AECNav: Active Evidence Consolidation for Efficient Zero-Shot Open-Vocabulary Object Navigation

2026-08-11

AI Summary

This work addresses the high latency and low accuracy in open-vocabulary zero-shot object navigation, which stem from redundant perception and insufficient evidence. The authors propose a training-free, evidence-driven perception-decision framework that explicitly distinguishes targets from distractors by treating non-detections as negative evidence. The approach integrates three key components: evidence-gated perception, cluster-based log-odds evidence fusion, and an information-gain-guided active exploration strategy. Evaluated on HM3D-v2, HM3D-OVON, and MP3D, the method achieves success rates of 84.7%, 57.3%, and 51.3%, respectively, with significantly reduced inference overhead. Notably, it demonstrates robust real-world performance on a quadruped robot, attaining a 95% success rate at approximately 5 Hz.

Abstract

Zero-shot object-goal navigation (ZSON) in open-vocabulary scenarios is challenging, as it requires a robot to locate an arbitrarily specified object in an unseen environment without task-specific training. Currently, the task still suffers from high latency and limited accuracy due to redundant perception pipelines and insufficient evidence for reliable target confirmation. In this letter, we reframe ZSON as an evidence-driven perception-to-decision problem and present AECNav, a training-free pipeline built on three components: i) Evidence-gated perception, which utilizes a shared encoding across all reasoning stages to establish a unified semantic basis and eliminate redundant computations; ii) Evidence consolidation, which aggregates detections into cluster-level log-odds beliefs. This explicitly separates genuine target support from the false confidence of visually similar distractors, while treating the absence of expected detections as negative evidence; and iii) Active evidence acquisition, which sustains productive exploration under weak semantic cues by selecting frontiers that maximize information gain at minimal traversal cost. As a result, AECNav significantly outperforms previous methods and achieves state-of-the-art success rates of 84.7%, 57.3%, and 51.3% on HM3D-v2, HM3D-OVON, and MP3D, respectively, with substantially lower inference overhead, and attains 95% success across 40 trials on a physical quadruped robot at roughly 5Hz. Code will be made publicly available upon acceptance.

Problem

• zero-shot object navigation
• open-vocabulary
• evidence consolidation
• perception efficiency
• robotic navigation

Innovation

• Zero-shot navigation
• Evidence consolidation
• Open-vocabulary object navigation
• Active exploration
• Training-free pipeline