Adham Enaya

Research

Subject

Transfer Learning for Adaptable and Transferable Urban Mobility Models

First and second supervisors

Abstract

Urban mobility modelling is vital for addressing the complex transportation challenges faced by cities worldwide. Traditional models, however, often struggle to generalise across different urban contexts due to the unique characteristics of each city. This research explores the potential of transfer learning to enhance the adaptability and transferability of urban mobility models across diverse cities. By identifying universal patterns such as population density, urban layout, and street networks, this study leverages deep learning techniques, including Convolutional Neural Networks (CNNs) to capture the non-linearity in urban features, Physics-Informed Neural Networks (PINNs) to integrate universal physical laws in urban mobility, such as distance decay, and deep learning fine-tuning to create models that can generalize while accounting for city-specific variations.

The framework integrates domain adaptation techniques to transfer knowledge from data-rich source cities to data-scarce target cities, enabling the creation of models that require less training data from the latter. Case studies involving cities like Coventry, Birmingham, Bogotá, and Amman demonstrate the approach’s ability to fine-tune universal models for individual city contexts, leading to more effective urban planning and mobility policy development. This research contributes to the efficient use of data in urban modelling and supports the transfer of policy insights across urban environments, thereby enhancing decision-making in urban planning and development.

Biography

Throughout my career, I've been driven by a passion for making complex systems accessible. As a software engineer and system analyst, I specialised in building adaptable, no-code platforms that empower non-technical users to create their own software solutions.

Alongside my work in tech, I've always held a deep fascination with mapping and the dynamics of urban spaces. This led me to pursue an MSc in Urban Analytics at the University of Warwick and, subsequently, a PhD at UCL focused on creating generalizable mobility models. My research aims to provide cities of all sizes and data capacities with the tools to understand and plan their transportation systems effectively.

Both my software work and my urban research share a common goal: democratising access to powerful tools. Just as no-code systems open up the world of software development, my research strives to ensure that even cities without extensive planning expertise can build urban models that meet is adaptable to their unique needs.

Funding

This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 949670), and from Economic and Social Research Council, UK (grant No. ES/Y010558/1).

Links

Image: Adham Enaya