Adaptive Opportunistic Beamforming in Ricean Fading Channels - Adaptive Signal Processing in Wireless Communications - page 187

1. As in the conventional opportunistic beamforming, the information on H k (t), or
equivalently R k (t), is fed back to the base station.
2. The base station estimates the DOAs* of the users using the values of H k (t).† he
estimated DOA of user k is denoted by θ ˆ k , k = 1, … , M.
3. The base station conducts the proportional fair scheduling to choose a user. Let
us assume that user k* is chosen.
4. When the base station transmits the data to user k* during a time slot excluding
the mini-slot, the artificial phase shift ϕ (t) of the weight coefficient is set to - θ ˆ k *
and this weight coefficient is multiplied to the transmitted data.
Note that this adaptive scheme forms the beams only in the directions where users
really exist, as opposed to conventional opportunistic beamforming, which forms the
beams blindly over the omnidirectional space. Therefore, the adaptive scheme can
improve the performance without wasting resources such as time and power. In adap-
tive opportunistic beamforming, the LOS-related component of H k ( t ), which is G k ( t ), is
maximized. Therefore, the performance heavily depends on the K -factors. Specifically,
larger K -factors { K k } result in larger performance improvement; smaller K -factors result
in smaller improvement. In an extreme case where the K -factors are zero, which is the
Rayleigh case, the adaptive scheme reduces to the conventional scheme because ϕ( t ) is
chosen as a sample of a random variable uniformly distributed over [0,2π).
The vital step of adaptive opportunistic beamforming is to estimate the DOA as accu-
rately as possible with the information available. In the previous publications, a number
of algorithms for DOA estimation have been studied, such as the multiple signal classifi-
cation (MUSIC) method [20], Root-MUSIC [21], and the estimation of signal parameters
via the rotational invariance technique [22] (also see [ 4 ] and the references therein).
Those previous methods, however, are not applicable to the system considered, because
the base station has only very limited channel information: the magnitude values H k ( t )
of the equivalent channels. In the next section, a new and efficient DOA estimation
algorithm requiring only H k ( t ) values is proposed for use in adaptive opportunistic
beamforming.
7.4.2 Estimation of Users' DOAs
In this section, a maximum-likelihood (ML) estimator of {θ k } k =1 is developed.‡ To this
end, the probability density function (PDF) and cumulative density function (CDF) of
H k ( t ) are first derived in the following theorem:
* To be precise, Θ k is the DOA of user k . However, since Θ k is uniquely determined by θ k , we simply
refer to θ k as the DOA of user k in this section.
† The DOAs must be estimated based only on H k ( t ) values to ensure that the channel estimation
algorithm of the receivers and the feedback overhead remain the same as in conventional oppor-
tunistic beamforming.
‡ Note that ML estimators can be asymptotically considered as minimum variance unbiased esti-
mators [23].