GuideNav: User-Informed Development of a Vision-Only Robotic Navigation Assistant For Blind Travelers

TL;DR

GuideNav develops a vision-only teach-and-repeat navigation system for blind travelers, leveraging a large open formative dataset (GuideData) to inform a modular VT&R architecture that repeats taught routes without GPS or LiDAR. It combines topological mapping with visual place recognition and direct pose estimation to achieve kilometer-scale outdoor guiding on a quadruped robot, validated with BLV users and professional trainers. Field results show reliable navigation, low cognitive load for users, and meaningful differences from real guide dogs that guide future refinements. By releasing open data and demonstrating a practical, camera-based approach, the work advances human-centered design for assistive mobility and broadens deployment potential across devices and environments.

Abstract

While commendable progress has been made in user-centric research on mobile assistive systems for blind and low-vision (BLV) individuals, references that directly inform robot navigation design remain rare. To bridge this gap, we conducted a comprehensive human study involving interviews with 26 guide dog handlers, four white cane users, nine guide dog trainers, and one O\&M trainer, along with 15+ hours of observing guide dog-assisted walking. After de-identification, we open-sourced the dataset to promote human-centered development and informed decision-making for assistive systems for BLV people. Building on insights from this formative study, we developed GuideNav, a vision-only, teach-and-repeat navigation system. Inspired by how guide dogs are trained and assist their handlers, GuideNav autonomously repeats a path demonstrated by a sighted person using a robot. Specifically, the system constructs a topological representation of the taught route, integrates visual place recognition with temporal filtering, and employs a relative pose estimator to compute navigation actions - all without relying on costly, heavy, power-hungry sensors such as LiDAR. In field tests, GuideNav consistently achieved kilometer-scale route following across five outdoor environments, maintaining reliability despite noticeable scene variations between teach and repeat runs. A user study with 3 guide dog handlers and 1 guide dog trainer further confirmed the system's feasibility, marking (to our knowledge) the first demonstration of a quadruped mobile system retrieving a path in a manner comparable to guide dogs.

Figures (2)

• Figure 1: GuideNav teach-and-repeat framework. In the teach phase, an expert performs remote control of the robot along the desired route. The robot records a sequence of images, and the topomap generator constructs a topological map by extracting keyframe images. In the repeat phase, the robot performs visual place recognition by matching current image features, $\mathbf{f}_t$, extracted by the feature extractor $\Phi$, with the topomap using cosine similarity, and then selects a subgoal. The pose estimator, $\mathcal{R}$, computes the relative pose between the current and subgoal images. GuideNav achieved kilometer-scale autonomous navigation under varying scene conditions. The system was also used by three guide dog handlers and one guide dog trainer, who provided positive feedback on its potential to support mobility assistance for BLV individuals.
• Figure 2: GuideNav outputs during guidance of H02, showing topological map node prediction and relative pose estimation. The system correctly predicts the node despite drastic lighting changes and unseen objects, and maintains accurate pose estimation.