Lifelong Topological Visual Navigation

TL;DR

This paper introduces a sampling-based graph construction and maintenance approach for topological visual navigation, enhancing lifelong navigation performance and robustness in real-world environments.

Contribution

It presents a novel sampling-based graph building method and maintenance strategies that improve navigation accuracy and adaptability over time.

Findings

Sparse graphs with higher navigation success

Effective removal of spurious edges

Significant performance improvements with lifelong maintenance

Abstract

Commonly, learning-based topological navigation approaches produce a local policy while preserving some loose connectivity of the space through a topological map. Nevertheless, spurious or missing edges in the topological graph often lead to navigation failure. In this work, we propose a sampling-based graph building method, which results in sparser graphs yet with higher navigation performance compared to baseline methods. We also propose graph maintenance strategies that eliminate spurious edges and expand the graph as needed, which improves lifelong navigation performance. Unlike controllers that learn from fixed training environments, we show that our model can be fine-tuned using only a small number of collected trajectory images from a real-world environment where the agent is deployed. We demonstrate successful navigation after fine-tuning on real-world environments, and notably show significant navigation improvements over time by applying our lifelong graph maintenance strategies.