To address the complex response requirements posed by multiple dynamic threats, bystanders, and graduated use of force in urban public-safety emergencies, conventional weapon-target assignment (WTA) models typically treat weapon effectiveness as predetermined parameters and cannot capture dynamic changes in actual engagement effectiveness. This paper proposes FC-TEWA, a dynamic multistage bi-level optimization framework for threat evaluation and weapon assignment with embedded fire-control computation. The framework employs finite-state machines to model target-state evolution, integrates TOPSIS with three-way decision theory to classify targets into engage, observe, and ignore categories, and constructs spatiotemporal feasibility sets through range and line-of-sight checks. Non-lethal fire-control computation is embedded within the outer-layer WTA process to generate firing parameters, threat-mitigation probabilities, and local personnel risks online for feasible shooter–target pairs, enabling joint optimization of threat suppression, collateral-damage risk, and force escalation. A threat-driven multi-operator adaptive large neighborhood search with simulated annealing (TM-ALSA) and an elite-led hierarchical collaborative black-winged kite algorithm (ELHCBKA) are developed for the outer and inner problems, respectively. ELHCBKA achieves the best or tied-best solutions across all six heterogeneous firing conditions. Across three scenarios, TM-ALSA yields the lowest mean total cost, with mission success rates of 100%, 96.67%, and 96.67%. Full-mission simulations further demonstrate a closed-loop decision process comprising ”situation update–target classification–weapon assignment–fire-control execution–state feedback,” while prioritizing non-lethal force, escalating force only when necessary, and performing proximity-aware assignment. FC-TEWA effectively balances operational effectiveness, personnel risk, and graduated use of force in dynamic multistage WTA.