Distributed Artificial Intelligence
Research Program - Results (9)
A platform for developing systems of asynchronous cognitive agents
In order to test our ideas it was found necessary to develop several platforms
for building systems of agents. The first one was called OSACA
(Open Systems of Asynchronous Cognitive Agents). OSACA was one
of the first environments to allow actual open systems easily, resulting
from the combination of AI techniques and of object architectures (OMG/CORBA).
OSACA constituted a significant improvement over systems like ARCHON. Furthermore,
we offered in OSACA a language for creating agents (LAG), which screened
out the difficult problems of negociation protocols from the user. It also
offered a number of development tools.
We designed a two step development method for creating agents.
-
The first step consisted of cloning the generic agent GAg and adding the
necessary expertise in an OSACA local environment. The user could test
the agent behavior and modify it using a simple text editor.
-
In a second step the agent was migrated to its operating machine, and thus
become truly independent.
The tools we developed (exerciser, monitor, control agents) were the first
tools of a more complex toolbox which will probably be necessary to routinely
design agent systems.
More recent advances have seen the development of two new platforms:
SMAS for simulating complex Multi-Agent
Systems and OMAS for a real time execution.
The designer of the MAS first debugs her system in the SMAS
environment, and then executes it in the real time distributed OMAS environment.
Last modified 8/4/02