Bio-Inspired Attention System for Dynamic Stimulus Management in a Pet Robot

Abstract

Pet robots are designed to promote bonding through physical contact, which makes them particularly sensitive platforms for tactile and movement stimuli. Therefore, this sensory richness can lead to overstimulation situations when multiple sensors are activated simultaneously, causing incoherent responses from the robot. This paper addresses this problem by presenting the Bio-inspired Attention System (SABi, from its initials in Spanish), a perception architecture designed to manage and prioritise stimuli dynamically in real time. The SABi is inspired by two human cognitive mechanisms: inhibition of return, which reduces the importance of previously attended stimuli, and attentional fatigue, which penalises the repeated occurrence of stimuli of the same type. This module constitutes the perception system of the Mia social robot, allowing it to select the most relevant stimulus at any given moment according to a ranking of priorities. The results of a real interaction scenario show that the SABi improves the robot’s ability to filter redundant signals and respond coherently to the affective context, favouring a more fluid and natural interaction.