Adaptive AR Channel Model Identification of
Time-varying Communication Systems
Zarko B.Krusevac
Australian National University
Canberra, ACT 2052 Australia
zarko.krusevac@rsise.anu.edu.au
Abstract This paper implements an adaptive identification
of autoregressive AR model coefficients for model-based filter-
ing over time-varying communication channels. The presented
approach does not require a-priori knowledge of the dynamics
of the system which overcomes the issue of determining model
coefficients that capture the dynamics of unknown time-varying
channels. Simulation MSE performance analysis in a multiuser
environment shows superior experimental performance of the
AR(2) model-based adaptive algorithm with adaptive model
identification, comparing to the AR(1) model-based adaptive
algorithm with adaptive model identification, the same algorithm
with fixed model coefficients and standard observation-only-
based LMS and RLS adaptive algorithms.
I. I NTRODUCTION
Background and Motivation - Model-based adaptive al-
gorithms exploit additional information of the propagation
conditions, in the form of a predetermined model, to improve
tracking performance over time-varying channels [1]. The dy-
namics of the channel time variations are generally described
in state equation form, in addition to the observation equation
[2], [3].
If the model captures the dynamics of the propagation
conditions, the model-based adaptive algorithms can provide
very good performance in non-stationary environments [4].
A stochastic modeling of time-varying channels helps to
overcome the complexity of the real propagation environment,
provides a focus on the critical broad attributes and tends to
average out nuisance parameters. Finite AR models [5], [6]
are often considered as proper stochastic description of the
channel dynamic in the form of linear difference (differential)
state equations.
The model-based approach to adaptive filtering usually
lends itself to a Kalman filter formulation or the problem
of jointly solving the process and measurement equations
[3]. However, the direct Kalman problem formulation for
the adaptive equalization over an unknown channel with ISI
requires information of the channel [7]. This can be achieved
by using observation-only based adaptive algorithms such as
LMS or RLS to provide the required channel estimate [8],
[9]. In [8] this technique is extended to adaptive equalization
over non-stationary communication channels. Even though the
model-based Kalman filtering theory is used and the system
is adaptive, the whole approach has the crucial drawback
of not recognizing the dynamics of time-varying channels,
978-1-4244-2204-3/08/$25.00 2008 IEEE
Predrag B. Rapajic
University of Greenwich at Med. Pemb.
KENT ME4 4TB United Kingdom
P.Rapajic@greenwich.ac.uk
neither by Kalman state (process) equation nor by observation-
only adaptive algorithms used for the channel estimation [4].
However, capturing the dynamics of channel time variations
by the Kalman process (state) equation is essential for the
overall performance of the system [4].
Since the channel output observation equation is linear in
the state process, assuming a known input sequence (training
sequence during the training period or sequence of tentative
decisions in decision directed mode), the Kalman filtering
problem can be directly formulated for channel process estima-
tion of an unknown time-varying channel [6], [10], [11]. Given
that the Kalman process equation is the channel state equation,
the dynamics of the channel time variations are explicitly
incorporated into the structure of the algorithm. However
this approach assumes separate explicit channel estimation
followed by equalization and data detection, which may not be
optimal in terms of information rate in the presence of channel
noise [12].
A specific approach to the Kalman equalization, originated
from [7] and named the Godard-Kalman approach [9], uses
the state vector to be the true (optimal) tap weights of
the equalizer. Optimal tap weights are calculated to minimize
the expected mean squared distortion of the filter output [7].
Since time-invariant channels are considered in [7], the optimal
tap weights are constant in time and the state transition
matrix is the identity matrix. However, if the channel is
time-varying then the optimal tap weights are time-varying
as well. Therefore, the Godard-Kalman approach can be
extended to time-varying channels by defining properly the
optimal tap weights transition matrix to implicitly capture
the dynamic of the time-varying channel process [4]. This
approach enables adaptive model-based (Kalman) equalization
without prior separate channel estimation. However, the issue
of determining model coefficients to capture the dynamics
of the optimal weights time-variations still remains. In this
paper, an adaptive identification of AR model coefficients is
performed to overcome this issue.
This paper extends and generalizes the Godard-Kalman
model based adaptive approach [7] to adaptive filtering over
time-varying communication channels. The extended Godard-
Kalman approach presented here utilizes RLS-based adaptive
identification of AR model coefficients. The adaptive model
identification, for a given model order, enables adaptive model
618
based equalization without any a-priori information of fading
conditions. Simulation analysis in a multiuser environment
shows the superior experimental MSE performance of Godard-
Kalman algorithm with adaptive model identification com-
pared to one with fixed model coefficients and to standard
observation-only-based LMS and RLS algorithms.
II. A DAPTIVE M ODEL - BASED (K ALMAN ) F ILTERING
T HEORY
A. Model-based Design
The main idea behind the model-based adaptive algorithm
design is to use the information of the propagation conditions
to stochastically describe the time-varying channel process
using the channel state equation [2][4]. The observation equa-
tion and the state equation define a state-space formulation
of a communication system over time-varying communication
channels. A discrete time, state-space model can be expressed
as follows
y n = Sn xn + nn
Sn+1 = Hn Sn + Qn
where y n is the observable output of the channel, Sn is the
channel process matrix, xn is the input of the channel, nn is
observation noise and Qn is channel process noise matrix, all
at the time instant n. Furthermore, Hn is the channel transition
(system) matrix.
The observation (measurement) equation (1a) describes the
observation vector and relates the observable output of the
channel y n to the channel process matrix Sn . The channel
process matrix Sn is defined as the minimal set of data
that is sufficient to uniquely describe the unforced dynamical
behavior of the channel by the state (process) equation (1b).
B. Statement of the Kalman Filtering Problem
III. G ODARD -K ALMAN A DAPTIVE A LGORITHMS FOR
T IME - VARYING C HANNELS
A. Godard-Kalman approach
A direct Kalman formulation for the adaptive filtering
for the communication system (1) (solution in xn ) over an
unknown time-varying channel with ISI is not possible due to
non linearity of the observation equation (1a) in channel input
xn . The straightforward solutions requires information of the
channel (channel estimation).
Alternatively, the Godard-Kalman approach [7] to the
Kalman equalization [9] uses the state vector to be the true
(optimal) tap weights of the equalizer. Optimal tap weights are
calculated to minimize the expected mean squared distortion
E[|e(n)|2 ] of the filter output x
n [7], where
en = xn x
n , (2)
and
n = H
x n yn . (3)
Vector n in (3) is the vector of the adaptive filter coefficients,
(1a)
y n is the observable output of the communication system/the
(1b) filter input sequence and xn in (2) is the training sequence
(or sequence of tentative decisions), a priori known to the
receiver.
We use opt
n to denote the optimal (true) filter coeffi-
cients, which minimize the E[|en |2 ] of the error (2). While in
time-invariant environments the optimal coefficients opt n =
opt are constant in time [7], if the channel is time-varying,
then opt
n becomes time-varying as well.
The main reason for adopting the Godard-Kalman algorithm
for adaptive model-based equalization over time-varying chan-
nels is the possibility to use the information of the propagation
conditions to describe how true filter coefficients optn vary
with time. Here we assume vector AR stochastic model for
the time-varying true coefficients opt
n

The Kalman filtering problem is the problem of jointly opt opt

n+1 = H n n + q n (4)
solving the state (process) and observation (measurement) where Hn is a system (transition) matrix, q n is a white noise
equations for the unknown state matrix in an optimal manner sequence with var(q) = q2 I.
[2], [3]. The filter output observation equation is given by
The Kalman filtering problem may be formally stated as
follows [3]: Use the entire observed data y 1 ,..., y n to find, for
each n 1, the minimum mean-square estimate of the state at n = H
x n y n + n (5)
the time i. The problem is called filtering if i = n, prediction where n = xn ( opt H
n ) yn is the true noise with the
if i > n and smoothing if 1 i < n. variance 2 .
The concept of state is fundamental for the Kalman filtering The observation equation (5) and the state equation (4) form
theory formulation [3] and it is not necessarily related to the the state-space model for the Kalman filtering problem state-
channel process Sn in (1). For instance, the Godard-Kalman ment. The iterative solution can be derived in the following
formulation [7] defines the state as the vector of optimal filter form [4]
coefficients which are calculated to minimize the expected n+1|n = Hn n+1|n + Kn n
mean squared distortion of the filter output. Furthermore, the
use of Kalman filtering presupposes that the system under n = xn y n n|n1
consideration is described by a set of linear difference (differ- Hn Pn|n1 y Tn
Kn = 2 (6)
ential) equations [7]. However, the observation equation (1a)  + y n Pn|n1 y n+1
is not necessarily linear in channel input xn . Thus, the state- 
Pn+1|n = Hn Pn|n1 HnT
space formulation (1) is more general than the Kalman filtering 
formulation. Kn (2 + y n Pn|n1 y Tn )KnT + q2 I

619
where where 0 = 1 is the forgetting factor. The time constant
for convergence is given by [2], [4]
n+1|n = E( n+1 |yn , yn1 , ..., y1 ) (7)
1
and = . (11)
log |1 0 |
Pn+1|n = E( n+1 n+1|n )( n+1 n+1|n )T (8)
The convergence criterion (10) and the time constant for
with initial conditions given by P (0) = P0 . convergence (11) do not depend on filter input y(n) statistics.
The main advantage of the Godard-Kalman adaptive ap- Consequently the RLS is near-far resistant. As a Newton
proach is adaptive filtering (equalization) without prior sepa- method algorithm, the RLS provides a fast initial convergence.
rate channel estimation. However, in non-stationary environments, this fast initial con-
vergence usually has to be traded-off to improve stability and
B. Adaptive Model Identification for the Kalman-Godard Fil- tracking, as it is shown in simulations, Section V.
ter
2) Second Order Stability and Tracking Analysis: The
While the Godard-Kalman adaptive approach implicitly relative steady state excess MSE or misadjustment for the RLS
captures the dynamics of time-varying channels, the issue of algorithm with time-varying true weights is [2], [4]:
determining model coefficients still remains.
E 2 p 0 2 2 y2 p
Mtot = = + + O(20 )
yk (n) xk (n)
2 2 0 2 2
Adaptive Decision
receiver device = Mnoise + Mlag + O(20 ) (12)

where E = limn E[|e(n)|2 ]; p = tr(Ry )/y2 ; y2 is

the variance of y(n); tr(Ry ) = trace(Ry ) is the sum of
AR(p) the eigenvalues of Ry = E[y(n)y H (n)]; 2 = q2 / 2 ; and
model-based
adaptive = I H is true speed of change of the time varying
algorithm true weights. The expression Mnoise in (12) is the excess
Adaptive AR(p) xk (n) MSE due to noise adaptation and Mlag is the excess lag
model coefficients Training
identification sequence terms due to the time-varying optimal weights.
The second order stability requires [2], [4]
Fig. 1. AR(p) model-based filtering with adaptive AR(p) coefficients 1
Mtot = Mnoise + Mlag < (13)
identification 3
In Fig. 1 we propose the Godard-Kalman filtering with Furthermore, 0,opt and Mopt can be found to be [2], [4]
the adaptive model coefficient Hn identification. Due to the y y
physical foundation of the channel process time-variations and Mopt = p for 0,opt = (14)
 
mass inertia effects, time variations of the system matrix Hn
are usually slow enough (at the time-scale of the higher rate B. Model-based Algorithm Performance Analysis
data). As shown in [4], the RLS algorithm is well suited 1) Convergence: Following [2], [4], the first order conver-
to adaptively estimate only very slow time-invariant system gence criterion is given by
matrix Hn , Fig. 1, as follows 1
0 < max
2
<2 (15)
Hn+1 = Hn + Kn n
n = xn (Hn n1 )H y n (9) where = 2 /2 and max = max (Ry ) is the maximum
eigenvalue of Ry .
1 Pn1 y n
Kn = The time constant for convergence for convergence is given
1 + 1 y H
n Pn1 y n by
Pn = Pn1 Kn y H
1 1
n Pn1 1
=  12 1  (16)
We emphasise that the RLS algorithm is ill-suited to adapting log 1
2 +1
(time-varying) opt
n , but through the state-space formulation
well-suited to adapting the very slow time varying system where = max /min is condition number of Ry and min is
matrix Hn . the minimum eigenvalue of Ry . Time constant for convergence
(16) depends on eigenvalue spread but much more weakly
IV. A DAPTIVE A LGORITHMS P ERFORMANCE A NALYSIS
than for LMS. For example, for for conditional number =
A. RLS Performance Analysis max /min = 16, the LMS time constant is equal to 8 [4] and
1) Convergence: The first order RLS convergence criterion the model-based algorithm time constant (16) is 3. Thus, the
is given by [2], [4] model-based adaptive algorithm is much less near-far sensitive
0 < 0 < 2 (10) than LMS.

620
 2) Second Order Stability and Tracking Analysis: Again
Adaptive MMSE receivers bank
following [2], [4], the relative steady state excess MSE or
misadjustment for the model-based adaptive algorithm with
Adaptive
time-varying true weights is receiver 1

E 2 p cy p cy 2
Mtot = 2
= + + O(2 )
 2 2 2 Adaptive Decision
y
= Mnoise + Mlag + O(2 ) (17) receiver 2
device
1
where cy = tr(Ry )/p and = 2 /2
2

Furthermore, k,opt and Mopt can be found

Adaptive
receiver K
Mopt = p cy for k,opt = . (18)
  Decision
directed
x sequence
The second order stability requires Mtot = Mnoise +Mlag <
1 AR(p) model-based algorithm
3 [2]. In addition
with
 Training
p p adaptive AR(p) coefficients estimation
x sequence
1 1 1  1 1  1 1 2 2
cy = tr(Ry )=
2
u
u2 p=
2
p x =x (19)
p p 1 p 1 p
Fig. 2. AR(p) model-based adaptive multiuser receiver with adaptive AR(p)
model coefficients estimation
leads to Mopt (Kalman) Mopt (RLS), (and
Mopt (Kalman) Mopt (LMS) [4]) and which means
that model-based adaptive algorithm offers superior tracking where y k,n is discrete time received samples of kth user at
to RLS [4]. the time n, k,n are the adaptive filter coefficients of kth user
at the time n.
V. S IMULATION S TUDY The coefficients are obtained adaptively during the train-
A. System Model for Simulation Study Adaptive MMSE Mul- ing period, by minimizing the MSE, E[|ek,n |2 ],
tiuser Receiver
ek,n = xk,n x
k,n , (22)
The adaptive MMSE multiuser receiver structure Fig. 2,
presented in [13], [14], consists of a bank of adaptive MMSE where xk,n is the kth user training sequence at the time n, a
FIR filters along with the centralized (ML) detector part of the priori known to the receiver.
receiver for data detection. The number of users in the cell of B. Simulation results
interest K is assumed to be known. However, the number
KI of intercell interferers (I1 , I2 , ...IKI in Fig. 2) is unknown
[14]. h=[1 0.25 0.5 0.125] RLS
While the non-linear centralized ML detection improves 2 LMS
n=0.1;
significantly the performance of the adaptive MMSE multiuser
receiver [14], it does not influence the tracking performance Number of users:6
1
of the adaptive filters during the training period (the training 10
Experimental MSE

Adaptive ident.
sequence is a priori known at the receiver). Since we focus on AR(2) GK
the tracking performance of adaptive algorithms assuming an
a priori known input sequence, without loss of generality, we
consider only the bank of adaptive MMSE filters (the linear 2
10
front-end of the receiver).
The discrete-time received sample vector or the observation
of adaptive filter inputs at time n, y n can be expressed as
follows
6 5 4

y n = Fn xn + nn
10 10 10
(20) Normalized fading rate fDTs

where Fn is the matrix of sampled system signatures which Fig. 3. Experimental MSE performance for AR(2) model-based algorithm
includes spreading sequences and channel signatures, nn is with adaptive model coefficients identification and observation-only based
AWGN and xn contains the transmitted symbols. LMS and RLS algorithms
The output of the MMSE filter for the kth user, k = 1, ..., K We use the MATLAB multiple path fading channel
at the time n is rayleighchan.m, 6-users, spreading gain of 6. The multi-
k,n = H
x k,n y k,n (21) path channel consists of four-spaced channel taps with initial

621
values h = [1 0.25 0.5 0.125]. The channel noise variance
is given by n2 = 0.25. The feedforward filter for each user
consists of 24 taps. User k = 1 is the user of interest.
Simulations are performed for a range of normalized fading
rates fD TS .
Fig. 3 based on the average square error performance (ex-
perimental MSE) corroborates the theoretical performance
analysis provided in Section IV. The AR(2) model-based
adaptive algorithm with adaptive model identification provides
superior tracking performance to LMS and RLS, especially for
the medium speed fading conditions and the LMS algorithm
outperforms the RLS which provides very limited tracking
performance at medium and high-medium speed fading con-
ditions. On the other hand, if the channel is very slow (quasi
time-invariant) with fD TS 0, then the superior initial
convergence speed of RLS provides a performance advantage
compared to LMS, given the reduced tracking requirements.
Even more, overall performance of the RLS approaches that of
the model-based algorithm, since the model-based algorithm
structure converges to the RLS algorithm.
data1nonadaptive iden. AR(1) GK
data2adaptive iden. AR(1) GK
data3adaptive iden. AR(2) GK
1 2
10
Experimental MSE
h=[1 0.25 0.5 0.125]
2 n=0.1;
10
6 5
10 10
Normalized fading rate f T
D S
Fig. 4. Experimental MSE performance for AR(2) and AR(1) model-based
algorithms with adaptive model coefficients identification and AR(1) model-
based algorithm with fixed model coefficients H = I, = 0.85
Fig. 4 based on the experimental MSE shows that the
AR(2) model-based approach with adaptive model identifi-
cation achieves notable performance improvement over the
AR(1) model-based approach with adaptive model identifi-
cation, especially at the fast fading rate. Furthermore, Fig. 4
confirms the superiority of the model-based approach with
adaptive model identification over the same approach with
fixed model parameters for a range of fD TS . At medium
fading rate fD TS = 5 104 , the adaptive approach with
a fixed model parameters H1 = I; = 0.85 provides
some performance improvement comparing to slow and high-
medium fading rate conditions, since the model is better suited
for particular fading conditions. However, the model-based
approach with adaptive model identification does not use any
a-priori information of the fading conditions.
R EFERENCES
[1] P. Newson and B.Mulgrew, Model based adaptive channel identifica-
tion algorithms for digital mobile radio, in Proc. IEEE International
Conference on Communications (ICC), vol. 3, May 1994, pp. 1531
1535.
[2] V.Solo and X. Kong, Adaptive Signal Processing Algorithms - Stability
and Performance, 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1995.
[3] S. Haykin, Adaptive Filter Theory, 4th ed. Upper Saddle River: Prentice
Hall, 2002.
[4] Z. B. Krusevac, P. B. Rapajic, and R. A. Kennedy, Model-based
adaptive algorithms for time-varying communication channels with
application to adaptive multiuser detection, in Proc. IEEE Vehicular
Technology Conference 2006-spring (VTC-spring06), Melbourne, VIC,
Australia, May 2006.
[5] K. E. Baddour and N. C. Beaulieu, Autoregressive modeling for fading
channel simulation, IEEE Transactions on Wireless Communications,
1
vol. 4, no. 4, pp. 16501662, July 2005.
[6] M.McGuire and M.Sima, Low-order Kalman filters for channel esti-
mation, in Proc. IEEE Pacific Rim Conference on Communications,
Computers and Signal Processing,(PACRIM), Aug. 2005, pp. 352355.
3
[7] D. Godard, Channel equalization using a Kalman filter for fast data
transimision, IBM J.Res.Dev., pp. 267273, 1974.
[8] S. McLaughlin, B. Mulgrew, and C. F. N. Cowan, A novel adaptive
equaliser for nonstationary communication channels, in Proc. IEEE
International Conference on Acoustics, Speech and Signal Processing,
(ICASSP), vol. 2, May 1989, pp. 944 947.
[9] S. McLaughlin, Adaptive equalization via Kalman filtering techniques,
2 IEE Proc.-F, vol. 138, no. 4, pp. 388396, Aug. 1991.
[10] C. Komninakis, C. Fragouli, A. Sayed, and R. Wesel, Channel estima-
Number of users:6 tion and equalization in fading, in Thirty-Third Asilomar Conference on
Signals, Systems, and Computers, Pacific Grove, CA, 1999, pp. 1159
4
10 1163.
[11] S. Haykin, A.H.Sayed, J. R.Zeidler, P.Yee, and P.C.Wei, Adaptive
tracking of linear time-variant systems by extended RLS algorithms,
IEEE Transactions on Signal Processing, vol. 45, no. 5, May 1997.
[12] Z. B. Krusevac, P. B. Rapajic, and R. A. Kennedy, Optimal implicit
channel state estimation for finite-state Markov communication chan-
nels, in Proc. IEEE Int. Symp. on Inform. Theory (ISIT06), Seattle,
Washington, USA, July 2006, pp. 26572661.
[13] P. B. Rapajic and B. Vucetic, Adaptive receiver structures for asyn-
chronous cdma systems, IEEE Journal on Selected Areas in Commu-
nications, vol. 12, no. 4, pp. 685697, May 1994.
[14] P. B. Rapajic and D. K. Borah, Adaptive MMSE maximum likelihood
CDMA multiuser detection, IEEE Journal on Selected Areas in Com-
munications, vol. 17, no. 12, Dec. 1999.
622