Enabling Self-expression: the Use of Roles to Dynamically Change Adaptation Patterns

Self-expression is the capability of a system of changing its adaptation pattern at runtime; this can lead to better performance still keeping the achievement of the global goal in a Collective Adaptive System (CAS). In this paper, we show how self-expression can be achieved by means of roles. Developers can embed the adaptation logic in pieces of software that represent roles that system units can play: by changing the roles at runtime, the adaptation pattern changes as a consequence. Exploiting a swarm robotics case study, we show the applicability of our approach and the improvement of the performances with respect to keeping the same pattern during the execution.