Adaptive Deep Energy-Aware Edge Caching for Multimedia IoT in Edge–Fog Networks
—Edge caching has emerged as a key enabler for latency-sensitive multimedia Internet of Things (IoT) applications by bringing content closer to end users. However, existing caching strategies often fail to adapt to dynamic network conditions and do not jointly optimize multiple performance objectives, particularly energy efficiency i n e dge–fog e nvironments. This paper proposes an adaptive deep reinforcement learning (DRL)- based energy-aware edge caching framework for multimedia IoT networks. The approach models caching as a sequential decision-making problem and dynamically learns optimal content placement policies using real-time network states and user demand patterns. A double deep Q-network (DDQN)-based framework is developed with a multi-objective optimization model that jointly improves cache hit ratio, reduces server access, and minimizes energy consumption. An energy-aware reward mechanism is designed to guide efficient caching decisions, while the edge–fog architecture enables scalable deployment. The model operates without prior knowledge of traffic distributions, making it suitable for heterogeneous IoT scenarios. Simulation results demonstrate that the proposed framework significantly outperforms baseline methods in terms of cache efficiency, energy utilization, and reduced server dependency, highlighting its effectiveness for intelligent edge–fog IoT networks.