Saved in:
Bibliographic Details
Main Author: Özkan Canay
Format: Recurso digital
Language:English
Published: Zenodo 2025
Online Access:https://doi.org/10.5281/zenodo.17530820
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • This study compares the classical asynchronous advantage actor-critic (A3C) algorithm with a quantum-assisted variant (QA3C) for three-dimensional loading of 20-ft containers in maritime logistics. Simulations used sets of 300, 350, and 400 boxes of varying sizes, and both agents were trained under four qubit configurations (0, 5, 10, and 15). Each configuration was evaluated over fifty trials, recording fill rate, skip rate, and training time. On average, QA3C improved the fill rate by 5.70 percentage points, peaking at 72.40% for 400 boxes with ten qubits. It also produced more uniform, layered stacks that reduced unused volume. Gains remained consistent across box counts, demonstrating QA3C's robustness under varying load sizes. Overall skip rate fell by about 3.00 percentage points and success rate reached 65.70% in the 300-box scenario. A3C training converged in around 1.40 s, whereas QA3C required between 6.60 and 21.50 s as qubit count rose. These findings indicate that adding parametrized quantum circuits to reinforcement learning can boost packing efficiency and produce more stable placement outcomes without extra training time. Such benefits could help optimize container space use by reducing unused volume and improving placement consistency across different box counts.