Methods and arrangements for CSI prediction
The present disclosure relates to methods and devices for Channel State Information, (CSI) prediction. More particularly the disclosure pertains to predicting CSI for a dynamic channel that is varying over time, e.g. because the receiver is moving. This object is obtained by a method performed in a first wireless node of predicting CSI of a dynamic wireless channel H between the first wireless node and a second wireless node. The method comprises deriving channel covariance estimates C k (n), . . . , C k (n−M) of the dynamic wireless channel H, estimating one or more channel properties of the dynamic wireless channel H, wherein one of the estimated channel properties defines a spectrum spread of the dynamic wireless channel H, and determining a covariance prediction filter, based on the estimated one or more channel properties. The method further comprises predicting one or more channel covariance estimates Ĉ k (n+N|n) by applying the determined covariance prediction filter to the derived channel covariance estimates C k (n), . . . , C k (n−M) and calculating a predicted CSI using the predicted covariance estimates Ĉ k (n+N|n). Hence, this disclosure proposes predicting CSI by predicting channel covariance using a methodology which implies deriving optimal covariance prediction filters.
1. A method performed in a first wireless node of predicting Channel State Information (CSI) of a dynamic wireless channel between the first wireless node and a second wireless node, the method comprising:
deriving channel covariance estimates C k (n), . . . , C k (n−M) of the dynamic wireless channel H,
estimating one or more channel properties of the dynamic wireless channel H, wherein one of the estimated channel properties defines a spectrum spread of the dynamic wireless channel H,
determining a covariance prediction filter, based on the estimated one or more channel properties,
predicting one or more predicted channel covariance estimates Ĉ k (n+N|n) by applying the determined covariance prediction filter to the derived channel covariance estimates C k (n), . . . , C k (n−M),
calculating predicted CSI using the one or more predicted channel covariance estimates Ĉ k (n+N|n), and
enabling the second wireless node to select optimal transmission parameters for transmissions to the first wireless node by transmitting the predicted CSI to the second wireless node, and
wherein:
k is a channel layer,
n is a sample number in time,
M is a number of past channel estimates,
N is a prediction length, and
(n+N|n) is a prediction time for sample time n+N, given signals measured up to sample time n.
2. The method of claim 1 , wherein the determining is further based on a Signal to Noise Ratio (SNR) of H.
3. The method of claim 1 , wherein the step of deriving involves estimating an effective channel H=H P W, where H P is the physical channel and wherein W is a possible precoding matrix applied by the second wireless node.
4. The method of claim 1 , wherein the determining implies selecting a covariance prediction filter g out of a set of pre-defined covariance prediction filters g 1 , g 2 , . . . , g F that are stored in the first wireless node.
5. The method of claim 1 , wherein the covariance prediction filter is a linear filter.
6. The method of claim 1 , wherein the covariance prediction filter is derived through an approximate IIR realization of an LMMSE filter.
7. The method of claim 1 , wherein when CSI is predicted for a channel layer k, the calculating comprises using a predicted covariance when calculating the level of channel layer k and an instantaneous covariance when calculating the level of co-channel interference.
8. The method of claim 1 , wherein when CSI is predicted for a channel layer k, the calculating comprises using a predicted covariance when calculating the level of channel layer k as well as when calculating the level of co-channel interference.
9. The method of claim 1 , wherein the calculating further comprises estimating a predicted Signal to Noise Ratio (SNR).
10. The method of claim 9 , wherein if the estimated SNR is negative, then SNR is set to a value larger than or equal to 0.
11. The method of claim 1 , wherein the CSI comprises at least one of the following: Modulation and Coding Scheme, Signal to Noise Ratio, Rank Indicator, Precoder, or functions thereof.
12. The method of claim 1 , wherein the channel properties comprises a Doppler frequency f max defining the Doppler spectrum spread of H.
13. The method of claim 1 , wherein the covariance prediction filter is a LMMSE covariance prediction filter g=R h −1 r d,h , where R h −1 is an inverse of the autocorrelation matrix of the total estimated covariance of the channel for all layers C and wherein r d,h is a cross correlation vector between the ideal channel covariance of the actual channel matrix D and the channel covariance of H.
14. The method of claim 13 , wherein the LMMSE filter is estimated by assuming that:
r h (τ)=4σ 4 δ(σ)+ J 0 2 (2π f max τ) and r d,h (τ)= J 0 2 (2π f max τ)
where I 0 is the zeroth order Bessel function of the first kind.
15. A nontransitory computer readable storage medium comprising computer program code which, when executed in a first wireless node, causes the first wireless node to execute a method of predicting Channel State Information (CSI) of a dynamic wireless channel between the first wireless node and a second wireless node, the method comprising:
deriving channel covariance estimates C k (n), . . . , C k (n−M) of the dynamic wireless channel H,
estimating one or more channel properties of the dynamic wireless channel H, wherein one of the estimated channel properties defines a spectrum spread of the dynamic wireless channel H,
determining a covariance prediction filter, based on the estimated one or more channel properties,
predicting one or more predicted channel covariance estimates Ĉ k (n+N|n) by applying the determined covariance prediction filter to the derived channel covariance estimates C k (n), . . . , C k (n−M),
calculating predicted CSI using the one or more predicted channel covariance estimates Ĉ k (n+N|n),
enabling the second wireless node to select optimal transmission parameters for transmissions to the first wireless node by transmitting the predicted CSI to the second wireless node, and
wherein:
k is a channel layer,
n is a sample number in time,
M is a number of past channel estimates,
N is a prediction length, and
(n+N|n) is a prediction time for sample time n+N, given signals measured up to sample time n.
16. A first wireless node configured for predicting Channel State Information (CSI) of a dynamic wireless channel H between the first wireless node and a second wireless node, the first wireless node comprising:
radio circuitry adapted to receive a radio signal transmitted over the wireless channel H and
processing circuitry adapted to:
derive channel covariance estimates C k (n), . . . , C k (n−M) of the dynamic wireless channel H,
estimate one or more channel properties of the dynamic wireless channel H, wherein one of the estimated channel properties defines a spectrum spread of the dynamic wireless channel H,
determine a covariance prediction filter, based on the estimated one or more channel properties,
obtain one or more predicted channel covariance estimates Ĉ k (n+N|n) by applying the determined covariance prediction filter to the derived channel covariance estimates C k (n), . . . , C k (n−M),
calculate a predicted CSI using the predicted channel covariance estimates Ĉ k (n+N|n),
enable the second wireless node to select optimal transmission parameters for transmissions to the first wireless node by transmitting the predicted CSI to the second wireless node, and
wherein:
k is a channel layer,
n is a sample number in time,
M is a number of past channel estimates,
N is a prediction length, and
(n+N|n) is a prediction time for sample time n+N, given signals measured up to sample time n.
17. The first wireless node of the preceding claim 16 , wherein the processing circuitry is further adapted to select a filter g out of a set of pre-defined filters g 1 , g 2 , . . . , g F that are stored in the first wireless node.
18. The first wireless node of claim 16 , wherein the covariance prediction filter is a linear filter.
19. The first wireless node of claim 16 , wherein the prediction filter g is derived through an approximate IIR realization of a LMMSE filter.
20. The first wireless node of claim 16 , wherein the first wireless node is a User Equipment.