Optimization of Task Offloading Using Deep Reinforcement Learning in Wireless Power Transfer and Edge Computing Systems

Internet of Things (IoT) devices often face limited computational capacity and energy resources. To address these challenges, Wireless Power Transfer (WPT) and Mobile Edge Computing (MEC) are widely used. However, existing heuristic algorithms struggle with real-time offloading decisions due to fast-fading channel fluctuations. This paper proposes a Deep Reinforcement Learning-based Adaptive Online Offloading (DRAOO) algorithm, which integrates Batch Normalized Deep Neural Networks (BNDNN) and Hierarchical Ordinal Quantization (HOQ) to optimize task offloading and resource allocation. HOQ efficiently quantizes task offloading strategies into binary decisions without significantly increasing computational complexity, enabling faster and more precise decision-making. DRAOO algorithm continuously optimizes task offloading decisions in dynamic wireless channel environments, thereby achieving optimal task processing speed. Experimental results demonstrate that DRAOO substantially reduces decision-making time while maintaining efficient computation, leading to improved overall system performance, especially in large-scale Wireless IoT Devices (WIoTDs) scenarios.