Self-organisation with Aggregate Computing: A Reflection Under the Lenses of Multi-agent Systems Engineering

Abstract

In multi-agent systems research, large-scale is traditionally addressed by bio-inspired self-organisation: starting from the paradigmatic case of ant colonies, potentially myriads of agents can be designed to bring about global complex behaviour by simple individual tasks and local interactions, by emergence. With that inspiration, the research on aggregate computing developed a theoretical and programming framework to express and engineer resilient collective behaviour in a compositional and declarative fashion, completely abstracting from population size and shape. This grounds on previous work on co-fields for agent coordination, macro-languages for amorphous computing, and environment design for multi-agent systems, and culminated in the field calculus computing model, novel programming languages, and libraries for complex and distributed situation recognition, swarm behaviour, agent-assisted crowd engineering, and the like. Overall, after a decade, several results have been provided in the area of coordination, programming languages, distributed algorithms and platforms, and self-adaptive and self-organising systems. In this chapter we seek to pave the way for bringing some of such results back to the agent community. We overview such results and elaborate on how they can provide contributions and research perspectives to traditional multi-agent systems research threads, especially in the design of distributed digital signs for environment-mediated agent coordination, collective plans for agents and swarms, and reinforcement learning frameworks for many-agent systems.