Computation Offloading and Resource Allocation in C-RAN Supporting Wireless Charging

In this article, we consider a cloud-radio access network (C-RAN) with multiple enhanced remote radio heads (RRHs), which support for wireless charging, computation offloading and information transmission. The employment of enhanced RRHs requires a redesign for efficient resource allocation. Our goal is to minimize the weighted sum of energy transfer time and computation time by jointly optimizing energy transmission time, energy beamforming vectors, offloading factors, information transmission time, and combining vectors, while satisfying the communication performance requirements of terminal nodes. The problem is formulated as a mixed-integer nonlinear programming, which can be solved by alternating optimization (AO) utilizing semidefinite relaxation (SDR) and nonlinear transformation. To reduce the online computation delay of AO algorithm, we propose an optimization method based on the deep learning framework, which can determine an optimal solution with a trained neural network in real-time application. To improve the performance of the learning algorithm, we use two methods to set labels for training deep neural network to obtain the optimal offloading decision: 1) preoptimization and 2) self-evolution. Simulation results show that with the trained neural network, the algorithms based on deep learning can achieve a similar performance to the AO algorithm with less computation time.