Optimal Energy Efficiency for Multi-MEC and Blockchain Empowered IoT: a Deep Learning Approach

Wireless Internet-of-Things (IoT) networks empowered by blockchain have became a promising architecture to establish trust and consensus mechanisms in a distributed manner. However, the computational complexity and limited on-board energy of wireless devices impose great challenges on applying blockchain into IoT networks. To address this issue, this work introduces multiple mobile edge computing (MEC) to provide sufficient computational power for miners (i.e., IoT devices). As such, the computation-intensive tasks of the miners can be either computed locally or offloaded to some certain MEC servers to fully exploit the computation resources. Moreover, to decrease the energy consumption of the IoT networks, an optimization problem is formulated to maximize the energy efficiency of IoT devices by jointly optimizing the computation mode selection and power allocation. Since the formulated problem is generally intractable with mixed-integer variables, an Fmincon-based algorithm is proposed, which guarantees a globally optimal solution. To further reduce the computational complexity of the proposed optimal method, a Deep Neural Network (DNN)-based deep learning method is applied to facilitate the computation of the proposed algorithm. Finally, numerical results demonstrate the advantages of the proposed network architecture and the algorithm in terms of both the energy and computational efficiency.