α set to 0.65. As can be seen in points 42, exponential filtering removes some of the local signal disturbances. For example, the exponential filter could use the following formula, where the observation starts at time t=0: so = xo, for t=0, st= axt + (1- a)st-i, for t>0 Equation (1) st is the output of the filtering algorithm and the raw data sequence is represented by xt starting at time t=0 ya is the aging parameter or filtering factor, with 0 < a < 1. In another approach, a Savitzky-Golay filter could be used to filter the motion profile of ΜΛ / a / ZUZZ / UU l ÓZO railway vehicle of a railway vehicle 5. Figure 7 shows the curve from Figure 5, which has been filtered using a Savitzky-Golay filter. The filtered rail vehicle motion profile is shown as the continuous curve 46. The circles correspond to the measurement points, which have already been shown in Figure 5. The index on the horizontal axis corresponds to the time domain. Each index point represents a specific time at which the transmitted time measurement is taken. The Savitzky-Golay (SGF) filter is a digital filter used to remove outliers and preserve the signal's trend. It employs a convolution procedure, placing the subset of adjacent points within a window of size 2w+1 using a polynomial of degree p. The parameter w specifies the number of samples used for filtering. This is referred to as the window parameter, as it specifies how many samples before and after the sample of interest are used for filtering (2w+1). The parameter p specifies the degree of the polynomial function used to approximate the trend of the points within the window. For example, if p=1, a linear function is used for filtering. Higher p values ​​are used to approximate samples with large variations.In practice, the curve filtering used by SGF is performed by applying the convolution coefficients and normalization parameter values ​​to the original dataset. In the example in Figure 7, a window size of 5 and a cubic polynomial (3) are used. Depending on the specific configuration, the window size p could be adapted to the respective track circuits. In another modality, a Hodrick-Prescott filter could be used to filter the rail vehicle movement profile of a rail vehicle 5. Figure 8 shows the curve from Figure 5, which has been filtered using a Hodrick-Prescott filter. A Hodrick-Prescott filter (HPF) is a digital filter used to remove any cyclical components from the time series sequence and to provide the signal trend. The filtered rail vehicle movement profile is shown with the continuous curve 48. The circles correspond to the measurement points, which have already been shown in Figure 5. The index on the horizontal axis corresponds to the time domain. Each index point represents a specific time at which the transmitted time measurement is taken. The present invention could use one of the above filters to filter the data provided by the transceiver 6. In other embodiments, no filtering filter is used or other filtering filters could be used. In some modes, the same filtering parameters and filters will be used for all rail vehicle movement profiles. For this purpose, the filtering parameter and / or filter are stored, for example, in memory, specifically controller 7's memory. ML / a / ZUZZ / UU l ózo Next, the second stage 32 is explained, in which the characteristics are extracted. This stage is divided into one or more substages. In a first sub-stage, the rail vehicle movement profiles, in particular the filtered rail vehicle movement profiles, are standardized. In a second sub-stage, the first and second derivatives of the normalized profiles of railway vehicle movement and an area between the curves are calculated. In a third sub-stage, the characteristics of interest that will be used in the third stage 34 are determined. Next, the first sub-stage is explained, namely, the normalization of the railway vehicle movement profiles of a railway vehicle, in particular, the filtered railway vehicle movement profiles. According to one modality, a z-normalization is applied to railway vehicle movement profiles, in particular, filtered railway vehicle movement profiles. It should be noted that other normalization methods could also be used. For example, other normalization methods that can be used here are average and minimum-maximum normalizations. Using minimum-maximum normalization, all original samples are scaled to the range [0,1]. For this purpose, the (filtered) rail vehicle motion profile is presented as a sequence of equidistant samples of the time-transmitted stream, Sk =<Ski,Sk2.....SkNk> where Skn represents the nth sample of the kth rail vehicle and Nk represents the total number of points on the kth rail vehicle moving along the track. For example, each sample corresponds to a value of the transmitted (filtered) current in an index in Figures 6-8. According to one modality, the average pk and standard deviation σk of the rail vehicle motion profile of the k-th rail vehicle (filtered) are calculated. Then, the following equation is used and applied to each sample Skn (with n=1 for Nk) aíEquation (2), where is the normalized value, pk is the average, Ok is the standard deviation and Skn is the nth sample of the k-th rail vehicle. Then, the normalized (filtered) rail vehicle movement profile is determined with the following formula: 3^=5^,5^,...,5^ , with ό'; which is the normalized value in the nth sample of the kth rail vehicle. It should be noted that normalization is applied to samples of the (filtered) rail vehicle motion profile of an individual rail vehicle in motion. This allows for the determination of the motion of rail vehicles whose original range might not be the same, for example, due to different transmitter gain values, etc. The second sub-stage is then described, namely, the calculation of the first and second derivatives of the normalized profiles of railway vehicle movement and an area between the curves. For the purpose of describing the shape of a railway vehicle movement profile, presented ML / a / ZUZZ / UU f ozo Equation (3) 7.7 . For example, on the curve of ML / a / ZUZZ / UU f ozo Equation (4) (n) . . Given yskn+i delan1 1 as the sequence of values ​​of the normalized rail vehicle movement profile, a first and second derivative are determined. In one example, the first and second derivatives are calculated using numerical procedures. However, other estimation or calculation methods could also be used. The first derivative, denoted as dn, at the point .7' , is calculated as dW _ («)un ^kn+l ^kn (n) where I > 1 represents the distance, in indices, of two sample points kn+i and the first derivative is calculated at every I point, starting with the first sample normalized (filtered) rail vehicle motion profile. Figure 9 shows a curve of a calculation of the samples of the first derivative of the normalized (filtered) motion profile values ​​with l=2. It should be noted that in the mode shown, the derivative is calculated only for the points represented as filled circles in the figure. In other modes, the derivative is calculated for all sample points or index values. The second derivative ddn at the point snes calculated with the following formula: ddW _ 2s(n>+ s(n)““n ¿kn—l ^kn ^kn+l (n) where I > 1 presents the distance, in two-point indices of samples kn-l nth sample q',','. Figure 10 shows a curve from a sample calculation of the second derivative of the normalized rail vehicle motion profile values ​​(filtered) with l=2. It should be noted that the second derivative is only calculated for the sample points or index values ​​presented as filled circles in the figure. In other modalities, the derivative is calculated for all sample points or index values. Furthermore, in some methods, the distance I might be different for the first and second derivatives. Ideally, the distance I is the same for both. Additionally, the number of points depends on the number of measured values. It should be noted that in Figures 9 and 10 and in the following figures, the upper index (n) used here earlier in equation (4) to indicate that the calculations are based on normalized values ​​and the lower index k used here earlier to indicate the k-th railway vehicle is not shown. Next, the area between the normalized (filtered) rail vehicle movement profile and the line connecting the first and last values ​​of the movement profile is determined, called the area between curves (ABC). To do this, the area between the x-axis and the normalized (filtered) rail vehicle movement profile, or the area under the curve (AUC), is determined first; see the shaded area in Figure 11. The AUC is estimated based on the sum of eight trapezoids Ai, A2, ... As. If the Sk sequence includes more or fewer points, the number of trapezoids could also vary accordingly. Figure 12 shows an area under the line (AUL) Al of the normalized (filtered) rail vehicle movement profile, which is defined by the area under a line defined by the first and last sample or index value, here si and sg. The area under the line is the shaded area in Figure 12. Figure 13 shows the area between curves (ABC) A of the normalized (filtered) rail vehicle movement profile, which is the difference between the AUL and the AUC. The ABC is shown as the shaded area in Figure 13. In alternative modalities, the ABC value could also be determined using other methods. For example, calculated derivatives could be used, or a numerical integration method could be employed. Next, the third sub-stage will be explained. In the next sub-step, one or more of the features are extracted based on the first derivative, the second derivatives, and / or the ABC determination. For example, one or more of the following features are extracted: • The minimum of the first derivative, • The maximum of the first derivative, • The minimum of the second derivative, • The maximum of the second derivative, • The variation of the first derivative, • The variation of the second derivative, • The median of the first derivative, • The median of the second derivative, • The average of the first derivative, • The average of the second derivative, • The standard deviation of the first derivative, • The standard deviation of the first derivative, • The percentage of the first derivative that is negative, • The percentage of the second derivative that is negative, and / or • The surface area value ABC. In one mode, the previous stages of feature extraction and filtering are performed in the same way as a reference, in particular, the simulated rail vehicle movement profiles and the actual measured rail vehicle movement profile as can be seen on the left and right sides of the flowchart in Figure 4. Next, the third stage 34 of the classification of railway vehicle movement profiles (filtering) is explained with respect to figures 14-19A. This is done firstly with the simulated profiles of railway vehicle movement in order to generate a model of the best representations of the different movement profiles ΜΛ / a / ZUZZ / UU 1ó Or railway vehicle. According to the modalities, a modified k-average clustering method is performed. A k-average clustering method is a vector quantization that aims to divide n observations into k clusters, where each observation belongs to a cluster with the nearest average. The nearest average is also a centroid and serves as a prototype for this cluster. The k-average clustering method is based on the motion profiles of a railway vehicle. As discussed previously, the k-average clustering method is divided into a training phase and a testing phase. As explained, during the training phase, an input dataset generated through simulations is used. In other cases, different reference profiles of railway vehicle movement could be used. Artificially, simulated rail vehicle movement profiles exhibit different, though known, characteristics such as acceleration, deceleration, and constant speed. At the end of this phase, as explained below, the most representative values ​​for each type of rail vehicle movement profile, called centroids, are determined. These centroids are used as a model or set of thresholds to classify each real rail vehicle movement profile as it occurs—for example, acceleration, deceleration, or constant speed—based on its distance from the centroids. Depending on the model, models are created for various track lengths. For example, a reference curve is generated, which establishes a relationship between distance and transmitted and measured current. Figure 14 again shows a reference curve (the solid line 50) that includes the respective points where a branch is positioned (see the circles). As a reference, a diagonal dashed line is shown connecting the endpoints of the reference curve 50. In a subsequent stage, several different rail vehicle motion profiles are created. Each profile could be different from the others. The rail vehicle motion profiles could include rail vehicles decelerating, rail vehicles at constant speed, and rail vehicles accelerating. For each rail vehicle motion profile, the velocity, acceleration (if applicable), and related location are assumed to be found. Then, using the velocity profiles and the previously determined reference curve, a simulated or reference rail vehicle motion profile is determined that shows the transmitted current with respect to time. The data could be stored as a sequence of simulated measurement values. Figure 15 shows sequences of the normalized (filtered) rail vehicle motion profiles for different simulated rail vehicles moving along the railway track. Rail vehicles decelerating are shown with a circle, rail vehicles moving at a constant speed are shown with a square, and rail vehicles ML / a / ZUZZ / UU Simulated accelerations are shown with a triangle. In one mode, normalization could be performed as discussed above using z-normalization. In addition, prior to normalization, the simulated or reference rail vehicle motion profiles could have been filtered. Figure 16 shows the first and second derivatives of the filtered and normalized rail vehicle motion profiles from Figure 15. The first row of graphs shows the first derivative, and the second row shows the second derivative. The area under curves (AUC) values ​​are also calculated. The first column shows the data for rail vehicle motion profiles that are decelerating. The second column shows the data for rail vehicle motion profiles that are at constant speed. The third column shows the data for rail vehicle motion profiles that are accelerating. The result of one or more of these calculations is used for the classification. This simulated dataset is used for training to generate average k-centroids, which will then be used to classify the actual rail vehicle motion profiles as accelerating, decelerating, or at constant speed. In general, many simulated rail vehicle motion profiles are created—for example, between 15 and 50, specifically 20—for a given type of rail vehicle motion profile (accelerating, decelerating, constant speed). However, for clarity, only two different simulations per rail vehicle motion profile type are used in these examples. In a first stage, the k-average centroids are calculated with respect to a given type of rail vehicle movement profile using the following formula: 1V And / ¿n-nnun typc^H. Equation (5), Where x is a vector of the characteristics of the railway vehicle motion profile, the profile G^ti.nn typi presents the association of the simulated railway vehicle motion profile as a function of its railway vehicle motion profile type (such as accelerating, decelerating, or constant speed), N is the number of train movements within a railway vehicle motion profile type. In one mode, the vector xn could include selected features as a vector, for example, the ABC value, the average of the first derivative, and the negative percentage of the second derivative. In other modes, one or more other features could be selected, in particular, from the list of possible features shown above. For each rail vehicle movement profile, a unique vector is created that includes the selected features. The value h¡> >tr.iin typi is selected from the values ​​0 or 1. For example, if centroids for railway vehicle motion acceleration profiles are created, the value is 1 for acceleration motion profiles and 0 for other motion profiles (deceleration or constant speed). In other ML / a / ZUZZ / UU f ózo words, this variable is used to select the correct feature vector for centroid calculation. Next, some of the extracted features are grouped using the k-average method. For example, in one mode, the ABC value, the first derivative variation, and the negative percentage of the second derivative are used as features. The first derivative variation is calculated as the difference between the first and last derivative values. The negative percentage of the second derivative is calculated because the percentage of time during the sequence of values ​​is negative. In other modes, other features could be chosen. Below, the centroids of a pair of the selected features above (the ABC value, the first derivative variation, and the negative percentage of the second derivative) are shown with their respective centroids. The centroids can be designated with pc and are shown as a star in the following figures. Figure 17 shows the variation of the first derivative with respect to the ABC value. Decelerating railway vehicles are shown with a circle, railway vehicles moving at a constant speed are shown with a square, simulated accelerating railway vehicles are shown with a triangle, and the three respective centroids are shown as stars 60, 62, and 64. Figure 18 shows the negative percentage of the second derivative with respect to the ABC value. Decelerating rail vehicles are shown with a circle, rail vehicles moving at a constant speed are shown with a square, simulated accelerating rail vehicles are shown with a triangle, and the three respective centroids are shown as stars 60, 62, and 64. Figure 19 shows the variation of the first derivative with respect to the negative percentage of the second derivative. Decelerating rail vehicles are shown with a circle, rail vehicles moving at a constant speed are shown with a square, simulated accelerating rail vehicles are shown with a triangle, and the three respective centroids are shown as stars 60, 62, and 64. Figure 19A shows a three-dimensional graph of the railway vehicle motion profile characteristic and the respective centroids. Rail vehicles decelerating are shown with a circle, rail vehicles moving at a constant speed are shown with a square, simulated accelerating rail vehicles are shown with a triangle, and the three respective centroids are shown as stars 60, 62, 64. Centroid 60 refers to the centroid of the rail vehicle motion profile type of decelerating rail vehicles, centroid 62 refers to the centroid of the rail vehicle motion profile type of rail vehicles at constant speeds, and centroid 64 refers to the centroid of the rail vehicle motion profile type of accelerating rail vehicles. Once the training is complete, the actual data from the railway vehicle regarding transmitted and measured currents could be used for the purpose of classification using the results. ML / a / ZUZZ / UU l ózo of the k-average classification training phase described above. The same preparation steps are used, for example, filtering, normalization, and feature extraction as described above. According to the modalities, the same characteristics are extracted for the actual profiles of railway vehicle movement in terms of training. Then the distance, in particular the Euclidean distance between the feature extracted from the railway vehicle motion profile as a vector xry the centroids pc, where c can designate a railway vehicle at constant speed, decelerating or accelerating. In one modality, the vector xr could include the selected features as a vector, for example, the ABC value, the variation of the first derivative, and the negative percentage of the second derivative. It should be noted that c depends on the number of different types of rail vehicle motion profiles with which the k-average classification method has been trained. In other words, it is not limited to the three types of rail vehicle motion profiles mentioned above. The rail vehicle is assigned to the nearest centroid in terms of the calculated (Euclidean) distance. Formally, the classification is determined as c* = arg niin||xrμ, ||_>. Equation (6), where <' represents the type of rail vehicle motion profile (e.g., a rail vehicle at constant speed, decelerating, or accelerating) with the nearest centroid in terms of the Euclidean distance. For the purpose of illustrating the classification process, Figure 20 shows a compilation of the actual rail vehicle movement profiles, which are filtered and normalized, in particular, with respect to an index that corresponds to the time samples. For each measurement series for a specific rail vehicle, the same set of features used for determining the centroids is defined, for example, in the form of the vector Xr. In one example, the vector Xr's set of features includes the ABC value, the average of the first derivative, and the negative percentage of the second derivative. In other configurations, other features could be selected and / or the number of features could be different. As detailed above, in particular with respect to equation 6, the classification of the measured series for each actual railway vehicle movement is based on the minimum distance between its characteristic vector xry and the k-average centroids. Figures 21, 22, 23, and 23A show the respective characteristics of the vehicles. Vehicles decelerating are shown with a circle, those moving at a constant speed with a square, and those accelerating with a triangle. The three respective centroids are shown as stars. The centroids are shown in the same positions as in Figures 17, 18, 19, and 19A. Centroid 60 refers to the centroid of the motion profile type of the ML / a / ZUZZ / UU f ózo decelerating railway vehicles, centroid 62 refers to the centroid of the motion profile type of railway vehicles with constant speeds and centroid 64 refers to the centroid of the motion profile type of accelerating railway vehicles. Figure 21 shows the variation of the first derivative with respect to the ABC value. Figure 22 shows the negative percentage of the second derivative with respect to the ABC value. Figure 23 shows the variation of the first derivative with respect to the negative percentage of the second derivative, and Figure 23A shows a three-dimensional representation. Figures 24, 25, and 26 show, respectively, the normalized motion profile with respect to the index, of a railway vehicle classified as decelerating (Figure 24), classified as traveling at constant speed (Figure 25), and classified as accelerating (Figure 26). For comparison, the derivation data, which correspond to the reference curve in Figure 14, are shown. This invention provides a method by which data from passing railway vehicles collected as transmitted current versus time are analyzed to determine whether a train was accelerating, decelerating, or had a relatively constant speed. The location or determination of a distance of the rail vehicle 5 with respect to the feed at points 9 can be improved based on the classification. Furthermore, the dynamic time distortion process for calibration could also be improved by removing those railway vehicles that are accelerating or decelerating. This disclosure can be used to overcome the limitations of the DTW process in order to avoid location errors due to rail vehicle acceleration or deceleration. This will allow for more accurate location and enable railways to use track circuits to provide real-time rail vehicle location with improved resolution. Figure 27 shows a flowchart of a method according to a modality. For example, the method could be implemented in controller 7 of a circuit system on track 1 shown in Figure 1. In a first stage 1000, the controller obtains a rail vehicle movement profile. For example, the controller could store a plurality of transmitted and measured currents that are measured by transceiver 6. The plurality of transmitted and measured currents are aligned in a sequence with respect to time to obtain the rail vehicle movement profile, in particular, when a rail vehicle is moved along the railway track that includes rails 3. In stage 1010, the rail vehicle movement profile is filtered. For example, as detailed previously, an exponential filter, a Savitzky-Golay filter, or a Hodrick-Prescott filter could be used for this purpose. Other suitable filters could also be used to remove or eliminate interleaving in the data. In stage 1020, the rail vehicle movement profile is normalized. For example, a z-normalization could be used for this purpose as explained above. ML / a / ZUZZ / UU l ózo In step 1030, one or more features of the standardized rail vehicle movement profile are extracted. For example, for this purpose, one or more derivatives of the standardized rail vehicle movement profile could be calculated. In step 1040, the distance of the extracted features with respect to each centroid of a given rail vehicle motion profile type in a classification process is calculated. If the rail vehicle motion profile type includes three types—namely, an accelerating rail vehicle, a rail vehicle at constant speed, and a decelerating rail vehicle—each rail vehicle motion profile type has a centroid, so three distances are calculated. In step 1050, the rail vehicle movement profile for the rail vehicle movement profile type with the nearest centroid is assigned. This makes it possible to obtain the rail vehicle movement profile type from a measured rail vehicle movement profile, which could then be used by other methods, for example, to calibrate the circuit system of track 1 or to determine the position of a rail vehicle on the track. The present invention could also refer to the following forms: According to one aspect, a computer-implemented method is provided for determining the type of rail vehicle motion profile from a rail vehicle motion profile, wherein the rail vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver with respect to time, comprising - obtain a rail vehicle movement profile; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. Additional features could refer to one or more of the following characteristics, which could be combined in any technically feasible combination: - the extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, where the features are calculated from the normalized rail vehicle movement profile; - Before the normalization stage, the rail vehicle movement profile is filtered; ML / a / ZUZZ / UU l ózo - the filtering is performed using one or more filters selected from the group of an exponential filter, a Savitzky-Golay filter and a Hodrick-Prescott filter; - the extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, where the features are calculated from the filtered and normalized rail vehicle movement profile; - The classification process for determining centroids comprises: - obtain a plurality of railway vehicle movement reference profiles for each type of railway vehicle movement profile, wherein for each railway vehicle movement reference profile, the type of railway vehicle profile is known; - to standardize each of the reference profiles for railway vehicle movement; and - extract one or more features for each standardized reference profile of railway vehicle movement; - Determine, for each type of railway vehicle movement profile, the centroid of the extracted features. - the plurality of railway vehicle movement profiles comprises an accelerating railway vehicle, a railway vehicle at a constant speed, and a decelerating railway vehicle; - obtaining a plurality of reference rail vehicle movement profiles for each type of rail vehicle movement profile includes simulating the respective rail vehicle movement profiles based on the characteristics of the track on which the rail vehicles move; - the characteristics of the track include the lengths and strength of the rail; - The characteristics extracted from the standardized reference profiles of railway vehicle movement are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value; According to another aspect, a method is provided for calibrating a track circuit system, wherein the track circuit system includes a transceiver that is connected to a pair of rails of a railway track and a controller that receives from the transceiver the transmitted and measured currents of a ML / a / ZUZZ / UU l ózo circuit of the track, the method comprising: - to recover, by means of the controller, a plurality of railway vehicle movement profiles as a function, respectively, of the sequence of transmitted currents and track circuit measurements with respect to time; - for each of the railway movement profiles, determine the type of railway vehicle movement profile according to a method described herein; - retain the rail vehicle movement profiles of rail vehicles at a constant speed; and - calibrate, by means of the controller, the track circuit system according to the retained movement profiles. Additional features could refer to one or more of the following characteristics, which could be combined in any technically feasible combination: - The calibration of the track circuit system uses a dynamic time-warping process to calculate the transmitted current with respect to the location relationship of the rail vehicle. According to another aspect, a method is provided for determining the position of a railway vehicle using a track circuit, wherein the track circuit system includes a transceiver that is connected to a pair of rails of a railway track at a feed point and a controller that receives from the transceiver the transmitted and measured currents of a track circuit, wherein the method comprises: - calibrate a track circuit system according to a method described herein; - Recover a railway vehicle movement profile based on a measured sequence of transmitted currents and track circuit measurements over time; - determine the distance of the nearest axle of the rail vehicle with respect to a feed at a point on the rail; and - Determine the position of the railway vehicle based on distance. According to an additional aspect, a non-transient, computer-readable storage medium is provided comprising instructions, which, when executed by a computer, cause the computer to perform the following steps: - obtaining a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver with respect to time; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. ML / a / ZUZZ / UU l ózo According to another aspect, a track circuit system controller is provided, the track circuit system including a transceiver that is connected to a pair of rails of a railway track and the controller receives the transmitted and measured currents from the transceiver, where the controller is adapted to: - obtaining a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver with respect to time; - normalize the rail vehicle movement profile; - extract one or more features from the standardized rail vehicle movement profile; - to determine the distance of the extracted features with respect to each centroid of a given type of railway vehicle movement profile in a classification process; and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid. Additional features could refer to one or more of the following characteristics, which could be combined in any technically feasible combination: - the processor is also adapted before the normalization stage to filter the railway vehicle movement profile; - the filtering is performed using one or more filters selected from the group of an exponential filter, a Savitzky-Golay filter and a Hodrick-Prescott filter; - The extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, where the features are calculated from the normalized rail vehicle movement profile. - The classification process for determining centroids comprises: - obtain a plurality of railway vehicle movement reference profiles for each type of railway vehicle movement profile, wherein for each railway vehicle movement reference profile, the type of railway vehicle profile is known; - to standardize each of the reference profiles for railway vehicle movement; and - extract one or more features for each standardized reference profile of railway vehicle movement; - Determine, for each type of railway vehicle movement profile, the centroid of the extracted features. - the plurality of railway vehicle movement profiles comprises a railway vehicle ML / a / ZUZZ / UU l OZO accelerating, a railway vehicle at a constant speed and a railway vehicle decelerating. The written description uses examples to illustrate the invention, including the best embodiment, and also to enable anyone skilled in the art to carry out and use the invention. While the invention has been described in terms of several specific embodiments, those skilled in the art will recognize that the invention can be practiced with modifications within the spirit and scope of the claims. The patentable scope of the invention is defined by the claims and could include other examples that occur to those skilled in the art. It is intended that such other examples fall within the scope of the claims.

Claims

1. A computer-implemented method for determining the rail vehicle motion profile type from a rail vehicle motion profile, wherein the rail vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver over time, comprising: - obtaining a rail vehicle motion profile; - normalizing the rail vehicle motion profile; - extracting one or more features from the normalized rail vehicle motion profile; - determining the distance of the extracted features from each centroid of a determined rail vehicle motion profile type in a classification process; and - assigning the rail vehicle motion profile to the rail vehicle motion profile type with the nearest centroid.

2. The method according to claim 1, wherein the extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, wherein the features are calculated from the normalized rail vehicle motion profile.

3. The method according to claim 1 or 2, wherein, prior to the standardization stage, the railway vehicle movement profile is filtered.

4. The method according to claim 3, wherein the filtering is performed using one or more filters selected from the group of an exponential filter, a Savitzky-Golay filter, and a Hodrick-Prescott filter.

5. The method according to claim 3, wherein the extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, wherein the features are calculated from the filtered and normalized rail vehicle motion profile.

6. The method according to claim 1 or 2, wherein the classification process for determining the centroids comprises: - obtaining a plurality of railway vehicle motion reference profiles for each type ML / a / ZUZZ / UU railway vehicle motion profile, wherein for each railway vehicle motion reference profile, the type of railway vehicle profile is known; - standardizing each of the railway vehicle motion reference profiles; and - extracting one or more features for each standardized railway vehicle motion reference profile; - determining for each type of railway vehicle motion profile, the centroid of the extracted features.

7. The method according to claim 6, wherein the plurality of railway vehicle movement profiles comprises an accelerating railway vehicle, a railway vehicle at a constant speed, and a decelerating railway vehicle.

8. The method according to claim 6, wherein obtaining a plurality of railway vehicle movement reference profiles for each type of railway vehicle movement profile includes simulating the respective railway vehicle movement profiles based on the characteristics of the track on which the railway vehicles move.

9. The method according to claim 8, wherein the track characteristics include the rail lengths and strength.

10. The method according to claim 6, wherein the features extracted from the standardized reference profiles of railway vehicle movement are one or more of the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value.

11. A method for calibrating a track circuit system, wherein the track circuit system includes a transceiver connected to a pair of rails of a railway track and a controller receiving from the transceiver the transmitted and measured currents of a track circuit, the method comprising: - retrieving, by means of the controller, a plurality of railway vehicle motion profiles as a function, respectively, of the sequence of transmitted and measured currents of the track circuit with respect to time; - for each of the railway motion profiles, determining the type of railway vehicle motion profile according to the method of claim 1; - retaining the railway vehicle motion profiles of railway vehicles at a constant speed; and - calibrating, by means of the controller, the track circuit system as a function of the retained motion profiles.

12. The method according to claim 11, wherein the calibration of the circuit system of ML / a / ZUZZ / UU l ózo the track uses a dynamic time-warping process to calculate the transmitted current with respect to the location relationship of the railway vehicle.

13. A method for determining the position of a railway vehicle using a track circuit, wherein the track circuit system includes a transceiver connected to a pair of rails of a railway track at a point feed and a controller receiving from the transceiver the transmitted and measured currents of a track circuit, wherein the method comprises: calibrating a track circuit system according to the method of claim 11; retrieving a railway vehicle motion profile as a function of a measured sequence of transmitted and measured currents of the track circuit with respect to time; determining the distance of the nearest axle of the railway vehicle with respect to a point feed on the rail; and determining the position of the railway vehicle as a function of the distance.

14. A computer-readable, non-transient storage medium comprising instructions which, when executed by a computer, cause the computer to perform the following steps: - obtain a rail vehicle motion profile, wherein the rail vehicle motion profile comprises a sequence of currents transmitted and measured by a transceiver of a track circuit with respect to time; - normalize the rail vehicle motion profile; - extract one or more features from the normalized rail vehicle motion profile; - determine the distance of the extracted features with respect to each centroid of a given rail vehicle motion profile type in a classification process; and - assign the rail vehicle motion profile to the rail vehicle motion profile type with the nearest centroid.

15. A track circuit system controller, the track circuit system including a transceiver connected to a pair of rails of a railway track, the controller receiving transmitted and measured currents from the transceiver, wherein the controller is adapted to: - obtain a railway vehicle motion profile, wherein the railway vehicle motion profile comprises a sequence of transmitted and measured currents from a track circuit transceiver over time; - normalize the railway vehicle motion profile; - extract one or more features from the normalized railway vehicle motion profile; - determine the distance of the extracted features from each centroid of a given railway vehicle motion profile type in a classification process;and - assign the rail vehicle movement profile to the rail vehicle movement profile type with the nearest centroid.; 16. The controller according to claim 15, wherein the processor is further adapted ML / a / ZUZZ / UU l ózo before the normalization stage to filter the railway vehicle motion profile.

17. The controller according to claim 16, wherein the filtering is performed using one or more filters selected from the group of an exponential filter, a Savitzky-Golay filter, and a Hodrick-Prescott filter.

18. The controller according to any one of claims 15-17, wherein the extracted features are one or more from the selected group of: the minimum of the first derivative, the maximum of the first derivative, the minimum of the second derivative, the maximum of the second derivative, the variation of the first derivative, the variation of the second derivative, the median of the first derivative, the median of the second derivative, the average of the first derivative, the average of the second derivative, the standard deviation of the first derivative, the standard deviation of the first derivative, the percentage of the first derivative that is negative, the percentage of the second derivative that is negative, and / or the ABC surface value, wherein the features are calculated from the normalized rail vehicle motion profile.

19. The controller according to any one of claims 14-17, wherein the sorting process for determining the centroids comprises: - obtaining a plurality of railway vehicle motion reference profiles for each type of railway vehicle motion profile, wherein for each railway vehicle motion reference profile, the type of railway vehicle profile is known; - normalizing each of the railway vehicle motion reference profiles; and - extracting one or more features for each normalized railway vehicle motion reference profile; - determining for each type of railway vehicle motion profile, the centroid of the extracted features.

20. The controller according to claim 19, wherein the plurality of railway vehicle movement profiles comprises an accelerating railway vehicle, a railway vehicle at a constant speed, and a decelerating railway vehicle.

Images