EPSRC Doctoral Studentship - Plans vs Reality: AI-Enhanced Modelling of Disrupted Urban Mobility
About the Project
Project details
Urban mobility models typically assume fixed daily activity schedules and stable transport networks. In reality, individuals continuously adapt their plans in response to disruptions such as delays, missed connections, time pressures, caring responsibilities, safety concerns, and changing priorities. This gap—between planned and realised behaviour—is poorly represented in current models, limiting their usefulness for designing robust, equitable, and sustainable transport systems.
This PhD will extend an existing agent-based and activity scheduling framework developed through the EPSRC AI4CI Hub. The framework integrates: (1) synthetic population generation grounded in demographic structures; (2) activity-based demand modelling; (3) multi-modal routing (walking, cycling, car, and bus); and (4) dynamic schedule adaptation under constraints. The project will advance this system by incorporating behavioural realism and disruption-aware modelling.
Two key innovations underpin the research. First, behavioural modelling using AI, including reinforcement learning (RL) and large language models (LLMs), will be explored to better represent decision-making under uncertainty, including preference evolution and context-aware adaptation. Second, disruption-aware scheduling and routing will be developed by embedding stochastic models of events such as delays, congestion, weather shocks, and safety perceptions, allowing analysis of how disruptions propagate through individual schedules and the wider system.
The overarching aim is to develop a framework for resilient urban mobility modelling that integrates machine learning with individual-based simulation. The project will:
- Compare behavioural modelling approaches (rule-based, RL, and LLM-assisted) for dynamic decision-making.
- Quantify how disruptions lead to plan breakdowns, rescheduling, and unequal impacts across populations.
- Generate actionable insights for transport planning, including reliability, service design, ans support for vulnerable users.
Key research questions include:
- How do different modelling approaches perform in representing adaptive behaviour under uncertainty?
- How do disruptions affect activity completion, mode choice, and network performance?
The project will build on and extend an existing modelling framework through four components:
- Synthetic populations and activity generation:A detailed population will be created using census and survey data (e.g. National Travel Survey), capturing household structure and care responsibilities.
- Network modelling: A unified, multi-modal, time-dependent network will be constructed, incorporating travel time, cost, comfort, and perceived safety, alongside dynamic routing capabilities.
- Dynamic activity scheduling:Individuals will follow baseline daily schedules subject to hard and soft constraints, with real-time adaptation triggered by disruptions such as delays or cancellations.
- Behavioural modelling:Agents will be implemented using rule-based approaches, RL policies, and LLM-guided decision-making, and evaluated based on realism, robustness, and task completion.
Model calibration and validation will be conducted throughout
The model will be applied to policy-relevant scenarios developed with stakeholders (e.g. Transport Scotland, local authorities, operators). Scenarios may include reliability improvements, safety interventions, and enhanced real-time information systems. The aim is to identify how interventions affect disruption propagation, accessibility, and equity.
About the School/Research Unit
The student will be embedded within a highly interdisciplinary and collaborative research environment spanning artificial intelligence, urban analytics, and social science. The project is situated within the EPSRC-funded AI for Collective Intelligence Hub (Smart City Design theme), where supervisors Heppenstall and Gamal are developing cutting-edge individual-based models, including housing and mobility systems with activity scheduling components. This ongoing work provides a strong methodological and technical foundation for the PhD. Through the Hub, the student will have access to a wide network of academic and non-academic stakeholders, offering opportunities for collaboration, data access, and real-world impact.
In addition, the student will be part of the Urban Analytics research group, led by the supervisors, which includes an active and growing community of national and international researchers. The student will also engage with the Urban Big Data Centre and the wider Urban Analytics group within the College of Social Sciences. This environment offers expertise in working with granular urban datasets and aligns closely with institutional research priorities, particularly around understanding social, economic, and health inequalities in cities, as reflected in initiatives such as Glasgow Changing Futures.
Eligibility
Applicants must meet the following eligibility criteria:
- A good Masters degree (or overseas equivalent)
- A demonstratable interest in the topic area under investigation
- Able to study on a full-time basis only
- Considered 'Home' or 'Rest of UK' for fee status
- Entry requirements for the Urban Studies, PhD
Funding Notes
The scholarship is available as a full-time +3.5 (3.5 year) PhD programme only. The programme will commence in October 2026. The full funding package includes:
Unlock this job opportunity
View more options below
View full job details
See the complete job description, requirements, and application process
