Optimization of Sensor Management/Sensor Registration

To provide multiple layers of defense against increasingly complex threats, the heterogeneous sensors of the Ballistic Missile Defense System (BMDS) must be carefully coordinated to make optimal use of their diverse capabilities. In particular, because information demands are tied to the timeline of various decision points in an engagement, and sensor dynamics impose limitations on the timing of sensing tasks, foresight and planning are crucial to ensure the required information is available in time for effective engagements. A sensor resource manager (SRM) for the BMDS must therefore employ sophisticated models of the sensors to satisfy long-term performance goals. The SRM algorithm proposed here is based on approximate solutions of a partially observable Markov decision process (POMDP), to properly account for dynamics and generate the desired farsighted behavior. Moreover, the algorithm does not depend on analytical sensor models, but rather uses simulation-based functional models for predicting complex local sensor behavior. The functional models are integrated with the performance prediction and search method in a systematic and intelligent way, making the algorithm much more efficient than a naive Monte Carlo approach, and providing several mechanisms for scaling the amount of computation to trade off performance and runtime.