Detecting Abnormal Speed of Marine Robots using Controlled Lagrangian Particle Tracking Methods

Abstract

The ability to detect abnormal conditions is of great importance for the survivability of marine robots. However, false alarms can occur and may lead to unnecessary interruption of robotic missions. This paper presents recent results on anomaly detection, which may reduce the rate of false alarms in the framework of controlled Lagrangian particle tracking (CLPT), a theoretical tool that analyzes interactions between robot motion and ocean flow. Adaptive learning algorithms extract vehicle speed as an indicator of anomalies from trajectory information using a predicted trajectory to identify when abnormal motion is detectable. The methods are verified by simulation results.