Fully automated train operation requires a technical system, capable of surveying the track in the same way as a human driver does and guaranteeing the same integral reduction of risk. Since the system has to operate in complex environments and under varying weather conditions, this is not an easy task. This paper will introduce a system that overcomes present limitations. A multi sensor obstacle detection system having a up to 400 m look-ahead range under typical operating conditions was developed. To get maximum information about a fact/object the concept of complementary physical principles and strategies was exploited. To get maximum confidence in interpretation the concept of redundancy is used. Concerning sensor front-end design this concept means to choose a multi sensor approach with stand alone sensor units. When choosing the individual sensors both aspects of the concept must be considered: complementarity and redundancy. The passive sensor unit is a multi focal assembly of one survey fixed camera and two remote cameras with a pan device. The viewing angle is controlled via a mirror that is rotated by a stepper motor/PPI combination.The active sensor unit consists of a fixed short distance LIDAR and a scanning long distance LIDAR. The sensor systems are designed to cover the whole possible surveillance space (considering maximum horizontal and vertical curvature). There is information from track and infrastructure databases available. Information about vehicle properties like position on the track, velocity etc. are also given. Furthermore a modular system architecture is chosen. Independent function modules with standardized interfaces were established, thus being able to add new modules to the system till the requirements are fulfilled or replace modules if new ones do the job more efficiently. The prototype system has successfully been installed on a test car of the Deutsche Bahn AG (DB-AG). It had to be robust and reliable enough to get over a two days acceptance test. Before that the system had been tested on trains driving up to 120 km/h over long distances across Germany. During these tests low false alarm rates, robustness against vibrations and electrical stability were shown. The overall detection performance has shown to be comparable to that of a human driver. Future steps are the optimization, miniaturization and the integration of the active and passive sensor components of the obstacle detection system. The computational optimization of the object detection algorithms is another important step in order to reduce necessary computing resources.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen


    Exportieren, teilen und zitieren



    Titel :

    Long range obstacle detection on automated railroad tracks


    Beteiligte:
    Scherer, F. (Autor:in) / Schuster, P. (Autor:in) / Möckel, S. (Autor:in)


    Erscheinungsdatum :

    2003


    Format / Umfang :

    7 Seiten, 7 Bilder, 7 Quellen


    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Datenträger


    Sprache :

    Englisch




    Automated Visual Inspection of Railroad Tracks

    Resendiz, E | Online Contents | 2013


    Automated Visual Inspection of Railroad Tracks

    Resendiz, Esther / Hart, John M. / Ahuja, Narendra | IEEE | 2013


    RAILROAD CROSSING OBSTACLE DETECTION DEVICE

    MAE TOMOAKI / MURAMATSU KOJIRO / NISHIMURA MAKI et al. | Europäisches Patentamt | 2020

    Freier Zugriff

    Modeling guided wave propagation with application to the long-range defect detection in railroad tracks

    Bartoli, I. / Lanza di Scalea, F. / Fateh, M. et al. | Tema Archiv | 2005


    Multi-sensor obstacle detection on railway tracks

    Mockel, S. / Scherer, F. / Schuster, P.F. | Tema Archiv | 2003