Decentralized control
Local information
Robots select actions from bounded local observations and feasible action masks.
Open research benchmark
AssemblyGrid v1
A reproducible benchmark for decentralized multi-robot production with local information, material handoffs, temporary robot coalitions, productive concurrency, and geometry-dependent feasibility.
Local decisionsShared production
What the benchmark brings together
AssemblyGrid v1 places several production decisions inside the same controlled task-level environment instead of studying them as isolated mechanisms.
Material flow
Robot-to-robot handoffs
Production objects can move through bilateral transfers while retaining recipe and provenance semantics.
Cooperation
Temporary coalitions
Recipe-defined operations can require transient groups of robots to execute an operation together.
Parallel production
Productive concurrency
Independent activities can proceed simultaneously when resource and geometry constraints permit.
Process semantics
Recipe-driven progression
Explicit production recipes define valid process paths, requirements, and terminal outputs.
Physical compatibility
Geometry-dependent feasibility
Task-level geometric constraints can invalidate actions that are otherwise valid in the production logic.
Nine official scenarios
The released benchmark organizes its official scenario suite around three workload families, each provided at easy, medium, and hard levels.
Flow
EasyMediumHard
Coalition
EasyMediumHard
Concurrency
EasyMediumHard
Interactive Three.js demonstration
This visualization uses the AssemblyGrid v1 UR10 and Robotiq 2F-85 demonstration build. The validated choreography and telemetry are preserved in a streamlined public interface with playback and camera controls. Engineering utilities remain internal so the presentation stays focused on the benchmark behavior.
Live Three.js scene
Open full screen
Cite AssemblyGrid v1
If you use the benchmark, experimental environment, or a derived implementation in research, please cite the accompanying paper.
@article{bahrpeyma2026assemblygrid,
title = {AssemblyGrid v1: A Benchmark for Multi-Robot Production with Temporary Coalitions, Local Information, and Geometric Constraints},
author = {Bahrpeyma, Fouad and Heik, David and Reichelt, Dirk},
journal = {arXiv preprint arXiv:2609.16075},
year = {2026},
doi = {10.48550/arXiv.2609.16075},
url = {https://arxiv.org/abs/2609.16075}
}