Channel Gain Prediction For Wireless Links With Kalman Filters And Expectation-Maximization

In non-reciprocal wireless channels, the performance of link adaptation heavily deteriorates with outdated measurement reports. To provide an effective solution to this problem, we construct an accurate channel gain predictor by combining the Kalman Filter (KF) with Expectation-Maximization (EM). This joint design minimizes the mismatch between the autoregressive model of the KF and the true channel gain. After deriving the filter parameters in closed form, we compare the accuracy of our solution to two relevant alternatives in a 3GPP-compliant scenario. Impressive gains at low computational complexity make our solution a promising candidate to improve the next generation of link adaptation functions.