Machine Reasoning for Effects-Based Operations: A Generic Architecture for Multi-Domain Workarounds Reasoning

Effects-based operations must determine how the enemy might respond to air strikes. Current approaches to predicting enemy response to target damage suffer from serious limitations: they typically do not consider how the enemy might repair or modify thestructure of a target system, they typically reason only about a single type of target system, they cannot adequately represent action duration, concurrent actions, and uncertainty, and their models are difficult for analysts to construct and maintain.In Phase I, we developed a machine reasoning architecture and associated algorithms that address these problems, and implemented a prototype to validate the approach. Our design integrates emerging knowledge acquisition technology with novel extensions tostate-of-the-art resource allocation and Hierarchical Task Network (HTN) planning technology, to provide a generic, configurable framework for computing near-optimal workarounds that reconstitute or otherwise modify network systems to accomplishobjectives.In Phase II, we will fully implement our design and demonstrate its application to computing structural workarounds in multiple network systems, including ground transportation, POL, and electric power. The resulting system will automatically computeworkaround options and predict enemy allocations of resources to achieve time-based, capacitated objectives. It will incorporate extensions to handle uncertainty regarding initial conditions and adversary capabilities, and it will provide knowledgeacquisition technology enabling analysts to readily build and maintain models of these network systems and associated workaround procedures.