Topological Graph

A Topological Graph Framework for Enhanced Scene Detection and Understanding in Robotic Navigation

In this work, we propose a topological graph framework to support semantic reasoning for MLLMs in navigation tasks. The environment is represented as a graph, where each node is attributed with a semantically meaningful region label (e.g., bedroom, kitchen), and a node classification model predicts these region types.

Project Report PDF here