par Sion, Antoine;Reina, Andreagiovanni ;Birattari, Mauro ;Tuci, Elio
Référence Lecture notes in computer science, 13499 LNAI, page (193-204)
Publication Publié, 2022-01-01
Article révisé par les pairs
Résumé : Self-organised aggregation is one of the basic collective behaviours studied in swarm robotics. In this paper, we investigate an aggregation problem occurring on two different sites. Previous studies have shown that a minority of robots, informed about the site on which they have to aggregate, can control the final distribution of the entire robot swarm on the sites. We reproduce this strategy by adapting the previous probabilistic finite-state machine to a new simulated robotic platform: the Kilobot. Our simulation results highlight that the update time (i.e., the amount of time a robot waits before making a decision on leaving a site) impacts the dynamics of the aggregation process. Namely, a longer update time lowers the number of robots wandering in the arena, but can slow down the dynamics when the target final distribution is far from the one initially formed. To ensure a low number of wandering robots while maintaining a quick convergence towards the target final distribution of the swarm, we introduce the concept of a dynamic update time increasing during the aggregation process.