Unsupervised machine intelligence for automation of multi-dimensional modulation

In this letter, we propose a new unsupervised machine learning technique for a multi-dimensional modulator that can autonomously learn key exploitable features from significant variations of multi-dimensional wireless propagation parameters, followed by a real-time prediction of the best multi-dimensional modulation mode to be used for the next resilient transmission. The proposed method aims to embrace the potential of the unsupervised $K$ -means clustering into the physical layer of non-coherent multi-dimensional transmission. Simulation results show that the proposed scheme can outperform the benchmarks at a cost of simple offline training.