AIA 3D scene graph designed to give robots persistent memory
A post describing the system says a vision-language model records what each interaction reveals, while linked keyframes retain details such as the name on a cup.
TLDR
A post describes a persistent-memory system for robots whose underlying 3D scene graph changes based on what the robot does. In the described system, a vision-language model reads each interaction and records what it reveals. Linked keyframes preserve details the graph itself does not store, such as the name on a cup.
Combined views
4.2K
1 Source, first seen 1d ago
A 3D scene graph designed to give robots persistent memory
A post describing the system says a vision-language model records what each interaction reveals, while linked keyframes retain details such as the name on a cup.
TLDR
A post describes a persistent-memory system for robots whose underlying 3D scene graph changes based on what the robot does. In the described system, a vision-language model reads each interaction and records what it reveals. Linked keyframes preserve details the graph itself does not store, such as the name on a cup.