EURO 2025 Leeds
Abstract Submission

3184. Real-Time Optimization of Intra-Hospital Transportation: A Deep Q-Learning Approach

Invited abstract in session TC-13: AI in healthcare, stream OR in Healthcare (ORAHS).

Tuesday, 12:30-14:00
Room: Clarendon SR 1.01

Authors (first author is the speaker)

1. Hossein Torkinezhadirani
Industrial Engineering, Koc University
2. Sibel Salman
Industrial Engineering, Koc University
3. Ali Hassanzadeh
Alliance Manchester Business School, University of Manchester
4. Ozgur Araz
Supply Chain Management and Analytics, University of Nebraska Lincoln

Abstract

Intra-hospital transportation is necessary for maintaining efficient hospital operations and ensuring timely and effective delivery of treatments in large healthcare facilities. This study investigates the Real-Time Intra-Hospital Transportation Problem (RIHTP), which seeks to improve the operational efficiency and responsiveness of the hospital transportation systems by optimizing the management of transporters (porters and service robots with different capacities). The goals are cost reduction, increased service, and equitable task distribution among staff. The static version of the problem, where all requests are known to the system, is modeled using the Dial-A-Ride Problem (DARP) framework. For the dynamic version, we utilize a two- phase approach that decomposes the problem into the assignment and routing problems. In the first phase, the assignment problem is solved. Transportation requests are dynamically assigned by a Deep Q-Learning (DQL) employing a Bayesian Neural Network (BNN). Subsequently, in phase two, the routing problem is solved. The route for the assigned transporter is reoptimized through a heuristic approach (Best Insertion combined with Variable Neighborhood Search).
Based on the preliminary computations, our two-phase approach demonstrates significant improvements in solution quality and computational efficiency compared to the benchmarks.

Keywords

Status: accepted

Back to the list of papers