Atria: cooperative agent-based recovery in a communications network

Abstract

Today's dynamic battlefield environment requires rapid reconfiguration and reorganization of warfighter assets, communication links, and computing resources in response to enemy attacks. The extent to which supporting operations can take place autonomously, and thus more rapidly, could be critical to neutralizing future enemy threats. This paper describes a software system developed at The Aerospace Corporation called Automated Threat Response using Intelligent Agents (ATRIA). ATRIA agents detect signals indicating an enemy theater ballistic missile attack and then automatically gather strategic data, generate internal plans, and task surveillance and strike resources to track and destroy the missiles. The agents discover and access data and resources distributed across a dynamic, wide area network. In earlier reported versions of ATRIA [Quan, 2001], we assumed that this network infrastructure was always available and operated without errors. However real networks, particularly in the battlefield, are subject to message traffic congestion as well as node and link outages. Accordingly, we have introduced into ATRIA a capability that allows failures and congestion in the network, and then demonstrates how such problems can be automatically detected and repaired by software (SW) agents through the use of unmanned aerial vehicles (UAV). Unlike the previous version of ATRIA, there is no central authority or hierarchical organization that determines problem priority or allocation of resources, such as which network nodes to repair and which UAVs to utilize. Rather, it is the cooperative activity of multiple SW agents, driven by measures of overall node value, that results in distributed, self-organizing problem solving. A prototype has been implemented to allow testing of the ATRIA system against various scenarios. TABLE OF CONTENTS