teal background triangular pattern
June 18, 2024

Embodied AI Workshop at CVPR 2024

Pacific Daylight Time (UTC -7)

Location: Seattle, Washington

About the workshop

Host conference:  (opens in new tab)The Conference on Computer Vision and Pattern Recognition (CVPR) (opens in new tab) June 17-21, 2024

Panel speakers: Ashley Llorens, Ade Famoti, Stevie Bathiche, Olivia Norton (Sanctuary AI)

Workshop scientific advisorAndrey Kolobov

Workshop organizerAde Famoti (opens in new tab) (see all workshop organizers (opens in new tab))

The Embodied AI 2024 workshop will be held in conjunction with CVPR 2024 in Seattle, Washington. The overarching theme of this year’s workshop is Open World Embodied AI: Being an embodied agent in a world that contains objects and concepts unseen during training. For truly effective embodied AI agents, agents should be able to deal with tasks, objects, and situations markedly different from those that they have been trained on.

Embodied AI Workshop at CVPR 2024 | left to right: Ashley Llorens, Ade Famoti, Steven Bathiche, Olivia Norton

This umbrella theme is divided into three topics:

  • Embodied mobile manipulation: We go places to do things, and to do things we have to go places. Many interesting embodied tasks combine manipulation and navigation to solve problems that cannot be done with either manipulation or navigation alone. This builds on embodied navigation and manipulation topics from previous years and makes them more challenging.
  • Generative AI for embodied AI: Generative AI isn’t just a hot topic, it’s an important tool that researchers are using to support embodied artificial intelligence research. Topics such as generative AI for simulation, generative AI for data generation, and generative AI for policies (e.g., diffusion policies and world models) are of great interest.
  • Language model planning: When we go somewhere to do something we do it for a purpose. Language model planning uses large language models (LLMs), vision-language models (VLMs), and multimodal foundation models to turn arbitrary language commands into plans and sequences for action – a key feature needed to make embodied artificial intelligence systems useful for performing the tasks in open worlds.