Dynamic computational offloading approaches for IoT devices in cloud computing

Dynamic computational offloading approaches for IoT devices in cloud computing

연구 분야: Networking

학회: The Journal of Supercomputing

Mobile devices (MDs) have become integral to everyday life due to advancements in the Internet of Things (IoT). Offloading complex tasks to cloud environments can leverage the cloud's computational resources to enhance the performance of IoT devices. However, challenges such as high communication costs, network congestion, and the need for low latency make real-time applications like image recognition and video streaming difficult to handle with cloud-only solutions. To address these challenges, we propose a novel computational offloading method specifically designed for IoT environments, which significantly improves efficiency and performance. The proposed approach incorporates a first-order Gauss–Markov and Minimum Mean Square Error estimation technique to optimize small- and large-scale fading estimation. To ensure efficient allocation of cloud resources, we introduce the Dynamic Collaborative Particle Swarm Offloading Optimization technique, which dynamically adjusts resource allocation based on real-time needs. Additionally, the method integrates multi-level blockchain-enabled data transmission to ensure secure and privacy-preserving data transfers. We also present the Lyapunov-Based Cooperative Deep Reinforcement Optimization (LCDRO) algorithm, which aims to minimize energy consumption and enhance decision-making. Furthermore, a mobility-aware Deep Reinforcement Learning technique is employed to optimize resource usage under varying conditions. The performance of the proposed method is demonstrated with qualitative data: The technique achieves a mean square error of 18.7 dB, a latency of 70 ms, energy consumption of 320 J, and a transmission delay of 84 ms for 50 tasks. When compared to existing methods such as DCOS and ADRLO, our approach shows superior efficiency, scalability, and reliability, making it well-suited for large-scale IoT applications. This study contributes to the development of cost-effective, resource-efficient offloading strategies for IoT devices in cloud computing environments.