# Learning a Decentralized Multi-arm Motion Planner

arXiv: https://arxiv.org/abs/2011.02608 · HTML: https://mlsanity.com/paper/2011.02608/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Nov 2020.

## Built on by (most cited first, 20 of 20)

- [RoCo: Dialectic Multi-Robot Collaboration with Large Language Models](https://mlsanity.com/paper/2307.04738/) (2023)
- [DualAfford: Learning Collaborative Visual Affordance for Dual-gripper Manipulation](https://mlsanity.com/paper/2207.01971/) (2022)
- [Distilling Motion Planner Augmented Policies into Visual Control Policies for Robot Manipulation](https://mlsanity.com/paper/2111.06383/) (2021)
- [Language-Conditioned Path Planning](https://mlsanity.com/paper/2308.16893/) (2023)
- [Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation](https://mlsanity.com/paper/2505.09144/) (2025)
- [Jacta: A Versatile Planner for Learning Dexterous and Whole-body Manipulation](https://mlsanity.com/paper/2408.01258/) (2024)
- [Planning-Guided Diffusion Policy Learning for Generalizable Contact-Rich Bimanual Manipulation](https://mlsanity.com/paper/2412.02676/) (2024)
- [Learning-based Motion Planning in Dynamic Environments Using GNNs and Temporal Encoding](https://mlsanity.com/paper/2210.08408/) (2022)
- [RoboBallet: Planning for Multi-Robot Reaching with Graph Neural Networks and Reinforcement Learning](https://mlsanity.com/paper/2509.05397/) (2025)
- [Reinforcement Learning with Prior Policy Guidance for Motion Planning of Dual-Arm Free-Floating Space Robot](https://mlsanity.com/paper/2209.01434/) (2022)
- [Bi-Manual Block Assembly via Sim-to-Real Reinforcement Learning](https://mlsanity.com/paper/2303.14870/) (2023)
- [DAIR: Disentangled Attention Intrinsic Regularization for Safe and Efficient Bimanual Manipulation](https://mlsanity.com/paper/2106.05907/) (2021)
- [V-MAO: Generative Modeling for Multi-Arm Manipulation of Articulated Objects](https://mlsanity.com/paper/2111.03987/) (2021)
- [R×R: Rapid eXploration for Reinforcement Learning via Sampling-based Reset Distributions and Imitation Pre-training](https://mlsanity.com/paper/2401.15484/) (2024)
- [Neural Marionette: Unsupervised Learning of Motion Skeleton and Latent Dynamics from Volumetric Video](https://mlsanity.com/paper/2202.08418/) (2022)
- [A Learning System for Motion Planning of Free-Float Dual-Arm Space Manipulator towards Non-Cooperative Object](https://mlsanity.com/paper/2207.02464/) (2022)
- [Diffusion-Guided Multi-Arm Motion Planning](https://mlsanity.com/paper/2509.08160/) (2025)
- [CLAMGen: Closed-Loop Arm Motion Generation via Multi-view Vision-Based RL](https://mlsanity.com/paper/2103.13267/) (2021)
- [An Efficient Multi-Robot Arm Coordination Strategy for Pick-and-Place Tasks using Reinforcement Learning](https://mlsanity.com/paper/2409.13511/) (2024)
- [RoboArm-NMP: a Learning Environment for Neural Motion Planning](https://mlsanity.com/paper/2405.16335/) (2024)

## Abstract (arXiv)

We present a closed-loop multi-arm motion planner that is scalable and flexible with team size. Traditional multi-arm robot systems have relied on centralized motion planners, whose runtimes often scale exponentially with team size, and thus, fail to handle dynamic environments with open-loop control. In this paper, we tackle this problem with multi-agent reinforcement learning, where a decentralized policy is trained to control one robot arm in the multi-arm system to reach its target end-effector pose given observations of its workspace state and target end-effector pose. The policy is trained using Soft Actor-Critic with expert demonstrations from a sampling-based motion planning algorithm (i.e., BiRRT). By leveraging classical planning algorithms, we can improve the learning efficiency of the reinforcement learning algorithm while retaining the fast inference time of neural networks. The resulting policy scales sub-linearly and can be deployed on multi-arm systems with variable team sizes. Thanks to the closed-loop and decentralized formulation, our approach generalizes to 5-10 multi-arm systems and dynamic moving targets (>90% success rate for a 10-arm system), despite being trained on only 1-4 arm planning tasks with static targets. Code and data links can be found at https://multiarm.cs.columbia.edu.