Artificial Intelligence has always been a very stimulating work field.
A lot of people try to "take off from cylinder" the magical algorithm
that blows life into the machine, but the problem is complex and has
many facets. A lot of developers and researchers gave their contribution,
usually concentrating on some partial aspects of the task, hoping that
those systems can be further integrated in a general-purpose
architecture. So, at this time we don't have an organic vision of what
we need to address "the big problem".
The ambitious goal of AKIRA is to create a development tool allowing to build virtual worlds and to populate them with agents manifesting high-level behaviour. The advanced underlying logic represents a general framework that allows to integrate all the relevant aspects of the interaction betweens agents and their world. Thanks to AKIRA it is possible to model Agents, with different intelligence degrees and computational requests, that are able to act in an external world interacting with the AKIRA's server through the AKIRA's XML Transmission Protocol.
The software itself is a run-time C++ multithreading environment able
to execute Agents and a web/system development platform to model their
behaviour. The system core is made up of a server daemon that responds to the network web programming and A.I. requests and that physically executes
Agent's instances, plus a programming interface based on a MACRO
language and some automated scripts that allows to create new Agents.
The whole system is written in C++ and integrates different C++ open
source software, implementing various aspects of the framework.
Two soft computing technologies are provided: Fuzzy Logic and Fuzzy
Cognitive Maps. A strong multithread model ensures great scalability and computational power. With respect to the other agent oriented software, AKIRA implements an innovative cognitive model, that is the result of an upcoming cognitive theory, studied at the ISTC-CNR in Rome.