Uplink-downlink channel transformation using an adaptive Kalman filter for multicarrier systems

In this paper, we consider uplink-downlink transformation of the channel transfer function in frequency division duplex multicarrier systems with time-varying frequency-selective mobile radio channels. The uplink-downlink transformation shall be performed at the base station by exploiting the current uplink and delayed downlink measurements. Based on a physically motivated geometric channel model, where different plane waves are assumed to superpose at the receiver antenna, the uplink-downlink transformation problem is shown to be a linear estimation problem. We propose a Kalman filter based uplink-downlink transformation algorithm. Commonly, the parameters of the Kalman filter, e.g., the state transition matrix, are obtained by using the channel autocorrelation matrix from stochastic channel models, e.g., the Jakes' fading model with truncated one-sided exponential power delay profile. However, the stochastic process for many channels of interest is not ergodic. Thus these stochastic channel models do not describe the statistics for a single realization. We thus propose to use these stochastic channel models for initialization only and then to periodically update the parameters of the Kalman filter by using the estimated channel coefficients. This results in an adaptive Kalman filter based uplink-downlink transformation algorithm. Performance of the proposed algorithm is assessed using Monte Carlo simulations.