EURO 2025 Leeds
Abstract Submission

2270. Operation STOR-i Time(tabling): Our entry to IHTC-2024

Invited abstract in session MB-10: Integrated Healthcare Timetabling Competition I, stream Automated Timetabling.

Monday, 10:30-12:00
Room: Clarendon SR 1.06

Authors (first author is the speaker)

1. Matthew Davison
Department of Mathematics and Statistics, Lancaster University
2. Graham Burgess
STOR-i Centre for Doctoral Training, Lancaster University
3. Rebecca Hamm
Maths and Stats, Lancaster University
4. Ben Lowery
Department of Mathematics and Statistics, Lancaster University
5. Adam Page
School of Mathematical Sciences, Lancaster University

Abstract

Nurse rostering is a highly complex optimisation problem where traditional exact methods become computationally infeasible for large instances. Our open-source hyper-heuristic solution for the IHTC-2024 employs a two-phase strategy to construct and refine schedules.

In the first phase, we generate feasible solutions using a greedy heuristic. Mandatory patients are prioritised and sequentially inserted into the schedule. If any remain, they are re-prioritised for the next pass, and the schedule is reset. This process continues until all patients are admitted or the algorithm times out. Nurses are then assigned to shifts in sequence to meet workload demands. The second phase refines the initial solution using a parallelized heuristic, applying diverse sequences of moves to improve schedule quality. Moves are selected according to either a simple random approach or one of two reinforcement learning approaches (Monte-Carlo or Q-learning). Solution acceptance follows either a record-to-record or simulated annealing scheme.

By experimenting with different methods, we gained insight into heuristic approaches for the IHTC-2024, leading to the results for competition submission. Our approach highlights the adaptability and capability of heuristic methods for addressing complex rostering challenges.

Keywords

Status: accepted

Back to the list of papers