Power-Location Optimization for Cooperative Nomadic Relay Systems Using Machine Learning Approach

The exact expressions and simple tight lower bounds for end-to-end average symbol error rate (ASER) and outage probability (OP) are derived in amplify-and-forward (AF) source-relay-destination cooperative link, provided that source-destination path is correlated with the source-relay path. Afterward, an optimum power allocation (PA) and relay location (RL) algorithm is presented. The effect of correlation factor and path-loss exponent (PLE) on the optimal nodes' power and location is investigated. The results show that optimizing relay location is more efficient than power allocation. Furthermore, a machine learning (ML) implementation of the proposed convex optimization-based algorithm is investigated to cop the computational burden. Specifically, the data set is obtained by using the proposed algorithm. Given the data set, the optimization algorithm can be translated into a regression problem, and feed-forward neural networks (FNNs) are then employed to solve this problem efficiently. The simulation results represented a compromise between accuracy and computation times for the ML-based joint PA-RL optimization.