The 3D bin packing problem for multiple boxes and irregular items based on deep Q-network

Irregular packing in e-commerce warehouses is a special case of a three-dimensional box packing problem (3DBPP). It is necessary to select the type and quantity of boxes and determine the location and orientation of the items to maximize the use of the loading space. In this paper, a spatial particl...

Full description

Saved in:
Bibliographic Details
Published in:Applied intelligence (Dordrecht, Netherlands) Netherlands), 2023-10, Vol.53 (20), p.23398-23425
Main Authors: Liu, Huwei, Zhou, Li, Yang, Jianglong, Zhao, Junhui
Format: Article
Language:English
Subjects:
Algorithms
Artificial Intelligence
Boxes
Computer Science
Convexity
Deep learning
Granulation
Hulls
Machine learning
Machines
Manufacturing
Mechanical Engineering
Operations research
Packing problem
Processes
Three dimensional models
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Irregular packing in e-commerce warehouses is a special case of a three-dimensional box packing problem (3DBPP). It is necessary to select the type and quantity of boxes and determine the location and orientation of the items to maximize the use of the loading space. In this paper, a spatial particle model of the 3DBPP for multiple boxes and irregular items is constructed using the three-dimensional (3D) point cloud and granulation method. In the model, the 3D point cloud is used to describe the shapes of irregular items, and the granulation method is used for the transformation from sparse and uneven point clouds to spatial particle convex hulls. In addition, we designed an empirical simulation algorithm (ESA) based on the combination of expert rules extracted in practical packing activities and empirical simulation, and an intelligent algorithm for 3DBPPs with irregular items combined with the framework of the deep Q network (DQN) algorithm in deep reinforcement learning. An instance generator is proposed based on industry data to generate realistic projects with representative attributes for the above two algorithms, such as types of boxes, irregular items, 3D spatial plane convex hulls, and spatially granular data. The numerical results show that the ESA can quickly obtain a high-quality packing scheme, and the intelligent DQN packing algorithm in deep reinforcement learning can avoid the limitation of expert rules and achieve a better scheme with a certain time for the training process.
ISSN:0924-669X
1573-7497
DOI:10.1007/s10489-023-04604-6