Overview
Learn about groundbreaking research from Carnegie Mellon University in a 32-minute video exploring Embodied-RAG, an innovative framework that revolutionizes how robots and autonomous systems process and utilize memory. Dive into the technical details of extending Retrieval-Augmented Generation (RAG) for embodied agents, examining how the framework constructs hierarchical memory systems through topological mapping and semantic forests. Explore practical applications in rescue missions using AI drones, understanding the integration of multimodal data and spatial-temporal correlations. Follow the detailed breakdown of memory construction processes, from building topological maps to organizing observations through agglomerative clustering. Discover how the semantic forest architecture enables different levels of abstraction in robotic perception and navigation, while learning about the DARPA-funded collaboration with Lockheed Martin Corporation. Master concepts in multi-sensor data fusion, geospatial semantic mapping, and tree traversal retrieval mechanisms that enhance robotic navigation and communication in real-world scenarios.
Syllabus
NEW AI Drones tech in rescue missions
New AI Research by Carnegie Mellon Univ
Hierarchical Geospatial Semantic Mapping
Two new components for AI Memory construction
What is a topological map and how to construct
The Hierarchical Semantic Forest in AI memory
Compare with idea from Stanford Univ
How to build Hierarchical Semantic Forest
Perform task w/ tree traversal retriever
DARPA financed w/ Lockheed Martin Corp
Multi sensor data fusion on-board
Taught by
Discover AI