Turn Right: Analysis of Rotation Errors in Turn-by-Turn Navigation for Individuals with Visual Impairments
Dragan Ahmetovic, Uran Oh, Sergio Mascetti, Chieko Asakawa · 2018 · Proceedings of the 20th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '18) · doi:10.1145/3234695.3236363
Summary
This paper examines a frequently overlooked problem in navigation assistance for blind and visually impaired (BVI) people: how accurately users execute rotation instructions during turn-by-turn guidance. While much research has focused on improving localization accuracy (knowing where the user is), this study addresses the complementary challenge of execution accuracy (how well users follow instructions once received). Even small rotation errors are amplified by subsequent forward movement, potentially leading users onto incorrect or dangerous paths. The researchers analyzed trajectory data from 11 blind participants who were guided along three routes through a multi-story shopping mall using NavCog, an open-source Bluetooth beacon-based indoor navigation assistant. The routes totaled approximately 400 meters with 26 turns, through a complex environment spanning a basement and multiple floors across three connected buildings. NavCog provided coarse verbal instructions (e.g., "Turn slightly left") plus audio-haptic feedback (chime and vibration) when the target orientation was reached. The study examined rotation errors across three factors: turn type (slight vs. ample), spatial layout (narrow vs. open), and participant characteristics.
Key findings
Participants over-rotated by an average of 17 degrees across all turns, consistently extending rotations beyond what was instructed. The error pattern was not proportional to the intended angle. "Slight turns" (22.5-60 degrees) produced significantly higher rotation errors (17.4 degrees, SD=4.0) than "ample turns" (60-120 degrees, 13.4 degrees, SD=3.7). Critically, ample turns were consistently approximated to 90 degrees regardless of the actual instructed angle — the average executed rotation was 90.4 degrees even when instructed angles averaged 77 degrees. This "90-degree bias" likely reflects both the prevalence of right-angle turns in built environments (the "Manhattan world" assumption) and the biomechanics of the human body. Video analysis showed that 12% of turns were incorrect: of those, 10% were self-corrected by participants, 47% were system-corrected via updated instructions, and 42% failed entirely (5% of all turns), sometimes requiring intervention to prevent dangerous situations like walking into escalators. Spatial layout (narrow vs. open areas) did not significantly affect turning performance. Prior smartphone and navigation app experience correlated with better accuracy. Surprisingly, participants over 45 performed more accurate turns than younger participants — they were more cautious starting turns and more responsive to stop instructions.
Relevance
This research has practical implications for anyone designing navigation assistive technology. The finding that users systematically over-rotate, particularly on slight turns, suggests that navigation systems should not simply tell users when to stop turning but should anticipate and compensate for over-rotation — for example, by issuing stop instructions earlier based on the expected overshoot. The strong 90-degree bias for ample turns suggests that navigation route planning could improve outcomes by favoring right-angle turns where possible, since users execute these most reliably. The 5% complete failure rate for turns — including potentially dangerous situations — underscores that rotation accuracy is a safety issue, not just a convenience concern. The study also challenges assumptions about user homogeneity: age, smartphone experience, and prior navigation app use all influenced performance, suggesting that adaptive systems that learn individual turning patterns could significantly improve guidance accuracy. The authors propose reinforcement learning approaches to personalize turn compensation per user.
Tags: navigation · blindness · orientation and mobility · indoor navigation · assistive technology · turn-by-turn navigation · Bluetooth beacons