ABSTRACT
This work investigates robust resource allocation for counter unmanned aircraft systems (UAS) defense under saturation attacks within the dynamic weapon-target assignment (DWTA) framework. A physics-based simulation of short-range air defense missiles and mixed threat sets is employed to formulate DWTA over discrete time slots, incorporating feasibility masks, time-varying threat priorities, and per-slot launch capacities. Several complementary solution approaches are examined, ranging from exact optimization techniques to reinforcement learning methods. In particular, a feasibility-aware Proximal Policy Optimization (PPO) policy with heuristic action correction is proposed and implemented as part of this study. The evaluation is conducted across multiple scenarios and assesses bi-criteria performance, considering both the achieved objective value relative to the ILP optimum and the computational runtime.
