A multi-agent-based approach for resolving real-time train rescheduling problems of large- scale railway networks

Liu, Jin (2020). A multi-agent-based approach for resolving real-time train rescheduling problems of large- scale railway networks. University of Birmingham. Ph.D.

[img] Liu2020PhD.pdf
Text - Accepted Version
Restricted to Repository staff only until 31 May 2025.
Available under License All rights reserved.

Download (2MB) | Request a copy


Train delays occur often in daily railway operations due to a variety of incidents in railway circumstance. As a result, train operations are usually affected by unforeseen disturbances in railway circumstance which lead to unpredicted minor train delays in the railway network. With the connections and conflicts in the journey of the delayed trains, minor delays are likely to be transferred into a series of knock-on delays which may influence all the trains across the whole network. In this thesis, a multi-agent-based system is proposed to tackle train rescheduling problems in large-scale railway networks which considers strong coupling of train timetables among junction areas. With the evaluation approaches in the UK railway network, the proposed multi-agent-based approach can generate optimised train dispatching solutions within reasonable computational time compared to the First Come First Served dispatching rule, which provides a step forward in tackling real-time train rescheduling in a distributed way.

The thesis starts with illustration of the current railway traffic management system and recent research on decision support systems which generate real-time actions to considered rail traffic, and discusses the main challenges of the state of the art, which are bottlenecks in the development of decision support systems for large-scale railway networks. A systematic hierarchical multi-agent system for generating real-time solutions for train rescheduling problems in large-scale railway networks is proposed, which covers a mathematic formulation for the train rescheduling problem, an approach for decomposing a large-scale railway network, a Genetic Algorithm for local searching of train dispatching solutions and collaborative strategies to achieve final solutions with global feasibility. Furthermore, train rescheduling problems in railway bottleneck sections are tackled by a two-agent system (P2P system) with a series of negotiation approaches between the two agents. To test the proposed approaches, UK rail infrastructures are applied to test the P2P and multi-agent systems, respectively. In a railway bottleneck, the computational results are compared with a traditional centralised algorithm and First Come First Served dispatching approach based on designed delay scenarios, and the pros and cons of the P2P system are studied. For a large-scale railway network, the solutions generated by the multi-agent system for different delay conditions are compared against a First Come First Served approach. The proposed multi-agent approach can generate train dispatching solutions better than First Come First Served in terms of proposed objective function.

The thesis also discusses the limitations of the proposed approaches and outlines the author’s future work plan on railway traffic management systems. With the proposed decentralised system for large-scale railway networks, more traffic actions such as the reallocation of crew members and rolling stock can be considered. And to satisfy the time horizon of traffic control in the industry, the author also discusses his ideas for boosting the searching algorithm.

Type of Work: Thesis (Doctorates > Ph.D.)
Award Type: Doctorates > Ph.D.
Licence: All rights reserved
College/Faculty: Colleges (2008 onwards) > College of Engineering & Physical Sciences
School or Department: School of Engineering, Department of Electronic, Electrical and Systems Engineering
Funders: None/not applicable
Subjects: T Technology > TF Railroad engineering and operation
URI: http://etheses.bham.ac.uk/id/eprint/10265


Request a Correction Request a Correction
View Item View Item


Downloads per month over past year