About

This is an OpenAI gym environment for the 3D Bin Packing game.

The game is like "3D tetris": to pack coming boxes to a container as space efficient as possible.

Game rules:

• You see 1 or more upcoming boxes.
• You choose an x,y position and rotation of the box
• You place the box into the container and get reward.
• The space to place the box should be accessible from the top, and the box should be physically stable after the placement
• The game ends when you make an invalid placement, either because you hit the top (like tetris) or your placement is unstable.

This "game" is commonly played in logistic, eg packing trucks with items from conveyor belt.

Code refactored from https://github.com/alexfrom0815/Online-3D-BPP-DRL

The "CUT-1", "CUT-2" box sequence generator, and the box stability judgement follows rules as specified in Ref[1]

Reference

1. Online 3D Bin Packing with Constrained Deep Reinforcement Learning

2. https://github.com/alexfrom0815/Online-3D-BPP-DRL (Repo of 1)

3. https://github.com/openai/gym/blob/master/docs/creating-environments.md

Install

git clone https://github.com/ylchan87/gym-BinPack3D.git
cd gym-BinPack3D
pip install -e .

Doc

See doc/MinimalExample.ipynb for example usage

See func step in gym_BinPack3D/envs/BinPack3DEnv.py to customize the reward.

See func check_box_placement_valid in gym_BinPack3D/envs/Container.py to customize the box stability check.