Collaborative Learning Agent (CLA) for Trident Warrior

Abstract : Collaborative Learning Agent (CLA) is a technology selected for Navy on Trident Warrior '08, which is an annual FORCEnet SEA Trial. The theme for '08 is "Maritime Domain Awareness". The objective is to demonstrate a set of CLAs in a distributed network to learn behavior patterns from historical MDA data and then apply them for search, prediction, and identification of anomalies and reasons that might cause the anomalies, e.g. weather or potential terrorist activities. We will show collaborating with three MDA participants (Navy, Coast Guard and Police) using unstructured data sources as the bases for normal behavior profiles. A new real-time observation is compared with the normal profiles. An anomaly meter reports and shows if the new observation is an anomaly and why. The TW08 effective attributes include "capable, accurate, usable and relevant" to evaluate CLA as follows: * capable: agent learning and prediction from unstructured data * accurate: compare with predictions with the ones from human analysts * usable: ease-of-use in interface, visualization and display * relevant: does CLA predict anomaly or interesting MDA behavior. We will summarize the evaluation results in terms of these attributes.