Machine learning assisted delay minimization in full duplex energy constrained cooperative communication network

Abstract

Cooperative communication is key in attaining high network throughput and better coverage area for a wireless powered communication network. The conventional solution methodologies proposed in the literature for these networks suffer from high run-time complexity. This paper presents an efficient polynomial-time machine learning approach for solving the total transmission time minimization problem through relay selection and scheduling of the users in a full duplex wireless powered cooperative communication network. Exploiting the simplicity offered by the tree ensemble models, a distributed gradient-boosted decision tree based model (XGBoost) is incorporated. The inputs of the model are the uplink channel coefficients and energy harvesting rates of the relays and the users, and outputs are the optimal completion time of the relays and users, relay selection and schedule of the users. Simulation results indicate that the proposed model very well approximates the true outputs, i.e., performs very close to the optimal solution of the optimization problem, while maintaining a very low run-time computational complexity. Specifically, for a network of 9 users, the run-time of the proposed model is just 25% of the optimal solution, and this ratio grows as the number of users increases in the network.