This work presents a comparative study of two advanced, communication-free voltage control strategies for islanded DC microgrids (DCMGs) comprising photovoltaic (PV) arrays and battery energy storage systems (BESS), each interfaced via buck-type DC/DC converters. The first method employs a reinforcement learning controller trained with the Deep Deterministic Policy Gradient (DDPG) algorithm to generate real-time voltage references using local observations. The second augments classical droop control by incorporating an Extended Kalman Filter (EKF) for state estimation and a Fuzzy Logic Controller (FLC) for adaptive virtual damping. Both strategies are validated in high-fidelity MATLAB/Simulink simulations under realistic conditions, including load variation, high line impedance, and plug-and-play operation. Results show that both approaches significantly improve voltage regulation, transient response, and current sharing compared to conventional droop control. A comparative analysis highlights trade-offs in dynamic performance, interpretability, and deployment complexity. While simulation-based, the study uses validated models reflecting practical DCMG architectures.