HiSPADA: Self-Organising Hierarchies for Large-Scale Multi-Agent Systems

Abstract

The formation of hierarchies within large-scale systems can solve problems of scalability and distributed control. In this paper, we suggest a self-organising partitioning control scheme that uses a distributed set partitioning algorithm to dynamically introduce and resolve hierarchy layers in a decentralised fashion according to the needs of the application at runtime. The partitioning control can work within a predefined organisational framework and is highly adaptable to application-specific needs. We demonstrate the approach with an application from the domain of distributed power management and provide evaluations that show that a self-organising hierarchy formation can increase scalability by simplifying control decisions with negligible overhead. Keywords-Autonomous agents; Hierarchical Systems; Adaptive Systems