TopoNav: Topological Graphs As a Key Enabler for Advanced Object Navigation
Yihao Qin, Hang Zhou, Jun Ma,, Renjing Xu,, Yiding Ji
AI summary
Problem
Current LLM-driven ObjectNav methods lack robust spatial memory, causing fragmented reasoning, goal confusion, and inefficient paths in complex, long-horizon tasks.
Approach
TopoNav constructs a dynamic topological graph that encodes room-level nodes and connectivity, integrating it with semantic point clouds and LLM reasoning to guide exploration and strategic backtracking.
Key results
- State-of-the-art performance on standard ObjectNav benchmarks
- Higher success rates and shorter navigation paths compared to baselines
- Effective long-horizon planning via topological memory recall and backtracking
- Successful real-world robotic deployment validating simulation results
Why it matters
Enables embodied agents to overcome memory bottlenecks for robust, scalable navigation in large-scale, dynamic real-world environments.
Abstract
Object Navigation (ObjectNav) has made great progress with large language models (LLMs), but still faces challenges in memory management, especially in long-horizon tasks and dynamic scenes. To address this, we propose TopoNav, a new framework that leverages topological structures as spatial memory. By building and updating a topological graph that captures scene connections, adjacency, and semantic meaning, TopoNav helps agents accumulate spatial knowledge over time, retrieve key information, and reason effectively toward distant goals. Our experiments show that TopoNav achieves state-of- the-art performance on benchmark ObjectNav datasets, with higher success rates and more efficient paths. It particularly excels in diverse and complex environments, as it connects temporary visual inputs with lasting spatial understanding.