This study aims to investigate the emergence and structural evolution of proto-linguistic items within a collective of socially learning real robots through different versions of a communication game. In the form of different versions of a communication game, the robots transmit, receive and copy proto-linguistic items modeled as collections of short and long beeps. Based on the dynamics of the auditory channel, several game configurations are designed to influence the structural evolution of the learned items. It is observed that, due to copying errors arising during embodied vocal imitation, items with higher structural organization can emerge and evolve within the robot collective. The structural evolution of the learned items is examined in terms of complexity, compressibility, expressibility and learnability.
This study presents a biologically inspired model of linguistic iterated learning that examines the emergence and propagation of structure in learned proto-linguistic items over the auditory channel within a group of socially learning real robots. The working hypothesis adopted in this research is that embodiment and noisy social learning constitute essential substrates for the evolution of communication systems.
It is shown that forgetting, noisy social learning and random item selection are the main mechanisms that allow the structural evolution of proto-linguistic items on a physical system.
This study introduces a novel approach in which a group of real robots is used to model the linguistic iterated learning of a basic proto-language. Within the proposed model, the robots iteratively converge on a Morse code-like system that enables them to categorize distinct colors. The primary innovation of this research is the demonstration that computational models implemented on real robots can serve as an alternative platform for investigating the evolution of proto-linguistic structure.
