Prädiktion der Übertragungsfunktion von Mobilfunkkanälen in Zeit- und Frequenzrichtung zum Gewinnen senderseitiger Kanalzustandsinformation
Final Report Abstract
Advanced transmission techniques like adaptive modulation and coding, bit-loading, pre-coders for spatial multiplexing, etc., require transmitter side channel state information (CSI). Unfortunately, only receiver side CSI can be obtained based on the transmission of a priori known training signals. In time division duplex (TDD) systems, the receiver side CSI may be used as transmitter side CSI in the next transmission instant, whereas in frequency division duplex (FDD) systems, the receiver side CSI can be fed back to the transmitter. However, as the mobile radio channel is time-varying, the receiver side CSI is potentially outdated when used as a transmitter side CSI. Furthermore, feedback of the receiver side CSI reduces spectral efficiency as resources which could otherwise be used for transmission of useful data are used to transmit the estimated receiver side CSI. Thus it is essential that the transmitter side CSI is predicted from the receiver side CSI. Two approaches to obtaining transmitter side CSI have been investigated. In the physically motivated approach, channel prediction in multiple-input multiple-output (MIMO) systems is considered using the double directional channel model which characterizes MIMO channels by the complex amplitudes, delays, Doppler shifts, angles-of-departure and angles-of-arrival of the propagation paths. With knowledge of the propagation path parameters, the unknown channel coefficients can be predicted via deterministic reconstruction of the channel transfer function as long as the propagation paths do not change significantly. In the filter based approach, the unknown channel coefficients are modeled as linear combinations, i.e., filtering, of other known channel coefficients. It has been shown that both approaches are mathematically equivalent as long as the filter order is at least equal to the number of propagation paths. Thus the filter based approach, which is computationally much less intensive than the physically motivated approach, is of high practical interest. In this project, linear channel prediction filters are considered. In single-input single-output (SISO) systems, the filter coefficients are estimated by exploiting the linear relation between the known channel coefficients. Thus the equivalence of the filter based and the physically motivated approaches can be used to not only determine the minimum filter order but also to optimize the estimation of the filter coefficients. For MIMO systems, owing to the narrowband assumption for array signal processing, the filter coefficients are the same for all SISO subchannels of a MIMO channel. Thus, in MIMO systems, the filter coefficients can be estimated with high reliability compared with SISO systems. Towards further performance improvement, different side information could be exploited for MIMO channel prediction. A readily available side information which could be exploited for MIMO channel prediction is the feedback information. The delayed and hence outdated transmitter side CSI obtained from the feedback channel can be used jointly with the receiver side CSI to obtain up-to-date transmitter side CSI in FDD systems. It has also been proposed that channel prediction techniques can be used to improve estimation of receiver side CSI. Especially in TDD systems, tracking the MIMO channel improves the performance of not only prediction of transmitter side CSI but also estimation of receiver side CSI. Commonly, towards tracking the MIMO channel, channel correlation estimates obtained from stochastic channel models are used to exploit the linear relation between the channel coefficients. However, as most channels of interest are not ergodic, the considered stochastic channel models do not describe the statistics of the individual channel realizations. As a result of this “surprising” observation, it is proposed that the stochastic channel models should be used for initialization only. After few iterations of the channel tracking, the linear relation between the estimated channel coefficients can be exploited towards tracking the channel. This results in an adaptive channel tracking setup which matches the actual mobile radio channel. Performance assessments using real world channel measurements and Monte Carlo simulations have shown that the proposed algorithms can be used to deliver satisfactory prediction performance.
Publications
-
Coordinated Multi-Point in Mobile Communications. ch. 9.1, pp. 193–208, Cambridge, Cambridge University Press, 2011 ISBN: 978-1-107-00411-5
Zirwas, W.; Thiele, L.; Weber, T.; Palleit, N.; Jungnickel, V.: Channel Estimation for CoMP. Marsch, P.; Fettweis, G. P. (Eds.)
-
Prediction of frequency selective SIMO channels. Proc. IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’11), pp. 1428–1432, Toronto, September 2011
Palleit, N.; Weber,T.
-
Time prediction of non flat fading channels. Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP’11), pp. 2752–2755, Prague, May 2011
Palleit, N.; Weber,T.
-
Channel prediction in multiple antenna systems. Proc. International ITG Workshop on Smart Antennas (WSA’12), pp. 1–7, Dresden, March 2012
Palleit, N.; Weber,T.
-
Time and frequency prediction of the channel transfer functions in multiple antenna systems. Proc. IEEE Wireless Communications and Networking Conference (WCNC’12), pp. 1049–1053, Paris, April 2012
Palleit, N.; Weber,T.
-
Channel prediction using an adaptive Kalman filter. Proc. International ITG Workshop on Smart Antennas (WSA’15), pp. 1–7, Ilmenau, March 2015
Shikur, B. Y.; Weber,T.
-
Uplink-Downlink channel transformation using an adaptive Kalman filter for multicarrier systems. Proc. IEEE International Workshop on Signal Processing Advances in Wireless Communications (SPAWC’15), pp. 465–469, Stockholm, June 2015
Shikur, B. Y.; Weber,T.
