Observer- based Adaptive Fuzzy Control for Nonlinear Systems with
Unknown Dead-zone Input
Tantan Yang, Bing Chen and Chong Lin
Institute of Complexity Sciences, Qingdao University, Qingdao 266071, China
E-mail: 326707812@qq.com
Abstract: This paper focuses on adaptive fuzzy output-feedback control for a class of single-input single-output(SISO)
strict feedback systems with dead zone input. In this research, fuzzy logic systems are utilized to approximate the un-
known nonlinear functions, and a state observer is constructed to estimate the unmeasurable state variables. Then adaptive
fuzzy output feedback is developed via bacstepping recursive technology. The proposed control strategy guarantees that
all the signals in the closed-loop system are semi-globally uniformly ultimately bounded (SUUB) by Lyapunov method
and simulation results.
Key Words: Dead zone, Fuzzy logic systems, State observer, SUUB
1 Introduction
In the past decade, adaptive fuzzy or neural network con-
trol has been paid a lot of attentions. Among them, direct
or indirect adaptive fuzzy control methods for SISO sys-
tems were proposed in [1]-[2] and extended to uncertain
multiple-input multiple-output (MIMO) systems in [3]. Es-
pecially, based on the traditional backstepping design tech-
nique, adaptive fuzzy or neural network has been devel-
oped further in [4]. The main contributions of these adap-
tive methods are that they can deal with nonlinear systems
which can not satisfy the matching condition or it is not
essential to have a linearization to the parameter for the un-
known nonlinear functions.
As is known to us, non-smooth nonlinear characteristics
such as dead zone, saturation and hysteresis are common
in industrial process and they have a bad influence on the
systems performance, thus many scholars have lots of re-
search on it in [5]-[9]. Among them, dead-zone inverses
are constructed to minimize the effect of dead zone in [7].
However, the method is only suitable for linear systems and
nonlinear systems in which nonlinear function is known or
it can be linearized. In addition, Lyapunov or integral-type
Lyapunov functions are used to prove the stability of the
closed-loop systems for the nonlinear systems with dead
zone and without satisfying the matching condition in [8]-
[9]. The main limitation in [8]-[9] is that the states of the
systems are required to be measurable. An adaptive fuzzy
output feedback control scheme is proposed in [10]-[11]
for nonlinear systems with immeasurable states. Unlike the
works in [10]-[11], a linear matrix inequality condition, is
proposed for the stability analysis of the observation er-
ror dynamics, and it makes the proposed control strategy is
easier to be implemented in practice. Based on this reason,
by combining Lyapunov method, backstepping design and
adaptive fuzzy control, the control design of strict feedback
systems with dead zone is considered in this paper and the
stability of the system is guaranteed.
2 Preliminaries
Consider the following SISO nonlinear strict-feedback sys-
tems with dead zone:
˙𝑥
𝑖
= 𝑥
𝑖+1
+ 𝑓
𝑖
(¯𝑥
𝑖
) ,𝑖 =1,...,𝑛− 1
˙𝑥
𝑛
= 𝑢 + 𝑓
𝑛
(¯𝑥
𝑛
)
𝑦 = 𝑥
1
(1)
where ¯𝑥
𝑖
=[𝑥
1
,𝑥
2
,...,𝑥
𝑖
]
𝑇
∈ 𝑅
𝑛
denotes the state vec-
tor, 𝑦 ∈ 𝑅 is the output, 𝑓
𝑖
(.),𝑖 =1, ..., 𝑛 is the unknown
smooth nonlinear function and 𝑓
𝑖
(0) = 0. And only output
variable 𝑦 can be measured directly.
𝑢
𝑖
is nonsymmetric dead-zone output which is defined in
the following form:
𝑢 = 𝐷(𝑣)=
𝑚
𝑟
(𝑣 − 𝑏
𝑟
),𝑣≥ 𝑏
𝑟
0,𝑏
𝑙
<𝑣<𝑏
𝑟
𝑚
𝑙
(𝑣 + 𝑏
𝑙
),𝑣≤ 𝑏
𝑙
(2)
Here, 𝑣 denotes the input of the dead zone. It is supposed
that 𝑚
𝑟
,𝑚
𝑙
,𝑏
𝑟
,𝑏
𝑙
are all unknown positive constants.
To facilitate the control design, we need the following as-
sumptions and lemma.
Assumption 1 There exists a large positive constant 𝑀
such that ∣𝑣∣≤𝑀.
Assumption 2 For 𝑓
𝑖
(.), there exist known constants ℎ
¯
𝑖𝑗
and
¯
ℎ
𝑖𝑗
such that for 1 ≤ 𝑖, 𝑗 ≤ 𝑛, ℎ
¯
𝑖𝑗
≤
∂𝑓
𝑖
∂𝑥
𝑗
≤
¯
ℎ
𝑖𝑗
.
Lemma 1: For 𝑧
𝑖
=ˆ𝑥
𝑖
− 𝛼
𝑖−1
(𝑖 =1, 2,...,𝑛), ∥ˆ𝑥∥≤
𝑛
𝑖=1
∣ 𝑧
𝑖
∣ 𝜙
𝑖
(
ˆ
𝜃
𝑖
), with 𝜙
𝑖
=(1+𝑘
𝑖
)+
1
𝑎
2
𝑖
ˆ
𝜃
𝑖
𝑆
𝑇
𝑖
𝑆
𝑖
.
Thus, the dead zone (2) can be rewritten as:
𝑢 = 𝑚(𝑡)𝑣(𝑡)+𝑔(𝑡) (3)
1698
978-1-4673-9714-8/16/$31.00
c
2016 IEEE