Activity sequence learning in mission critical tasks

In this paper, we present an interface agent, Commander's Learning Agent (CLearn), that aims to learn a commander's mission model and intent in the form of sequences of activities based on frequency. The major benefits of CLearn include automation of repetitive actions performed by the user and near real-time monitoring of available critical resources leading to alert generation when resource degradation will impact missions dependent upon those resources. We develop and implement passive and active monitoring techniques in conjunction with attribute-based learning and sequential pattern mining. While CLearn is relevant to many applications, including information security management, we describe our sequence learning approach in the context of how it can be used to provide IPB (Intelligence Preparation of the Battlespace) support for planning, operations, and assessment missions in Air and Space Operation Centers (AOCs). We describe in detail our plan to enhance the proposed sequence learning approach to a more robust HMM-based approach to learn sequences of activities at various levels of abstractions. The approach is in line with the network centric command and control structure within the military.