Underwater robot localized sources with 88 mm error using fluorescent whiskers
The system achieved an average localization error of 88 mm in a 600 mm test area.
A new paper posted on arXiv details how a fluorescence‑enhanced whisker array can help an underwater robot locate hydrodynamic sources.
The sensor consists of five whiskers. Each whisker has a nitinol core for flexibility and a fluorescent urethane shell that glows under ultraviolet light. A single monocular camera records the glow.
Image processing first amplifies the fluorescence, then segments each whisker to track its deformation over time.
A lightweight convolutional neural network analyses 2‑second video clips. It extracts temporal changes and cross‑whisker patterns to predict the source direction and distance.
Pool tests reported a mean spatial error of 88 mm, with a radial error of 73.5 mm and an angular error of 2.5°, across a 600 mm region spanning ±30°.
A moving thruster was tracked in real time, showing the method works when the source changes position.
The authors suggest the approach could be added to underwater robots that need to sense flow without relying on acoustic or visual imaging.
Limits: Experiments were limited to a controlled pool environment. The system depends on UV illumination and has only been validated with five whiskers, so performance in open‑water conditions remains unknown.
Source: arXiv cs.RO