DynaNav: Dynamic Feature and Layer Selection for Efficient Visual Navigation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Jiahui, Chen, Changhao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909808616013824
author Wang, Jiahui
Chen, Changhao
author_facet Wang, Jiahui
Chen, Changhao
contents Visual navigation is essential for robotics and embodied AI. However, existing foundation models, particularly those with transformer decoders, suffer from high computational overhead and lack interpretability, limiting their deployment in resource-tight scenarios. To address this, we propose DynaNav, a Dynamic Visual Navigation framework that adapts feature and layer selection based on scene complexity. It employs a trainable hard feature selector for sparse operations, enhancing efficiency and interpretability. Additionally, we integrate feature selection into an early-exit mechanism, with Bayesian Optimization determining optimal exit thresholds to reduce computational cost. Extensive experiments in real-world-based datasets and simulated environments demonstrate the effectiveness of DynaNav. Compared to ViNT, DynaNav achieves a 2.26x reduction in FLOPs, 42.3% lower inference time, and 32.8% lower memory usage, while improving navigation performance across four public datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2509_21930
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DynaNav: Dynamic Feature and Layer Selection for Efficient Visual Navigation
Wang, Jiahui
Chen, Changhao
Computer Vision and Pattern Recognition
Robotics
Visual navigation is essential for robotics and embodied AI. However, existing foundation models, particularly those with transformer decoders, suffer from high computational overhead and lack interpretability, limiting their deployment in resource-tight scenarios. To address this, we propose DynaNav, a Dynamic Visual Navigation framework that adapts feature and layer selection based on scene complexity. It employs a trainable hard feature selector for sparse operations, enhancing efficiency and interpretability. Additionally, we integrate feature selection into an early-exit mechanism, with Bayesian Optimization determining optimal exit thresholds to reduce computational cost. Extensive experiments in real-world-based datasets and simulated environments demonstrate the effectiveness of DynaNav. Compared to ViNT, DynaNav achieves a 2.26x reduction in FLOPs, 42.3% lower inference time, and 32.8% lower memory usage, while improving navigation performance across four public datasets.
title DynaNav: Dynamic Feature and Layer Selection for Efficient Visual Navigation
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2509.21930