DR. TAOMEI ZHU, EXTRAORDINARY AWARD. MADRID TECHNICAL UNIVERSITY DOCTORAL THESIS

CITEF’s best assets are its human resources, and a fine example of this is the Doctoral Thesis by Ms. Taomei Zhu “Traffic Conflict Detection and Resolution (CDR) in ERTMS System”, which earned her a Ph.D in Industrial Engineering and which was produced at CITEF under the guidance of José Manuel Mera Sánchez de Pedro.

The thesis has now been acknowledged by the Madrid Technical University – UPM’s Extraordinary Doctoral Thesis Award.

Research work in the thesis focuses on the problem of rail traffic conflicts. Even though in the modern world rail traffic can be partially managed by a traffic control system, this system is insufficient to take operational decisions when the disturbances become frequent, especially on railway nodes with significant volumes of traffic. In ERTMS/ETCS systems, while all advanced technologies are employed to increase safety, train speed and to reduce headway interval, increasing challenges of real-time conflict resolution are also encountered.

This thesis focuses on the methodological realisation of conflict detection and resolution, and aims to develop a CDR support system for railway traffic control and management.

Research work in the thesis gave rise to the following publications and presentations at congresses:

• T. Zhu, J.M. Mera, B. Suarez, and J. Maroto. “An agent-based support system for railway station dispatching,” Expert Systems With Applications 61 (2016) 39–52. JCR impact factor: 2.571 Q1
• T. Zhu and J.M. Mera. “Railway traffic conflict detection via a state transition prediction approach,” IEEE Transactions on Intelligent Transportation Systems, Volume: 18, Issue: 5 May 2017. JCR impact factor: 3.039 Q1
• Zhu, J.M. Mera. A 3D simulation approach for railway conflict detection in Traffic Management System layer, Computers in Railways XIV, 2014.
• Zhu, J.M. Mera, E. Castellote, J. López. Feasibility evaluation and critical factor analysis for subway scheduling, Computers in Railways XV, 2016.

The details of the work produced in the Thesis which won the UPM award are as follows:

Modelling

To describe the context of the conflict detection and resolution problem, railway traffic is depicted using models of the infrastructure, rolling stock and scheduled timetables. These partial models are data-based.

• An original timetable or an updated (rescheduled) timetable is represented for the purposes of providing common goals for the entities involved in the railway system and organising the related resources in real-time traffic control.
• The train model consists of the relative routes of the trains and their derivatives and dynamic train movement information, such as train position, train speed, occupied track circuits, operation times, movement authorities etc.
• The infrastructure model is established both from a line (or a network) range and a region range. The former is described by stations and corridors between stations, which are known as sections. The latter is mainly represented by routes and segments. Track description data for dynamic train control are also based on instantaneous route states and segments in the infrastructure model.

Signs of conflict in the early stages of potential conflicts are expressed on the basis of the proposed models.

Conflict detection (CD)

Conflict detection procedure and feasible approximations are studied separately (from conflict resolution approximations). Based on the train model and the infrastructure model, respectively, the approximations of train trajectories and infrastructure state transition are proposed for short-period traffic prediction and conflict detection.

• In train trajectory approximations, the detailed train movement information is monitored. The correlations of each of the two dimensions are supervised cooperatively. In particular, the deviations of distance, the tracks used and the speeds are detected in the multi-dimensional trajectory profiles.



• The infrastructure state transition approximation is based on precise state transition maps and corresponding relation matrices. Historical segment and route states (state vectors) are stored to obtain empirical state transition maps. These maps are applied to the detection of abnormalities in the dynamic rail traffic environment, and the structural constraints of infrastructure topology and route compatibilities are in turn applied to the prediction of potential situations of conflict.