On the Dynamic Evolution of Distributed Computational Aggregates

Engineering and programming approaches for collective adaptive systems often leverage ensemble-like abstractions to characterise a subset of devices as a domain for a given task or computation. In this paper, we address the problem of programming the dynamic evolution of distributed computational aggregates, through neighbour-based coordination. This is a problem of interest, since several situated activities (especially in large-scale settings) require decentralised collaboration, and need to be sustained by limited subsets of devices. These subsets may vary dynamically due to delegation, completion of local contributions, exhaustion of resources, failure, or change in the device set induced by the openness of system boundaries. In order to study and develop how distributed aggregates progressively take form by local coordination, we build on the field-based framework of aggregate processes, and extend it with techniques to support more expressive evolution dynamics. We propose novel algorithms for more effective propagation and closure of the boundaries of dynamic aggregates, based on statistics on the information speed and a notion of progressive closure through wave-like propagation. We verify the proposed techniques by simulation of a paradigmatic case study of multihop message delivery in mobile settings, and show increased performance and success rate with respect to previous work.