Authors
Mitrakov A.
|
Degree
|
Master of Computer Science, Perm State National Research University |
|---|---|
|
E-mail
|
mitrakov-artem@yandex.ru |
|
Location
|
Perm |
|
Articles
|
Knowledge applications for agent synchronization in the parallel discrete-event simulationThis paper is devoted to synchronization algorithms of agent-based parallel discrete-event simulation.
There are a lot of algorithms for such distributed simulation modeling, and most of them are
fallen into two groups: conservative and optimistic ones.
However the sphere of the research is simulation of agent-based systems. The agents are known
to be «smart»: they have some kind of knowledge that helps them reasoning in different situations.
So the idea is that we could use this knowledge base not only for artificial intelligence, but also for
synchronization algorithms in parallel simulation.
This paper provides the family of algorithms called KBASA (Knowledge Based Agent Synchronization
Algorithms) that result to significant improvement of standard optimistic algorithms. They are
designed exclusively on agent-based systems and are not compatible with usual distributed simulation
because of lack of knowledge.
Ontology was introduced as a knowledge base of an agent. The specification of the ontology
is also provided in the paper.
Eventually KBASA were designed and implemented on Akka (for Scala programming language)
and tested on real models. Theoretical description and practical results are also provided as well as
comparison with the standard optimistic algorithms on the same model.
Read more...
|
