Multimodal Dataset¶

HERCULES ships a dataset-collection pipeline that exports time-synchronized multimodal data for heterogeneous robot teams. For each robot it captures RGB, depth, semantic segmentation, and LiDAR along a coverage or leader–follower trajectory, in standard formats: KITTI-style layouts and ROS 2 bags.

City Block sequence Forest sequence Australia — center coverage Australia — perimeter coverage

Download¶

The released sequences across the desert, forest, and city worlds are hosted on Hugging Face:

huggingface.co/datasets/GeorgiaTech/HERCULES

Collecting your own¶

Trajectories are generated offline and executed with the unified heterogeneous autonomy stack, with complementary coverage and leader–follower patterns. Sensor streams are configured through the settings file and captured via the Image APIs, LIDAR, and the optimized ROS 2 wrapper.

The same exported sequences drive the collaborative SLAM and cooperative perception benchmarks.