The integration of Internet of Things (IoT), wireless communications and Artificial Intelligence (AI) has become essential for enhancing IoT service efficiency. Autonomous Flying Base Stations (AFBSs)—such as unmanned aerial vehicles (UAVs) equipped with wireless base stations—have emerged as next-generation technologies that provide dynamic, on-demand coverage to ground users. However, determining optimal AFBS placement remains challenging, as its position directly impacts both energy consumption and the number of connected devices. To address this, providing reliable connectivity while balancing coverage and power efficiency is critical. In this paper we integrate the IoT, wireless communication and AI approaches and propose a deep reinforcement learning (DRL)-based approach to optimize AFBS positioning for ground IoT networks. We formulate mathematical air-to-ground channel models to inform location optimization and evaluate our proposed scheme across three distinct ground IoT distribution scenarios—Gaussian, random and grid. Experimental results demonstrate that the proposed scheme effectively guides the AFBS to stable convergence and best-found 3D coordinates across the three spatial deployment topologies, while also highlighting the significant impact of IoT device distribution on optimal placement decisions.