Skip to content

HERCULES

An Open-Source Simulation Framework for Heterogeneous Multi-Robot SLAM, Collaborative Perception, and Exploration

HERCULES is a simulation platform for heterogeneous multi-robot autonomy that lets unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs) collaboratively explore and understand large-scale, complex environments. It is an advanced fork of AirSim / Cosys-AirSim built on Unreal Engine 5.

Getting Started Code arXiv Dataset

HERCULES system overview

What HERCULES adds

HERCULES extends Cosys-AirSim with a unified heterogeneous autonomy stack, new sensor modalities, dynamic environments, and ready-to-run multi-robot benchmarks:

It inherits the full Cosys-AirSim / AirSim sensor suite and API surface — cameras, depth, semantic segmentation, LiDAR, GPU-LiDAR, echo/radar, IMU, GPS, instance segmentation, annotation, skid-steer vehicles, and more — documented under Simulator.

Environments

Three photorealistic worlds — desert, forest, and city — each stress a different class of perception and planning failure, plus georeferenced real-world scenes via Cesium for Unreal.

  • City
  • Forest

Citation

@misc{garimella2026hercules,
  title         = {HERCULES: An Open-Source Simulation Framework for Heterogeneous Multi-Robot SLAM, Collaborative Perception, and Exploration},
  author        = {Garimella, Sandilya Sai and Butterfield, Daniel Chase and Wilson, Sean and Gan, Lu},
  year          = {2026},
  eprint        = {2606.22756},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url           = {https://arxiv.org/abs/2606.22756}
}

Acknowledgements

HERCULES builds on AirSim (Microsoft) and Cosys-AirSim (Cosys-Lab, University of Antwerp), and benchmarks ROMAN (MIT-ACL) and DAIR-V2X. Developed at LunarLab, Georgia Institute of Technology. Released under the MIT License.