RISUD EFA Final Progress Research Salon - Towards a Smart System for Post-Windstorm Tree Debris Cleanup and Transportation Network Restoration in Hong Kong (只有英文版本)
研究院/研究中心講座
摘要
This project presents innovative tools for enhancing urban infrastructure and traffic management in Hong Kong. We integrate multi-source urban data to develop an AI-enabled parking vacancy prediction framework. A large-scale traffic simulation tool models traffic dynamics and simulates vehicle trajectories, using data-driven calibration to replicate real conditions and assess emergent events. We propose a deep learning-based framework, TCNSurv, for predicting failures in civil infrastructure by integrating survival and time series analysis. Additionally, a deep reinforcement learning framework manages traffic through coordinated ramp metering and perimeter control. Lastly, a multi-agent reinforcement learning approach optimizes emergency vehicle dispatching during typhoons. These advancements enhance intelligent urban management in Hong Kong.
