Machine Learning Techniques for Channel Estimation in Free Space Optical Communication Systems

In the free space optical (FSO) communication system, condition of optical channel changes continuously. In this kind of channel, a prior channel state information (CSI) at the receiver can help in the data recovery and lead to a significant improvement of the bit error rate (BER) performance. In this work, study on an experimental FSO link with optical turbulence generating (OTG) chamber as an optical channel has been carried out. For estimating the channel coefficients, maximum likelihood estimation (MLE) and Bayesian estimation techniques are used. It has been seen analytically and verified experimentally that estimated channel coefficients in both the cases are almost same. However, due to lower complexity MLE will be preferred over the Bayesian. Further, it is observed that for a given transmitted power level, increase in the pilot symbol length leads to better BER irrespective of turbulence level. These are significant and practically useful results.