cagataydev/

scout-earthrover-ecot

A multi-agent Embodied Chain-of-Thought (ECoT) capture from an Earth Rover Mini Plus driving around a real apartment. Three Strands agents drove the same physical rover concurrently, each recording its own self-contained LeRobot v3 dataset plus a time-aligned record of why it acted (language reasoning + tool calls bound to the video by frame_index).

Not just vision→action (VLA). Vision → reasoning →… See the full description on the dataset page: https://huggingface.co/datasets/cagataydev/scout-earthrover-ecot.