Towards Agent Aggregates: Perspectives and Challenges

Recent works in the context of self-organisation foster the idea of engineering large-scale situated systems by taking an aggregate stance: system design and development are better conducted by abstracting away from individuals’ details, rather directly engineering (designing, programming, verifying) the overall system behaviour, as if it were executed on top of a single, continuous-like machine. As a consequence, concerns like interaction protocols, self-organisation, adaptation, and large-scaleness, get automatically hidden “under the hood” of the platform supporting aggregate computing, with notable advantages in raising the abstraction level and scaling with behaviour complexity. This paper provides an initial exploration of potentials and challenges of using aggregate computing techniques in the context of multi-agent systems, considering impact on large-scale reactive MASs, environment engineering and its cognitive exploitation, and on collective team-work by the notion of aggregate plan.