Rethinking Embodied Navigation via Relational Inductive Bias

2026-06-08

AI Summary

This work addresses key challenges in embodied navigation—namely, false detections from open-vocabulary perception, outdated priors, and the absence of embodied verification—by introducing the first relational inductive bias framework that integrates both activation and inhibition mechanisms. The proposed method constructs a relation-based exploration graph through target-centric relation decomposition and perceptual confusion analysis, dynamically modulating frontier exploration via action-level falsification and online observation-driven graph updates. Crucially, it explicitly models and filters unreliable semantic cues without requiring online vision-language model inference. Experimental results demonstrate that the approach significantly outperforms existing methods on the ObjectNav benchmark, achieving state-of-the-art performance in both Success Rate (SR) and Success weighted by Path Length (SPL), while also enhancing navigation robustness and interpretability.

Abstract

Object navigation requires an agent to locate a target in an unknown environment through visual observations. Existing methods typically rely on open-vocabulary detectors or vision-language models (VLMs) to answer where to search, but often overlook what not to trust - which semantic cues are unreliable. Open-vocabulary perception is prone to systematic misleading evidence: false positives, outdated static priors, and repeated failed exploration due to lack of embodied verification, which contaminates mapping and decision-making. Such errors are rooted in structured object relations in real-world scenes. To address this, we propose DB-Nav, a framework that reshapes the search space via dual relational biases. It factorizes target-centric relations into an Activation Bias (propagates contextual evidence) and an Inhibition Bias (suppresses unreliable regions via perceptual confusion and action-level falsification). These biases are unified into a Relational Activation-Inhibition Exploration Graph that modulates frontier exploration values using online observations and failed accesses. Experiments on ObjectNav benchmarks show that DB-Nav significantly outperforms existing methods in success rate (SR) and Success weighted by Path Length (SPL), offering a lightweight, interpretable, and robust navigation framework without costly online VLM reasoning.

Problem

• embodied navigation
• object navigation
• relational inductive bias
• open-vocabulary perception
• semantic reliability

Innovation

• relational inductive bias
• embodied navigation
• activation-inhibition mechanism
• object navigation
• perceptual reliability