Probabilistic Multi-Objective Strike Planning Methodology: STORM on the Pareto Frontier

ABSTRACT:Currently, the Synthetic Theater Operations Research Model (STORM) utilizes a strike planning methodology that provides sub-optimal, and often wasteful, munitions assignments greatly reducing the utility of this vital tool. Numerica Corporation proposes to develop a novel strike planning algorithm for integration into STORM that provides guarantees on mission objectives while simultaneously minimizing fuel expenditure as well as unnecessary losses of munitions and aircraft. Numerica's solution will utilize multi-objective optimization to directly address the balance that must be struck between high value strike planning assignments and the incurred cost of those assignments. Our approach fully and completely represents the options available during difficult scenarios, rather then pretending one perfect solutions exists. Numericas solution also allows for uncertain future events to be considered in the current optimization. This results in solutions that are optimal across time, rather than suboptimal greedy solutions. Leveraging recent results in sampling theory allows our solutions to be scalable and tractable even for the large problem sizes posed by difficult engagement scenarios. The Phase I effort will include a proof-of-concept prototype, compared alongside the existing greedy solution as well as a report on the integration path for the prototype into STORM's command and control manager.BENEFIT:The main product of this Phase I effort will be a novel strike planning algorithm for integration into the Synthetic Theater Operations Research Model (STORM). In contrast to the current solution, this algorithm will simultaneously consider the value of the strike planning assignments as well as the incurred cost of those assignments. Additionally, this solution will consider uncertain future events resulting in a solution that is optimal across time. This will provide a considerable improvement in the quality of strike plans when compared to the greedy solution currently used in STORM. A supplemental benefit of this technology is the ability to leverage Numerica's existing uncertainty quantification tools to learn performance bounds for the output of STORM simulations when the optimal trade off between rewards and costs are not known a priori. The primary transition path for the technology in this proposal is into the command and control manager within STORM. However this solution is not specific to the STORM simulation and so could be applied to any offline planning algorithm.