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

arXiv: https://arxiv.org/abs/2212.02094 · HTML: https://mlsanity.com/paper/2212.02094/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- Builds on 31 of its 90 references that are in the map.
- First posted on arXiv 5 Dec 2022.

## Builds on (most cited first, 30 of 31)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation](https://mlsanity.com/paper/1612.00593/) (2016)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [Convex optimization](https://mlsanity.com/paper/2106.01946/) (2021)
- [OpenAI Gym](https://mlsanity.com/paper/1606.01540/) (2016)
- [Prioritized Experience Replay](https://mlsanity.com/paper/1511.05952/) (2015)
- [Rainbow: Combining Improvements in Deep Reinforcement Learning](https://mlsanity.com/paper/1710.02298/) (2017)
- [Dueling Network Architectures for Deep Reinforcement Learning](https://mlsanity.com/paper/1511.06581/) (2015)
- [A Distributional Perspective on Reinforcement Learning](https://mlsanity.com/paper/1707.06887/) (2017)
- [Benchmarking Deep Reinforcement Learning for Continuous Control](https://mlsanity.com/paper/1604.06778/) (2016)
- [Transporter Networks: Rearranging the Visual World for Robotic Manipulation](https://mlsanity.com/paper/2010.14406/) (2020)
- [ABC: A Big CAD Model Dataset For Geometric Deep Learning](https://mlsanity.com/paper/1812.06216/) (2018)
- [Distributed Prioritized Experience Replay](https://mlsanity.com/paper/1803.00933/) (2018)
- [Noisy Networks for Exploration](https://mlsanity.com/paper/1706.10295/) (2017)
- [Distributed Distributional Deterministic Policy Gradients](https://mlsanity.com/paper/1804.08617/) (2018)
- [Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation](https://mlsanity.com/paper/1708.05144/) (2017)
- [Equivariant Transporter Network](https://mlsanity.com/paper/2202.09400/) (2022)
- [A formal proof of the Kepler conjecture](https://mlsanity.com/paper/1501.02155/) (2015)
- [Online 3D Bin Packing with Constrained Deep Reinforcement Learning](https://mlsanity.com/paper/2006.14978/) (2020)
- [Learning 3D Shape Completion under Weak Supervision](https://mlsanity.com/paper/1805.07290/) (2018)
- [A Dataset for Improved RGBD-based Object Detection and Pose Estimation for Warehouse Pick-and-Place](https://mlsanity.com/paper/1509.01277/) (2015)
- [Solving a New 3D Bin Packing Problem with Deep Reinforcement Learning Method](https://mlsanity.com/paper/1708.05930/) (2017)
- [Stable bin packing of non-convex 3D objects with a robot manipulator](https://mlsanity.com/paper/1812.04093/) (2018)
- [Learning Practically Feasible Policies for Online 3D Bin Packing](https://mlsanity.com/paper/2108.13680/) (2021)
- [TAP-Net: Transport-and-Pack using Reinforcement Learning](https://mlsanity.com/paper/2009.01469/) (2020)
- [PackIt: A Virtual Environment for Geometric Planning](https://mlsanity.com/paper/2007.11121/) (2020)
- [Planning Irregular Object Packing via Hierarchical Reinforcement Learning](https://mlsanity.com/paper/2211.09382/) (2022)
- [Toward Fast and Optimal Robotic Pick-and-Place on a Moving Conveyor](https://mlsanity.com/paper/1912.08009/) (2019)
- [Graph-based Reinforcement Learning meets Mixed Integer Programs: An application to 3D robot assembly discovery](https://mlsanity.com/paper/2203.04120/) (2022)

## Built on by (most cited first, 12 of 12)

- [PIN-WM: Learning Physics-INformed World Models for Non-Prehensile Manipulation](https://mlsanity.com/paper/2504.16693/) (2025)
- [DISCO: Efficient Diffusion Solver for Large-Scale Combinatorial Optimization Problems](https://mlsanity.com/paper/2406.19705/) (2024)
- [Deliberate Planning of 3D Bin Packing on Packing Configuration Trees](https://mlsanity.com/paper/2504.04421/) (2025)
- [RASP: Revisiting 3D Anamorphic Art for Shadow-Guided Packing of Irregular Objects](https://mlsanity.com/paper/2504.02465/) (2025)
- [GFPack++: Improving 2D Irregular Packing by Learning Gradient Field with Attention](https://mlsanity.com/paper/2406.07579/) (2024)
- [ASAP: Exploiting the Satisficing Generalization Edge in Neural Combinatorial Optimization](https://mlsanity.com/paper/2501.17377/) (2025)
- [Learning Polynomial Representations of Physical Objects with Application to Certifying Correct Packing Configurations](https://mlsanity.com/paper/2312.06791/) (2023)
- [HERB: Human-augmented Efficient Reinforcement learning for Bin-packing](https://mlsanity.com/paper/2504.16595/) (2025)
- [OPA-Pack: Object-Property-Aware Robotic Bin Packing](https://mlsanity.com/paper/2505.13339/) (2025)
- [One4Many-StablePacker: An Efficient Deep Reinforcement Learning Framework for the 3D Bin Packing Problem](https://mlsanity.com/paper/2510.10057/) (2025)
- [PARSE: Part-Aware Relational Spatial Modeling](https://mlsanity.com/paper/2603.07704/) (2026)
- [Differentiable Packing of Irregular 3D Objects with Adaptive Container Estimation](https://mlsanity.com/paper/2606.16333/) (2026)

## Abstract (arXiv)

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.