Skip to content

kwpoon/SDF-Pack

Repository files navigation

SDF-Pack: Towards Compact Bin Packing with Signed-Distance-Field Minimization

Jia-Hui Pan1, Ka-Hei Hui1, Xiaojie Gao1,3, Shize Zhu3, Yun-Hui Liu2,3, Pheng-Ann Heng1 and Chi-Wing Fu1

1Department of Computer Science and Engineering, 2Department of Mechanical and Automation Engineering,

The Chinese University of Hong Kong.

3Hong Kong Centre for Logistics Robotics

Introduction

This repository is for our IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2023 paper 'SDF-Pack: Towards Compact Bin Packing with Signed-Distance-Field Minimization'. In this paper, we present a novel packing method named SDF-Pack which leverages the truncated signed distance field to model the container’s geometric condition. We also developed the SDF-minimization heuristic to effectively evaluate spatial compactness and find compact object placements.


The current version contains the key implementation for the signed distance field (SDF) construction given the container's top-down heightmap and that of the SDF-Minimization packing heuristics. We are still updating the repository.

Dataset Preparation

We performed experiments on 1000 packing sequences of 96 types of objects collected from the YCB dataset and the Rutgers APC RGB-D dataset. Please download the processed dataset from Google Drive and extract the files in the folder ./dataset/. The object IDs forming the packing sequences can be found at 1000_packing_sequences_of_80_objects.npy.

|-- 1000_packing_sequences_of_80_objects.npy
|-- dataset  
|   |-- our_oriented_dataset
|   |   |-- 00000003_072-c_toy_airplane-processed.ply
|   |   |...
|   |-- our_oriented_decomp
|   |   |-- 00000003_072-c_toy_airplane-processed.obj
|   |   |...
|   |-- our_oriented_occs
|   |   |-- 00002777_cheezit_big_original-processed_objocc.npy
|   |   |-- 00002777_cheezit_big_original-processed_depth.npy
|   |   |...

The subfolder ./dataset/our_oriented_dataset/ contains the object meshes that are simplified and processed to be watertight. These meshes are further processed through V-HACD convex decomposition for collision simulation, and the processed collision models are presented in the folder ./dataset/our_oriented_decomp/. We also provide the voxelization results of the objects in ./dataset/our_oriented_occs/.

Key Implementation

Our key implementation of the GPU-based SDF construction and the SDF-Minimization packing heuristics can be found at v11_heuristic_packing.py.

Installation

  • Environment
    conda env create -f environment.yml
    conda activate sdf_pack
    

Citation

@inproceedings{pan2023sdf,
  title={SDF-Pack: Towards Compact Bin Packing with Signed-Distance-Field Minimization},
  author={Pan, Jia-Hui and Hui, Ka-Hei and Gao, Xiaojie and Zhu, Shize and Liu, Yun-Hui and Heng, Pheng-Ann and Fu, Chi-Wing},
  booktitle={2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={10612--10619},
  year={2023},
  organization={IEEE}
}

Recent Updates

Please refer to our updated version PPN-Pack

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages