Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
In diesem Kapitel wird die Digitalisierung der Zugdisposition für einen sicheren und wirtschaftlichen Personenverkehr auf nicht ausgelasteten industriellen Eisenbahninfrastrukturen untersucht, wobei der Schwerpunkt auf dem Ansatz des "Zugdispatchers in der Cloud" (ZLiC) liegt. Der Text diskutiert die Herausforderungen des Strukturwandels in Kohlebergbaugebieten und das Potenzial für den Ausbau des Schienenverkehrs, um diesen Veränderungen zu begegnen. Es stellt die ZLiC-Systemarchitektur vor, die Komponenten für Verriegelungslogik, Sprachprozeduren, Spracheingabe und Sprachausgabe umfasst. Die Implementierung von ZLiC beruht auf ontologiebasiertem Engineering, und der Text enthält eine detaillierte Beschreibung der Systemkomponenten und ihrer Interaktionen. Die Evaluierung des ZLiC-Konzepts wird diskutiert, einschließlich Simulationen, Prototypentests und zukünftiger Einsatzpläne. Die Schlussfolgerung unterstreicht das Potenzial von ZLiC, Lücken zu schließen und Probleme effizient zu lösen, sowie die Notwendigkeit weiterer Risikobewertungs- und Zulassungsaktivitäten.
KI-Generiert
Diese Zusammenfassung des Fachinhalts wurde mit Hilfe von KI generiert.
Abstract
To mitigate the adverse effects of climate change, greenhouse gas emissions need to be minimised. The FlexiDug project investigates sustainable transport perspectives for structurally transforming areas where coal mining phases out. This work presents a safe, economical, and extendable approach on reusing industrial railways for passenger transport. We have digitalised the Zugleitbetrieb, a mode of operation that requires no trackside equipment. Our Train Dispatcher in the Cloud (German Zugleiter in der Cloud, henceforth ZLiC) has been developed ontology- and model-based. It also takes a railway network model as input, e.g., for the generic interlocking logic. Speech to text, naturallanguage understanding, and text to speech recreate the established speech interface towards conductors. A state machine ensures the prescribed voice procedure. Custom voice-activated recording allows using COTS radio devices. ZLiC has been evaluated successfully in simulations and field tests. We plan further improvements, evaluations, and a risk assessment.
1 Introduction
To reduce the adverse effects of climate change, multiple countries reduce burning of fossil fuels. Coal mining areas in Germany thus face structural transformation. Unused infrastructure need to be removed and former mining areas rehabilitated [2]. For example, open-pit mines are turned into lakes and raise the recreational attractiveness of the areas.
Funded by the Federal Ministry for Digital and Transport (Bundesministerium für Digitales und Verkehr, BMDV), the research project FlexiDug investigates transport perspectives for structurally transforming mining areas. Example region is the German part of Lusatia. The structural transformation together with new work models and the proximity to metropolitan areas (e.g., Dresden, Berlin) lead to foreseeable changes in settlement structures. We suggest a climate-friendly take on addressing these changes and look at possibilities to extend rail transport.
Anzeige
In Lusatia coal is hauled by train. Almost 400 km of railway tracks are still in use, additional lines have been closed but partly still exits. FlexiDug explores possibilities for reusing those infrastructures to save and reduce emissions.
However, infrastructures of the mining railways cannot be used for passenger transport without further ado. Special circumstances, such as isolated networks and fixed connections instead of couplings between wagons, have led to technical and operational differences. Technical differences include train detection, train protection, and signalling systems of low density and custom design. Operational differences include low maximum speeds (i.e., 60 km/h) and push-pull trains operated from the end of train. The mining railways do not fulfil the strict regulations for passenger transport and rebuilding to today’s standards is costly.
This paper presents an approach on enabling safe, economical, and extendable passenger transport on yet little-used industrial railway infrastructures.
We build on the German Zugleitbetrieb [7] (ZLB, comparable to North-American Track Warrant Control). In this mode of operation, train movements are controlled by a Zugleiter (train dispatcher special to Zugleitbetrieb, henceforth dispatcher for readability). The dispatcher notes track occupations in a Belegblatt (graphical occupation sheet special to Zugleitbetrieb, henceforth occupation sheet for readability). On each train, operational duties are with a Zugführer (train conductor special to Zugleitbetrieb, henceforth conductor for readability). Conductors and the dispatcher communicate via radio and a prescribed voice procedure. Trackside equipment is optional but can increase performance.
Anzeige
2 Approach
We propose the digitalisation of the Zugleitbetrieb by introducing the Train Dispatcher in the Cloud (German Zugleiter in der Cloud, henceforth ZLiC). While ZLiC implements the dispatcher in software, the speech communication for the human conductors remains the same. The digitalisation addresses the shortage of skilled workers in all areas of railway transport and also allows for future scalability. Due to existing experiences among personnel (including approval1, bodies) and little trackside requirements, we assume economical realisation.
We currently justify safety by the approval of the non-digitalised Zugleitbetrieb. In perspective, the introduction of components requires a safety re-analysis. Given a safe ZLiC, eliminating chances for human error can also increase overall safety.
Fig. 1.
ZLiC central (large box), distributed, input, and related (dashed) components.
Figure 1 depicts the software architecture of the ZLiC. The core of the system are components for an interlocking logic, the voice procedure, speech input, and speech output. A digital rail network planning is initially required to derive the internal network model. Once created, the network topology is extracted to be used by the interlocking logic. Extracted routes are required by the state machine to adapt the generic voice procedure for the particular network. Incoming communication is converted from speech to text (STT) and then interpreted using natural-language understanding (NLU). Outgoing communication is converted from text to speech (TTS). An audio gateway implements the interface to the trackside. For communicating with conductors, the system interacts with commercial off-the-shelf (COTS) radio devices via the radioIO component. In contrast to other components, the two radio components might need to be spatially close to the rail network. Depending on the radio technology used and the area to cover, it can be necessary to deploy radio components multiple times.
Our design and implementation have been guided by ontology-based engineering. Figure 2 shows excerpted and simplified example models and artefacts. We have first defined ontology models for each concept, such as software components, domain-specific data models, and generic artefact models. Because we utilise the widely used Resource Description Framework (RDF), we are able to reuse models, e.g., for state machines [11]. Using these concepts, we modelled the system itself, e.g., the voice procedure, occupation sheet, and final artefacts.
Fig. 2.
Exemplified and simplified path from the modelled system to the implementation. All models relate to the ontology. T* represent some implemented transformations.
The component model and the modelled Zugleitbetrieb were discussed and revised with domain experts. Because the component model also describes the data flow, we are able to add the component interfaces and a connecting middleware programmatically. We also define domain-specific data models, e.g., for railway vocabulary and infrastructure models. The infrastructure model refers to digital network planning documents, e.g., PlanPro [10], which contain information about a specific railway network. Data models can contain supplementary data, such as hypothetical train routes and hypothetical train stations.
Artefacts, such as source code, are generated in two steps. First, we implemented transformations that create artefact models from the modelled system. Second, further transformations create the artefacts from the artefact models. The second step also uses information from the overall modelled system, e.g., from the component model for Docker and Kubernetes configurations.
The indirections in the modelling, generation, and implementation can only be briefly outlined here. The thereby achieved modularity allows modifying concepts (e.g., domain-specific semantics), models (e.g., processes), their interaction (e.g., communication patterns), inputs (e.g., infrastructure models), and artefacts. For example, implementations currently relying on Python could be replaced with SPARK [6], an Ada subset for safety-critical programming, prior to approval.
3 Implementation
From the conductors’ spoken language, intentions are recognised and converted to parameterised events using STT and NLU. For STT we use OpenAI Whisper off-the-shelf which performs well in our scenarios. For NLU we use Rasa, which needs to be trained per network using a small text model. TTS is done with SVOX via Pico. We use EclipseDDS as middleware. For auditability and customisability, the chosen third-party components and ZLiC itself [1] are open-source. For, e.g., availability, ZLiC works without an Internet connection.
The voice procedure of the Zugleitbetrieb is used to define the state machine that interacts with the interlocking logic and the occupation sheet. As Sect. 2 describes, we use the state machine’s model for its transformation to source code.
We have designed central components to be able run on both, Infrastructure as a Service (IaaS) platforms and COTS computers. Compared to specialised hardware, this reduces costs and eases procurement. The design decisions made, however, require external mechanisms for dependability. Especially IaaS platforms usually offer fault tolerance mechanisms. Since we consider the component model during artefact generation, redundancy structures can be derived systematically.
Our implementation currently allows trackside equipment to be mocked (e.g., for development), simulated (e.g., for evaluations), or compatible to EULYNX [3] and the Rail Safe Transport Application network protocol (RaSTA) [8].
For interfacing between the ZLiC’s audio gateway and COTS radio devices, we implemented the program radioIO. Because no suited framework could be identified, we have implemented custom digital signal processing (DSP).
Fig. 3.
Processing chain for automatically forwarding audio from the radio to the ZLiC.
Figure 3 shows the audio input processing we implemented. After a band-pass filter removed frequencies irrelevant for voice, a peak-tracking gate triggers the recording based on a threshold. A ring buffer between the gate and the recording implements a lookahead, so the beginning of the speech is captured completely. After no detected voice for the duration of the gate’s hold time, the recording is stopped, normalised, and sent to the audio gateway (WebSocket). Audio received from the gateway is normalised and a remotely inaudible audio signal is prepended to trigger the voice-operated transmission feature (VOX) of radio devices.
Our DSP design requires analogue audio wiring only. Because no control connections etc. are needed, virtually any COTS radio device can be used.
For execution efficiency, our implementation of radioIO pre-allocates required memory and processes data zero-copy as far as possible. We chose the Nim programming language, which proved to be a well-working compromise between development efficiency, effortless interfacing with C libraries, runtime efficiency, and predictable timing due to deterministic memory management.
4 Evaluation
Our evaluations examine the feasibility of our proof of concept. In a next step, the feedback gathered from practitioners can be included in a also more formalised implementation. Both would support potential approval activities.
The evaluation is done in several tiers, each adding complexity. As the first tier, we realised simulations (e.g., [5, 9]) and discussed them with railway experts. The verdict has been, that our approach is generally feasible but needs to be more specific to match with regulations. We implemented changes accordingly. As the second tier, we tested a first prototype with real trains as part of a demonstration in the Ore Mountains. Here, we focused on occupation tracking (monitored via the occupation sheet), speech communication in a field environment, and varying qualities of the radio signal. The third tier will be the deployment of the ZLiC on a model railway. Following the idea of serious gaming [4], we will then ask railway personnel to perform realistic railway operations using our system. We hope to obtain detailed feedback on the overall experience to identify room for improvement, such as fallback scenarios. The last tier will demonstrate the revised ZLiC implementation in Lusatia with real trains (but no passengers).
5 Conclusion
Due to the challenges related to climate change, there is a demand to increase rail transport. In Germany, nonetheless, there are hardly any solutions that allow for an economical reuse and operation of industrial railway infrastructures for passenger transport. Compared to the requirements for main lines, the presented concept ZLiC requires minimal trackside infrastructure and human resources. Additionally, the system can be upgraded gradually for increased reach and performance. The concept of ZLiC cannot be applied to routes that are part of the European corridor as defined in the Technical Specifications for Interoperability (TSIs). However, this is unlikely to be a relevant issue in rural areas.
New approaches in the railway sector only have prospects for approval if they consider regulations from the beginning on. The focus on model support and automation not only supports development efficiency but also the approval process, as it embodies structured design and well-defined processes. Small-scale projects like ZLiC can be helpful prototypes to familiarise approval bodies with concepts new to the domain. Such projects also help to gather feedback and experiences with the general approach of interweaving computer science and other domain-specific practices and cultures. With FlexiDug and ZLiC, we demonstrate that interdisciplinary research can bridge gaps and solve problems efficiently.
Besides technical improvements and more evaluations, future work includes a risk assessment to derive safety and functional requirements. As necessary for approval, the results can then guide a compliant re-engineering of ZLiC.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
In this work, approval, authorisation,admission, certification, and, in that sense,assessment are not distinguished and can thus be used interchangeably.
Abt, C.C.: Serious Games. University press of America (1987)
5.
Boockmeyer, A., et al.: Processing digital railway planning documents for early-stage simulations of railway networks. In: Proceedings of the Transport Research Arena, TRA 2024, 15-18 April 2024, Dublin (2014). To appear
6.
Carré, B., Garnsworthy, J.: SPARK – an annotated ADA subset for safety-critical programming. In: Proceedings of the TRI-ADA 1990 conference, pp. 392–402 (1990)
7.
DB Netz AG: Richtlinie 436 (Ril 436) – Zug- und Rangierfahrten im Zugleitbetrieb durchführen. (Regulations on Zugleitbetrieb by Deutsche Bahn)
8.
Electric signalling systems for railways – part 200: safe transmission protocol according to DIN EN 50159 (VDE 0831-159) (2015)
9.
Kamp, A., et al.: Evaluating possibilities of flexible train schedules to allow high-priority sporadic traffic (2023). Poster presented at SUMO User Conference
10.
Maschek, U., Klaus, C., Gerke, C., Uminski, V., Girke, K.J.: Planpro - durchgängige elektronische datenhaltung im estw-planungsprozess. Signal und Draht 104, 22–26 (2012)
11.
Schneider, G.F., Pauwels, P., Steiger, S.: Ontology-based modeling of control logic in building automation systems 13(6), 3350–3360 (2017)
