Learning Physically Realizable Skills for Online Packing of General 3D Shapes

TL;DR

This paper introduces a reinforcement learning approach for online packing of irregular 3D shapes, focusing on physical realizability and efficient training, resulting in a policy that outperforms existing methods across multiple datasets.

Contribution

The paper presents a theoretically-grounded candidate action generation method and an efficient RL training pipeline for physically realizable 3D packing tasks.

Findings

Outperforms state-of-the-art baselines by at least 12.8% in packing utility

Trains a packing policy within 48 hours in a physics-based simulation environment

Demonstrates effectiveness across diverse real-world shape datasets

Abstract

We study the problem of learning online packing skills for irregular 3D shapes, which is arguably the most challenging setting of bin packing problems. The goal is to consecutively move a sequence of 3D objects with arbitrary shapes into a designated container with only partial observations of the object sequence. Meanwhile, we take physical realizability into account, involving physics dynamics and constraints of a placement. The packing policy should understand the 3D geometry of the object to be packed and make effective decisions to accommodate it in the container in a physically realizable way. We propose a Reinforcement Learning (RL) pipeline to learn the policy. The complex irregular geometry and imperfect object placement together lead to huge solution space. Direct training in such space is prohibitively data intensive. We instead propose a theoretically-provable method for candidate action generation to reduce the action space of RL and the learning burden. A parameterized policy is then learned to select the best placement from the candidates. Equipped with an efficient method of asynchronous RL acceleration and a data preparation process of simulation-ready training sequences, a mature packing policy can be trained in a physics-based environment within 48 hours. Through extensive evaluation on a variety of real-life shape datasets and comparisons with state-of-the-art baselines, we demonstrate that our method outperforms the best-performing baseline on all datasets by at least 12.8% in terms of packing utility.