The proposed study outlines the key objectives of strengthening the cyber-physical infrastructure through the infusion of deep reinforcement learning (DRL)-powered adaptive resource management. The proposed foundation taps the flexible learning power of DRL for continuous monitoring, forecasting, and fast reaction in the event of uncertainties, ensuring that resources are assigned optimally in systems that are complex and interdependent. Here, the method becomes intelligent through the decision-making algorithm to see environmental changes and operational feedback and by the way learn to adapt strategy automatically. The system's design fuses a perception module for system state analysis with a DRL controller that modifies its policies in an ongoing way according to the changing state of the environment. This then allows the system to take a proactive role in mitigating any potential failures, implementing recovery strategies that are also efficient, and also ensuring sustainable operational continuity. The proposed approach is a new chapter in the self-healing and adaptive deftness of cyber-physical systems, offering a highly flexible and knowledgeable method to infrastructure resilience challenges faced in uncertain and dynamic environments.