Large-Scale MIMO Pilot Contamination: Deep Learning-Assisted Pilot Assignment Scheme

The addition of a large number of antennas to a conventional MIMO system results in significant improvement in the performance of multicellular communication systems. The performance of this so-called massive MIMO system however suffers from pilot contamination. This interference to a user communication by a nearby cell base station causes significant limitation in the performance of the system. In this work, we propose a pilot allocation scheme as a careful allocation of pilots sequences can mitigate the diverse effect of pilot contamination. We train convolutional neural networks to discover the best set of users that can share the same pilot sequences such that contamination does not occur. The simulation results show that our proposed solution is capable of pilot assignment to avoid pilot contamination.