Search papers, labs, and topics across Lattice.
This paper introduces DB-Nav, a novel framework for object navigation that incorporates dual relational biases to enhance search efficiency in unknown environments. By distinguishing between reliable and unreliable semantic cues, DB-Nav utilizes an Activation Bias to propagate contextual evidence and an Inhibition Bias to suppress misleading information, thereby improving decision-making. Experimental results demonstrate that DB-Nav significantly outperforms existing navigation methods in both success rate and path efficiency, providing a more robust approach without the need for complex vision-language model reasoning.
DB-Nav redefines object navigation by effectively filtering out misleading cues, leading to a dramatic increase in navigation success rates.
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 Open-vocabulary perception is prone to systematic misleading evidence: false positives, outdated static priors, and repeated failed exploration due to lack of embodied verification, which contaminates mapping and decision-making.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.
