"\"\\\"Modern signaling systems play a major role to provide safety net to prevent accidents due to human errors, and the management of railway operation to maximize the use of scarce resources and optimize the investment in infrastructure. With the technological evolution of the ERTMS-ETCS radio-based signaling system, that is envisaged on the new MoU signed beginning of 2012 by the major European railways stake-holders, satellite positioning technology will become a priority. In fact the GNSS technology matches primarily unmet demand for cost-efficient train control systems such those required on low traffic and regional lines and it matters as much as ever for drastically reducing the life cycle costs of the train signaling systems. The main challenge to face in the introduction of GNSS-based localization systems on the train signaling systems is to not impacting the high safety levels reached so far by the current systems which rely on reliable but costly odometry-based and wayside equipments to estimate the train position. Since the GNSS constellations are rapidly evolving towards a redundant and resilient global infrastructure we have developed a novel multi-constellation PVT algorithm for train localization determination, specifically designed to handle the case of multiple track scenarios. This solution is compatible with the ERTMS-ETCS train control system platform and compliant with the SIL-4 safety requirements of CENELEC railways norms i.e., Tolerable Hazard Rate (THR), 10E-9 ? THR ?10E-8. Railways applications are referred as safety-related systems, a sub case of the well known safety-of-life GNSS application and they requires a higher performance especially in terms of availability, continuity, integrity and accuracy. Conventional, stand-alone, GPS systems are unable to provide the positioning information with an error bounded by a protection level compliant with the safety requirements of railways applications. On the other hands, recent developments of GNSS prove to be inspiring for safety-of-life applications: for instance, modernized GNSS signals are broadcast with increased power and enhanced characteristics for multipath mitigation, while the presence of multiple constellations may potentially increase the overall availability along a rail line. To meet the SIL 4 Requirements, we adopted an architecture charactized by a dedicated integrity monitoring and augmentation network and by the use of multi-constellation receivers. More in detail, in our system, the train is equipped with the GNSS Location Determination System On-Board Unit (LDS OBU), which provides the PVT estimate to the existing localization system, and the Authorization To Proceed (ATP). Each GNSS LDS OBU is equipped with (i) two GNSS receivers, (ii) a local processor performing the PVT estimation starting from local measures, (iii) a track DataBase and the augmentation data received from a Track Area LDS Safety (TALS) server, and (iv) a communication module. The PVT algorithm estimates the train location by explicitly accounting for the fact that the train is constrained to lie on a railway track. Basically, exploiting this constraint allows to estimate train location even when only two satellites are in view. Effective reduction in the number of required satellites to make a fix, when track constraint is applied, depends on the track-satellite geometry. In essence, satellites aligned along the track give more information than those at the cross-over. Satellites in excess can then be employed either to increase accuracy or to increase integrity and availability. In [1] the authors presented a SIL-4 solution for PVT train estimation, valid for single-track scenarios. However, multiple-track scenarios have to be faced to account for real situations and those cases are more challenging due to the fact that inter-track separation is rather smaller compared to the inter-train separation along the same track. As a matter of fact, the cross-track protection level (1.5 m) is lower than the along track protection level (15 m) of one order of magnitude. In this paper, we address the problem of, PVT estimation in presence of multiple tracks. This can be formulated as a combination of hypothesis testing (i.e., which is the current track where the train lies on, or, better, which is the probability of a train lying on a given track?) and parameter estimation (i.e., given a track, which is the curvilinear abscissa of the train receiver?). Then, assuming that the train can be located along one of M tracks, and considering the k-th hypothesis as corresponding to the k-th track, proceeding as in [2], we first (i) •\\\\tEstimate, for each candidate track, the curviliner abscissa of the receiver by means of a Weighted Least Square Estimator (WLSE), assuming that the corresponding hypothesis is true, and (ii) then, we use these conditional estimates to compute the measurement residuals conditioned to each hypothesis, and from them the a posteriori probability of each track. Those a posteriori probabilities can be combined in generalized log-likelihood ratio tests to detect the current track. In fact, assuming that the hypotheses are uniformly distributed, the Bayesian (optimal) track detection rule selects the hypothesis corresponding to the largest of them. Moreover multiple observations can be combined observing that track occupancy can be modeled as a Markov time series. Since the generalized log-likelihood ratio magnitudes provide information about the reliability of the decision, their values are compared to thresholds to verify that enough information has been acquired before a decision on which track the train is lying on is transmitted to the ATP processor. For each track, the conditional PVT estimate is computed by solving a set of non-linear equations relating the observables (e.g. pseudoranges and carrier phases) to the receiver curvilinear abscissa and clock offset, by means of an iterative WLSE procedure, accounting for the different statistics of the equivalent measurement noise, due to both satellite elevation and signal characteristics specific of each constellation. In principle, extended Kalman filter algorithms could also be considered, then accounting for train dynamics. Recently, particle filters have been proposed in place of extended Kalman filters to solve the pseudorange nonlinear equations; nevertheless, their computational complexity qualifies them as not mature for high integrity receivers. In this contribution, a detailed description of the overall processing is given. The performance of the track detector versus inter-track distance, observation time duration, signal-to-noise ratio and observables (pseudoranges and carrier phase) are discussed. Then, the impact of satellite failures on the PVT error magnitude and track detection is exploited in terms of Hazardous Misleading Information for both cases i.e., autonomous and augmented OBU operational modes. Finally, the assessment of the performance is also provided by means of simulation results making use of both synthetic data (Monte Carlo simulations) and measures recorded on a railway test bed environment. Reference [1] A. Neri, A. Filip, F. Rispoli, and A.M. Vegni, “An Analytical Evaluation for Hazardous Failure Rate in a Satellite-based Train Positioning System with reference to the ERTMS Train Control Systems,” in Proc. of ION GNSS 2012, September 17-21, 2012, Nashville, TN, USA [2]\\\\tVan Trees, “Detection, Estimation, and Modulation Theory”, Part I, p. 92, John Wiley & Sons, 2001\\\"\""

Neri, A., Vegni, A.M., Rispoli, F. (2013). A PVT Estimation for the ERTMS Train Control Systems in presence of Multiple Tracks. In Proceedings of the 26th International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS+ 2013) (pp.631-644). The Institute of Navigation.

### A PVT Estimation for the ERTMS Train Control Systems in presence of Multiple Tracks

#####
*NERI, Alessandro;VEGNI, ANNA MARIA;*

##### 2013-01-01

#### Abstract

"\"\\\"Modern signaling systems play a major role to provide safety net to prevent accidents due to human errors, and the management of railway operation to maximize the use of scarce resources and optimize the investment in infrastructure. With the technological evolution of the ERTMS-ETCS radio-based signaling system, that is envisaged on the new MoU signed beginning of 2012 by the major European railways stake-holders, satellite positioning technology will become a priority. In fact the GNSS technology matches primarily unmet demand for cost-efficient train control systems such those required on low traffic and regional lines and it matters as much as ever for drastically reducing the life cycle costs of the train signaling systems. The main challenge to face in the introduction of GNSS-based localization systems on the train signaling systems is to not impacting the high safety levels reached so far by the current systems which rely on reliable but costly odometry-based and wayside equipments to estimate the train position. Since the GNSS constellations are rapidly evolving towards a redundant and resilient global infrastructure we have developed a novel multi-constellation PVT algorithm for train localization determination, specifically designed to handle the case of multiple track scenarios. This solution is compatible with the ERTMS-ETCS train control system platform and compliant with the SIL-4 safety requirements of CENELEC railways norms i.e., Tolerable Hazard Rate (THR), 10E-9 ? THR ?10E-8. Railways applications are referred as safety-related systems, a sub case of the well known safety-of-life GNSS application and they requires a higher performance especially in terms of availability, continuity, integrity and accuracy. Conventional, stand-alone, GPS systems are unable to provide the positioning information with an error bounded by a protection level compliant with the safety requirements of railways applications. On the other hands, recent developments of GNSS prove to be inspiring for safety-of-life applications: for instance, modernized GNSS signals are broadcast with increased power and enhanced characteristics for multipath mitigation, while the presence of multiple constellations may potentially increase the overall availability along a rail line. To meet the SIL 4 Requirements, we adopted an architecture charactized by a dedicated integrity monitoring and augmentation network and by the use of multi-constellation receivers. More in detail, in our system, the train is equipped with the GNSS Location Determination System On-Board Unit (LDS OBU), which provides the PVT estimate to the existing localization system, and the Authorization To Proceed (ATP). Each GNSS LDS OBU is equipped with (i) two GNSS receivers, (ii) a local processor performing the PVT estimation starting from local measures, (iii) a track DataBase and the augmentation data received from a Track Area LDS Safety (TALS) server, and (iv) a communication module. The PVT algorithm estimates the train location by explicitly accounting for the fact that the train is constrained to lie on a railway track. Basically, exploiting this constraint allows to estimate train location even when only two satellites are in view. Effective reduction in the number of required satellites to make a fix, when track constraint is applied, depends on the track-satellite geometry. In essence, satellites aligned along the track give more information than those at the cross-over. Satellites in excess can then be employed either to increase accuracy or to increase integrity and availability. In [1] the authors presented a SIL-4 solution for PVT train estimation, valid for single-track scenarios. However, multiple-track scenarios have to be faced to account for real situations and those cases are more challenging due to the fact that inter-track separation is rather smaller compared to the inter-train separation along the same track. As a matter of fact, the cross-track protection level (1.5 m) is lower than the along track protection level (15 m) of one order of magnitude. In this paper, we address the problem of, PVT estimation in presence of multiple tracks. This can be formulated as a combination of hypothesis testing (i.e., which is the current track where the train lies on, or, better, which is the probability of a train lying on a given track?) and parameter estimation (i.e., given a track, which is the curvilinear abscissa of the train receiver?). Then, assuming that the train can be located along one of M tracks, and considering the k-th hypothesis as corresponding to the k-th track, proceeding as in [2], we first (i) •\\\\tEstimate, for each candidate track, the curviliner abscissa of the receiver by means of a Weighted Least Square Estimator (WLSE), assuming that the corresponding hypothesis is true, and (ii) then, we use these conditional estimates to compute the measurement residuals conditioned to each hypothesis, and from them the a posteriori probability of each track. Those a posteriori probabilities can be combined in generalized log-likelihood ratio tests to detect the current track. In fact, assuming that the hypotheses are uniformly distributed, the Bayesian (optimal) track detection rule selects the hypothesis corresponding to the largest of them. Moreover multiple observations can be combined observing that track occupancy can be modeled as a Markov time series. Since the generalized log-likelihood ratio magnitudes provide information about the reliability of the decision, their values are compared to thresholds to verify that enough information has been acquired before a decision on which track the train is lying on is transmitted to the ATP processor. For each track, the conditional PVT estimate is computed by solving a set of non-linear equations relating the observables (e.g. pseudoranges and carrier phases) to the receiver curvilinear abscissa and clock offset, by means of an iterative WLSE procedure, accounting for the different statistics of the equivalent measurement noise, due to both satellite elevation and signal characteristics specific of each constellation. In principle, extended Kalman filter algorithms could also be considered, then accounting for train dynamics. Recently, particle filters have been proposed in place of extended Kalman filters to solve the pseudorange nonlinear equations; nevertheless, their computational complexity qualifies them as not mature for high integrity receivers. In this contribution, a detailed description of the overall processing is given. The performance of the track detector versus inter-track distance, observation time duration, signal-to-noise ratio and observables (pseudoranges and carrier phase) are discussed. Then, the impact of satellite failures on the PVT error magnitude and track detection is exploited in terms of Hazardous Misleading Information for both cases i.e., autonomous and augmented OBU operational modes. Finally, the assessment of the performance is also provided by means of simulation results making use of both synthetic data (Monte Carlo simulations) and measures recorded on a railway test bed environment. Reference [1] A. Neri, A. Filip, F. Rispoli, and A.M. Vegni, “An Analytical Evaluation for Hazardous Failure Rate in a Satellite-based Train Positioning System with reference to the ERTMS Train Control Systems,” in Proc. of ION GNSS 2012, September 17-21, 2012, Nashville, TN, USA [2]\\\\tVan Trees, “Detection, Estimation, and Modulation Theory”, Part I, p. 92, John Wiley & Sons, 2001\\\"\""I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.