Evolutionary Agent Societies

Abstract

Evolutionary Agent Societies (EAS) expand current agent-based IT solutions, by further emphasizing reliability, scalability, and adaptability of intelligent software. Reliability is addressed by seeking societies of similar agents, capable of cooperating and intelligently dividing complex tasks into manageable units. Autonomous, mobile agents capable of seeking each other out, based on their inherent problem solving qualities, and adept at forming loose societies, are sought. Scalability is addressed by both providing mobile agents and having masses of agents collaborate with one another in a parallel distributed manner. Adaptability is addressed by supplying individual agents with the capability to reproduce. Encoding agent capabilities in the form of agent persona's (defining an agent's capabilities, problem solving strategies, and biases), and allowing for natural selection among individual agents within a society, aims to harness the adaptability long observed and sought after in biological systems. This paper reports ongoing R&D efforts in exploiting EAS under the larger goal of constructing Autonomously Learning Intelligent Agents (ALIA) at ANSER. Tackling the tasks of automatic Facial Recognition from various video sources and exploitation of intranet/internet located data sources, is at the heart of this effort.