Stability Analysis of Weighted Multiple Model
Adaptive Control of a Linear Time-Invariant
Discrete-Time System
Weicun Zhang
1
, Qing Li
1∗
, Xiong Luo
2
, Tao Feng
1
1. School of Automation Science and Electrical Engineering, University of Science and Technology Beijing
2. School of Computer and Communication Engineering, University of Science and Technology Beijing
(weicunzhang@ustb.edu.cn; liqing@ies.ustb.edu.cn; xluo@ustb.edu.cn; snowdancing2008@gmail.com)
Abstract—The paper is concerned with the long-standing
problem of the stability of weighted multiple model adaptive
control (MMAC). Based on virtual equivalent system (VES)
concept and methodology, a positive answer to the problem is
presented under a unified framework which is independent of
specific ’local’ control strategy and specific algorithm for weight
calculation. The obtained results indicate that the stability of
a weighted MMAC system depends mainly on the convergence
of posterior probability weights, i.e., the posterior probability
corresponding to the model closest to the true plant converges to
1, and others converge to 0.
Index Terms—Stability, weighted multiple model adaptive
control, Virtual equivalent system
I. INTRODUCTION
Weighted multiple model adaptive estimation and control
appeared in initial 70’s, where multiple Kalman filter (KF)
based models were studied to improve the accuracy of the
state estimate in estimation and control problems [1-3]. This
was followed in later years by several practical applications
[4-7]. This kind of multiple model adaptive control is always
produced as the probability-weighted average of elemental
controller outputs. In recent years, a new type of weighted M-
MAC, i.e., robust multiple model adaptive control (RMMAC)
architecture was put forward with convictive experiment results
[8-10], which arouses once again the enthusiasm of researchers
in the field of adaptive control. Some useful convergence
results on the probabilistic weighting have been obtained under
suitable assumptions [11-14]. Specifically, as Fekri, Athans
and Pascoal pointed out in [9] that under certain ergodicity
and stationarity assumptions, one of the posterior probabilities
will converge almost surely to unity and will ’identify’ the
model closest to the true plant, i.e. the one with smallest
Baram Proximity Measure (BPM). In spite of decades of
theoretical and experimental research, it is widely accepted that
the stability of weighted MMAC system is difficult to prove.
Actually, to the best of the authors’ knowledge, no stability
results were reported up to now, regarding any type of weighted
MMAC.
This work was supported by Major State Basic Research Development
Program (973 Program) (No. 2012CB821200); National Nature Science Foun-
dation of China (No. 61174103); Fundamental Research Funds for the Central
Universities (No. FRF-TP-11-002B).
To address this issue, VES concept and methodology is
adopted in this paper to give a positive answer. VES is an
artificial system that is equivalent to the original adaptive
control system in the input-output sense. It was originated
from [15] and gradually adapted later in [16-18]. VES has
been successful in understanding and judgment of the stabil-
ity and convergence of a general self-tuning control (STC)
system, which consists of arbitrary control strategy, arbitrary
parameter estimation algorithm, and a deterministic/stochastic
linear time-invariant (LTI) plant. More than that, the necessary
conditions required for global stability and convergence of
a general STC system are relaxed, i.e., the convergence of
parameter estimates is removed for both deterministic and
stochastic STC schemes, for more details, the reader is referred
to [18].
As we will see later, VES concept is also suitable for
the stability analysis of a weighted MMAC system. Roughly
speaking, the stability of weighted MMAC system can be
guaranteed if only one of the posterior probabilities will
converge almost surely to unity and will ’identify’ the model
closest to the true plant.
Although, this paper is motivated by [9], the proposed anal-
ysis method is not limited to RMMAC architecture. Besides,
we focus on the discussion of weighted MMAC for discrete-
time systems because most practical systems are controlled by
computers which are discrete in nature. As Narendra pointed
out in [19], the presence of random noise can be dealt with
more easily in the case of discrete-time systems. Since most
practical systems have to operate in the presence of noise,
the stability and performance of MMAC systems in such
contexts has to be well understood, if the theory is to find
wide application in practice.
The rest of the paper is organized as follows. Section 2
gives the description of weighted MMAC. Section 3 introduces
two kinds of VESs of a general weighted MMAC system in
two different situations. The main results are then developed
in Section 4. Finally, some conclusions and future works are
drawn in Section 5.
202
978-1-4577-2119-9/12/$26.00
c
2011 IEEE