Learning to Pack: A Data-Driven Tree Search Algorithm for Large-Scale 3D Bin Packing Problem

Qianwen Zhu,Xihan Li,Zihan Zhang,Zhixing Luo,Xialiang Tong,Mingxuan Yuan,Jia Zeng

DOI: https://doi.org/10.1145/3459637.3481933

2021-01-01

Abstract:ABSTRACTThe 3-dimensional bin packing problem (3D-BPP) is not only fundamental in combinatorial optimization but also widely applied in real world logistics. In the modern logistics industry, the complexity of constraints, heterogeneity of cargoes and scale of orders are dramatically increased, leading to great challenges to devise packing plans up to standard. While the tree search algorithm is proved to be a successful paradigm to solve the 3D-BPP, it is too time-consuming to be applied in the aforementioned large-scale scenarios. To overcome the limitation, we propose a data-driven tree search algorithm (DDTS) to tackle the 3D-BPP. The solution space with complicated constraints is explored by a tree search algorithm, and a convolutional neural network trained with historical data guides pruning the tree so as to accelerate the search process. Computational experiments on real-world datasets show that our algorithm outperforms the state-of-the-art approach with a loading rate improvement of 2.47%. Moreover, the deep learning technique increases searching efficiency by 37.14% with only 0.04% performance loss. The algorithm has been deployed in Huawei Logistics System, which increases the loading rate by 3% and could reduce the logistics cost by millions of dollars per year. To the best of our knowledge, we are the first to embed pruning networks into tree search for the large-scale 3D-BPP.