Peer-reviewed optical-fibre sensor uses a ResNet-LSTM model for underwater robot perception

In a peer-reviewed journal article dated 16 August 2026, Haiqing Li, Boxue Sun, Xiaonan Bian, Yujia Wang and Xianping Fu present a U-shaped plastic optical-fibre sensor with a cilium-like cantilever modelled on a fish's lateral-line sense. It records flow components in two directions. Performance figures are author-reported.

A hybrid ResNet-LSTM model estimates velocity and direction; the paper reports a mean error of 0.0105 m/s and 1.51 degrees.

The sensor could aid underwater navigation and marine inspection, but field robustness, calibration drift, fouling and independent replication are still needed.