Abstract

Urbanization has given rise to increasingly complex systems that sustain and manage the lives of urban residents. As populations grow, the challenges of managing traffic, overcrowding, and environmental challenges escalate. These complexities place stress on urban ecosystems, demanding innovative solutions to ensure smooth and sustainable operations. We explore critical aspects of urban management concerning human behaviour and environmental challenges. By leveraging state-of-the-art machine learning techniques, including reinforcement learning, vision-language foundation models, and diffusion models, our research addresses pressing issues in pedestrian behaviour prediction, crowd management, and urban air temperature analysis. First, in the context of autonomous driving, an emerging solution to traffic pressure and road safety, we focus on improving the accuracy and reliability of pedestrian behaviour prediction. By incorporating behavioural uncertainty, our work aims to enhance the safety of vulnerable road users in dynamic urban environments. Second, we advance crowd analysis through a novel task: referring expression counting. Beyond traditional counting, this task incorporates attribute based insights, such as distinguishing crowd flow in different directions or identifying gender-specific customer counts. This enriched understanding provides more actionable data for urban planners and businesses. Lastly, we address the environmental challenges of urban areas. Urban heat islands (UHI), significantly influencing human well-being and energy demand, manifest through air temperature variations linked to land use and land cover configurations. Our research highlights the high-resolution air temperature prediction to empower urban planners in mitigating UHI effects. Collectively, our work underscores the importance of integrating advanced machine learning solutions to tackle urban challenges, contributing to safer, more efficient, and environmentally sustainable cities.

Speaker’s Profile

Dai Siyang received her B.Eng. degree in Electrical and Electronic Engineering and M.Sc degree in Computer Control and Automation from Nanyang Technological University (NTU) in 2013 and 2017 respectively. She is currently pursuing a Ph.D. in Information Systems Technology and Design (ISTD) at Singapore University of Technology and Design (SUTD) under the Industrial Postgraduate Program (EDB-IPP). She’s currently working at ST Engineering on machine learning operations (MLOps). Her research focuses on exploring ways to improve urban solutions through advanced machine learning techniques. Her broader research interests include machine learning and computer vision.