SIMPAG - Simple Agents for Complex Worlds

Dmitri Kondratiev
d_kondr@yahoo.com


SIMPAG MAS
simpagmas.pdf
Abstract: SIMPAG MAS is a multi-agent system (MAS) where autonomous agents represent real entities in modeling experiment. SIMPAG MAS is a discrete event simulation system that allows build agent federations according to different decomposition principles from simple agents or simpAgs. SIMPAG MAS uses notion of agent context to represent particular agent activity. Context defines attributes, behaviors and communications specific to some activity. Contexts allow developer program complex agent behavior as a set of more simple ones. Context attributes are sensors and traits. Agent uses these attributes to communicate with other agents in the same context. This paper describes ideas behind simple agents and provides preliminary design of SIMPAG MAS framework.

Simpag CogSpace
cogspace-simpag-ppt.pdf
Abstract: These slides illustrate deployment of SIMPAG MAS simulation on 2D grid with Simple Agents. Also CogSpace construction environment is presented using Scout character as an example. Scout is an actor with a goal to reach some target in 2D grid space. While approaching the target Scout must walk around different obstacles, such as walls, buildings, etc . Scout actor is implemented as Simpag MAS that models an artificial being with the locomotion, vision and orientation capabilities, all implemented by separate simpags.

Constructing Federations from Simple Agents (simpAgs ) for Pengi World
Dmitri O. Kondratiev, Vasily V. Suvorov
R&D, Luxoft
dkondratiev@luxoft.com, vsuvorov@luxoft.com
simpAgsPoster.pdf
Abstract: Pengi simpAg Federation controls penguin in Pengo Game according to deictic representation principles, first presented by Philip E. Agre in his work "The Dynamic Structure of Everyday Life" (Artificial Intelligence Laboratory, Cambridge, MA). Pengi can focus on its environment with deictic entities implemented with simpAgs sensors and traits.

Constructing Federations from Simple Agents and Environment Contexts
(Extended Abstract)
Dmitri O. Kondratiev, Vasily V. Suvorov
dkondratiev@ luxoft.com, vsuvorov@luxoft.com
Submited to Seventh Annual Swarm Researchers Meeting, SwarmFest 2003
simpAgs.pdf
Abstract: This paper attempts to define the core set of agent / environment features and investigate the possibility to realize them with elementary building blocks that agents can be built from. Proposed core agent architecture should allow building agent federations according to different decomposition principles from simple agents. Federations of interest for this research include agent swarms and hierarchies of recursive and self-similar agents or "holons". Minimal set of agent features can be used across various domains to construct agents of arbitrary complexity and intelligence. This work next foresees a set of general modeling primitives to be developed for specialized programming languages. In fully reflective architecture that we develop agents themselves will use these modeling languages to specify their behavior and thus become learning agents.

1 1