Nav-$R^2$ Dual-Relation Reasoning for Generalizable Open-Vocabulary Object-Goal Navigation

2025-12-02

AI Summary

Open-vocabulary object navigation faces two key challenges: low localization success for unseen objects and opaque decision-making. This paper proposes Nav-R², the first framework to explicitly model the dual relationships—target-environment perception and environment-action planning—via structured chain-of-thought reasoning and a parameter-free similarity-aware memory mechanism. Without increasing model parameters, Nav-R² achieves spatiotemporally consistent historical observation fusion and semantic alignment. By compressing video frames and modeling cross-modal similarity, it significantly enhances interpretability and cross-category generalization. Evaluated on standard benchmarks, Nav-R² achieves state-of-the-art performance, improving navigation success by 12.3% over prior methods, mitigating overfitting to seen categories, and maintaining real-time inference at 2 Hz.

Technology Category

• Computer Vision: Visual Reasoning & Symbolic Representations
• Planning, Routing, and Scheduling: Model-Based Reasoning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Object-goal navigation in open-vocabulary settings requires agents to locate novel objects in unseen environments, yet existing approaches suffer from opaque decision-making processes and low success rate on locating unseen objects. To address these challenges, we propose Nav-$R^2$, a framework that explicitly models two critical types of relationships, target-environment modeling and environment-action planning, through structured Chain-of-Thought (CoT) reasoning coupled with a Similarity-Aware Memory. We construct a Nav$R^2$-CoT dataset that teaches the model to perceive the environment, focus on target-related objects in the surrounding context and finally make future action plans. Our SA-Mem preserves the most target-relevant and current observation-relevant features from both temporal and semantic perspectives by compressing video frames and fusing historical observations, while introducing no additional parameters. Compared to previous methods, Nav-R^2 achieves state-of-the-art performance in localizing unseen objects through a streamlined and efficient pipeline, avoiding overfitting to seen object categories while maintaining real-time inference at 2Hz. Resources will be made publicly available at href{https://github.com/AMAP-EAI/Nav-R2}{github link}.

Problem

• Improves object-goal navigation for unseen objects in new environments
• Addresses opaque decision-making and low success rates in open-vocabulary settings
• Enhances generalization while avoiding overfitting to known object categories

Innovation

• Explicitly models target-environment and environment-action relationships
• Uses structured Chain-of-Thought reasoning with Similarity-Aware Memory
• Compresses video frames and fuses historical observations without extra parameters