# Rearrangement with Nonprehensile Manipulation Using Deep Reinforcement Learning

arXiv: https://arxiv.org/abs/1803.05752 · HTML: https://mlsanity.com/paper/1803.05752/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 Mar 2018.

## Built on by (most cited first, 9 of 9)

- [Visual Room Rearrangement](https://mlsanity.com/paper/2103.16544/) (2021)
- [PIN-WM: Learning Physics-INformed World Models for Non-Prehensile Manipulation](https://mlsanity.com/paper/2504.16693/) (2025)
- [Semantically Grounded Object Matching for Robust Robotic Scene Rearrangement](https://mlsanity.com/paper/2111.07975/) (2021)
- [Imitation Learning for Nonprehensile Manipulation through Self-Supervised Learning Considering Motion Speed](https://mlsanity.com/paper/2206.10289/) (2022)
- [Efficient and High-quality Prehensile Rearrangement in Cluttered and Confined Spaces](https://mlsanity.com/paper/2110.02814/) (2021)
- [A Simple Approach for Visual Rearrangement: 3D Mapping and Semantic Search](https://mlsanity.com/paper/2206.13396/) (2022)
- [Semi-Autonomous Teleoperation via Learning Non-Prehensile Manipulation Skills](https://mlsanity.com/paper/2109.13081/) (2021)
- [Hierarchical Policy for Non-prehensile Multi-object Rearrangement with Deep Reinforcement Learning and Monte Carlo Tree Search](https://mlsanity.com/paper/2109.08973/) (2021)
- [Push Smarter, Not Harder: Hierarchical RL-Diffusion Policy for Efficient Nonprehensile Manipulation](https://mlsanity.com/paper/2512.10099/) (2025)

## Abstract (arXiv)

Rearranging objects on a tabletop surface by means of nonprehensile manipulation is a task which requires skillful interaction with the physical world. Usually, this is achieved by precisely modeling physical properties of the objects, robot, and the environment for explicit planning. In contrast, as explicitly modeling the physical environment is not always feasible and involves various uncertainties, we learn a nonprehensile rearrangement strategy with deep reinforcement learning based on only visual feedback. For this, we model the task with rewards and train a deep Q-network. Our potential field-based heuristic exploration strategy reduces the amount of collisions which lead to suboptimal outcomes and we actively balance the training set to avoid bias towards poor examples. Our training process leads to quicker learning and better performance on the task as compared to uniform exploration and standard experience replay. We demonstrate empirical evidence from simulation that our method leads to a success rate of 85%, show that our system can cope with sudden changes of the environment, and compare our performance with human level performance.