Exploiting Side Information for Intelligent and Reconfigurable PHY: Experiments on LTE Transceivers

In the wireless physical layer (PHY), the bit-error-rate (BER) and throughput achieved depends on several factors like channel fading, hardware impairments, interference level, and frequency location. In existing wireless PHY, some of these values are explicitly communicated by the receiver to the transmitter as feedback. Online learning algorithms use such feedback to learn frequencies/channels that offer lower BER and higher throughput. The instantaneous values of some of these factors can be measured at the transmitter itself and can be used as additional information (referred to as side-information, i.e. SI) over the feedback given by the receiver. In certain cases, there is a strong correlation between SI and PHY performance. For example, higher interference adjacent to active transmission implies a high BER. In this paper, we incorporate such SI in the upper confidence bound (UCB) algorithm (referred to as UCBwSI) in a realistic LTE PHY setup. We demonstrate that on-the-fly reconfiguration of various PHY parameters using SI can result in significant improvement in throughput and BER achieved in the network for a wide range of channel and interference conditions. Additionally, simulation results demonstrating the superiority of the UCBwSI algorithm even when SI is erroneous or insignificant highlight the need for further theoretical analysis.