Demo Overview

Scene 1: Office Tables

Unconscious workers from gas exposure. Robot identifies casualties via face recognition.

Scene 2: Stacked Boxes

Conscious trapped worker with compound hazard. Robot assesses scene via VLM.

Scene 3: Chemical Spill Source

Deceased worker, hazmat containers, vapor plume. Robot identifies root cause.

Objective

3-scene free-play indoor demo at Onyx Lab. Ghost Vision 60 robot dog autonomously navigates, identifies casualties, assesses scenes, and streams to a remote operator at the Milipol venue.

Key Capabilities

  • LLM-driven plan generation (DYNO)

  • Dynamic replanning on sensor input

  • Real-time face recognition on the move

  • VLM scene analysis (Qwen3-Next-80B)

  • Remote operator dashboard with live video