This study presents a robust Integrated Guidance and Control architecture tailored for nuclear-capable Intercontinental Ballistic Missiles (ICBMs) operating in hostile electronic warfare and GPS-denied environments. Addressing the critical challenge of maintaining precision under severe sensor noise and deliberate fault injection, the proposed solution synergizes an Adaptive Continuous Higher-Order Sliding Mode Controller with a predictive Long Short-Term Memory (LSTM) neural network. The LSTM module provides real-time compensation for inertial navigation errors, enabling the adaptive controller to ensure finite-time convergence and stability. Validation is conducted using a high-fidelity, industrial-grade six-degree-of-freedom simulation that incorporates variable mass/inertia dynamics, Mach-dependent aerodynamics, and realistic actuator constraints. Comparative simulations across four operational scenarios demonstrate that the proposed framework outperforms classical PID, sliding mode, and H∞controllers, achieving a 40% reduction in Circular Error Probable (CEP), a 30–35% decrease in settling time, and a 15% saving in energy consumption. Furthermore, parametric sensitivity analysis confirms robust performance under ±20% uncertainty. These findings establish the proposed architecture as a resilient and scalable solution for next-generation strategic missile guidance.