| Abstract |
Complex, natural, social, technological and economic systems have
recently given rise to the need of a new paradigm for computational
systems that are adaptive, can self-organise and exhibit emergent
behaviour. The design of such systems concerns a homogeneous set of
agents in which each agent receives an input and has to map it to a `good'
output, and where self-organisation emerges from the interaction between
agents. Although general and simple, this concept is representative
for a very wide spectrum of applications such as protocol design for large
computer networks, design of collective robotics, and automative
traffic engineering. Surprisingly, only a handful of recent research is
aimed at a domain-independent (or: general) design of such systems.
We propose as a first step towards a solution for the design-problem
a framework that tackles the local (agent) level formally and
the global (system) level empirically. This allows us to do
rigorous formal verification of the behaviour of the individual
agents, as well as large-scale empirical validation of the system as a
whole. Besides, it exploits the specific advantages of the approaches
regarding the scale of the system: formalisation is good for small
systems, while simulation works well for (very) large systems. |