par Trabattoni, Marco
;Valentini, Gabriele
;Dorigo, Marco 
Référence Lecture notes in computer science, 11172 LNCS, page (57-70)
Publication Publié, 2018-01-01
;Valentini, Gabriele
;Dorigo, Marco 
Référence Lecture notes in computer science, 11172 LNCS, page (57-70)
Publication Publié, 2018-01-01
Article révisé par les pairs
| Résumé : | The design and control of swarm robotics systems generally relies on either a fully self-organizing approach or a completely centralized one. Self-organization is leveraged to obtain systems that are scalable, flexible and fault-tolerant at the cost of reduced controllability and performance. Centralized systems, instead, are easier to design and generally perform better than self-organizing ones but come with the risks associated with a single point of failure. We investigate a hybrid approach to the control of robot swarms in which a part of the swarm acts as a control entity, estimating global information, to influence the remaining robots in the swarm and increase performance. We investigate this concept by implementing a consensus achievement system tasked with choosing the best of two resource locations. We show (i) how estimating and leveraging global information impacts the decision-making process and (ii) how the proposed hybrid approach improves performance over a fully self-organizing approach. |



