The project investigates whether Large Language Models (LLMs) and related AI-assisted coding tools can help generate optimisation formulations, heuristic design, and solver implementations for railway scheduling problems, and whether those outputs can be rigorously verified and benchmarked against established operations research methods.
The project is grounded in optimisation and decision-science perspectives, with strength in mathematical modelling, combinatorial optimisation, heuristic design, and computational benchmarking. LLMs are treated as a research instrument and experimental tool rather than a primary research focus, with emphasis on evaluating their feasibility, solution quality, scalability, robustness, and computational efficiency.
The project does not aim to develop new LLMs. Instead, it focuses on using existing AI tools to generate candidate models and algorithms, and systematically assessing them through solver validation, constraints checking, and benchmark comparisons. The key contribution is a rigorous framework for verifying, repairing, and benchmarking AI-assisted optimisation artefacts in a demanding railway scheduling context.
While railway optimisation methods are well established and LLMs show promising capabilities in code generation and automated mathematical modelling, these two streams have not yet been integrated sufficiently. This work addresses this gap by developing a systematic framework for generating, validating, repairing, and evaluating AI-assisted optimisation models and algorithms for railway scheduling problems.
The framing of this project makes it suitable for two types of applicants: candidates with operations research or mathematical optimisation backgrounds who want to learn how AI tools can support modelling and algorithm design, as well as candidates with AI, computer science, or software backgrounds who want to apply their skills to a challenging real-world optimisation problem.
The project will be supervised by Associate Professor Dr Elham Mardaneh and Dr Yefei Zhang at Curtin University, within the Curtin Centre for Optimisation and Decision Science. Dr Mardaneh will serve as principal supervisor, and Dr Zhang as co-supervisor. The supervisory team provides expertise in optimisation, mathematical modelling, heuristic algorithm design, computational experimentation, and decision science. The AI-specific skills will be developed through the candidate’s research training, current literature, available tools, and potential interdisciplinary collaboration where appropriate.
The selected PhD candidates will be embedded within the centre, ensuring that the candidates receive the best support from high-profile, rich-industry-experienced researchers.
Aim
The aim of this project is to develop and evaluate a rigorous AI-assisted optimisation framework for train scheduling and rescheduling, with a particular focus on verifying whether automatically generated mathematical models, solver implementations, and heuristic algorithms can produce feasible, scalable, and high-quality schedules when compared with established optimisation approaches.
Objectives
Review the railway scheduling and AI-assisted optimisation literature.
- Implement baseline train scheduling and rescheduling models.
- Create or adapt benchmark instances for computational experiments.
- Use existing AI tools to generate candidate optimisation formulations and algorithms.
- Design validation tests to detect missing constraints, infeasible schedules, and weak formulations.
- Run computational experiments comparing generated methods with established baselines.
- Complete thesis and prepare manuscripts for publishing in operations research journals, including Computers & Operations
Research, Transportation Research Part B: Methodological, Transportation Research Part C: Emerging Technologies, European Journal of Operational Research, Transportation Science, and Journal of Rail Transport Planning & Management.
Significance
Railway scheduling is a high-impact optimisation problem. Passenger services, freight logistics, mining supply chains, manufacturing networks, and intermodal transport systems all depend on reliable train schedules. Even small disruptions can propagate across a network and create substantial delays, asset underutilisation, missed connections, and operational cost.
Traditional railway scheduling algorithms require careful mathematical modelling and deep domain expertise. Mixed-integer programming, constraint programming, dispatching rules, decomposition methods, tabu search, large neighbourhood search, and matheuristics have all been used to address variants of train timetabling and rescheduling. However, developing and adapting these methods can be slow when railway operators face new infrastructure layouts, service patterns, disruption scenarios, or operational policies.
AI tools may accelerate parts of the optimisation workflow by helping generate candidate formulations, solver code, local-search moves, or algorithmic ideas from structured problem descriptions. However, railway scheduling is safety-critical and highly constrained. AI-generated outputs may omit essential constraints, produce infeasible schedules, or appear plausible while failing under realistic scenarios. This project, therefore, asks a timely and important question: how can emerging AI tools support optimisation without sacrificing rigour, reliability, or operational validity?
Ideal Candidate
We seek a highly motivated PhD candidate interested in computational problem solving and applied optimisation problems with emerging AI relevance. We particularly welcome applicants with one or more of the following backgrounds:
- mathematical optimisation, operations research, applied mathematics, computer science, data science, or engineering;
- programming experience in Python, C++, or a similar language;
- interest in railway systems, transport, logistics, mining supply chains, or decision support;
- experience with optimisation solvers such as Gurobi, CPLEX, OR-Tools, or related software.
Prior experience with LLMs is welcome but not essential. The project is especially suitable for a candidate who has strong optimisation or programming foundations and is willing to develop AI-assisted modelling skills during the PhD. Additionally, the applicants should meet the eligibility criteria for entry into a PhD program at Curtin University.
This project is open to International and Domestic applicants.
Scholarship
If you are identified as the preferred candidate for this project, you may be considered for an RTP scholarship.
Enquires and How to Apply
For enquires about this opportunity contact Dr Andy Zhang at Andy.Zhang1@curtin.edu.au
To formally apply submit an Expression of Interest to Dr Andy Zhang during the Central Scholarship round (July 1st – July 31st 2026)