Sunteți pe pagina 1din 22

Available online at www.sciencedirect.

com

Journal of the Franklin Institute 356 (2019) 2668–2689


www.elsevier.com/locate/jfranklin

Novel integral sliding mode control for small-scale


unmanned helicopters
Tao Jiang∗, Defu Lin, Tao Song
Beijing Key Laboratory of UAV Autonomous Control, Beijing Institute of Technology, Haidian District, Beijing
100081, People’s Republic of China
Received 22 July 2017; received in revised form 14 December 2018; accepted 24 January 2019
Available online 1 February 2019

Abstract
Integral sliding mode (ISM) control which consists of a nominal control and a sliding-mode motion
control, provides a nice framework for high tracking performance and good disturbance reduction. Our
work develops ISM to attenuate the adverse effect of mismatched perturbations. By properly choosing
sliding-manifold surface, the elimination of disturbances on control outputs enables to be achieved.
Additionally, the chattering of sliding-mode control part is attenuated based on second-order sliding
mode idea. Then, the proposed novel ISM control scheme is applied to address trajectory tracking
problem for helicopters under perturbations. Approximated input-output linearization is implemented,
such that the obtained linearized model is suitable for applying the proposed ism control. The stability
of the closed-loop system for helicopter and its convergence to zeros of tracking errors are demonstrated
by Lyapunov theory analysis. Several comparison simulations illustrate the effectiveness and superiority
of the proposed methods.
© 2019 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Unmanned helicopters have received wide interest over the past decades due to their ca-
pabilities, such as vertical taking off and landing from unprepared sites, broad envelope of

∗ Corresponding author.
E-mail addresses: jiangtao_1992@outlook.com (T. Jiang), lindf@bit.edu.cn (D. Lin), 10901034@bit.edu.cn
(T. Song).

https://doi.org/10.1016/j.jfranklin.2019.01.035
0016-0032/© 2019 The Franklin Institute. Published by Elsevier Ltd. All rights reserved.
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2669

flight that ranges from hovering to cruising, potential to fly at low altitudes, and highly agile
maneuvering in tightly constrained environments [1,2]. Unmanned helicopters have been
widely applied in defense, rescue, surveillance, supervision, and other fields. However, con-
trolling unmanned helicopters, especially small-scale ones, which are more sensitive to ma-
nipulation and external disturbances, is difficult. Such difficulty arises from their complex
aerodynamics, strong dynamic couplings, and significant parameter and model uncertainties
[1–4]. In the field of unmanned aircraft vehicle, designing high-performance controller for
helicopters bears importance.
Traditional linear control method is based on the model linearized around the trim points,
including PID [1], linear quadratic regulation [5], H∞ control theory [4–6], and gain schedul-
ing controllers [7], that may utilize synthesis techniques. These techniques are advantageous to
project realization, whereas they suffer from performance degradation when helicopters leave
away from their designed trim points or execute aggressive maneuvers. In the last few decades,
various nonlinear control approaches were developed to tackle these limitations, which are
divided into full nonlinear and linearized methods [3]. Full nonlinear methods mainly re-
fer to backstepping (BS) based control [8] due to the hierarchical framework of helicopters’
dynamics. For guaranteeing the robustness performance, BS controllers are often combined
with other disturbance compensation method, such as integral control [9], adaptive control
[10,11], and disturbance observer (DO) [12,13], etc. Compared with full nonlinear methods,
linearized methods provide a framework to connect linear control methods with nonlinear
models. When applying linearized methods to a helicopter model, approximate input-output
feedback linearization is initially achieved through choosing position and yaw angle as out-
puts [14]. However, unmatched disturbances occur in the linearized dynamics, causing some
difficulties in implementation of exact trajectory tracking. In [15], model predictive control
(MPC) combined with DO guarantees the global asymptotic stability of tracking errors. Fang
et al. [16] developed a composite controller based on H∞ control and feedforward technique,
which can attenuate the mismatched disturbances from the state variables. A novel sliding
mode control (SMC) based on the estimation of disturbances was proposed in [17], which
provides excellent tracking performance and robustness.
As a variable structure control technique, SMC can eliminate the effect of the parameter
uncertainties and perturbations by applying a switching control law to alter the plant dynamics
[18]. In the sliding phase, the system response remains invariant for disturbances and the
nominal control performance is recovered. However, during the reaching phase, the invariance
of SMC is not guaranteed and the system response is sensitive to perturbations. Additionally,
the imperfect implementation of high-frequency switching in SMC results in chattering at
control responses, which may cause the tracking performance degradation and the loss of
actuators. Integral sliding mode (ISM) enables to eliminate the reaching phase by enforcing
the sliding mode in an entire system response so that the invariance of SMC is ensured
from the initial time instant [19–21]. Due to these advantages, ISM has been widely used
in practical applications, such as helicopters [22–26], mobile robot [27], hypersonic vehicles
[28], etc.
Conventional ISM [19–21] methods select sliding manifolds based on projection matrix
approach such that no amplification of unmatched disturbances occurs. Then other robust
techniques, such as H∞ control [20], or MPC [29,30], can be combined with ISM method to
robustify against perturbations. Unfortunately, given the existence of unmatched perturbations,
tracking performance degradations still occur. For helicopter control, the unmatched distur-
bances (the external force disturbances generated by wind gusts) have much more effect on
2670 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

the position tracking of helicopters. To tackle these limitations, the previous works [23,25]
usually combine it with the recursive design to eliminate the adverse effect of unmatched
disturbances. Our work develops another way to solve these problem through eliminating
mismatched perturbations from the output channels. By properly choosing a sliding-manifold
surface based on the relative degree idea, mismatched disturbance attenuation is achieved,
such that high tracking performance and robustness are guaranteed. A systematic method
is developed for the sliding-manifold gain design. Moreover, to attenuate the chattering of
the SMC signals, a second-order sliding mode controller [31,32] is introduced to achieve
sliding-mode motion.
Additionally, to apply ISM for flight control, the linearized model around the trim points is
considered [23,26]. Our work applies approximately feedback-linearized for a small-scale heli-
copter mathematical model subjected to external perturbations and uncertainties. The tracking
error dynamics with lumped matched and unmatched disturbances are obtained, which is suit-
able to larger flight envelop. Then, the proposed ISM method enables to be applied to design
the helicopter controller, where the sliding-manifold gain is properly chosen to eliminate the
influence of disturbances on output channels. A second-order SMC is introduced to attenuate
the chattering of control inputs. The main contributions are listed as follows:

(1) A novel form of ISM, by designing, is developed to eliminate the influence of the un-
matched perturbations. Its sliding-manifold is properly designed through relative degree
idea, such that the effect of disturbances is attenuated on the output channels. Moreover,
chattering-free SMC is implemented based on the second-order SMC to drive the state
trajectory of the closed-loop system onto the sliding surface.
(2) The trajectory tracking problem for small-scale unmanned helicopters subjected to ex-
ternal disturbances and uncertainties are tackled by applying the proposed ISM con-
trol. Approximate feedback-linearization technique is conducted to obtain the linearized
model which is suitable to applying ISM control. This composite control system builds
a feasible control framework to guarantee high tracking performance and robustness.
(3) A rigorous proof of stability of the closed-loop system is derived by stability analysis.
Some comparative simulation results of helicopter trajectory tracking have illustrated
the effectiveness and superiority of our proposed control system.

This paper is arranged as follows. Section 2 presents the dynamic model of a small-
scale helicopter, and the linearized model is obtained based on the approximated feedback
linearization. Section 3 proposes the control method and provides performance analysis. Then,
the helicopter control problem is addressed. Section 4 discusses the performed simulations.
Finally, Section 5 draws the conclusions.

2. Helicopter model

This section presents the nonlinear dynamic model of a small-scale unmanned helicopter.
The helicopter is considered a six-degree-of-freedom rigid body model with simplified force
and moment generation process. Model uncertainties and external perturbations are considered
in the modeling phase. First, two reference frames are defined as follows: earth reference frame
(ERF) = {Oxyz}, which is fixed to the earth; body reference frame (BRF) = {Ob xb yb zb }, whose
origin is located at the helicopter center of gravity [3,4]. Fig. 1 shows the direction of the
two reference frames.
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2671

Tm

Tt
b

a
O
y (east)
zb Ob
xb
x (north) yb
z (down)

Fig. 1. Simple illustration of a small-scale helicopter model.

The dynamic model of the small-scale unmanned helicopter can be described as follows
[3,4]:
P˙ = V (1)

1
V˙ = ge3 + R (!)F (2)
m

R˙ (!) = R (!)S (ω ) (3)

J ω˙ = −ω × J ω + M (4)
# #
where P = [x y z]T and V = [u v w]T refer to the helicopter’s position and velocity
vector in the ERF, respectively; m is helicopter mass, and g is gravitational acceleration;
e3 = [0 0 1]T is a unitary vector; S(·) is a skew-symmetric matrix that corresponds to
the vector (·) [8,9]; J represents the approximate inertia matrix written in the following form:
⎡ ⎤
Jxx 0 −Jxz
J=⎣ 0 Jyy 0 ⎦ (5)
−Jxz 0 Jzz
The rotation matrix from BRF to ERF is given as follows:
⎡ ⎤
Cθ Cψ Sφ Sθ Cψ − Cφ Sψ Cφ Sθ Cψ + Sφ Sψ
R (!) = ⎣Cθ Sψ Sφ Sθ Sψ + Cφ Cψ Cφ Sθ Sψ − Sφ Cψ ⎦, (6)
−Sθ Sφ Cθ Cφ Cθ
#
where C(·) and S(·) are shorts for cos(·) and sin (·), respectively. ! = [φ θ ψ ]T represents
#
the Euler angles, which include roll, pitch, and yaw angles, respectively. ω = [ p q r ]T
denotes the angular rates in the BRF. Attitude kinematics is expressed as follows:
˙ = '(!)ω
!
2672 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689
⎡ ⎤
1 Sφ Cθ Cφ Sθ /Cθ
'(!) = ⎣0 Cφ −Sφ ⎦ (7)
0 Sφ /Cθ Cφ /Cθ
In Eqs. (2) and (4), F and M are the external forces and torque exerted on fuselage in
BRF, respectively.
F = Fb + F# (8)

M = Mb + M# (9)
⎡ ⎤
0
Fb = m⎣ 0 ⎦, Mb = J (Aω + Buc )
−g + Zw w + Zcol δcol
uc = [δcol δlon δlat δ ped ]T is the control input vector, whose elements denote the collec-
tive pitch of the main rotor, longitudinal cyclic, and lateral cyclic and collective pitch of the
tail rotor, respectively. F# and M# represent the lumped force and moment disturbance in BRF,
respectively, involving external perturbations, parameter variations, and model uncertainties.
Constant matrices A and B are expressed as follows [12,15–17]:
⎡ ⎤
−τ Lb −τ La 0
A = ⎣−τ Mb −τ Ma 0⎦
0 0 Nr
⎡ ⎤
0 Llon Llat 0
B=⎣ 0 Mlon Mlat 0 ⎦,
Ncol 0 0 Nped
where the coefficients Zw , Zcol , Lb , La , Mb , Ma , Llon , Llat , Mlon , Mlat , Ncol , Nr , and Nped depend
on the helicopter structure, which can be obtained via system identification technique. Along
Eqs. (8) and (9), helicopter dynamics (2) and (4) can be rewritten as follows:
V˙ = ge3 + R (!)e3 (−g + Zw w + Zcol δcol ) + #F (10)

ω˙ = −J −1 (ω × J ω ) + (Aω + Buc ) + #M, (11)


1 −1
where #F = m
R(!)F# , and #M = J M# .
#
Given the desired smooth trajectory Pd = [xd yd zd ]T and yaw angle ψd , this work aims
to develop control signals [δcol δlon δlat δ ped ]T for helicopter model with uncertainties
and external disturbances, such that its practical trajectory P and yaw angle ψ converge
asymptotically to Pd and ψd .

2.1. Approximate feedback linearization

Feedback linearization is initially carried out to facilitate controller design and perfor-
mance analysis. In the helicopter model, exact input–output linearization fails to linearize
the complete system and results in unstable zero dynamics [14]. In this subsection, the
simplified helicopter model is developed based on approximate feedback linearization. It
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2673

requires the second-time derivative of the main rotor thrust T̈m as the new input, where
Tm = m(g − Zw w − Zcol δcol ). The approximated input–output feedback linearization procedure
is performed as follows:
First, we obtain eP1 = P − Pd . The derivative of e p1 from Eq. (1) is given as follows:
e˙ P1 = V − P˙d (12)
Next, let eP2 = V − P˙d along Eq. (10); its derivative is expressed as follows:
1
e˙ P2 = ge3 − R (!)e3 Tm + #F − P̈d (13)
m
Selecting eP3 = ge3 − m1 R(!)Tm − P̈d , the derivative of eP3 is derived from Eq. (3) as
follows:
1 1 ...
e˙ P3 = − R (!)S (ω )e3 Tm − R (!)e3 T˙m − P d (14)
m m
Transition control variables are defined as follows: [Mφ Mθ Mψ ]T = −J −1 (ω × J ω ) +
...
(Aω + Buc ). Let eP4 = − m1 R(!)S(ω)e3 Tm − m1 R(!)e3 T˙m − P d along Eq. (11); its time
derivative is obtained as follows:
e˙ P4 = fPe (t ) + uP + #M p, (15)
where
1 2
fPe (t ) = − R (!)S 2 (ω )e3 Tm − R (!)S (ω )e3 T˙m − Pd(4 )
m ⎡ ⎤ m ⎡ ⎤
Tm Mθ Tm #M (2 )
1 1
uP = − R (!)⎣−Tm Mφ ⎦, #M p = − R (!)⎣−Tm #M (1 )⎦
m T̈m m 0

The position error subsystem is approximately feedback-linearized by applying T̈M as input.


The dynamics of yaw angle errors eψ1 = ψ − ψd is described from Eq. (7) as follows:
Sφ Cφ
e˙ ψ1 = q+ r − ψ˙ d (16)
Cθ Cθ
Sφ Cφ
Selecting eψ2 = Cθ
q + Cθ
r − ψ˙ d , its derivative is yielded along Eq. (11) as follows:
e˙ ψ2 = fψe (t ) + uψ + #Mψ , (17)
where
% &
˙ + Sφ Mθ Cθ − Sφ φr
Cφ φq ˙ Cθ + Sφ qSθ θ˙ + +Cφ r Sθ θ˙
fψe (t ) = − ψ̈d
Cθ2

uψ = Mψ

Sφ Cφ
#Mψ = #M (2 ) + #M (3 )
Cθ Cθ
Combining Eqs. (12)–(17), the complete input-output feedback linearization of the heli-
copter dynamics is given as follows:
e˙ P1 = eP2
2674 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

e˙ P2 = eP3 + #F
e˙ P3 = eP4
e˙ P4 = fPe (t ) + uP + #M p
e˙ ψ1 = eψ2
e˙ ψ2 = fψe (t ) + uψ + #Mψ , (18)

where uP and uψ are considered new inputs for the dynamics system (18). Our objective is
changed to design the input [uPT uψ ]T to ensure that the trajectory and yaw-angle errors
will converge to zero.
Remark 1. In Eq. (18), #F , #M p, and #Mψ are lumped disturbances that act on the he-
licopter model. #M p and #Mψ are matched disturbances, whereas #F is the unmatched
disturbance. The unmatched disturbance #F significantly affects tracking precision. Given
the existence of unmatched disturbances, various techniques, such as the classic SMC and
adaptive control, are restricted.

3. Controller design

The proposed controller is attained based on ISM technique, which allows a 2-degrees-of-
freedom design. Through the sliding-mode motion control, disturbance effects can be com-
pletely counteracted in the output. Then, the nominal controller is employed to guarantee the
ideal performance of the closed-loop system [20,21]. The section is organized as follows.
First, the design process and corresponding performance analysis are presented. Then, the
implementation of this method, which aims to solve the problem of unmanned helicopter
trajectory tracking, is shown.

3.1. Novel ISM

The traditional ISM design only compensates the matched perturbations, but the unmatched
one is replaced by another equivalent disturbance [20,21]. Through selecting the suitable slid-
ing manifolds, the equivalent disturbance equals the unmatched one, that is, no amplification
of unmatched disturbance exists. In this subsection, a novel ISM is proposed for systems
with matched and unmatched perturbations. Sliding manifolds are designed to eliminate dis-
turbances in output channels.
A simplified perturbed linear system is considered as follows:
x˙ (t ) = Ax (t ) + Bu (t ) + Bd d (t )
, (19)
y (t ) = Cx (t )
where x(t ) ∈ Rn is the state; u(t ) ∈ Rm denotes the input; d (t ) ∈ Rl expresses the unknown
disturbance vector due to model uncertainties or external disturbances. y(t ) ∈ Rm represents
the output; A, B, Bd , and C are system matrices with the appropriate dimensions. The fol-
lowing assumptions are considered:
Assumption 1. The pair (A, B ) is controllable, and matrices B and Bd are of full ranks, i.e.,
rank(B) = m, rank(Bd ) = l.
Assumption 2. Full state feedback is available for system (19).
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2675

A control law of the form is considered as follows:


u (t ) = u0 (t ) + u1 (t ), (20)
where u0 (t ) = K x is the nominal control, which is responsible for the performance of the
nominal system (i.e., without any perturbations); u1 (t ) denotes the control signal for conver-
gence of ISM. ISM surface is designed as follows:
' ( t )
s (x, t ) = G x (t ) − x (0 ) − (Ax (t ) + Bu0 (t ))dt , (21)
0
m×n
where G ∈ R is the constant matrix, which is designed to eliminate disturbance Bd d (t )
in the output channels. In this condition, the matrix (GB ) should be invertible. Note that at
t = 0, s(x(0), 0 ) = 0. Thus, the system consistently starts at the sliding mode.
Along Eq. (19), the derivative of the sliding mode surface is expressed as follows:

s˙(x, t ) = G[(Ax (t ) + Bu (t ) + Bd d (t )) − (Ax (t ) + Bu0 (t ))]


= GBu1 (t ) + GBd d (t ) (22)

Assumption 3. Disturbances d (t ) satisfy the condition that their time derivatives are bound
by known constant values, i.e., d˙(t ) ≤ Cd , where Cd > 0 is the known constant vector.
To guarantee the establishment and maintenance of a sliding manifold M = {x |s(x , t ) = 0},
u1 is selected as follows:
( t
−1
u1 (t ) = (GB ) u1s (t )dt
*0 1
+
u1s (t ) = −α · sign s˙ + β|s| 2 sign (s ) , (23)

where α > 0, β > 0 is the controller constant.


Theorem 1. With controller Eq. (23), the reachability condition of sliding motion can be
guaranteed, that is, state trajectory of the closed-loop system will be globally driven onto the
sliding mode in finite time.
Proof. Considering controller Eq. (23), the second-order time derivative of sliding mode sur-
face s is given as follows:

s¨(x, t ) = GBu˙ 1 (t ) − GBd d˙(t )


= u1s (t ) − GBd d˙(t )
∈ u1s (t ) − GBd [−Cd , Cd ] (24)

Following [31], the second-order SMC u1s with prescribed convergence law is completed.
A brief analysis is presented in the following section, and additional detailed proofs are
illustrated in [31,32].
1
First, we prove that the trajectory of inclusion inevitably hits the curve s˙ + β|s| 2 sign (s) = 0
1
due to geometrical reasons. Function fs = s˙ + β|s| 2 sign (s) is differentiated along (23) and
(24) as follows:
1 1
f˙s (t ) = u1s (t ) − GBd d˙(t ) + β|s|− 2 s˙ · sign (s )
2
2676 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689
* 1
+ 1 1
∈ −α · sign s˙ + β|s| 2 sign (s ) − GBd [−Cd , Cd ] + β|s|− 2 s˙ · sign (s )
2
1 − 21
∈ −α · sign ( fs ) − GBd [−Cd , Cd ] + β|s| s˙ · sign (s )
2
Selecting α and β satisfying α − GBd Cd − 21 β 2 > 0 imply that f˙s sign ( fs ) < const < 0 ex-
ists in a vicinity of each point on curve fs = 0. Thus, the sliding-mode motion inevitably hits
and stays on curve fs = 0.
1
Then, 1-sliding mode s˙(x, t ) = −β|s| 2 sign (s) holds; this mode guarantees that s converges
to the origin in finite time.!
In classic ISM, control signal u1 is discontinuous and is selected with the following form:
(GB )T s (x, t )
u1 (t ) = −ρ ,
(GB )T s (x, t )
where ρ is a gain sufficiently high to enforce sliding motion. Control command chattering
can degrade the performance of the closed-loop system. Chattering-free second-order SMC
can attenuate chattering.
Remark 2. The derivative of sliding mode motion s˙ in Eq. (23) is unavailable, but it can be
approximately expressed as follows:
s (t ) − s (t − T ) s (t ) − s (t − Ts )
s˙ = lim ≈ ,
T →0 T Ts
where Ts is sample time. The parameter α in Eq. (23) is responsible for disturbance elimi-
nation, which should be selected according to the amplitude of perturbations. β in Eq. (23)
determines the convergence rate of s when the sliding-mode motion stays on curve fs = 0.
However, too large values of these parameters may cause violent chattering of control signals.
Thus, it is a trade-off process to properly choose the parameters.
Next, equivalent control method is employed to determine motion equations at the sliding
manifold . Equivalent control is yielded by solving the equation s˙(x, t ) = 0 in Eq. (22) for
u1 :
u1eq (t ) = −(GB )−1 (GBd )d (t ) (25)
Remark 3. Note that for the matching case, Bd = B. In Eq. (25), u1eq (t ) = d (t ) can be
calculated. Therefore, any matrix G, which guarantees det (GB ) ̸= 0, can remove the matched
disturbances through ISM technique.
Substituting u1eq (t ) for u1 (t ) in Eq. (19), sliding motion is given with the following form:
x˙ (t ) = Ax (t ) + BK x (t ) − B (GB )−1 (GBd )d (t ) + Bd d (t )
y (t ) = Cx (t ) (26)
In traditional ISM, G is selected as G = BT , which can counteract the matched distur-
bances completely but guarantee no amplification of the effect of unmatched disturbances.
Mismatched uncertainties cannot be eliminated completely from the state equation regard-
less of controller design [33]. We primarily aim to eliminate the influence of disturbances
(matched and mismatched) on the output via ISM method. Sliding manifold gain G is designed
as follows:
G = C (A + BK )−1 (27)
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2677

The following lemma and assumptions provided the condition for the state stability and
output convergence.
Lemma 1. For system with the following form:
x˙ (t ) = Ax (t ) + Bu (t ) (28)
if A is a Hurwitz matrix, and u(t ) features constant values in the steady state (i.e.,
limt→∞ u(t ) = U∞ ), then state x(t ) converges to a constant vector −A−1 BU∞ .
Assumption 4. Disturbances d (t ) is bound, i.e., d (t ) ≤ Cb .
Assumption 5. Disturbances d (t ) satisfy the condition that they possess constant values in
the steady state, i.e., limt→∞ d˙(t ) = 0 and limt→∞ d (t ) = C∞ .
Note that Assumption 5 is usually stronger than Assumption 4. When Assumptions 3
and 5 are satisfied, Assumption 4 can be obtained. The following theorem illustrates the main
results of our method.
Theorem 2. Given that Assumptions 1–3 are satisfied, controller gain K is designed such that
matrix (A + BK ) is Hurwitz. The sliding mode surface gain G is given in Eq. (27). Then, the
system (19) under control law (20) obtains the following properties.

(1) If Assumption 4 is satisfied, all the states are ultimately bounded with tunable ultimate
bounds.
(2) If Assumption 5 is satisfied, then the state of the closed-loop system is bound and
asymptotically stable. Lumped disturbance d (t ) can be completely eliminated from the
output channels in the steady state.

Proof. Before we provide analysis of convergence of properties, we will prove that the state
of the closed-loop system is bound in time interval [0 T ]. During this stage, the closed-loop
system is described as follows:
x˙ (t ) = Ax (t ) + BK x (t ) + Bu1 (t ) + Bd d (t )
From Eq. (23), the upper bound of u1 (t ) when t ∈ [ 0 T ] is (GB )−1 αT , i.e., u1 (t ) ≤
(GB )−1 αT , t ∈ [ 0 T ]. d (t ) ≤ Cd T given that time derivative of d (t ) is bound. Consid-
ering that the matrix (A + BK ) is Hurwitz, the state will not escape to infinity in finite time
T . Theorem 1 states that the sliding mode motion is achieved in finite time T , that is, the
systems will reduce to the sliding mode motion (26) in finite time.

(1) Then, considering system (26), the proof of conclusion (1) is provided. For the Hurwitz
matrix (A + BK ), positive matrices P and Q exist, satisfying the following equation:
(A + BK )T P + P (A + BK ) = −Q
The Lyapunov function V (x) = x T Px ≥ 0 is selected. Along Eq. (26) and Assumption 4,
one obtains the following:
, - , -
V˙ (x ) = x T (A + BK )T P + P (A + BK ) x + 2x T P −B (GB )−1 (GBd ) + Bd d (t )
, -
= −x T Qx + 2x T P −B (GB )−1 (GBd ) + Bd d (t )
2678 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

≤ −λmin (Q )x 2 + 2λmax (Gc )xd (t )


= x (−λmin (Q )x + 2λmax (Gc )Cb ),

where λmin (∗) and λmax (∗) denote the minimum and maximum singular values, respec-
tively; matrix Gc = P[−B (GB )−1 (GBd ) + Bd ]. Thus, we obtain the following:
2λmax (Gc )Cb
V˙ (x ) ≤ 0, if x ≥
λmin (Q )
Following [18], the state x is ultimately bounded with boundary {x ≥ 2λλmax (Gc )Cb
min (Q)
}.
(2) Applying Lemma 1 and Assumption 5, the steady value of state x(t ) is obtained as
follows:
, -
limt→∞ x (t ) = −(A + BK )−1 −B (GB )−1 (GBd ) + Bd C∞
Output is defined as y(t ) = Cx(t ). Hence, its steady value is given as follows:
, -
lim y (t ) = −C (A + BK )−1 −B (GB )−1 (GBd ) + Bd C∞
t→∞

Applying the expression of G in Eq. (27), the above equation can be rewritten as
follows:
, -
lim y (t ) = −G −B (GB )−1 (GBd ) + Bd C∞
t→∞
= (GBd − GBd )C∞ = 0
!
Remark 4. In Assumption 5, disturbance is assumed to feature a limiting value. Then, the
property of asymptotic convergence of the closed-loop system is established. This disturbance
case is impossible in numerous practical engineering systems, such as random gust distur-
bances. Although many actual cases are out of the scope of our assumptions, the effectiveness
of our method in such a case is illustrated via numerical simulation (i.e., complex disturbances
including wind disturbances and model uncertainties are introduced in our helicopter simula-
tion).
Remark 5. The design of nominal controller K x can be implemented based on many control
theories, such as pole placement, H∞ method [20], or MPC [29]. Our work mainly focuses
on the sliding-mode motion control design. Hence, the design process of controller gain K is
omitted to highlight our main contributions.
Remark 6. The function of ISM control term is similar to that of DO [33,34]. Estimations
of lumped disturbances in the ISM framework are obtained based on the equivalent control
law. Moreover, there exist lots of results about unmatched disturbance elimination, including
BS control combined with DO based control [12,35] or adaptive control [10], high-order
SMC [17,36], and relative degree approach [37]. Actually, the selection of sliding-manifold
surface (27) is a special application for relative degree approach. By properly choosing the
sliding-manifold gains, the effect of disturbances is removed from output channels.

3.2. Helicopter controller design

In this subsection, the proposed ISM is applied to deal with the unmanned helicopter
control problem.
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2679

Assumption 6. The attitude of helicopters constantly lies inside the region |φ| < π
2
and
|θ (t )| < π2 .
Remark 7. Assumption 6 ensures that the attitude kinematic matrix '(!) in (7) is not
singular. This assumption is valid because our helicopter is supposed to operate in hovering
or low-velocity fight condition.
In Section 2, the helicopter model has been approximately feedback-linearized in Eq. (18).
The model comprises two parts, namely, the position error dynamics and yaw-angle error
dynamics.
Position error dynamics is computed as follows:
, -
e˙ P = APe eP + BPe fPe (t ) + uP + BPd dP
yeP = CPe eP (29)

Yaw-angle error dynamics is computed as follows:


, -
e˙ ψ = Aψe eψ + Bψe fψe (t ) + uψ + Bψd dψ
yeψ = Cψe eψ (30)

where
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
03×3 I3×3 03×3 03×3 03×3 03×3 03×3
⎢0 03×3 I3×3 03×3 ⎥⎥ ⎢0 ⎥ ⎢ I 03×3 ⎥⎥
APe = ⎢⎣ 3×3 ⎦ , BPe = ⎢⎣ 3×3 ⎥⎦, BPd = ⎢⎣ 3×3 ,
03×3 03×3 03×3 I3×3 03×3 03×3 03×3 ⎦
03×3 03×3 03×3 03×3 I3×3 03×3 I3×3
' ) ' ) ' )
, - #F 0 1 0
CPe = I3×3 03×3 03×3 03×3 , dP = , Aψe = , Bψe = Bψd = ,
#MP 0 0 1
, -
dψ = #Mψ , and Cψe = 1 0 .

Applying the proposed ISM technique, the control command signals for unmanned heli-
copters are given as follows:
Position error subsystem:
( t
uP = − fPe (t ) + KPe e p + (GP BPe )−1 uP1s (t )dt
* 0 +
1
uP1s (t ) = −αP sign s˙P + βP |sP | 2 sign (sP ) (31)

Yaw-angle error subsystem:


(
% &−1 t
uψ = − fψe (t ) + Kψe eψ + Gψ Bψe uψ1s (t )dt
* 0 +
0 01 % &
uψ1s (t ) = −αψ sign s˙ψ + βψ 0sψ 0 2 sign sψ (32)

where Kpe and Kψe are controller gain matrices, (APe + BPe KPe ) and (Aψe + Bψe Kψe ),
respectively, which are Hurwitz matrices. The poles of the closed-loop sys-
tems constantly lie on the left-half plane. The sliding mode surface sP =
GP [eP − eP (0) − ∫t0 (APe eP + BPe ( fPe (t ) + KPe eP ) )dt ] and sψ = Gψ [eψ − eψ (0) − ∫t0 (Aψe eψ
+ Bψe ( fψe (t ) + Kψe eψ ))dt]; GP = CPe (APe + BPe KPe )−1 and Gψ = Cψe (Aψe + Bψe Kψe )−1 .
2680 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

Table 1
Parameters of small-scale unmanned helicopter.

Symbol Description Value


m Mass of unmanned helicopter 8.2 kg
g Acceleration of gravity 9.8 m/s2
J The moment of inertia matrix of helicopter diag(0.18 0.34 0.28) kg m2
Zw Linkage gain ratio of Tm to w −0.7615 s−1
Zcol Linkage gain ratio of Tm to δcol −131.4125 m/(rad s2 )
Coefficient matrix of / in (9) diag( −1
A ⎡ −48.1757 −25.5048 −0.9808) s ⎤
0 0 1689.5 0
B Coefficient matrix of u in (9) ⎣ 0 894.5 0 0 ⎦s−2
−0.3705 0 0 135.8

T
Kψe is selected, Gψ = [0 1] = Bψe , regardless of the controller value. Next, practical
control signals for the helicopter can be calculated from uP and uψ . According to their
definitions, one computes the following:
⎡ ⎤ ⎡ ⎤
Mφ 0 Tm 0
⎣Mθ ⎦ = −m · ⎣−Tm 0 0⎦ · R−1 (!) · uP (33)
T̈m 0 0 1


Mψ = uψ (34)

Then, control commands [δcol δlon δlat δ ped ]T can be obtained according to the defi-
11 t
nition of [Mφ Mθ Mψ ]T and TM = 0 T̈m dt.
2 3
Tm
δcol = − − g + Zw w /Zcol (35)
m
⎡ ⎤ ⎛⎡ ⎤ ⎞
δlon Mφ
⎣ δlat ⎦ = B (2 )\⎝⎣ Mθ ⎦ + J −1 (ω × J ω ) − Aω − B (1 )δcol ⎠, (36)
δ ped Mψ

where B(1) and B(2) are the first- and second-forth column of matrix B, respectively.
The following result is given directly by applying Theorem 2.

Theorem 3. Considering helicopter dynamic models (1)–(4) under Assumptions 1–6, if the
controller is designed as Eqs. (35) and (36), then position and yaw-angle tracking errors of
the closed-loop system asymptotically converge to zero.

4. Simulation analysis

In this section, numerical simulation results are presented to illustrate the efficiency of our
proposed ISM method for small-scale unmanned helicopters. Table 1 summarizes the corre-
sponding helicopter model parameters [12,38], where diag(x1 , x2 , . . . , xn ) denotes diagonal
matrix with diagonal elements x1 , x2 , . . . , xn .
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2681

Following [12], lumped disturbances in Eqs. (10) and (11) are introduced in simulation.
These disturbances include model uncertainties and external disturbances.
⎡ ⎤
' ) V
#F
= # · ⎣!⎦ + Bwind dwind , (37)
#M
ω

where constant matrix # ∈ R6×9 represents model uncertainty, and Bwind denotes the trans-
formation matrix from airspeed to force and moment. All elements of # are pseudorandom
values within the open interval (−0.5 0.5 ). Matrices # and Bwind are given as follows:
⎡ ⎤
−0.239 0.164 −0.055 0.456 0.013 −0.362 −0.096 0.071 −0.185
⎢ 0.203 0.351 0.419 0.351 −0.135 0.342 0.491 −0.359 −0.387⎥⎥

⎢ 0.035 0165 0.232 0.265 −0.384 0.254 −0.044 0.357 0.031 ⎥⎥
# = ⎢⎢
⎢ 0.311 0.045 0.369 −0.250 −0.303 −0.368 −0.352 0.012 −0.011⎥⎥
⎣−0.055 −0.465 0.062 0.290 0.048 −0.451 0.426 −0.454 0.209 ⎦
0.267 −0.409 −0.088 −0.482 0.057 0.459 0.138 −0.028 0.343
⎡ ⎤
−0.0505 0
⎢ 0 −0. 151 ⎥⎥

⎢ 0 0 ⎥
Bwind = m × ⎢⎢ ⎥
⎢ −0.144 0.143 ⎥⎥
⎣−0.0561 −0.0585⎦
0 0.0301

Wind disturbance dwind is assumed to possess wind components along [xb yb ] in BRF.
This variable is composed of constant and random components, i.e. dwind = dc + dr . The con-
stant component is set as [5 5]T . Stochastic wind disturbance is modeled by independently
excited correlated Gauss–Markov processes [6]:
' ) ' )' ) ' )
d˙u −1/τc 0 du q
˙
dr = ˙ = +ρ u ,
dv 0 −1 / τc d v qv
where qu and qv (σqu = σqv = 20ft/s) are independent with zero mean, and τc = 3.2 s is the
correlation time of the wind. ρ = 1/2 is the scalar weighting factor. Fig. 2 shows the wind
disturbance dwind used in the helicopters’ simulation.
To provide sufficient fidelity in the simulation, the flapping dynamics model is considered.
Flapping dynamics model is equivalent to the additional dynamics in the servo loop [38]:

τ f δ̄˙ lon = −τ f q − δ̄lon + δlon


τ f δ̄˙ lat = −τ f p − δ̄lat + δlat , (38)

where δ̄lon and δ̄lat are longitudinal and lateral applied control signals; τ f = 0.1 is the main
rotor’s dynamics time constant.

4.1. Description of control laws

To demonstrate the superiority of our method, several previous methods are compared with
our ISM design. These implementations are illustrated as follows.
2682 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

Fig. 2. Wind disturbance.

4.1.1. Pole placement (PP)


PP method is one of the most popular methods in control field. According to the flight
characteristics of small-scale helicopters, the parameters of the PP controller are selected as
follows:
⎡ ⎤
1800 0 0 2430 0 0 679 0 0 50 0 0
KPe = −⎣ 0 1800 0 0 2430 0 0 679 0 0 50 0 ⎦
0 0 1800 0 0 2430 0 0 679 0 0 50
, -
Kψe = − 450 45

where the corresponding poles in the position error subsystem and yaw-angle error subsystem
are selected as (−1 −4 −15 −30 ) and (−15 −30 ), respectively.

4.1.2. PP+ISM
This classic method is proposed in [15]. PP+ISM controller architecture is presented in
Eqs. (31) and (32). The nominal controller is the same as the PP controller. Sliding manifolds
are designed based on the projection matrix approach such that equivalent disturbances are
T T
equal to the unmatched ones. Sliding manifold gains are given as GP = BPe and Gψ = Bψe .
The second-order sliding mode method is used to achieve sliding-mode motion. The corre-
sponding ISM control parameter is designed as follows:

αP = diag(2, 2, 2 ), βP = diag(2, 2, 2 )
αψ = 2, βψ = 2
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2683

4.1.3. PP+output ISM


PP+output ISM is our proposed ISM method. Implementation of this method is the same
as that of the PP+ISM controller, except for the sliding manifold gains. In our method,
disturbances in output channels can be completely cancelled out, guaranteeing exact trajectory
tracking. The ISM manifold gains GP and Gψ can be obtained by applying Eq. (27).

4.1.4. H∞ +ISM
This technique uses an ISM controller to reject the matched perturbation and design the
nominal control using H∞ techniques to attenuate the unmatched ones [21]. The ISM com-
ponent is the same as the PP+ISM controller. The nominal component applies H∞ control
in the position error subsystem (unmatched disturbances occur only in this subsystem). Cost
matrix is selected as follows:
Q = diag(100, 100, 100, 100, 100, 100, 50, 50, 50, 30, 30, 30 ).
H∞ controller is solved as follows:
⎡ ⎤
865 0 0 1576.4 0 0 654.5 0 0 50.1 0 0
⎣ −7 −7
KPe = 0 865 0 0 1576.4 −4.1 × 10 0 654.5 −1.5 × 10 0 50.1 0 ⎦
0 0 865 0 −4.3 × 10−6 1576.4 0 −1.6 × 10−6 654.5 0 0 50.1

4.1.5. 1st output ISM


The difference between the 1st output ISM and our method is the sliding-mode motion
control component. Comparison demonstrates that by applying the second-order SMC, chat-
tering phenomenon of control signals can be improved. In this case, classic SM control under
boundary layer modification is utilized.


⎪ (GB )T s
⎨−ρ · T
, i f (GB )T s ≥ ε
u1 (t ) = ( GB ) s ,

⎪ (GB )T s
⎩−ρ · , i f (GB )T s < ε
ε
where constant ρ = 10000 and ε = 10−3 .

4.1.6. Backstepping + DO (BSDO)


This is one of the classical robust control methods to eliminate the adverse effect of
matched and unmatched disturbances, which has been widely used in the practical applications.
Following the system (18), applying BSDO method, it gives
eP2d,bs = −k1,bs eP1
% &
eP3d,bs = −k2,bs eP2 − eP2d,bs − # ˆF
% &
eP4d,bs = −k3,bs eP3 − e p3d,bs
% &
u p = − fPe (t ) − k4,bs e p4 − e p4d,bs − # ˆM
eψ2d,bs = −k1ψ,bs eψ1
% &
uψ = − fψe (t ) − k2ψ,bs eψ2 − eψ2d,bs − # ˆψ

Where # ˆ F = k f (e˙ P2 − eP3 ), #


ˆ M = k f (e˙ P4 − fPe (t ) − uP ), #
ˆ ψ = k f (e˙ ψ2 − fψe (t ) − uψ )
s+k f s+k f s+k f
are the estimation of the unknown disturbance items #F , #M p and #Mψ , respectively. The
control and filter gains are set as: k1,bs = 1, k2,bs = 4, k3,bs = 15, k4,bs = 30, k1ψ,bs = 15,
k2ψ,bs = 30 and k f = 2.
2684 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

Fig. 3. Trajectory tracking of helicopter.

Table 2
the RMS of positon tracking errors.

Index PP PP+ISM PP+outpout ISM Hinf +ISM BSDO


RMS (m) 2.151 1.954 0.453 3.114 0.584

4.2. Simulation results

The desired “8-shape” reference trajectory is described as follows:


, % &-T
Pd (t ) = 0 0 −7 1 − e−0.3t for t ≤ 7s
⎡ 2 3⎤

⎢ 20 1 − cos 23 (t − 7 ) ⎥
⎢ 2 3 ⎥
Pd (t ) = ⎢⎢ 4π ⎥ for t > 7s

⎣ 10 sin (t − 7 ) ⎦
% 23 −0.3t &
−7 1 − e
ψd = 0 (39)
Simulation results are illustrated in Figs. 3–8. Fig. 3 shows the three-dimension position
response curves of the closed-loop system based on the above methods. Compared with other
methods, the proposed ISM guarantees high tracking of the reference trajectory in the presence
of model uncertainties and wind disturbance. In Table 2, The root-mean-squares (RMS, eP1 )
of tracking errors are given to quantitatively display tracking performance of these control
systems. Given that introducing the traditional ISM shows no evident improvement on track-
ing performance, we can observe that the unmatched disturbance (lumped force disturbance)
exerts more influence on helicopter controller design. H∞ design slightly affects suppression
of unmatched disturbances and causes a decline in performance due to difficulties in select-
ing appropriate performance function. Additionally, the proposed method has the comparable
performance with BSDO, even smaller steady errors than BSDO. It proves that by properly
choosing the sliding manifold, the effect of mismatched perturbation is attenuated.
Fig. 4 shows the three-dimensional position trajectory intuitively, and results agree with this
analysis. Figs. 5–7 show the velocity, angle, and angular velocity response. The responses of
PP+output ISM fluctuate heavily to rapidly eliminate disturbances. Violent chattering exists
at t = 7s due to switching of reference trajectory. Fig. 8 displays the response curves of
controller inputs. The control signal curves of our method are smoother than the 1st output
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2685

Fig. 4. Position response of the closed-loop helicopter system.

Fig. 5. Velocity response of the closed-loop helicopter system.


2686 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

Fig. 6. Angular response of the closed-loop helicopter system.

Fig. 7. Angular velocity response of the closed-loop helicopter system.


T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2687

Fig. 8. Control inputs of helicopters.

ISM with low vibration amplitude. This result confirmed that control signals can be improved
by the second-order SM technique.

5. Conclusion

This paper proposes a novel ISM control method to deal with the trajectory tracking
problem for helicopter control. The controller is developed based on the approximate feedback-
linearized model with lumped matched and unmatched disturbances. Bounded disturbances
can be completely eliminated in the output channels by selecting the integral sliding mode
manifold suitably. The second-order chattering-free SM control is adopted to attenuate the
chattering phenomenon. Some simulation results demonstrate that through the ISM control
approach presented in this work, a model-scale unmanned helicopter will exhibit excellent
tracking performance in the presence of model uncertainties and gust perturbation.
The nominal component of the ISM controller, which guarantees ideal performance of
the closed-loop system, can be achieved through mature techniques, such as pole placement,
H∞ control method, and MPC. Future work will focus on developing the compound control
approach based on ISM and other advanced control techniques.

References

[1] B. Mettler, Identification Modeling and Characteristics of Miniature Rotorcraft, Kluwer Academic Pub, Boston,
2003.
[2] F. Kendoul, Survey of advances in guidance, navigation, and control of unmanned rotorcraft systems, J. Field
Robot 29 (2) (2012) 315–378.
[3] J. Alvarenga, N.I. Vitzilaios, K.P. Valavanis, M.J. Rutherford, Survey of unmanned helicopter model-based
navigation and control techniques, J Intell. Robot. Syst. 80 (1) (2015) 87–138.
[4] G.W. Cai, B.M. Chen, Unmanned Rotorcraft Systems, Springer, Heidelberg, 2011.
[5] O. Tanner, Modeling, Identification, and Control of Autonomous Helicopters, ETH, 2003.
2688 T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689

[6] J. Gadewadikar, B.M. Chen, K. Subbarao, Structured command and control-loop design for unmanned heli-
copters, J. Guid. Control Dyn. 31 (4) (2008) 1093–1102.
[7] M.L. Civita, G. Papageorgiou, W.C. Messner, T. Kanade, Design and flight testing of a gain-scheduled H_inf
loop shaping controller for wide-envelope flight of a robotic helicopter, in: Proceedings of the American Control
Conference, Denver, USA, 2003, pp. 4195–4200.
[8] I.A. Raptis, K.P. Valavanis., Linear and Nonlinear Control of Small-Scale Unmanned Helicopters, Springer,
London, 2011.
[9] C.T. Lee, C.C. Tsai, Adaptive backstepping integral control of a small-scale helicopter for airdrop missions,
Asian J. Control 12 (2010) 531–541.
[10] B. Zhu, W. Huo, Robust nonlinear control for a model-scaled helicopter with parameter uncertainties, Nonlinear
Dyn. 73 (1–2) (2013) 1139–1154.
[11] Z. Yao, H. Wei, Singularity-free backstepping controller for model helicopters, ISA Trans. 65 (11) (2016)
133–142.
[12] Y.B. He, H.L. Pei, T.R. Sun, Robust tracking control of helicopters using backstepping with disturbance ob-
servers, Asian J. Control 16 (6) (2014) 1–16.
[13] H. Lu, C. Liu, L. Guo, W.H. Chen, Flight control design for small-scale helicopter using disturbance-ob-
server-based backstepping, J. Guid. Control Dyn. 38 (11) (2015) 1–6.
[14] T.J. Koo, S. Sastry, Output tracking control design of a helicopter model based on approximate linearization, in:
Proceedings of the Thirty-seventh IEEE Conference on Decision & Control, Tampa, USA, 1998, pp. 3635–3640.
[15] C.J. Liu, W.H. Chen, J. Andrews, Tracking control of small-scale helicopters using explicit nonlinear MPC
augmented with disturbance observers, Control Eng. Pract. 20 (3) (2012) 258–268.
[16] X. Fang, A.G. Wu, Y.J. Shang, N. Dong, Robust control of small-scale unmanned helicopter with matched and
mismatched disturbances, J. Frankl. Inst. 353 (9) (2016) 4803–4820.
[17] X. Fang, A. Wu, Y. Shang, N. Dong, A novel sliding mode controller for small-scale unmanned helicopters
with mismatched disturbance, Nonlinear Dyn. 83 (1–2) (2016) 1053–1068.
[18] H.K. Khalil, Nonlinear Systems, 3rd ed., Prentice Hall, Upper Saddle River, NJ, 2002.
[19] W.J. Cao, J.X. Xu, Nonlinear integral-type sliding surface for both matched and unmatched uncertain systems,
IEEE Trans. Autom. Control 49 (8) (2004) 1355–1360.
[20] F. Castaños, L. Fridman, Analysis and design of integral sliding manifolds for systems with unmatched pertur-
bations, IEEE Trans. Autom. Control 51 (5) (2006) 853–858.
[21] M. Rubagotti, F. Castaños, A. Ferrara, L. Fridman, Integral sliding mode control for nonlinear systems with
matched and unmatched perturbations, IEEE Trans. Autom. Control 56 (11) (2011) 2699–2704.
[22] B.X. Mu, K. Zhang, Y. Shi, Integral Sliding mode flight controller design for a quadrotor and the application
in a heterogeneous multi-agent system, IEEE Trans. Ind. Electron. 64 (12) (2017) 9389–9398.
[23] B.X. Mu, Y.Y. Pei, Y. Shi, Integral sliding mode control for a quadrotor in the presence of model uncertainties
and external disturbances, in: Proceedings of the 2017 American Control Conference, Seattle, WA, USA, 2017.
[24] K. Nonaka, H. Sugizaki, Integral sliding mode altitude control for a small model helicopter with ground effect
compensation, in: Proceedings of the 2011 American Control Conference, San Francisco, CA, USA, 2011.
[25] H. Ramirez-RodriguezEmail, V. Parra-Vega, A. Sanchez-Orta, O. Garcia-Salazar, Robust Backstepping control
based on integral sliding modes for tracking of quadrotors, J. Intell. Robot. Syst. 73 (2014) 51–66.
[26] Q. Qu, F.Y. Chen, B. Jiang, G. Tao, Integral sliding mode control for helicopter via disturbance observer and
quantum information technique, Math. Probl. Eng. 2015 (2015) 1–7.
[27] G. Gui, Vukovich, Adaptive integral sliding mode control for spacecraft attitude tracking with actuator uncer-
tainty, J. Frankl. Inst. 352 (12) (2015) 5832–5852.
[28] J. Xu, Z. Guo, T.H. Lee, Design and implementation of integral sliding-mode control on an underactuated
two-wheeled mobile robot, IEEE Trans. Ind. Electron. 61 (7) (2014) 3671–3681.
[29] M. Rubagotti, D.M. Raimondo, A. Ferrara, Robust model predictive control with integral sliding mode in
continuous-time sampled-data nonlinear systems, IEEE Trans. Autom. Control 56 (3) (2010) 556–570.
[30] R. Errouissi, J. Yang, W.H. Chen, A. Al-Durra, Robust nonlinear generalised predictive control for a class of
uncertain nonlinear systems via an integral sliding mode approach, Int. J. Control 89 (8) (2016) 1698–1710.
[31] A. Levant, Principles of 2-sliding mode design, Automatica 43 (4) (2007) 576–586.
[32] A. Levant, Sliding order and sliding accuracy in sliding mode control, Int. J. Control 58 (6) (1993) 1247–1263.
[33] S.H. Li, J. Yang, C. Wen Hua, C. Xi Song, Generalized extended state observer based control for systems with
mismatched uncertainties, IEEE Trans. Ind. Electron. 59 (12) (2012) 4792–4802.
[34] W.H. Chen, J. Yang, G. Lei, L.S. Hua, Disturbance-observer-based control and related methods—an overview,
IEEE Trans. Ind. Electron 63 (2) (2016) 1083–1095.
T. Jiang, D. Lin and T. Song / Journal of the Franklin Institute 356 (2019) 2668–2689 2689

[35] S.H. Li, H.B. Sun, J. Yang, X.H. Yu, Continuous finite-time output regulation for disturbed systems under
mismatching conditon, IEEE Trans. Autom. Control 60 (1) (2015) 277–282.
[36] J. Yang, S.H. Li, J.Y. Su, X.H. Yu, Continuous nonsingular terminal sliding mode control for systems with
mismatched disturbances, Automatica 49 (7) (2013) 2287–2291.
[37] J. Yang, S.H. Li, W.H. Chen, Nonlinear disturbance observer-based control for multi-input multi-output nonlinear
systems subject to mismatching condition, Int. J. Control 85 (8) (2012) 1071–1082.
[38] V. Gavrilets, Autonomous Aerobatic Maneuvering of Miniature Helicopters, MIT, 2003.

S-ar putea să vă placă și