01

Atlass &
Isacc Sims

  • With
  • WPP, Boston Dynamics, Canon
  • Scope
  • Robotics, Synthetic Dataset
  • Skills
  • Virtual Production, Nvidia Cosmos, Isaac Sims

This collaborative project between WPP, Boston Dynamics, and Canon reimagines the future of content production bycombining robotics, AI, and virtual production. For the firsttime, Boston Dynamics’ Atlas robot was used behind thecamera—pushing creative boundaries in cinematography whileimproving precision and safety on set. By leveraging roboticmovement and AI-driven workflows, the project demonstrateshow emerging technologies can capture dynamic, previously unreachable shots in challenging or hazardous environments.

  • NVIDIA COSMOS Synthetic Data Workflow
  • Set up and executed the NVIDIA COSMOS pipeline to generate synthetic data from a virtual production shoot rendered in Omniverse. Providing foundational data needed for training RL agents in tasks like object manipulation and navigation in the shoot.
  • Technical Environment Setup with Docker, WSL, and VMs
  • Investigated, configured, and maintained the full runtime environment using Docker, WSL, and virtual machines to ensure stable and reproducible workflow execution.
  • Documentation and Workflow Integration for Robotics and AI
  • Created comprehensive technical documentation to guide internal teams in replicating and scaling the workflow. Validated the integration of COSMOS synthetic data pipelines as a foundation for future projects, including AI-driven robotics and reinforcement learning applications.
  • Future Steps: Reinforcement learning through NVIDIA Isaac Sims
  • The next step involves integrating NVIDIA Isaac Sim into the workflow to enable training of robots using reinforcement learning. By leveraging Isaac Sim’s advanced physics simulation, robots can be trained in highly realistic virtual environments, accelerating development and improving performance before real-world deployment.
© 2026.Eunah Lee All Right Reserved