Automated battle planning for combat models with maneuver and fire support

Abstract

Production combat models and simulations do not have an automated planning capability for multiple units in a command structure, and they lack tools for forward reasoning about effects such as suppression by firepower. These gaps limit modeled battle planning to the speed of meticulous human input. We present an architectural framework for automated battle planners, advocating a separation-of-duties approach—for example, between maneuver and fires—for the design and management of complex planning systems. We then describe a conceptual model for an automated fire support planning component, using greedy best-first search in continuous-time plan-space to reduce the risk of a given maneuver plan in polynomial time. We present an implementation of this component in an architecture conformant to the planning framework. We then describe a quantitative and qualitative approach to verification and validation of a planning model, and apply it to the fire support planner implementation. The results demonstrate that the automated fire support plans are effective at improving simulated combat results after a reasonable running time and have some realistic emergent properties. In addition to the novel planning algorithm, this research provides design principles, evaluation techniques, and promising results to guide improvements in behavior automation for combat models.