Sandro Mehic — Università di Firenze #
Role of information and topology in agent-based competitive systems with limited resources #
The problem of competition, cooperation and the emergence of collective behaviour in the presence of 
limited resources is quite general, and one of cornerstones of evolutionary dynamics both in the natural 
and the in the artificial worlds. For instance, trading is nowadays mainly performed by algorithms which 
act autonomously and almost form ecosystem. These models exhibit complex phenomena, governed by various 
parameters that describe the quantity and the quality of agents involved. We study how these parameters 
influence the efficiency of the models, measured as efficient distribution of limited resources, and how 
the additional information like vicinity, it's structure, the number and the cognitive abilities of 
participants modify these models efficiency. We extend the classical prototype of such environment, i.e, 
minority games, by allowing agents to process additional information. We analyse how this exchange 
information affects the dynamics of the system. The main focus is put on the role of the information from 
each agents vicinity and study the influence of different community structures on the model. We 
investigate several topologies like simple patch vicinity, von Neumann vicinity, small world, scale-free 
networks and a hierarchical small world network. We have found that there is a distinct relation between 
the structure and dimension of the vicinity, ie. the information given to each agent, and the efficiency 
of the model. These results could be used to optimise any kind of distributed algorithmic ecosystem that 
has a finite resources and needs an efficient use of it. We further investigate how these finding can be 
used in algorithmic ecosystems. The context within which we elaborate some of our ideas is high-frequency 
financial markets, that are run by an enormous number of trading algorithms. Among other possible 
applications we consider optimizing the routing protocols for Delay Tolerant Networks, more efficient 
smart-grid energy systems and so on.

<br/><br/>
In collaboration with F. Bagnoli