Stop Wandering: Efficient Vision-Language Navigation via Metacognitive Reasoning

TL;DR

MetaNav is a novel metacognitive navigation agent that enhances efficiency and robustness in vision-language navigation by integrating spatial memory, history-aware planning, and reflective correction using LLMs.

Contribution

The paper introduces MetaNav, a new approach that incorporates metacognitive reasoning into VLN agents, improving exploration efficiency and adaptability over existing methods.

Findings

MetaNav achieves state-of-the-art performance on multiple benchmarks.

Reduces VLM queries by 20.7% compared to prior methods.

Demonstrates improved robustness and efficiency through metacognitive strategies.

Abstract

Training-free Vision-Language Navigation (VLN) agents powered by foundation models can follow instructions and explore 3D environments. However, existing approaches rely on greedy frontier selection and passive spatial memory, leading to inefficient behaviors such as local oscillation and redundant revisiting. We argue that this stems from a lack of metacognitive capabilities: the agent cannot monitor its exploration progress, diagnose strategy failures, or adapt accordingly. To address this, we propose MetaNav, a metacognitive navigation agent integrating spatial memory, history-aware planning, and reflective correction. Spatial memory builds a persistent 3D semantic map. History-aware planning penalizes revisiting to improve efficiency. Reflective correction detects stagnation and uses an LLM to generate corrective rules that guide future frontier selection. Experiments on GOAT-Bench, HM3D-OVON, and A-EQA show that MetaNav achieves state-of-the-art performance while reducing VLM queries by 20.7%, demonstrating that metacognitive reasoning significantly improves robustness and efficiency.