cs.RONov 6, 2026

It Takes Two to Tangram: Multi-Agent Object Centric Reinforcement Learning for Assembly

Authors: Paul Knaußt, Kay Pompetzki, Nico Bohlinger, Niklas Funk, Cristiana de Farias, Rolf Findeisen, Jan Peters

Organizations: Intelligent Autonomous Systems Lab, Technical University of Darmstadt, Germany. · Control and Cyber-Physical Systems Laboratory, Technical University of Darmstadt · Hessian.AI, Germany. · German Research Center for AI (DFKI), SAIROL, Germany. · 

Abstract

Geometric assembly requires fine spatiotemporal coordination to place parts without causing collisions or deadlocks under tight clearances. To address this, we present an object-centric reinforcement learning framework that treats each part as an agent and learns a shared policy for continuous assembly control. A graph neural network represents the pieces, their target poses, and their spatial relationships, and predicts translational and angular velocities for each component. Per-agent proximal policy optimization with a centralized graph critic directly learns motion and temporal coordination within a single feedback policy. We study this formulation through planar Tangram assembly, where pieces must slide and rotate into a given configuration while remaining on the assembly surface. Procedural generation of target configurations and randomized initial arrangements support training and evaluation across diverse assembly problems. Our experiments isolate the roles of relational structure and multi-agent optimization, and compare learned control with sampling-based planning. We demonstrate concurrent assembly in simulation and transfer the policy to sequential execution on a real robot, as our object-centric formulation decouples learned assembly coordination from low-level physical control. Project page: https://tangrambench.org/

Figures & tables

Explore similar work

CardsList
  1. TanCRAFT: Context-Reasoned Anticipatory Feedback for Assembly Tasks

    Nov 10, 2026Sassan Mokhtar, Lars Doorenbos, Marius Bock +4Robotic Assembly

  2. RL-Lock: Reinforcement Learning for Generating Interlocking Assemblies

    Aug 3, 2026Xuyang Ma, Chaewoon Kim, Haonan Zhang +33D Shape GenerationRL for Combinatorial Optimization

  3. Assembling Two Parts in One Hand

    Sep 9, 2026Liuao Pei, Tianyue Wu, Hui Zhang +2RL for RoboticsDexterous Manipulation