Attentional Mechanisms for Lateral Line Sensing through Spectral Analysis

Abstract. For autonomy in underwater robotics it is essential to develop context-driven controllers, capable of leading from perception to action without human intervention. One of the key challenges in this area is to extract reliable information from noisy sensor signals in a fast and efficient manner. In this context, we present a novelty-detection mechanism for lateral line sensing; this mechanism is meant to highlight interesting stimuli and separate them from the background, by bringing into focus new frequencies appearing in the environment. The method is fast and computationally cheap; additionally, it paves the way for characterization and classification of detected novelties. We present a testing framework to explore how to integrate frequency-related, temporal and spatial information and we demonstrate the viability of this approach in a multiple dipole-source environment.

• Type: Journal
• Year: 2010
• Where: SAB
• Authors: Otar Akanyeti, Maria-Camilla Fiazza, Paolo Fiorini
• Keywords: Multiple Dipole-source Environment, Noisy Sensor, Optimization, SAB 2010, Underwater Robotics