Automation in Construction 119 (2020) 103342

Contents lists available at ScienceDirect

Automation in Construction
journal homepage: www.elsevier.com/locate/autcon

Adaptive sway reduction for tower crane systems with varying cable lengths T
a,⁎ a b c a
Menghua Zhang , Yongfeng Zhang , Bing Ji , Changhui Ma , Xingong Cheng
a
School of Electrical Engineering, University of Jinan, Jinan 250022, China
b
School of Control Science and Engineering, Shandong University, Jinan 250061, China
c
State Grid Shandong Electric Power Research Institute, Jinan 250002, China

A R T I C LE I N FO A B S T R A C T

Keywords: For practical underactuated tower cranes, vertical payload motion is always involved to increase the working
Underactuated system efficiency, making the control problem more complex. Most existing control methods are designed based on
Tower crane linearized crane dynamics and require exact plant parameters. To solve these problems, the dynamic model of
Modeling tower cranes with varying cable lengths is established firstly, on whose basis, an adaptive sway reduction control
Adaptive control
method is proposed. To our knowledge, the designed control method yields the first closed-loop control solution,
Sway reduction
Robust
with guaranteed theoretical analysis, to successfully address the tower crane anti-swing and positioning problem
in the presence of varying cable lengths. The model and control method are simulated by MATLAB/Simulink,
and the simulation results show that the anti-swing performance is significantly enhanced by more than 63.22%,
compared with PD method. Considering there are several control gains in the designed law, optimization of
control gains based on optimization methods will be conducted.

1. Introduction loop control methods will be badly attenuated and even lead to in-
stability [14,15]. In this case, closed-loop control methods, which is
For the past several decades or so, tower cranes, as powerful well-known to be insensitive to disturbances, might provide better
transportation tools, have gained tremendous popularity on construc- control performance. To be more precise, to drive the trolley and the jib
tion sites. Like other cranes, tower cranes have less control inputs than to their desired destinations within only one oscillation cycle of the
the degrees-of-freedom (DOFs) to be controlled [1–5], which are typical payload, Omar and Nayfeh design a gain scheduling feedback control
underactuated nonlinear systems whose control problems are challen- method for tower cranes [16]. In [17], Lee et al. develop a laser-tech-
ging and still open. Till now, almost all tower cranes are operated by nology-based lifting-path tracking system for a tower crane system
manual, which have the shortcomings concerning low working effi- under various outdoor conditions, and examine its feasibility. To move
ciency, poor anti-swing ability, high risk of human injury accidents, as the payload along a predefined trajectory and be able to account for
well as long time to train skilled operations personnel, and so forth system constraints, Böck and Kugi propose a model predictive control
[6–8]. Therefore, it is urgent to design automatic control methods for (MPC) method for tower crane systems [18]. In [19], by linearizing the
tower crane systems. tower crane system, Koumboulis et al. design a dynamic measurable
Control problems for tower crane systems have aroused extensive output multi delay control law to provide satisfactory performance
public concern for the past several decades, in the meantime, a large despite the nonlinearities of the dynamic model and the time delays of
quantity of effective control measures have been taken. Based on the wireless network. Ilešet al. develop a sequential distributed MPC
whether they require realtime state feedback, the existing control laws scheme for a 3D tower crane; and empty three locally stabilizing MPC
can be roughly divided into two categories: one is open-loop control control method to prove the stability of the controlled system [20]. In
methods, the other is closed-loop control methods. Among all the open- [21], based on the analysis of the mechanical behavior of a two wired
loop control methods, two more ubiquitous ones with simple and hammerhead tower crane, a linear feedback control method is de-
handy-implemented characteristics are input shaping [9–11] and op- signed. In [22], Escalona et al. propose a discretization procedure for
timal velocity control [12,13]. However, open-loop control methods are multibody modeling of reeving systems including transverse vibrations;
so sensitive to external disturbances and parametric uncertainties that and simulate the dynamics of a 3D tower crane system. The above-
when disturbances present, the overall control performance of open- mentioned control laws for tower crane systems are designed based on

⁎
Corresponding author.
E-mail address: [email protected] (M. Zhang).

https://doi.org/10.1016/j.autcon.2020.103342
Received 17 July 2019; Received in revised form 14 June 2020; Accepted 28 June 2020
Available online 10 July 2020
0926-5805/ © 2020 Elsevier B.V. All rights reserved.
M. Zhang, et al. Automation in Construction 119 (2020) 103342

simplified dynamics together with exact model knowledge. To solve represent the trolley translation displacement and the jib slew angle,
those problems, adaptive robust control algorithms are designed for 4- respectively, θ1 and θ2 denote the payload swing angles. All the
DOF tower cranes [23–25]. In addition to the aforementioned model- movements (cable, trolley, and jib motions) are generated by ac motors,
based control laws, the fuzzy method [26–27] and recurrent neural which are controlled by servo actuators connected to a software control
network-based control method [28] are also proposed for tower crane system. With regard to the software control system, the MATLAB/Si-
systems. mulink 2014b Real-Time Windows Target which runs in the environ-
After summarizing the existing control methods for tower crane ment of the Windows XP operating systems is used to implement the
systems, there are still a lot of issues to be addressed, which can be designed control scheme in real time. The Lagrange approach [29] is
outlined as follows. utilized to derive the dynamic equations of tower crane systems with
varying cable lengths. As can be seen from Fig. 1, the position vectors of
1) Existing control methods are designed for tower cranes with con- the trolley and the payload are given as follows:
stant cable length. In many cases, to improve the working efficiency,
pM = [x 0 0]T (1)
tower cranes lift/lower the payload while the trolley and jib travel
with the cable-suspended payload. Therefore, it is of great practical pm = [ x + lS1 C2 lS2 lC1 C2 ]T (2)
significance to design effective control laws incorporating the con-
sideration for varying cable lengths. where S1, S2, C1, and C2 stand for the abbreviations of sinθ1, sinθ2,
2) Most currently available control methods for tower cranes require cosθ1, and cosθ2, respectively.
exact plant parameters (e.g. trolley mass, payload mass, friction- The angular velocity vector of the tower crane systems is
related parameters, coefficients of air resistance). However, it is T
usually difficult to obtain these parameters, which may deteriorate Ω = [ 0 0 ϕ ̇] (3)
the control performance and even lead to instability. The velocity vectors of the trolley and the payload are given by
3) Most existing control methods for tower cranes are designed and
analyzed based on either linearization or approximation operations dpM
vM = + Ω × pM
to the original nonlinear dynamics. The control performance would dt
T
be badly degraded if the system states are far away from the equi- =[ x ̇ 0 0 ]T + [ 0 0 ϕ ̇] × [ x 0 0]T
librium point. T
=[ x ̇ xϕ ̇ 0 ] (4)
Therefore, to tackle the aforementioned problems, the dynamic
dpm
model of the tower crane systems is firstly established by using vm = + Ω × pm
dt
Lagrangian method. And on this basis, an adaptive control method with
payload swing suppression is proposed for the crane system, which is
used to provide adaptive control performance with respect to un- ⎡ x ̇ + lθ1̇ C1 C2 − lθ2̇ S1 S2 + lṠ 1 C2 ⎤
certain/unknown friction-related parameters and parametric un- =⎢ lθ2̇ C2 + lṠ 2 ⎥
⎢ ⎥
certainties. After that, Lyapunov method and LaSalle's invariance the- ⎢ ̇ ̇ ̇
⎣ −lθ1 S1 C2 − lθ2 C1 S2 + lC1 C2 ⎥ ⎦
orem are used to prove the stability and convergence. Simulation T
results show the proposed method can achieve superior control per- +[ 0 0 ϕ̇ ]
formance. × [ x + lS1 C2 lS2 lC1 C2 ]T
The main contribution of this paper can be concluded as follows.

1) The dynamic model of the tower crane system with varying cable
lengths is established, providing a model for realizing positioning ⎡ x ̇ + lθ1̇ C1 C2 − lθ2̇ S1 S2 − lϕṠ 2 + lṠ 1 C2 ⎤
and swing suppression and elimination. =⎢ lθ2̇ C2 + ϕ ̇ (x + lS1 C2) + lṠ 2 ⎥
⎢ ⎥
2) The proposed control method shows strong robustness against un- ⎢
⎣ −lθ1̇ S1 C2 − lθ2̇ C1 S2 + lC ̇ 1 C2 ⎥
⎦ (5)
certain/unknown system dynamics and external disturbances.
3) It is no need for the designed control method to linearize/approx- The kinematic energy of the tower crane systems is obtained as
imate the tower crane dynamics or neglect any nonlinear terms. A follows:
major advantage is that, even when the state variables are not close
1 1 1 2
enough to the equilibrium point, the suggested control scheme can T= T
Mt vM vM + mp vmT vm + Jϕ ̇
still work well. 2 2 2
1 1 2 1 2
= (Mt + mp ) x ̇ + (Mt + mp ) x 2ϕ ̇ + Jϕ ̇
2
2 2 2
The rest of this paper is organized as follows. In Section 2, the dy- 2 2 2
̇ ̇ 2̇ 2 2̇ 2 2 2
namical equations of tower crane systems with varying cable lengths ⎛ l θ1 C2 + l θ2 + l ϕ S2 + l ϕ S1 C2 + l ̇ + 2lxθ
2 2 2 2 ̇ 1̇ C1 C2 − 2 ⎞
are derived. The main results including the adaptive control with ⎜ lxθ ̇ 2̇ S1 S2 − 2lxϕṠ ̇ 2 ⎟
1
payload swing suppression design and stability analysis are discussed in + mp ⎜ 2
⎟
̇ ̇ ̇
2 ⎜ −2l ϕθ̇ 1 C1 S2 C2 + 2l ϕθ̇ 2 S1 + 2lxϕθ̇ 2 C2 + 2lxϕ ̇ S1 C2 + 2xl̇ ̇ ⎟
2 2
Section 3. In Section 4, simulation results are given to verify the control ⎜ S C + 2lϕxS ⎟
̇ ̇ 2
performance of the designed control strategy. Section 5 concludes the ⎝ 1 2 ⎠
whole paper. (6)
with J being the inertia moment of the tower and the jib about Z-axis,
2. Modeling for tower crane systems with varying cable lengths
Mt and mp standing for the trolley mass and the payload mass, respec-
tively.
In this section, the dynamic model of tower crane systems with
varying cable lengths is established. Remark 1. As can be seen from Fig. 1, we know that the trolley moves
Fig. 1 shows the schematic diagram of the tower crane systems. The along the jib in X-axis, the jib rotates around the mast in Y-axis, and the
trolley moves along the jib in X-axis, the jib rotates around the mast in cable is lowered and lifted in Z-axis. In other words, the trolley just
Y-axis, and the cable is lowered and lifted in Z-axis. In Fig. 1, O is the moves in the two-dimension (XY) plane; hence, the problem of the
origin of the coordinate system, l denotes the cable length, x and ϕ trolley attitude doesn't exist. Therefore, existing models for tower

2
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Fig. 1. Schematic diagram of tower crane systems.

cranes didn't consider the inertia moment to the attitude of the trolley, d ⎛ ∂V ⎞ 2 + m l 2 (S 2 + S 2 C 2 ) + 2m lxS C + J )
⎜ ⎟ = ((Mt + mp ) x p p 1 2
for example: references [8, 23–25, 33–35]. dt ⎝ ∂ϕ ̇ ⎠
2 1 2

The potential energy is ¨ 2 + mp xl¨S2

ϕ¨ − mp lxS

P = mp gl (1 − C1 C2) (7)
− mp l 2θ¨1 C1 S2 C2 + mp lθ¨2 (xC2 + lS1) + 2(Mt + mp ) xxϕ
̇ ̇ + 2mp lxϕ ̇
with g representing the gravitational acceleration.
θ1̇ C1 C2
Define the Lagrangian V as 2
̇ ̇ 1 C2 + mp l 2θ1̇ S1 S2 C2 + 2mp l 2θ1̇ θ2̇ C1 S22 − mp lxθ2̇
+ 2mp lxϕS
V=T−P (8)
(θ2̇ + 2ϕṠ 1 ) S2
The partial derivative of V with respect to ϕ is obtained as follows: + mp l 2ϕθ̇ 1̇ sin(2θ1 ) C22 − mp l 2ϕθ̇ 2̇ S12 sin(2θ2) + mp xlS
̇ ̇ 2 + mp xlθ̇ 2̇ C2
(11)
∂V
=0
∂ϕ (9) By using the Lagrange equations,

In a similar way, the partial derivative of V with respect to ϕ ̇ is d ⎛ ∂V ⎞ ∂V

⎜ ⎟ − = Fϕ − Ffϕ
calculated as dt ⎝ ∂ϕ ̇ ⎠ ∂ϕ (12)

∂V 2 2 2 2 2
̇ 2 − l 2θ1̇ C1 S2 C2 ⎞
⎛ l ϕṠ 2 + l ϕṠ 1 C2 − lxS it can be obtained that
= (Mt + mp ) x 2ϕ ̇ + Jϕ ̇ + mp ⎜ ⎟
∂ϕ ̇ 2 ̇ ̇ ̇ ̇
⎝+ l θ2 S1 + lxθ2 C2 + 2lxϕS1 C2 + lxS2 ⎠
(10) ¨ 2 + mp xl¨
((Mt + mp ) x 2 + mp l 2 (S22 + S12 C22) + 2mp lxS1 C2 + J ) ϕ¨ − mp lxS
S2
Taking the time derivative of Eq. (10), it is derived that
̇ ̇ + 2mp lxϕθ̇ 1̇ C1 C2
− mp l 2θ¨1 C1 S2 C2 + mp lθ¨2 (xC2 + lS1) + 2(Mt + mp ) xxϕ
2
̇ ̇ 1 C2 + mp l 2θ1̇ S1 S2 C2 + 2mp l 2θ1̇ θ2̇ C1 S22 − mp lxθ2̇ (θ2̇ + 2ϕṠ 1 ) S2
+ 2mp lxϕS
+ mp l 2ϕθ̇ 1̇ sin(2θ1 ) S22 − mp l 2ϕθ̇ 2̇ S12 sin(2θ2) + mp xlS
̇ ̇ 2 + mp xlθ̇ 2̇ C2
= Fϕ − Ffϕ
(13)

where Fϕ is the jib slew control torque, Ffϕ denotes the friction torque.
Similarly, by calculating the partial derivatives of V in regard to x,
ẋ, and l, l ,̇ respectively, Eqs. (14) and (15) can be accordingly obtained

3
M. Zhang, et al. Automation in Construction 119 (2020) 103342

¨ 2 + (Mt + mp ) x¨ + mp l¨S1 C2 + mp lθ¨1 C1 C2 − mp lθ¨2 S1 S2

− mp lϕS 1 T
VE (t ) = q ̇ M (q) q ̇ + mp gl (1 − C1 C2)
2 2 2 2 2 (22)
− (Mt + mp ) xϕ ̇ − 2mp lθ1̇ θ2̇ C1 S2 − mp lC2 (S1 (θ1̇ + θ2̇ + ϕ ̇ ) + 2ϕθ̇ 2̇ )
where M(q) and q̇ represent the inertia matrix, and the state vector,
− mp l (̇ ϕṠ 2 − θ1̇ C1 C2 + θ2̇ S1 S2) = Fx − Ffx
respectively. Detailed expressions of the matrix and the vector are given
(14) as follows:
2 2 2̇
mp l¨ + mp ϕxS
¨ 2 + mp xS
¨ 1 C2 − mp lθ1̇ C22 − mp lθ2̇ − mp ϕ ⎡ m11 − mp lS2 mp xS2 m14 m15 ⎤
⎢− mp lS2 Mt + mp mp S1 C2 mp lC1 C2 − mp lS1 S2 ⎥
(lS22 + lS12 C22 − xS1 C2) ⎢ ⎥
̇ ̇ 2 + 2mp lϕ ̇ (θ1̇ C1 S2 C2 − θ2̇ S1) − mp gC1 C2 = Fl − dl l ̇ M (q) = ⎢ mp S2 mp S1 C2 mp 0 0 ⎥
+ 2mp ϕxS (15) ⎢ m ⎥
m lC C 0 m l 2C 2 0
⎢ 14 p 1 2 p 2 ⎥
Additionally, the following two equations can be derived by ana- ⎢ m15 − mp lS1 S2 0 0 mp l 2 ⎥
⎣ ⎦
lyzing the partial derivatives of V with respect to θ1, θ1̇ , and θ2, θ2̇ ,
respectively: q = [ ϕ x l θ1 θ2 ]T
2̇
− mp l 2ϕC ¨ 1 C2 + mp l 2θ¨1 C22 − mp l (lS1 C2 + x ) ϕ C1 C2
¨ 1 S2 C2 + mp lxC with
− mp xlĊ 1 C2 − 2mp l C2 (θ1̇ S2 + ϕC
̇ 2 ̇ 1 C2 ) θ2̇ + mp glS1 C2 = −df 1 l 2C22 θ1̇
m11 ≜ (Mt + mp ) x 2 + mp l 2 (S22 + S12 C22) + 2mp lxS1 C2 + J ,
(16) m14 ≜ − mp l 2C1 S2 C2, m15 ≜ mp l (xC2 + lS1)

¨ 1 S2 + mp l 2θ¨2 + 2mp lxϕC

mp l (lS1 + xC2 ) ϕ¨ − mp lxS ̇ ̇ 2 + mp l Taking the time derivative of Eq. (22), it can be obtained that
2
S2 (xS1 − lC12 C2 ) ϕ ̇
VĖ (t )
2
+ 2mp l 2ϕθ̇ 1̇ C1 C2 + mp l 2θ1̇ S2 C2 + mp l (̇ xS
2 ̇ 2) + mp glC1 S2
̇ 1 S2 − ϕxC 1 T ̇
= qṪ M (q) q¨ + ̇ − C1 C2) + mp glθ1̇ S1 C2 + mp glθ2̇
q ̇ M (q) q ̇ + mp gl (1
= −df 2 l 2θ2̇ 2
C1 S2
(17)
where Fx and Fl stand for the control forces imposed on the trolley and 2
cable, respectively, Ffx and dl l ̇ represent the friction forces in directions = (Fϕ − Ffϕ ) ϕ ̇ + (Fx − Ffx ) x ̇ + (Fl − dl l ̇ + mp g ) l ̇ − df 1 l 2C22 θ1̇ − df 2 l 2
2
X and Z, df1 and df2 refer to the coefficients of air resistance. The friction θ2̇
torque Ffϕ and the friction force Ffx have the following forms [30–33]:
(23)
ϕ̇ indicating that the tower crane system with Fϕ, Fx, and Fl as the inputs,
Ffϕ = Ff 0ϕ tanh ⎛⎜ ⎞⎟ + kfϕ ∣ϕ∣̇ ϕ ̇
ε (18) ϕ ,̇ ẋ, and l ̇ as the outputs, is passive. It is shown that the system energy
⎝ ϕ⎠
VE(t) can merely be damped out by the actuated motion ϕ ,̇ ẋ, and l .̇
ẋ In consideration of Eqs. (18) and (19), Eq. (23) can be rewritten as:
Ffx = Ff 0x tanh ⎛ ⎞ + kfϕ ∣x ∣̇ x ̇
⎜ ⎟

⎝ εx ⎠ (19) VĖ (t )
Without loss of generality, the following assumption is reasonable. = (Fϕ − ηϕT ωϕ ) ϕ ̇ + (Fx − ηxT ωx ) x ̇ + (Fl − ηl ωl + mp g ) l ̇ − df 1 l 2
2 2
Assumption 1. For cranes working in practice, the payload is always C22 θ1̇ − df 2 l 2θ2̇ (24)
beneath the trolley and the jib, therefore, the payload swing angles
where the vectors are defined of the following forms:
always satisfy [34–37]:
T
π π π π ⎡ ϕ̇ ⎤
−
2
≤ θ1 ≤ , − ≤ θ2 ≤
2 2 2 (20) ηϕ ≜ ⎢ tanh ⎛⎜ ⎟⎞ ∣ϕ∣̇ ϕ ̇⎥ , ωϕ ≜ [ Ff 0ϕ kfϕ ]T
⎝ εϕ ⎠ (25)
⎣ ⎦
T
ηx ≜ ⎡ ⎛ ẋ ⎞ ⎤ F k T
⎢ tanh εx ∣x ∣̇ x ̇ ⎥ , ωx ≜ [ f 0x fx ]
⎜ ⎟

3. Control method development and closed-loop stability analysis ⎣ ⎝ ⎠ ⎦ (26)

ηl ≜ l ,̇ ωl ≜ dl (27)
In this section, an adaptive control method with payload swing re-
duction is designed for five DOFs tower crane systems, which achieves To facilitate the following analysis and control law development, we
simultaneous accurate positioning performance and effective payload define the following three error signals as:
swing suppression and elimination.
eϕ = ϕ − pdϕ , ex = x − pdx , el = l − pdl (28)

3.1. Adaptive control method development In view of the structure of Eq. (24), the adaptive control law is
proposed as follows:
The main control objective is to drive the trolley and the jib to their ϕ
Fϕ = −kpϕ eϕ − kdϕ ϕ ̇ + ηϕT ω (29)
desired position pdx and angle pdϕ while changing the cable length from
the initial value to the target value pdl. At the same time, the payload x
Fx = −kpx ex − kdx x ̇ + ηxT ω (30)
swing should be suppressed and eliminated during the transportation,
in the sense that Fl = −kpl el − kdl l ̇ − mp g + ηl ω
l (31)
lim x (t ) = pdx , lim ϕ (t ) = pdϕ , lim l (t ) = pdl , lim θ1 (t ) = 0, lim θ2 (t ) ϕ ,
with kpϕ, kdϕ, kpx, kdx, kpl, and kdl representing positive control gains, ω
t →∞ t →∞ t →∞ t →∞ t →∞
l being the estimation of ωϕ, ωx, and ωl, respectively. The
x , and ω
ω
=0 (21)
update law for ω ϕ , ω
x , and ωl are defined as
The tower crane system energy including the kinematic energy and
ϕ̇ = −Γϕ ηϕT ϕ,̇ ω
ω ̇x = −Γx ηxT x ̇, ω
l̇ = −Γl ηl l ̇ (32)
the potential energy is denoted by

4
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Fig. 2. Block diagram for the overall control system.

Table 1
Control gains.
Control methods Control gains

PD control method kpϕ = 15, kdϕ = 20, kpx = 8, kdx = 10, kpl = 20, kdl = 5
Proposed control method kpϕ = 15, kdϕ = 25, kpx = 9, kdx = 13.2, kpl = 15, kdl = 5, kϕ = 200, kx = 200, kl = 200

where Γϕ ∈ R2×2, Γx ∈ R2×2 denote positive definite diagonal matrixes, 1 T

Vall (t ) =
1 1
q ̇ M (q) q ̇ + mp gl (1 − C1 C2) + kpϕ eϕ2 + kpx ex2
Γl ∈ R1 represents a positive control gain. 2 2 2
As can be seen from Eqs. (29)–(31), the payload swing-related in- 1 1 ∼T −1 ∼ 1 ∼T −1 ∼ 1 ∼ −1 ∼
+ kpl el2 + ω ϕ ϕ Γ ωϕ + ω Γ
x x ω x + ω Γ
l l ωl
formation is not included in the control torque/forces. To solve this 2 2 2 2 (37)
problem and attenuate the payload swing, the final control method is where ω∼,ω ∼ , and ω ∼ stand for the estimation error vectors of ω , ω ,
ϕ x l ϕ x
revised as and ωl, respectively, which are defined as
2 2
ϕ − k ϕ (θ1̇ + θ2̇ ) ϕ ̇
Fϕ = −kpϕ eϕ − kdϕ ϕ ̇ + ηϕT ω ∼ =ω
ω ϕ − ωϕ , ω ∼ =ω ∼ =ω
x − ωx , ω l − ωl
(33) ϕ x l (38)

2 2 Differentiating Eq. (37) with respect to time, and inserting the re-
x − k x (θ1̇ + θ2̇ ) x ̇
Fx = −kpx ex − kdx x ̇ + ηxT ω (34) sults of Eqs. (24) and (33)–(35), one yields
2 2 ̇ (t )
l − kl (θ1̇ + θ2̇ ) l ̇
Fl = −kpl el − kdl l ̇ − mp g + ηl ω (35) Vall
2 2 2 2 2 2 2
= −kdϕ ϕ ̇ − kdx x ̇2 − kdl l ̇ − k ϕ (θ1̇ + θ2̇ ) ϕ ̇ − k x (θ1̇ + θ2̇ ) x ̇2 − kl
where kϕ, kx, kl refer to positive control gains. The overall control
2 2 2 2 2
system is shown in Fig. 2. (θ1̇ + θ2̇ ) l ̇ − df 1 l 2C22 θ1̇ − df 2 l 2θ2̇
≤0
3.2. Stability analysis (39)
implying that the closed-loop system is Lyapunov stable [21], i.e.
Theorem 1. The proposed adaptive sway reduction control method (Eqs.
∼,ω
Vall (t ) ∈ L∞ ⇒ ϕ,̇ x ̇, l ,̇ θ1̇ , θ2̇ , eϕ, ex , el , ω ∼,ω∼, ω
(33)–(35)), along with the update law ((Eq. (32)), can guarantee the trolley ϕ x l ϕ , ω
x , ω
l ∈ L∞
and the jib reach their desired position and angle, and damp out the payload (40)
swing simultaneously, in the sense that
To prove Eq. (36), let S as
T
lim [ ϕ x l θ1 θ2 ϕ ̇ x ̇ l ̇ θ1̇ θ2̇ ] ̇ (t ) = 0}
S ≜ { (q, q)̇ | Vall
t →∞ (41)
T
= [ pdϕ pdx pdl 0 0 0 0 0 0 0 ] (36) Then, define M as the largest invariant set included in S. It is easy to
conclude that in M
ϕ ̇ = 0, x ̇ = 0, l ̇ = 0 (42)
Proof: By using energy shaping method [38–39], the following
Lyapunov function candidate is constructed as It follows from Eq. (42) that

5
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Fig. 3. Simulation results for simulation group 1. (a) Jib slew angle, trolley translation displacement, cable length, payload swing angles. (b) Control torque/forces in
directions Y, X, Z.

ϕ¨ = 0, x¨ = 0, l¨ = 0, eϕ = λ ϕ , ex = λ x , el = λl , Ffϕ = 0, Ffx = 0, Fϕ = −kpϕ λ ϕ , Fx = −kpx λ x , Fl = −kpl λl − mp g (44)

dl l ̇ = 0, ϕ = λ ϕ + pdϕ , x = λ x + pdx , l = λl + pdl (43) which implies that Fϕ, Fx, and Fl keep constant in M.
Substituting the results of Eqs. (42)–(44) into Eq. (13), one has
where λϕ, λx, and λl are constants to be determined.
It can be obtained from Eqs. (42)–(43) that

6
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Table 2
Performance indices of simulation group 1.
Control methods θ1max θ2max θ1res θ2res Fϕmax Fxmax Flmax

PD control method 20.27 deg 6.68 deg 2.69 deg 3.78 deg 15.85 N 15.13 N 29.65 N
Proposed control method 5.64 deg 2.45 deg 0.5 deg 0.1 deg 15.52 N 14.45 N 24.83 N
(↓72.17%) (↓63.32%) (↓81.41%) (↓97.35%)

2 g
− lθ¨1 C1 S2 C2 + θ¨2 ((λ x + pdx ) C2 + lS1) + lθ1̇ S1 S2 C2 θ¨1 C2 − 2S2 θ1̇ θ2̇ + S1 = 0
l (56)
2 −kpϕ λ ϕ
+ 2lθ1̇ θ2̇ C1 S22 − (λ x + pdx ) θ2̇ S2 = 2 g
mp l (45) θ¨2 + θ1̇ S2 C2 + C1 S2 = 0
l (57)
Applying the fact that S22 = 1 − C22, lθ1̇ θ2̇ C1 S22 can be expanded
It follows from λx = 0, Eqs. (56)–(57), Eq. (52) can be rewritten as
into the following manner:
g 2 2
lθ1̇ θ2̇ C1 S22 = lθ1̇ θ2̇ C1 − lθ1̇ θ2̇ C1 C22 (46) S1 C2 ⎛− C1 C2 − θ1̇ C2 − θ2̇ ⎞ = 0
⎝ l ⎠ (58)
By using Eq. (45) and after some arrangements, Eq. (45) can be
According to Assumption 1, C1, C2 > 0, therefore, it is obtained
rewritten as
from Eq. (58) that
2
− lθ¨1 C1 S2 C2 + lθ1̇ S1 S2 C2 + lθ1̇ θ2̇ C1 S22 − lθ1̇ θ2̇ C1 C22 + (λ x + pdx ) C2 θ¨2 S1 = 0 ⇒ θ1 = 0, θ1̇ = 0, θ¨1 = 0 (59)
2 −kpϕ λ ϕ
− (λ x + pdx ) θ2̇ S2 + lS1 θ¨2 + lθ1̇ θ2̇ C1 = Substituting Eq. (59) into Eq. (50), it is derived that
mp l
d −kpϕ λ ϕ (λ x + pdx ) C2 θ2̇ = λ1 (60)
⇔ (−lθ1̇ C1 S2 C2 + (λ x + pdx ) C2 θ2̇ + lS1 θ2̇ ) =
dt mp l
Integrating Eq. (60) with respect to time, one has
(47)
(λ x + pdx ) S2 = λ1 t + λ3 (61)
Integrating Eq. (47) with respect to time, it can be easily calculated
that If λ1 ≠ 0, as t → ∞, S2 → ∞, which contradicts with ∣S2 ∣ < 1,
therefore, λ1 = 0. From λ1 = 0 and Eq. (59), we have
−kpϕ λ ϕ
− lθ1̇ C1 S2 C2 + (λ x + pdx ) C2 θ2̇ + lS1 θ2̇ = t + λ1
mp l (48) θ2̇ = 0 → θ¨2 = 0 (62)

with λ1 being a constant to be determined. As can be seen from Eq. (48), It follows from Eqs. (57), (59), and (62) that
if λϕ ≠ 0, then as t → ∞, g
C1 S2 = 0 → S2 = 0 → θ2 = 0
l (63)
∣ − lθ1̇ C1 S2 C2 + (λ x + pdx ) C2 θ2̇ + lS1 θ2̇ ∣ → ∞ (49)
Substituting Eqs. (42)–(43), (59), (62)–(63) into Eq. (15), it can be
which contradicts with the conclusions of Eq. (40). Therefore, λϕ = 0. concluded that
Eq. (48) can be rewritten as
− kpl λl = 0 ⇒ λl = 0, el = 0, l = pdl (64)
− lθ1̇ C1 S2 C2 + (λ x + pdx ) C2 θ2̇ + lS1 θ2̇ = λ1 (50)
By gathering the conclusions of Eqs. (42)–(43), (58), and (61)–(63),
By using λϕ = 0, we can obtain that we can conclude that the largest invariant set M contains only the
Fϕ = 0, eϕ = 0, ϕ = pdϕ (51) equilibrium point
T T
Similar with Eq. (45), applying the conclusions of Eqs. (42)–(43), [ ϕ x l θ1 θ2 ϕ ̇ x ̇ l ̇ θ1̇ θ2̇ ] = [ pdϕ pdx pdl 0 0 0 0 0 0 0 ]
Eq. (14) can be reduced to
therefore, by using LaSalle's invariance theorem [29], theorem 1 is
2 2 −kpx λ x proven.
θ¨1 C1 C2 − θ1̇ S1 C2 − θ1̇ θ2̇ C1 S2 − θ¨2 S1 S2 − θ1̇ θ2̇ C1 S2 − C2 S1 θ2̇ =
mp l
d −kpx λ x 4. Simulation results
⇔ (θ1̇ C1 C2 − θ2̇ S1 S2) =
dt mp l
In this section, to evaluate the control performance of the designed
(52) adaptive control method with payload swing suppression, the tower
The integral of Eq. (52) with respect to time is obtained as crane system with varying cable lengths is numerically simulated by
using Eqs. (13)–(19) in conjunction with the control law described in
−kpx λ x
θ1̇ C1 C2 − θ2̇ S1 S2 = t + λ2 Eqs. (32)–(35). In this study, the tower crane system parameters are set
mp l (53) as Mt = 3.5 kg, mp = 1 kg, J = 6.8 kg ⋅ m2, g = 9.8 m/s2, Ff0ϕ = 4.4,
where λ2 stands for a constant yet to be determined. if λx ≠ 0, then as kfϕ = − 0.5, Ff0x = 6.8, kfx = − 1.2, dl = 3.5, εϕ = εx = 0.01,
t → ∞, df1 = 0.1, df2 = 0.15, respectively. The initial values of the jib slew
angle, trolley position, and cable length are ϕ(0) = 0 deg, x(0) = 0 m,
∣θ1̇ C1 C2 − θ2̇ S1 S2 ∣ → ∞ (54) and l(0) = 1.5 m, respectively. The desired jib slew angle, trolley po-
which contradicts with the conclusions of Eq. (40). Therefore, λx = 0. sition, and cable length are taken as pdϕ = 45 deg, pdx = 1 m, and
Then, it can be obtained from λx = 0, Eqs. (43)–(44) that pdl = 0.5 m, respectively. The initial values of ω ϕ , ω
x , ω
l are set as
ϕ (0) = ω
ω x (0) = diag (0, 0) , ω
l = 0 .
θ1̇ C1 C2 − θ2̇ S1 S2 = λ2 , Fx = 0, x = pdx (55)
Remark 2. The crane operation process can be roughly divided into the
By substituting the conclusions of Eqs. (42)–(43) into Eqs. following three steps: (1) hoisting the payload to a safety height, (2)
(16)–(17), and after some reductions, one has moving the payload, hooked up to the trolley, to the position directly

7
M. Zhang, et al. Automation in Construction 119 (2020) 103342

(caption on next page)

8
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Fig. 4. Simulation results for proposed control method with respect to case 1. (a) Jib slew angle, trolley translation displacement, cable length, payload swing angles.
(b) Control torque/forces in directions Y, X, Z. (c) Parameters estimation.

over the desired location, and (3) lowering the payload to the desired the maximum payload swing amplitudes after the jib, the trolley,
location. Existing control methods for tower cranes are designed only and the cable stop, respectively.
for the second step. To achieve fast precise trolley positioning and 3) Maximum control torque/force amplitudes: Fϕmax, Fxmax, and Flmax,
payload hoisting/lowering as well as rapid payload swing elimination, which are defined as the Fϕmax ≜ maxt{| Fϕ(t)| }, Fxmax ≜ maxt{| Fx
we attempt to design an effective control strategy for the first, second, (t)| }, and Flmax ≜ maxt{| Fl(t)| }, respectively.
and third steps in this paper. Nonetheless, when there exists vertical
payload motion, the cable length turns from a parameter into a Fig. 3 displays the time evaluation of the jib slew angel ϕ, trolley
variable, which becomes a system state should be controlled. Existing translation displacement x, cable length l, payload swing angles θ1, θ2,
control methods can't control the cable length; hence, the cable length control torque/forces Fϕ, Fx, Fl in directions Y, X, and Z with the PD
can't reach its desired length. Tower crane systems with fixed cable control method and the proposed control method. The quantified re-
length and varying cable lengths are two different systems. Therefore, sults are shown in Table 2. It is noted that the designed control law
we didn't compare our control method with the conventional approach shows better payload swing suppression effects and less maximum
without consideration of varying cable lengths. To verify the superior control torque/force amplitudes compared with the PD control method.
control performance of the proposed control method, we compare our More precisely, the payload swing angles are suppressed and eliminated
control method with the traditional PD method. within a smaller range by the proposed adaptive control scheme with
payload swing suppression (θ1max: 5.64 deg, θ2max: 2.45 deg, θ1res:
Remark 3. It should be noticed that the PD control method is the most
0.5 deg, θ2res: 0.1 deg) than by the PD control law (θ1max: 20.27 deg,
widely used one in industrial cranes because it has a simple structure,
θ2max: 6.68 deg, θ1res: 2.69 deg, θ2res: 3.78 deg). Compared to the PD
hence, it is easy to implement [40–41]. As will be seen in the simulation
control method, θ1max, θ2max, θ1res, and θ2res of the designed control
results, the positioning performance of the PD control method is good
strategy are reduced about 72.17%, 63.32%, 81.41%, and 97.35%, re-
while the anti-swing performance is not satisfactory [40]. In addition,
spectively. In addition to this, the payload keeps swing back and forth
inevitable external disturbances and parametric uncertainties may
after the jib, the trolley, and the cable stop for the PD control method,
deteriorate the control performance and even lead to instability. To
while the payload is much more steady for the designed control algo-
solve these problems, we design the adaptive reduction control method
rithm. Moreover, the maximum control torque/force amplitudes of the
in this paper.
proposed control scheme (Fϕmax: 15.52 N, Fxmax: 14.45 N, Flmax:
To verify the superior control performance of the proposed control 24.83 N) are less than those of the PD control method (Fϕmax: 15.85 N,
law, two groups of numerical simulation tests are carried out. More Fxmax: 15.13 N, Flmax: 29.65 N). These results have evidently proven that
precisely, we compare the designed control law with the PD control the designed control method shows superior control performance.
method [29] in the first group. Subsequently, the robustness of the
proposed control method with respect to different unfavorable condi-
tions is confirmed. 4.2. Simulation group 2

4.1. Simulation group 1 The robustness of the designed control law is further verified in this
group. Towards this end, the following two cases of ‘perturbations’ are
In this group, to better show the control performance of the de- considered.
signed control strategy, the well-known PD control method [29] is se-
Case 1. Uncertain friction-related parameters. The friction-related
lected for comparison. The PD control method is of the following form:
parameters are changed to Ff0ϕ = 5.2, kfϕ = − 1, Ff0ϕ = 5.4, and
Fϕ = −kpϕ eϕ − kdϕ ϕ ̇ + Ffϕ (65) kfϕ = − 1.5.

Fx = −kpx ex − kdx x ̇ + Ffx (66) Case 2. External disturbances. In order to emulate external
disturbances, such as winds, different external perturbations are
Fl = −kpl el − kdl l ̇ − mp g + dl l ̇ (67) added to the payload swing angles. More precisely, impulsive
disturbances with amplitudes of 3 and 4 deg are induced to θ1
where kpϕ, kdϕ, kpx, kdx, kpl, and kdl stand for positive control gains. between 2 and 3 s, and between 6 and 7 s, respectively; impulsive
To be fair, the friction torque/forces are assumed to be known for disturbances with amplitudes of 2.5 and 3 deg are added to θ2 between
the PD control method and the proposed control strategy. Then, the 3 and 4 s, and between 7 and 8 s, respectively.
designed control method can be written as
2 2
In this group, the control gains of the designed control method are
Fϕ = −kpϕ eϕ − kdϕ ϕ ̇ + Ffϕ − k ϕ (θ1̇ + θ2̇ ) ϕ ̇ (68) tuned as kpϕ = 28, kdϕ = 0.8, kpx = 24.8, kdx = 8, kpl = 15, kdl = 5,
2 2 kϕ = 200, kx = 200, kl = 200, Γϕ = diag (5,5), Γx = diag (2,2), Γl = 2.
Fx = −kpx ex − kdx x ̇ + Ffx − k x (θ1̇ + θ2̇ ) x ̇ (69) Figs. 4–5 show the simulation results of the jib slew angle ϕ, trolley
2 2 translation displacement x, cable length l, payload swing angles θ1, θ2,
Fl = −kpl el − kdl l ̇ − mp g + dl l ̇ − kl (θ1̇ + θ2̇ ) l ̇ (70) control torque/forces Fϕ, Fx, Fl in directions Y, X, and Z, parameters
After careful tuning, the control gains of the proposed control law estimation with the proposed control method for the above three cases.
and the PD control method are provided in Table 1. The following By comparing Fig. 4 with Fig. 3, it is observed that the overall control
performance indices are introduced to better exhibit the simulation performance of the suggested control method, including positioning
results. and swing suppression and elimination, is not affected much by un-
certain friction-related parameters, which implies that it is adaptive to
1) Maximum payload swing amplitudes: θ1max and θ2max, which are system parametric changes/uncertainties. From Fig. 5, it is shown that
defined as θ1max ≜ maxt{| θ1(t)| }, and θ2max ≜ maxt{| θ2(t)| }, re- the unexpected external disturbances are attenuated and damped out
spectively. by the designed control strategy; as can be seen from the curves of the
2) Residual payload swing angles: θ1res and θ2res, which are defined as jib slew angle and the trolley displacement, it is observed that the jib
and the trolley move back to eliminate the external disturbances. The

9
M. Zhang, et al. Automation in Construction 119 (2020) 103342

(caption on next page)

10
M. Zhang, et al. Automation in Construction 119 (2020) 103342

Fig. 5. Simulation results for proposed control law with respect to case 3. (a) Jib slew angle, trolley translation displacement, cable length, payload swing angles. (b)
Control torque/forces in directions Y, X, Z. (c) Parameters estimation.

evolution of the parameters estimation is illustrated in Figs. 4c–5c, pendulums with periodic disturbances, IEEE Trans. Cybern. 50 (5) (2020)
which clearly shows that all the estimations converge in about 3 s. It 1867–1876, https://doi.org/10.1109/TCYB.2018.2884707.
[9] J. Vaughan, D. Kim, W. Singhose, Control of tower cranes with double-pendulum
can be concluded from the above analysis, the designed control method payload dynamics, IEEE Trans. Control Syst. Technol. 18 (6) (2010) 1345–1358,
shows strong robustness. https://doi.org/10.1109/TCST.2010.2040178.
[10] J. Vaughan, D. Kim, W. Singhose, Comparison of robust input shapers, J. Sound Vib.
315 (4/5) (2008) 797–815, https://doi.org/10.1016/j.jsv.2008.02.032.
5. Conclusion [11] X. Zhao, X. Wang, L. Ma, G. Zong, Fuzzy-approximation-based asymptotic tracking
control for a class of uncertain switched nonlinear systems, IEEE Trans. Fuzzy Syst.
To tackle the control problem of tower cranes when the cable length 28 (4) (2020) 632–644, https://doi.org/10.1109/TFUZZ.2019.2912138.
[12] H. Pan, H. Li, W. Sun, Z. Wang, Adaptive fault-tolerant compensation control and its
is varying, the dynamic model of tower cranes with varying cable application to nonlinear suspension systems, IEEE Trans. Syst. Man Cybern. Syst. 50
lengths is established in this paper. Then, an adaptive sway reduction (5) (2020) 1867–1876, https://doi.org/10.1109/TSMC.2017.2785796.
control method is proposed based on the original nonlinear dynamic [13] H. Pan, X. Jing, W. Sun, Z. Li, Analysis and design of a bioinspired vibration sensor
system in noisy environment, IEEE/ASME Trans. Mechatronics 23 (2) (2018)
model, where parametric uncertainties are present. Compared with
845–855, https://doi.org/10.1109/TMECH.2018.2803284.
existing control methods for tower cranes, the designed control method [14] Z. Wu, Q. Li, W. Wu, M. Zhao, Crowdsourcing model for energy efficiency retrofit
has both theoretical and practical significance. From a theoretical point and mixed-integer equilibrium analysis, IEEE Transactions on Industrial Informatics
of view, the stability of the equilibrium point of the closed-loop system 16 (7) (2020) 4512–4524, https://doi.org/10.1109/TII.2019.2944627.
[15] M. Zhang, Y. Zhang, B. Ji, C. Ma, X. Cheng, Modeling and energy-based sway re-
is proved rigorously without any linearizing operations to the original duction control for tower crane systems with double-pendulum and spherical-
dynamics. Additionally, the designed control method does not need the pendulum effects, Meas. Control 53 (1–2) (2020) 141–150, https://doi.org/10.
exact values of the system model. From a practical point of view, the 1177/0020294019877492.
[16] H.M. Omar, A.H. Nayfeh, Gain scheduling feedback control of tower cranes with
adaptive sway reduction control method is designed for tower cranes friction compensation, J. Vib. Control. 10 (2) (2004) 269–289, https://doi.org/10.
with varying cable lengths, which is much closer to actual tower crane 1177/1077546304035610.
systems. In addition, simulation results are illustrated to show that the [17] G. Lee, H.H. Kim, C.J. Lee, S.I. Ham, S.H. Yun, H. Cho, B.K. Kim, G.T. Kim, K. Kim, A
laser-technology-based lifting-path tracking system for a robotic tower crane,
designed control method is more beneficial in anti-swing performance, Autom. Constr. 8 (7) (2009) 865–874, https://doi.org/10.1016/j.autcon.2009.03.
with an improvement of more than 63.22%, compared with the PD 011.
control method. [18] M. Böck, A. Kugi, Real-time nonlinear model predictive path following control of a
laboratory tower crane, IEEE Trans. Control Syst. Technol. 22 (4) (2014)
1461–1473, https://doi.org/10.1109/TCST.2013.2280464.
Declaration of competing interest [19] F.N. Koumboulis, N.D. Kouvakas, G.L. Giannaris, D. Vouyioukas, Independent
motion control of a tower crane through wireless sensor and actuator networks, ISA
Trans. 60 (2016) 312–320, https://doi.org/10.1016/j.isatra.2015.11.011.
Compliance with ethical standards, the authors declare that they [20] Š. Ileš, J. Matuško, F. Kolonic, Sequential distributed predictive control of a 3D
have no conflict of interest concerning the publication of this manu- tower crane, Control. Eng. Pract. 79 (2018) 22–35, https://doi.org/10.1016/j.
script. conengprac.2018.07.001.
[21] I.G. Carmona, J. Collado, Control of a two wired hammerhead tower crane,
Nonlinear Dyn. 84 (4) (2016) 2137–2148, https://doi.org/10.1007/s11071-016-
Acknowledgements 2634-3.
[22] J.L. Escalona, G. Orzechowski, A.M. Mikkola, Flexible multibody modeling of re-
eving systems including transverse vibrations, Multibody Syst. Dyn. 44 (2018)
This work is supported by the National Key R&D Program of China 107–133, https://doi.org/10.1007/s11044-018-9619-6.
under Grant No. 2018YFB1305400, the Key Research and Development [23] N. Sun, Y. Fang, H. Chen, B. Lu, Y. Fu, Slew/translation positioning and swing
(Special Public-Funded Projects) of Shandong Province under Grant No. suppression for 4-DOF tower cranes with parametric uncertainties: design and
hardware experimentation, IEEE Trans. Ind. Electron. 63 (10) (2016) 6407–6418,
2019GGX104058, the National Natural Science Foundation for Young https://doi.org/10.1109/TIE.2016.2587249.
Scientists of China under Grant No. 61903155, and the Natural Science [24] H. Chen, Y. Fang, N. Sun, An adaptive tracking control method with swing sup-
Foundation of Shandong Province under Grant No. ZR2019QEE019. pression for 4-DOF tower crane systems, Mech. Syst. Signal Process. 123 (2019)
426–442, https://doi.org/10.1016/j.ymssp.2018.11.018.
[25] Y. Qian, Y. Fang, B. Lu, Adaptive robust tracking control for an offshore ship-
References mounted crane subject to unmatched sea wave disturbances, Mech. Syst. Signal
Process. 114 (2019) 556–570, https://doi.org/10.1016/j.ymssp.2018.05.009.
[26] T.S. Wu, M. Karkoub, W.S. Yu, C.T. Chen, M.G. Her, K.W. Wu, Anti-sway tracking
[1] Z. Zhang, L. Li, Y. Wu, Disturbance-observer-based antiswing control of under-
control of tower cranes with delayed uncertainty using a robust adaptive fuzzy
actuated crane systems via terminal sliding mode, IET Control Theory Appl 12 (18)
control, Fuzzy Sets Syst. 290 (2016) 118–137, https://doi.org/10.1016/j.fss.2015.
(2018) 2588–2594, https://doi.org/10.1049/iet-cta.2018.5344.
01.010.
[2] M. Zhang, Y. Zhang, H. Chen, X. Cheng, Model-independent PD-SMC method with
[27] T.S. Wu, M. Karkoub, H. Wang, H.S. Chen, T.H. Chen, Robust tracking control of
payload swing suppression for 3D overhead crane systems, Mech. Syst. Signal
MIMO underactuated nonlinear systems with dead-zone band and delayed un-
Process. 129 (2019) 381–393, https://doi.org/10.1016/j.ymssp.2019.04.046.
certainty using an adaptive fuzzy control, IEEE Trans. Fuzzy Syst. 25 (4) (2017)
[3] Q. Liu, H. Zhang, J. Leng, X. Chen, Digital twin-driven rapid individualised de-
905–918, https://doi.org/10.1109/TFUZZ.2016.2586970.
signing of automated flow-shop manufacturing system, Int. J. Prod. Res. 57 (12)
[28] S.C. Duong, E. Uezato, H. Kinjo, T. Yamamoto, A hybrid evolutionary algorithm for
(2019) 3903–3919, https://doi.org/10.1080/00207543.2018.1471243.
recurrent neural network control of a three-dimensional tower crane, Autom.
[4] J. Leng, H. Zhang, D. Yan, Q. Liu, X. Chen, D. Zhang, Digital twin-driven manu-
Constr. 23 (2012) 55–63, https://doi.org/10.1016/j.autcon.2011.12.005.
facturing cyber-physical system for parallel controlling of smart workshop, J.
[29] H.K. Khalil, Nonlinear Systems, 3rd ed., Prentice-Hall, Englewood Cliffs, NJ, USA,
Ambient. Intell. Humaniz. Comput. 10 (3) (2019) 1155–1166, https://doi.org/10.
2002 (ISBN-13: 978-0130673893).
1007/s12652-018-0881-5.
[30] B. Lu, Y. Fang, N. Sun, X. Wang, Antiswing control of offshore boom cranes with
[5] H. Chen, N. Sun, Nonlinear control of underactuated systems subject to both ac-
ship roll disturbances, IEEE Trans. Control Syst. Technol. 26 (2) (2018) 740–747,
tuated and unactuated state constraints with experimental verification, IEEE Trans.
https://doi.org/10.1109/TCST.2017.2679060.
Ind. Electron. 67 (9) (2020) 7702–7714, https://doi.org/10.1109/TIE.2019.
[31] H. Pan, X. Jing, W. Sun, H. Gao, A bioinspired dynamics-based adaptive tracking
2946541.
control for nonlinear suspension systems, IEEE Trans. Control Syst. Technol. 26 (3)
[6] Y. Sun, H. Qiang, J. Xu, G. Lin, Internet of things-based online condition monitor
(2018) 903–914 (doi:10.11-09/TCST.2017.2699158).
and improved adaptive fuzzy control for a medium-low-speed maglev train system,
[32] D. Qian, H. Ding, S.G. Lee, H. Bae, Suppression of chaotic behaviors in a complex
IEEE T. Ind. Inform. 16 (4) (2020) 2629–2639, https://doi.org/10.1109/TII.2019.
biological system by disturbance observer-based derivative-integral terminal sliding
2938145.
mode, IEEE/CAA J. Automatica Sin. 7 (1) (2020) 126–135, https://doi.org/10.
[7] X. Liang, Y. Fang, N. Sun, H. Lin, A novel energy coupling based hierarchical control
1109/JAS.2019.1911834.
approach for unmanned quadrotor transportation systems, IEEE/ASME
[33] H. Ouyang, X. Xu, G. Zhang, Energy-shaping-based nonlinear controller design for
Transactions on Mechatronics 24 (1) (2019) 248–259, https://doi.org/10.1109/
rotary cranes with double-pendulum effect considering actuator saturation, Autom.
TMECH.2019.2891083.
Constr. 111 (2020) 103054, https://doi.org/10.1016/j.autcon.2019.103054.
[8] W. Sun, S. Su, J. Xia, Y. Wu, Adaptive tracking control of wheeled inverted

11
M. Zhang, et al. Automation in Construction 119 (2020) 103342

[34] M. Zhang, Y. Zhang, H. Ouyang, C. Ma, X. Cheng, Adaptive integral sliding mode [38] N. Sun, Y. Fang, X. Wu, An enhanced coupling nonlinear control method for bridge
control with payload sway reduction for 4-DOF tower crane systems, Nonlinear cranes, IET Control Theory A. 8 (13) (2014) 1215–1223, https://doi.org/10.1049/
Dyn. 99 (2020) 2727–2741, https://doi.org/10.1007/s11071-020-05471-3. iet-cta.2013.0584.
[35] L. Tuan, H.M. Cuong, P. Van Trieu, L.C. Nho, L. Anh, Adaptive neural network [39] N. Sun, Y. Fang, X. Zhang, Energy coupling output feedback control of 4-DOF un-
sliding mode control of shipboard container cranes considering actuator backlash, deractuated cranes with saturated inputs, Automatica 49 (2013) 1318–1325,
Mech. Syst. Signal Process. 112 (2018) 233–250, https://doi.org/10.1016/j.ymssp. https://doi.org/10.1016/j.automatica.2013.01.039.
2018.04.030. [40] Y. Fang, W.E. Dixon, D.M. Dawson, E. Zergeroglu, Nonlinear coupling control laws
[36] N. Sun, D. Liang, Y. Wu, Y. Chen, Y. Qin, Y. Fang, Adaptive control for pneumatic for an underactuated overhead crane system, IEEE/ASME Transactions on
artificial muscle systems with parametric uncertainties and unidirectional input Mechatronics 8 (3) (2003) 418–423, https://doi.org/10.1109/TMECH.2003.
constraints, IEEE Trans. Industrial Informatics 16 (2) (2020) 969–979, https://doi. 816822.
org/10.1109/TII.2019.2923715. [41] S. Asad, M. Salahat, M.A. Zalata, M. Alia, A.A. Rawashdeh, Design of fuzzy PD-
[37] X. Huo, X. Zhao, G. Zong, Event-triggered adaptive fuzzy output feedback control of controlled overhead crane system with anti-swing compensation, Engineering 3 (7)
MIMO switched nonlinear systems with average dwell time, Appl. Math. Comput. (2011) 755–762, https://doi.org/10.4236/eng.2011.37091.
365 (15) (2020) 124665, https://doi.org/10.1016/j.amc.2019.124665.

12