Agent-based Modelling for Resilient Cities

Project overview

Urban areas are increasingly exposed to extreme weather and shifting mobility demands. These disruptions arise from complex interactions between transport systems, human behaviour, and environmental conditions, making them difficult to predict and manage with traditional approaches. This project develops an agent-based model that simulates how individuals, vehicles, and infrastructure interact under disruption. A software platform is being developed to help researchers and policymakers test interventions and assess their impacts on congestion, emissions, and social vulnerability at city scale.

The challenge

Urban mobility during extreme events is highly dynamic and behaviour driven. Four key challenges are: (1) Demand surges such as evacuation leading to nonlinear congestion and cascading effects; (2) uneven infrastructure disruptions reducing network capacity; (3) adaptive human decision-making under uncertainty, including rerouting and behavioural change; and (4) congestion that evolves gradually rather than instantaneously. Most existing models either assume steady-state conditions or lack integration of behavioural and environmental data, limiting their ability to support planning under disruption.

Our approach

We integrate multiple real-world data sources into a unified simulation, including transport networks, human mobility data (including extreme weather scenarios), infrastructure and environmental conditions, and socio-economic and behavioural signals, calibrated against observed data for realism and interpretability. We introduce a time-delayed congestion mechanism where agents make routing decisions based on past perceived travel costs rather than instantaneous information, allowing congestion to build and dissipate gradually in a realistic way. The model is formulated as a scalable agent-based congestion game, where agents optimise routes over a shared network while interventions modify infrastructure, incentives, or policies, and agents dynamically respond to these changes. This enables both microscopic behavioural realism and city-scale scalability within a single framework. The system is designed as an intervention-focused simulation platform supporting policy experimentation, including infrastructure upgrades, pricing mechanisms, behavioural nudges, and work policy changes, as well as mitigation and adaptation strategies.

Outcomes

A modelling framework for urban mobility and resilience; integration of multi-source data into a unified simulation pipeline; a proof-of-concept city-scale simulation under disruption scenarios; and modular software that can be adapted across cities and policy contexts.

Impact

This project supports a shift from reactive to more forward-looking urban planning by enabling simulation of evacuation and emergency response, identification of infrastructure vulnerabilities under extreme events, and testing of robust infrastructure and contingency strategies. It also supports sustainable mobility analysis, including congestion pricing, green transport policies, public transport optimisation, and emissions reduction, while helping assess social equity impacts across different groups and inform more inclusive policy design. By combining AI, agent-based modelling, and simulation methods, the project provides decision-support tools for urban stakeholders and contributes to the Digital, AI, and Sustainability agendas at the University of Warwick.

Next steps

Future work includes exploring additional funding opportunities, and developing stakeholder engagement activities, including interactive workshops with UK city councils and policy partners.

For more info