Modeling Swarm Intelligence Algorithms for CPS Swarms

Abstract

Swarms of cyber-physical systems (CPSs) have a high potential for innovative and successful applications. Swarm intelligence algorithms are one approach to handle the increased complexity that comes with the high number of CPSs in a swarm. In such algorithms, individual CPSs follow simple rules that lead to an emergent behavior. This has advantages such as adaptability, scalability, and robustness without relying on a central control. Nevertheless, the design of such systems is still a hard problem. In this paper, we propose an approach to model the local behavior of individual CPSs using swarm intelligence algorithms from a software engineering perspective. Therefore, we introduce a two-level hierarchy: The first level models the swarm intelligence algorithms as opaque blocks which are detailed in the second level by individual actions as activity diagrams. The individual actions allow the designer to customize existing and create new swarm intelligence algorithms. This is done by extracting single actions from original swarm intelligence algorithms and assembling them in the activity diagrams. Furthermore, the two-level approach allows us to link the modeling of the CPS hardware with the modelling of the behavior of each CPS in a swarm. Finally, we demonstrate the modeling technique by applying the swarm intelligence algorithm BEECLUST to the robotic platform Spiderino.