Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces ZONDA, a novel framework for zero-shot object navigation that effectively operates in multi-floor environments while addressing dynamic pedestrian avoidance. By integrating heuristic multi-floor planning, multi-view target verification, and dynamic pedestrian tracking, ZONDA significantly enhances navigation capabilities beyond traditional single-floor methods. Evaluated on both a real biped robot and extensive simulations, ZONDA demonstrates superior performance in dynamic environments, outperforming existing baselines in robustness and accuracy.
ZONDA achieves robust zero-shot navigation across multi-floor environments while dynamically avoiding pedestrians, setting a new standard for real-world deployment.
In Object Goal Navigation task, existing methods are typically restricted to static and single-floor environments, ignoring cross-floor topologies and dynamic pedestrian, which limits their real-world deployment. To address these limitations, we propose ZONDA, a zero-shot object navigation with dynamic avoidance framework. In particular, ZONDA integrates three core components: (i) Heuristic multi-floor planning: from height-difference traversable maps, enables stair traversal and cross-floor exploration without a platform-specific learned controller; (ii) Multi-view target verification: cross-checks multi-scale observations with a vision-language model, significantly reducing false positives; and (iii) Dynamic pedestrian avoidance: explicitly tracks and predicts moving pedestrians to generate anticipatory behaviors. Evaluated on a real Direct Drive Tech TITA biped robot and extensive simulations on HM3D and MP3D, ZONDA achieves significantly improved results. Moreover, ZONDA can maintain robust navigation on the dynamic benchmark HM3D-DYNA compared to the existing baseline.