An Agent-Based Simulation API for Speculative PDES Runtime Environments

Andrea Piccione

Abstract:

Agent-Based Modeling and Simulation is an effective paradigm to model systems which exhibit complex interactions. The goal is studying them in the hope of devising their emergent behavior, if it exists. Applications of this methodology range from modeling agent decisions in the stock market, supply chains, and consumer markets, to predicting the spread of epidemics, the threat of bio-warfare, and the factors responsible for the fall of ancient civilizations.
While Agent-Based Modeling and Simulation has been effectively used in many disciplines, most successful models are still run only sequentially, causing a potential waste of the computing resources offered by modern multi-core architectures. The high reliance of the model developers community on simple and easy-to-use languages such as NetLogo places a limit on the possibility to benefit from more effective runtime paradigms, such as Parallel Discrete Event Simulation (PDES). This is a significant problem since the required size of Agent-Based Models simulations is increasing everyday: traditional implementations are not up to the challenge.
The aim of this thesis is to somewhat bridge the gap between efficient simulation runtime paradigms, in particular Speculative PDES, and AgentBased Modeling and Simulation. For this purpose we propose a semanticallyrich API which allows to implement Agent-Based Models in a simple and effective way.
We also describe the critical points which should be taken into account when implementing this API in a speculative Parallel Discrete Event Simulation environment, in order to scale up simulations on distributed massively-parallel clusters. We include in this thesis a description of the implementation we developed, with a focus on the various optimizations we devised.
Our experimental assessment, carried on our reference implementation, shows that our API allows to implement complex interactions between agents and the surrounding environment with a reduced complexity, while delivering a non-negligible performance increase. This is a first important step in finally making powerful simulation tools accessible to the practitioners, independently of their computer science knowledge.

BibTeX Entry: