odometry algorithms for railway applications

28 downloads 0 Views 67KB Size Report
Universit`a degli Studi di Bologna. Dottorato di Ricerca in Meccanica Applicata. XV Ciclo. ODOMETRY ALGORITHMS. FOR RAILWAY APPLICATIONS.
Universit` a degli Studi di Bologna Dottorato di Ricerca in Meccanica Applicata XV Ciclo

ODOMETRY ALGORITHMS FOR RAILWAY APPLICATIONS PhD Thesis

Monica Malvezzi

Synthesis The aim of this work is the development of the odometry algorithm of a new ATP (Automatic Train Protection, [1], [2], [3], [4], [5], [6], [7]) system. An Automatic Train Control system provides an high level of train safety by automatically guarding against the consequences of driver errors. Automatic Train Protection is a system that ensures that trains comply with speed restrictions and prevents them passing red signals. It intervenes automatically to slow or stop the train where trains exceed speed limits or pass signals at danger. However, even if it improves the safety, ATP will not prevent all railway crashes. It can’t prevent incidents caused by signalling errors or by defects in the tracks or trains, for example. Automatic Train Protection is a part of the so-called Automatic Train Control (ATC), a more complex system that includes ATP, ATO (Automatic Train Operation) and ATS (Automatic Train Supervision). This term is adopted to describe the architecture of the automatically operated railway. The ATP system that will be analysed in this thesis is named SCMT (Sistema di Controllo della Marcia dei Treni), and will equip most of Italian trains. In this thesis, the odometry algorithm of the ATP system is developed. The developed algorithm will be also used as reference algorithm in the testing of ATP system implementation, which will be performed by means of a proper test rig ([8], [9]). Tipically an ATP system includes three components: • the wayside equipment, which generates codes to be transmitted to the train; • a track-to-train transmission system, which transmits information from the wayside equipment to the train; • train on board equipment, which elaborates the information received from the track and from sensors mounted on the train and decides the actions to be performed. The track subsystem gives to the train a series of information, for instance: • the current position; • the distance to targets (point in correspondence of which speed restrictions have to be achieved); • the target speed (i.e. the speed that has not to be exceeded when the train passes through a target point); These information are communicated to the train with the aid of fixed balises or another kind s of absolute information, for example GPS. Between two succeeding information acquisitions from the ground subsystem, the on-board subsystem continuously calculates the following data: • the minimum distance that allows to respect the speed restrictions at the next objective points: this value depends on the actual train speed, on the braking parameters and on the objective speed;

• the distance to the next information points. If the difference between the distance to one or more of the next objective points and the distance that allows to respect the speed restriction is smaller than a fixed value, the on-board subsystem intervenes, for instance by activating the emergency braking. Correct estimate of distance to target and actual velocity is crucial to evaluate residual braking resources, in terms of available deceleration, in order to reach the targets at the required speeds. Different methods have been developed for estimating the actual train speed and position. For example, the control systems described in [10], [11], [12] are based on GPS (Global Positioning System). The ERTMS (European Railways Traffic Management System, the system under development for the European railways) [13], [14] will probably use a set of sensors including: two encoders positioned on two independent wheels, a radar sensor positioned on the locomotive case, and a longitudinal accelerometer. Odometry techniques based on sensors located on one or more axles of the train may be used for dead reckoning between two subsequent exact position assessments. In this case the current speed is obtained from the measure of the angular speeds of train wheels, while the distance estimation is calculated by integration of speed measures. [15], [16], [17], [18], [19]. Dead reckoning by means of odometry may fail if degraded adhesion in the wheel-rail interaction occurs, due to rain, fog, ice, leaves, and so on, and the train is accelerating or braking, i.e. when pure rolling conditions do not hold anymore, and macroscopic sliding occurs on one or more of the axles equipped with odometry sensors. The SCMT (Sistema di Controllo della Marcia dei Treni) is an ATP system that is being realized for the Italian railways. In this system the track subsystem, named SST, communicates its information to the train by means of fixed balises. In the SCMT system [23], the evaluation of the actual train speed and of the distance to the next objective points are obtained by elaborating the measures of two incremental encoders positioned on two independent axles. The SCMT system contains a module that estimates train speed and position from the data measured by the sensors. This module is named odometry system and includes the so-called Odometry Algorithm, a procedure that elaborates the data from the track and from the sensors mounted on the train in order to estimate the current train speed and position, either when poor adhesion conditions between the wheel and the rail occur. The odometry algorithm for SCMT system was designed by the researchers of the Applied Mechanics Section of the Department of Energetics of University of Florence, and the technical personnel of Trenitalia S.p.A..

Original contribution of the thesis The on board components of the control systems devoted to the management of the train traffic on the railways need a as more precise as possible estimation of

train speed and position, in order to guarantee the safety of the whole system and to improve its efficiency. The control system SCMT, developed in Italy, uses as information for train speed and position estimation between two subsequent fixed balises, the measures from two incremental encoders mounted on two independent axles of the train locomotive. Odometry algorithms based on wheel velocities measurements for railway applications usually have a simple structure. For example, the systems used for train speed estimation included in the anti-skid and anti-slip devices, are usually based the measure of four wheel velocities ([24] [25], [26], [25]). In the anti-skid devices, train speed is evaluated as the maximum tangential wheel velocity, by introducing a saturation on the upper value of the deceleration. In other words, train speed is set equal to the maximum wheel tangential velocity, if its acceleration is lower than a maximum fixed value. If this condition is not satisfied, the system estimates train speed by integration, using a fixed value for the deceleration. In the Anti-slip devices, the speed estimation procedure is made by means of an analogous procedure. The odometry algorithm used by the ATP system has to be more carefully designed because: • it has to give a more precise as possible estimation of train speed in all the running conditions (acceleration, braking phases, constant speed etc.); • it uses only two wheel velocities measures, while anti-slip and anti-skid devices are usually based on four or more speed measures; • in this case, also the position estimation is required, it is obtained by integration of the speed estimation, then either a little, but characterized by a constant sign, error on speed estimation may produce an high error on train position. The developed odometry algorithm permitted to reach, in all the running conditions, acceptable precision on the estimation of both speed and position. This result was obtained by a quite elaborated procedure, in which each single rule was carefully tuned. Its performance were compared with those obtained by using procedures based on soft computing techniques (fuzzy inference systems and neural networks), trained by means of automatic procedures on the basis of experimental tests. The application of soft computing techniques gives interesting results in terms both of precision and computational load of the procedure.

Structure of the thesis This thesis resumes the activities that allowed to define the structure of the odometry algorithm, the realization of a prototype of the algorithm and the testing phase.

The thesis starts with an initial chapter in which the main features of the ATP system SCMT are summarized. In particular, the construction of the socalled braking curves will be shown. This first general introduction is necessary in order to understand the boundary conditions in which the odometry algorithm will operate and then its scope and importance. In the final part of the chapter, in particular, the effect of the odometry algorithm precision on ATP performance and safety will be shown with the aid of some simple examples. In the following chapter the odometry algorithm that was designed for SCMT and that now is in the testing phase is described. The third chapter resumes the results of the algorithm testing on a series of experimental test runs. The fourth and fifth chapters show some possible improvements of the odoemtry algorithm, based on soft computing techniques. In chapter 6 the results obtained with the different versions of the algorithm will be compared and discussed.

References [1] Bin Ning ”Analysis of train braking accuracy and safe protection distance in automatic train protection (atp) systems,” 5th COMPRAIL , pp. 111-118, 1996. [2] M. Miyachi and H. Irinatsu ”The development of an cost effective automatic train protection system for safety operation on suburban lines in the Tokyo metropolitan area,” 5th COMPRAIL , pp. 119-128, 1996. [3] Y. Sato, T. Takashige, and I. Watanabe ”Advanced automatic train protection system,” 5th COMPRAIL , pp. 333-342, 1996. [4] T. Nagata, S. Murata, Y. Naka, O. Sakashita, and N. Shimutzu ”A train operating method with real time adaptability to performance variations of rolling stock,” [5] B. Ning ”Absolute braking and relative distance braking train operation control modes in moving block systems”, 6th COMPRAIL , pp. 991-1001, 1998. [6] N. Cortial and M. Krueger ”Simulation test tool for validation of speed and distance on-board component of the European train control system,” 7th COMPRAIL , pp. 42-52, 2000. [7] N. Kumagai, S. Uchida, I. Hasegawa, and K. Watanabe ”Wheel slip rate control using syncronized speed pulse computing,” 7th COMPRAIL , pp. 623632, 2000. [8] B. Allotta, M. Malvezzi, L. Pugi, M. Rinchi, A. Rindi, P. Toni, A. Amore, R. Cheli, G. Cocci, P. Presciani, G. Puliatti, “A test rig for evaluating odometry algorithms,” Proc. of the Eight International Conference on Computers in Railways (COMPRAIL 2002), 1063-1072, Lemnos, Greece 12 - 14 June 2002.

[9] B. Allotta, M. Malvezzi, L. Pugi, M. Rinchi, A. Rindi, P. Toni, “Un banco prova per l’omologazione di sistemi antipattinanti ferroviari,” Atti del XV Congresso AIMETA di Meccanica Teorica ed Applicata, Taormina, 26-29 settembre 2001. [10] B. Cai, X. Wang “Train positioning via integration and fusion of GPS and inertial sensors ,” 7th COMPRAIL , pp. 1217–1226, 2000. [11] A. Filip, L. Bazant, H. Mocek, J. Cach “GPS/GNSS based train position locator for railway signalling ,” 7th COMPRAIL , pp. 1227,1242, 2000. [12] G.B. Palmerini, P.M. Mascucci, “Train positioning using integrated satellite/inertial navigation systems,” Proceedings of the International Conference on Railway Traction systems , vol. 3, pp. 146-164, Capri, Italy May 2001. [13] ERRI B 126 RP 30 “Existing and future train control and command systems on the European railways - Models for deceleration curves ,” April 2001. [14] ERTMS-SRS, Version 4a dated 14.07.97 [15] B. Allotta, M. Malvezzi, G. Cocci, P. Presciani, “Train Speed Estimation from Wheel Velocity Measurements,” Proceedings of International Conference on Railway Traction Systems, Capri, Italy, May 14-16, 2001. [16] M. Malvezzi, P. Toni, B. Allotta, V. Colla, “Train speed and position estimation using wheel velocity measurements,” Proc. of 2001 IEEE/ASME International Conference on Advanced Intelligent Mechatronics Vol. 1, pp. 220-224, Como, Italy, July 2001. [17] B. Allotta, P. Toni, M. Malvezzi, P. Presciani, G. Cocci, V. Colla, “Distance and speed evaluation from odometric measurements,” Proc. of World Congress on Railway Research 2001, (oral presentation) Koeln, Germany 25-29 November 2001. [18] B. Allotta, M. Malvezzi, V. Colla, M.C. Valigi, “Ricostruzione della velocita`e della posizione del rotabile a partire dalle misure tachimetriche relative a due assi,” Atti del XV Congresso AIMETA di Meccanica Teorica ed Applicata, Taormina, 26-29 settembre 2001. [19] B. Allotta, V. Colla, M. Malvezzi, “Train position and velocity estimation using wheel velocity measurements,” Proc. of the IMechE, Journal of Rail and Rapit Transit, vol.216, part F, pp. 207-225. [20] P. Presciani, M. Malvezzi, G. L. Bonacci, M. Balli, “Development of a braking model for speed supervision systems”, WCRR 2001, World Congress on Railways Research, K¨oln, Germany, 25–29 November 2001. [21] P. Presciani, “Studio di un modello di frenatura per sistemi di controllo della velocit`a,” Ingegneria Ferroviaria, Agosto 2000. [22] Trenitalia,UTMR, Direzione Tecnica “Modello di frenatura per SCMT ed aspetti correlati,” Specifica N. 372387 esp. 02, 25/06/2001.

[23] ANSALDO SIGNAL, “Blocco Procedurale Applicazione del Controllo Marcia Treni,” Internal Report,19/10/2001. [24] S. Kent J. Tunley. The evaluation of wheel slide protection equipment. AEA Technology Rail, 1997. [25] S. Papini. “Un banco prova per i sistemi antipattinanti ferroviari: modelli del rotabile”. Master’s thesis, School of Mechanical Engineering, University of Florence, April 2001. [26] T. Watanabe, “Application of anti-slip re-adhesion control with adhesion prediction to commuter electric multiple units” Proceedings of the RTS, Railway Traction Systems Conference, Capri, Italy 15-17 May 2001, Vol. 1, pp. 71-86. [27] F. Bassi, O. Bruno, A. Landi, P. Masini, L. Sani, A.G. Violi, “Active antiskidding control strategy on electric locomotives”, Proceedings of the RTS, Railway Traction Systems Conference, Capri, Italy 15-17 May 2001, Vol. 1, pp. 87-113. [28] Trenitalia,UTMR, Direzione Tecnica “FS general Braking Model for ERTMS/ETCS,”,Summary of the Specification UTMR n. 373622 rev. 01, 0901-2003. [29] Leaflet U.I.C. “Brakes - Braking power” 544-1, 3rd edition - Reprint dated 1979. [30] R.Panagin “La dinamica del veicolo ferroviario”, Editrice Universitaria Levrotto e Bella, Torino 1997. [31] M. Malvezzi, P. Presciani, B. Allotta, P.Toni, “Probabilistic analysis of braking performance in railways,” accepted for publication, Proc. of the IMechE, Journal of Rail and Rapit Transit, 2003. [32] TRENITALIA S.p.A. Unit`a Tecnologie Materiale Rotabile “SCMT Progetto dell’algoritmo per il calcolo della velocit`a istantanea del treno e dello spazio percorso,” UTMR/DT.S.PS, 31/10/2000. [33] UIC, ERTMS users group, “Glossary of terms and Abbreviations,” Subset -023, issue 2.0.0, 20 March 2000. [34] TRENITALIA S.p.A. Unit`a Tecnologie Materiale Rotabile “Odometria SCMT:Principi generali dell’algoritmo per il calcolo della velocit`a stimata in caso di pattinamento o slittamento degli assi di misura: Specifica Requisiti Funzionali.”Specifica N. 372574 esp. 03, 28/11/2002. [35] TRENITALIA S.p.A. Unit`a Tecnologie Materiale Rotabile “Odometria SCMT:Principi generali dell’algoritmo per il calcolo della velocit`a stimata in caso di pattinamento o slittamento degli assi di misura: Specifica Requisiti Software” Bozza 4 Specifica N. 373450 esp. 00, 29/11/2002.

[36] Q. Zhang, A. Benveniste: “Wavelet Networks,” IEEE Transactions on Neural Networks, Vol. 3, No. 6, pp. 889–898, November 1992. [37] L. Davis, “The handbook of Genetic Algorithms,” Van Nostrand Reingold, New York 1991. [38] D. Goldberg, “Genetic algorithm in search, optimization and machine learning”, Addison Wensley, 1989. [39] J. Joines, C. Houck, “On the use of non stationary penalty functions to solve constrained optimization problems with genetic algrithms,” 1994 IEEE Symposium evolutionary Computation, Orlando, Florida, pp. 579–584, 1994. [40] C. Houck, J. Joines, M. Kay: “A Genetic Algorithm for Function Optimization: A Matlab Implementation,” NCSU-IE TR 95-09, 1995. [41] Lee C. C., “Fuzzy logic in control systems, fuzzy logic controller parts 1 and 2 ,” IEEE Transactions on Systems, Man, and cybernetics , Vol. 20, No. 2, pp. 404–435,1990. [42] Jyh Shing, Roger Jang, “ANFIS: Adaptive-Network-Based Fuzzy Inference System ,” IEEE Transactions on Systems, Man, and cybernetics , Vol. 23, No. 3, pp. 665–685, May/June 1993. [43] G. Cibenko, “Approximation by superposition of a sigmoidal function,” Mathematics of Control, signals and systems, Vol. 2, pp. 303–314, 1989. [44] M. Norgaard, “Neural Network Based System Identification Toolbox,” Tech. Report. 00-E-891, Department of Automation, Technical University of Denmark, 2000. [45] D.S. Broonhead, D. Lowe, “Multivariable functional interpolation and adaptive networks,” Complex systems, Vol. 2, pp. 321–355, 1988. [46] D. Marquardt: “An algorithm for least–square estimation of nonlinear parameters,” SIAM Journal of Applied Mathematics, No. 11, pp. 164–168. [47] L.M. Reyneri, “Unification of Neural and Wavelet Networks and Fuzzy Systems”, IEEE Trans. on Neural Networks, Vol. 10, No. 4, pp. 801–814, July 1999. [48] Kohonen, T. Self-Organization and Associative Memory, 2nd Edition, Berlin: Springer-Verlag, 1987. [49] T. Kohonen: “The self-organizing map,” Proceedings of the IEEE, Vol. 78, No. 9, pp. 1464-1480, 1990.