Lessons Learned from an Evolutionary Algorithm-Based Approach to Impact Assessment Processing

摘要

The contemporary operating environment (COE) imposes significant challenges for military personnel and systems. The asymmetric nature of the threat and the proliferation of sophisticated weapon systems to rogue nations and trans-national terrorist organizations only further the need to ensure friendly forces maintain a high-level of war-fighter readiness. Additionally, the COE impacts directly on military acquisition of weapons and C4ISR systems and technologies and on the development of emerging tactics, concepts, and doctrine to meet these new challenges. As this threat evolves the importance of higher-level fusion processing and the associated development of automated fusion-based decision support systems is seen as a key enabler and facilitator for addressing the challenging requirements of the modern battle-space. Recent years has seen a focused emphasis placed on level 2 (situation assessment) and level 3 (impact assessment) data fusion processing, as defined within the Joint Directors of Laboratories (JDL) model (Steinberg, Bowman, and White). Higher levels of abstraction and inference dominate these level 2 and 3 fusion operations. Within level 2, situation abstraction is the construction of a generalized situation representation from incomplete data sets to yield a contextual interpretation of lower-level fusion products (e.g. entity track data). This level of inference is concerned with deriving knowledge from some type of pattern analysis of level 1 data. The distinction between levels 2 and 3 is that level 3 products attempt to quantify the threat's capability and predict its intent by projecting into the future, whereas level 2 results seek to indicate current hostile behavior patterns. A key component of level 3 processing is the prediction and analysis of likely enemy courses of action (eCOAs). This prediction and evaluation of enemy actions, or impact or threat assessment, is a key element of the Intelligence Preparation of the Battle-space (IPB) process. eCOA prediction and analysis is a time-consuming process requiring battle-staff personnel to manually search the space of eCOAs by generating high-level sketches (eCOA alternatives), analyzing and evaluating each alternative, and then wargaming via informal mental simulations. With the diversity of opposition facing our military forces of the future and the digitization of the battle-space, experienced staff may not exist. eCOAs developed for a particular enemy by a particular commander and staff over a period of time may be over idiosyncratic or stereotyped. Mission success depends on the ability to leverage what information is available about the enemy forces and capabilities, both from experts and from military intelligence sources, and use that to predict the likely enemy COAs for the current, or anticipated, military situation.