Aggregate plans for multiagent systems

We introduce the notion of 'aggregate plan', as a specification of the collective behaviour of a team of agents embedded in space. Aggregate plans are managed by self-organisation: they can be spread from a source agent, get iteratively diffused to the whole MAS, and be dynamically selected and carried on by each agent based on its local deliberation process. This approach is rooted on the aggregate computing paradigm, recently introduced to address open, dynamic, and large-scale scenarios of pervasive computing and IoT. We define the approach, describe the underlying computational mechanisms, discuss its expressiveness, and present an example of people rescuing in disaster scenarios.