Seminar by Micael Couceiro (preparation for Ph.D. defense)
(Supervisors: Prof. Rui P. Rocha, Prof. Nuno M. F. Ferreira (ISEC) )
This seminar will present a complete swarm robotic solution that can be applied to real-world missions. Although the proposed methods do not depend on any particular application, search and rescue (SaR) operations are considered as the main case study due to their inherent level of
complexity. The contributions that will be highlighted revolve around an extension of the Particle Swarm Optimization (PSO) to real Multi-Robot Systems (MRS), denoted as Robotic Darwinian Particle Swarm Optimization (RDPSO). The RDPSO is a distributed swarm robotic architecture that benefits from the dynamical partitioning of the whole swarm of robots by
means of an evolutionary social exclusion mechanism based on Darwin’s survival-of-the-fittest. Nevertheless, although currently applied solely to the RDPSO case study, the applicability of all concepts herein proposed is not restricted to it, since all parametrized swarm robotic algorithms may benefit from a similar approach. The proposed approaches are extensively validated in benchmarking tasks, in simulation, and with real robots. On top of that, and due to the limitations inherent to those (e.g., number of robots, scenario dimensions, real-world constraints), the presented work further contributes to the state-of-the-art by proposing a macroscopic model able to capture the RDPSO dynamics and, to some extent, analytically estimate the collective performance of robots under a certain task. It is the authors’
expectation that this seminar may shed some light into bridging the reality gap inherent to the applicability of swarm strategies to real-world scenarios, and in particular to SaR operations.