This paper explores the application of Reinforcement Learning (RL) to the two-dimensional rectangular packing problem. We propose a reduced representation of the state and action spaces that allow us for high granularity. Leveraging UNet architecture and Proximal Policy Optimization (PPO), we achieved a model that is comparable to the MaxRect heuristic. However, our approach has great potential to be generalized to nonrectangular packing problems and complex constraints.
@article{arxiv.2409.09677,
title = {Mitigating Dimensionality in 2D Rectangle Packing Problem under Reinforcement Learning Schema},
author = {Waldemar Kołodziejczyk and Mariusz Kaleta},
journal= {arXiv preprint arXiv:2409.09677},
year = {2024}
}