Student Seminar by Dr. Yimo Yan on Sep 29, 2025, 3PM

Title: Event-driven policy optimization for dynamic ambulance dispatch: An attention-based reinforcement learning approach

Speaker: Dr. Yimo Yan (Department of Data and Systems Engineering)

Date: Sep 29, 2025 (Monday)

Time: 3:00 pm - 4:00 pm

Venue: Room 8-28, Haking Wong Building, The University of Hong Kong

Registration Link: https://forms.office.com/r/nvWGefbq5n

ITS Student Committee will provide light refreshments and drinks for registered participants.

Abstract: The global healthcare strain from events such as the COVID-19 pandemic has intensified ambulance shortages, leading to prolonged patient waiting times and increased mortality. In this respect, efficient ambulance dispatch presents a backbone for providing timely patient care. This paper tackles the problem of dynamic ambulance dispatch as a semi-Markov decision process, with the objective of minimizing severity-weighted waiting times through en-route re-dispatching and event-driven decisions. We propose a policy gradient algorithm with a self-attention approximator, which enables inter-task (patients) and inter-agent (ambulances) communications under uncertainty and variable inputs. To further enable interpretability, we distill the learned policy into a decision tree with theoretically grounded features. Experiments on synthetic and real-world cases demonstrate reduced average waiting time and its variance for patients, improved ambulance operational efficiency, and effective use of strategic withholding. Our approach contributes to the development of interpretable, rule-based ambulance dispatch systems in resource-constrained medical environments.

Bios: Dr. Yimo Yan graduated from the Department of Data and Systems Engineering at The University of Hong Kong in Aug 2025. Yimo’s work focuses on transportation and logistics optimisation, incorporating methods like mixed integer linear programming, reinforcement learning, deep learning and large language model. His research, published in various academic journals, addresses challenges in last-mile delivery and scheduling.