Nonlinear Dyn (2022) 110:1437–1453

https://doi.org/10.1007/s11071-022-07682-2

ORIGINAL PAPER

Finite-time tracking control for nonholonomic wheeled

mobile robot using adaptive fast nonsingular terminal
sliding mode
Hao Xie · Jinchuan Zheng · Zhe Sun ·
Hai Wang · Rifai Chai

Received: 27 January 2022 / Accepted: 28 June 2022 / Published online: 10 July 2022
© The Author(s) 2022

Abstract System uncertainties and external distur- practically implementable. The finite-time stability of
bances are the major causes of the trajectory tracking the proposed tracking-error function is also proved
performance degradation in nonholonomic wheeled using the Lyapunov function. Finally, circular trajec-
mobile robots (NWMRs). In this article, an adap- tory tracking experiments are conducted to validate
tive fast nonsingular terminal sliding mode dynamic the robustness and convergence rate of the proposed
control (AFNTSMDC) method is proposed to pro- AFNTSMDC scheme in comparison with the exist-
vide enhanced robust and finite-time tracking perfor- ing methods including classic kinematic control, robust
mance for the NWMR. The proposed AFNTSMDC sliding mode kinematic control, and conventional slid-
is a systematic design method based upon both the ing mode dynamic control in the presence of uncertain-
kinematic and dynamic model of the NWMR. The ties and external disturbances.
proposed controller has a simple form without sin-
gularity issue in the control input, which makes it Keywords Trajectory tracking · Mobile robot ·
Finite-time control · Sliding mode control
Jinchuan Zheng, Zhe Sun, Hai Wang and Rifai Chai have
contributed equally to this work.

H. Xie (B) · J. Zheng · R. Chai 1 Introduction

School of Science, Computing and Engineering
Technologies, Swinburne University of Technology,
Melbourne, VIC 3122, Australia
Trajectory tracking control of nonholonomic wheeled
e-mail: [email protected] mobile robots (NWMRs) has attracted much attention
J. Zheng
in past decades due to their wide use in various applica-
e-mail: [email protected] tions [1]. In addition, they pose a challenge to control
R. Chai
practitioners since the mechanism of NWMRs is char-
e-mail: [email protected] acterized by nonholonomic constraints and the inher-
Z. Sun
ent nonlinearity limits the effectiveness of linear con-
Department of Automation, Zhejiang University of trollers [2].
Technology, Hangzhou 310023, Zhejiang, China Achieving a more robust and faster trajectory track-
e-mail: [email protected] ing performance is a challenging task due to the system
H. Wang nonlinearities, model uncertainties and disturbances,
Discipline of Engineering and Energy, Centre of Water, and many efforts have been devoted to this research
Energy, and Waste, Murdoch University, Perth, WA 6150,
Australia
field [3–7]. Fast response and strong robustness against
e-mail: [email protected] system uncertainties and disturbances are crucial objec-

123
1438 H. Xie et al.

tives in tracking control tasks which are also the key fea- control of nonholonomic mobile robots quickly became
tures of sliding mode control (SMC) techniques [8,9]. an emerging topic [20–22]. In [21], a cascaded con-
Although many studies on applying SMC techniques trol system is formed by splitting the error dynamics
for tracking control of the NWMR have been published, of the mobile robot into two subsystems. Based on the
it is worth paying attention to the question of how to controller developed, a finite-time convergence is guar-
utilize more advanced SMC to achieve better control anteed but only when the velocities and their deriva-
performance. The work in [10] proposed an SMC law tives are within a limited range. Additional modeling
for trajectory tracking of an NWMR using computed parameters are considered in the work [22] to improve
torque technique and representing the posture in polar the control accuracy. Likewise, two subsystems are
coordinate, which is subjected to constraints on the formed with two adaptive sliding mode controllers to
heading angle and desired velocities. This was further strengthen the robustness. However, the control system
extended in [11], which loosened the constraint of the only guarantees that the tracking errors can converge to
former study in regard to heading angles and desired a region instead of zero and the desired angular veloc-
velocities for the mobile robot by designing three sepa- ity cannot reach zero, which means that a straight-path
rate controllers under three operating conditions. How- following task is not achievable. In addition, both stud-
ever, there is an inherent drawback of the control input ies [21,22] only illustrated simulation results and thus
singularity around the origin in both aforementioned the practical effectiveness of their controllers is still
papers. questionable.
Meanwhile, attempts have been made to apply SMC Motivated by [15,22,23], we proposed an
law in Cartesian coordinates [12–18]. Among those AFNTSMDC method in this paper, whose main contri-
studies, cascaded control systems, namely inner-outer bution is to provide a unified control scheme compared
loop control structures [13–15], are adopted. The inner to the classic cascaded control structure. Thus, the
loop controller is targeted at velocity following con- assumption of perfect velocity tracking is not needed.
trol for the NWMR while the outer loop focuses on This also simplifies the design process as there is no
designing a model-free kinematic controller. Accord- need to design a kinematic controller and a dynamic
ing to the posture tracking errors, the outer loop con- velocity controller separately. This also leads to a
troller generates corresponding velocity commands to reduction in the number of tuning parameters and tun-
the inner loop which guarantees actual velocities to ing processes. In addition, the AFNTSMDC method
converge to the velocity commands. The capability for guarantees finite-time convergence of the tracking error
the NWMR to achieve the desired performance is built towards zero. It is also more robust against model
upon the assumption of perfect velocity tracking [19] uncertainties and external disturbances. By properly
which, however, may not hold in practice depending on designing the tracking-error function, an alternative
the accuracy of the inner-loop control. In [13], a super- solution to eliminating the control input singularity is
twisting SMC method is developed with a proportional- also proposed.
derivative (PD) controller which increases the robust- The remaining part of the paper is organized as
ness by mitigating the influence of neglected dynamics, follows. The plant model of the NWMR system con-
but singularity issue also arises based on the designed sisting of parametric uncertainties and external distur-
sliding surface. With a similar PD controller, the work bances is formulated in Sect. 2. Section 3 describes
[14] proposed an adaptive fuzzy SMC method that the AFNTSMDC design method, the stability analysis
reduces the system chattering by replacing the adaptive and parameters selection are also elaborated. Section
fuzzy logic with the traditional discontinuous portion in 4 presents the experimental performance of the devel-
SMC. By combining an event-triggered structure with oped controller on the NWMR platform in comparison
a robust SMC, the control input singularity was avoided with other existing methods. The conclusion is drawn
in [15]. in Sect. 5.
Note that most of the aforementioned papers only The following notations are used in paper: for
 T
guarantee the asymptotic stability of the mobile robot  ∈ R f , sgn() denotes sign(ψ1 ), . . . , sign(ψ f ) ;
system which means that they may achieve conver-  T
sig() represents |ψ1 |sign(ψ1 ), . . . , |ψ f |sign(ψ f ) ,
gence in infinite settling time and a fast convergence and diag([]) denotes the diagonal matrix with diag-
rate may not be accomplished. Finite-time tracking

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1439

τ , τd and λ are given as follows:

⎡ ⎤ ⎡ ⎤
m 0 m̄ 1 −sin(φg )
M̄( p) = ⎣ 0 m −m̄ 2 ⎦ , A T ( p) = ⎣ cos(φg ) ⎦
m̄ 1 −m̄ 2 I 0
⎡ ⎤ ⎡ ⎤
cos(φg ) cos(φg ) 0 0 m̄ 2 φ̇g
1⎣
B̄( p) = sin(φg ) sin(φg ) ⎦ , V̄ ( p, ṗ)=⎣0 0 m̄ 1 φ̇g ⎦
r
b −b 00 0

λ = −m ẋ g cos(φg ) + ẏg sin(φg ) φ̇g ,
τ1 τ
τ= , τd = d1
τ2 τd2
m̄ 1 = mhsin(φg ), m̄ 2 = mhcos(φg ) (2)

where m is the total mass of the NWMR including load

uncertainty; I is the moment of inertia of the mobile
Fig. 1 Top view of a nonholonomic wheeled mobile robot robot; τ1 and τ2 indicate the torques generated by the
right and left wheel; τd1 and τd2 denote the external
disturbances.
onal elements of ψ1 , . . . , ψ f . For  ∈ R f ×q ,  The nonholonomic kinematic constraints are described
denotes the element-wise absolute value of matrix . by
A( p) ṗ = 0. (3)
The kinematic model for the NWMR can be described
as
2 Problem formulation
ṗ = S( p)z (4)
In this section, the dynamics and kinematics of NWMRs
are discussed. A schematic top view is shown in Fig. 1,
where
where the mass center of the mobile robot is denoted
by Om and the distance from Om to the middle of the ⎡ ⎤
cos(φg ) 0
two front driving wheels Ol is denoted as h. Thus, v
S( p) = ⎣ sin(φg ) 0⎦ , z = (5)
{Og X g Yg } is the reference coordinate system, whereas ω
0 1
{Ol X l Yl } is the coordinate system fixed to the mobile
robot. The position of the robot is completely specified with v and ω denote the forward velocity and angular
by the vector p = [x g , yg , φg ]T . Then, a dynamic velocity, respectively. From (4), it follows that
model for the NWMR can be described as [22]
p̈ = S( p)ż + Ṡ( p)z. (6)
M̄( p) p̈ + V̄ ( p, ṗ) ṗ + τd = B̄( p)τ − A ( p)λ
T
(1)
Substituting (6) into (1) yields

where M̄( p) ∈ R3×3 is a positive symmetric definite 

M̄( p) S( p)ż + Ṡ( p)z + V̄ ( p, ṗ)S( p)z + τd
inertia matrix; τd ∈ R2 denotes the lumped compo-
nents including surface friction, external disturbances, = B̄( p)τ − A T ( p)λ. (7)
and unmodeled dynamics; τ ∈ R2 is the input vec-
tor. V̄ ( p, ṗ) ∈ R3×3 is the centripetal and coriolis Multiplied (7) by S T ( p) on the left, the system dynam-
matrix; A T ( p) ∈ R3×1 is the matrix associated with ics (1) can be rewritten as
constraints, and λ ∈ R is the vector of Lagrange mul-
tipliers. The matrices M̄( p), A T ( p), B̄( p), V̄ ( p, ṗ) , M ż = Bτ + d (8)

123
1440 H. Xie et al.

where trajectory command accurately in the presence of sys-

tem uncertainties and external disturbances in a finite
m0 1 1 1 d time. To achieve this goal, we will first define a
M= ,B = ,d = 1
0 I r b −b d2 tracking-error function. Then, a fast nonsingular ter-
= −S T ( p)τd . (9) minal sliding surface with an adaptive reaching law
will be developed, based on which the finite-time con-
In addition, the following parametric uncertainties are vergence is achieved, the singularity issue is resolved,
considered: and the unified control scheme is developed. Thus, the
AFNTSMDC law is constructed with both the velocity
M = M0 + M (10) and position as feedback signals and the motor torques
as the control input.
with

m0 0 m 0
M0 = , M = 3.1 Construction of the AFNTSMDC
0 I0 0 I
First, define a tracking error vector [19] for the NWMR
where m 0 and I0 denote the nominal model parameters as
and m and I denote the corresponding uncertain-
ties, respectively. ⎡ ⎤ ⎡ ⎤
xe cos(φg ) sin(φg ) 0
Combining (10) and (9), we can obtain the dynamic pe = ⎣ ye ⎦ = ⎣−sin(φg ) cos(φg ) 0⎦ ( pr − p) (14)
model for the NWMR as follows: φe 0 0 1
M0 ż = Bτ − δ (11)
where δ = [δ1 , δ2 ]T = M ż + d represents the refor- where pr = [xr , yr , φr ]T is the desired trajectory
matted uncertainty to system (9). posture and its kinematics can be modeled as

Assumption 1 In the following, the lumped uncer- ṗr = S( pr )zr (15)

tainty δ is assumed to be bounded by
where zr = [vr , ωr ]T is the desired velocity, vr denotes
|m −1
0 δ1 | < c1 + c3 |v| (12) the desired forward velocity, and ωr denotes the desired
|I0−1 δ2 | < c2 + c4 |ω| (13) angular velocity which can be calculated by

vr = ẋr2 + ẏr2 (16)

where ci (i = 1, 2, 3, 4) are unknown but bounded
positive numbers. The terms c1 and c3 denote the upper ẋr ÿr − ẍr ẏr
ωr = φ̇r = . (17)
bound of the external disturbance and surface friction. ẋr2 + ẏr2
Meanwhile, c2 and c4 denote the upper bound of the
dynamic impact caused by modeling imprecision. Combining (14) and the kinematic model described in
(4), one can obtain the error model for the trajectory
In the sequel, the control design for the AFNTSMDC tracking control as
will be constructed according to the kinematic model
in (4) and dynamic model in (11) formulated for the
ṗe = F + Gz (18)
NWMR.

in which
3 Control design ⎡ ⎤ ⎡ ⎤
vr cos(φe ) −1 ye
In this section, the AFNTSMDC design method will be F = ⎣ vr sin(φe ) ⎦ , G = ⎣ 0 −xe ⎦ . (19)
presented such that the NWMR can track the desired wr 0 −1

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1441

Differentiating (14), we have with η = [η1 , η2 ]T = αξ  + βξ γ . Meanwhile,

taking derivative of the sliding surface s, one has
p̈e = Ḟ + Ġz + G ż. (20)
ṡ = ξ̈ + α ξ̇ + βγ ξ γ −1 ξ̇ . (29)
Combining (11) with (20), we obtain
p̈e = Ḟ + Ġz + G M0−1 (Bτ − δ). (21) Substitute (26) into (29), we obtain

Second, we shall introduce a new tracking-error ṡ = Ḣ ṗe + H Ḟ + Ġz + G M0−1 (Bτ − δ)
function
+ α ξ̇ + βγ ξ γ −1 ξ̇ . (30)
ξ1 xe
ξ= = ρ −1 (22)
ξ2 φe + |xe |+ρ tan ye
Last, the AFNTSMDC law will be constructed based
on the tracking-error function and sliding surface pro-
where 0 < ρ < 1 is the control parameter to be tuned. posed above. Let δ = 0 and replace τ with τeq . Then,
Note that for 0 < ρ < 1 and any xe ∈ R, the expression solving (30) for ṡ = 0 leads to
ρ 
|xe |+ρ ∈ (0, 1], which can be regarded as a weighting τeq = −(H G M0−1 B)−1 Ḣ ṗe + H ( Ḟ + Ġz) + α ξ̇
parameter to adjust the convergence rate of ye and φe .
+βγ ξ γ −1 ξ̇ . (31)
It will be shown later that once ξ converges to zero, the
tracking error vector pe will converge to zero accord- Furthermore, a reaching control input τr is designed as
 
ingly. Hence, our control objective is now converted to τr = −(H G M0−1 B)−1 K 1 s + K 2 sig(s)μ
enabling the finite-time convergence of ξ through the
proposed method. −B −1 M0 L Ĉ (32)
Taking the first-order derivative of the tracking-error with
function (22) yields
ξ̇ = H ṗe (23) L = diag([sgn(s1 ), sgn(s2 )]) (33)
with C = [cˆ1 + cˆ3 |v|, cˆ2 + cˆ4 |ω|] T
(34)
 K 1 = diag([k1 , k2 ]) (35)
1 0 0
H = −ρ(|x | + ρ)−2 sgn(x )tan−1 y K 2 = diag([k3 , k4 ]) (36)
e e e
ρ
1 .
(|xe |+ρ)(1+ye2 )
(24) where ki (i = 1, 2, 3, 4) are positive control param-
eters, 0 < μ < 1; and the adaption gain cˆi (i =
Based on (23), we can obtain 1, 2, 3, 4) is updated by the following adaptive law:
ξ̈ = Ḣ ṗe + H p̈e . (25)
cˆ˙1 = ζ1 |s1 |, cˆ˙2 = ζ2 |s2 | (37)
Substituting (21) into (25) yields the tracking-error
dynamic equation as follows: c˙ˆ3 = ζ3 |v||s1 |, c˙ˆ4 = ζ4 |ω||s2 | (38)

ξ̈ = Ḣ ṗe + H Ḟ + Ġz + G M0−1 (Bτ − δ) . (26) with ζi ≥ 0, cˆi (0) ≥ 0 (i = 1, 2, 3, 4) are control
Furthermore, we define a sliding surface s as parameters to be designed. Thus, the complete form of
the AFNTSMDC law can be obtained as
s = ξ̇ + αξ + βsig(ξ )γ (27) τ = τeq + τr (39)

where α = diag([α1 , α2 ]) > 0, β = diag([β1 , β2 ]) > where τeq and τr are given in (31) and (32), respectively.
0, and 1 < γ < 2. It has been proved in Appendix B
that when the sliding variable s reaches to zero, for any 3.2 Stability analysis
initial values of ξ and ξ̇ , the tracking-error function ξ
can converge to zero in a finite time tξ bounded by The result for the proposed AFNTSMDC law is sum-
marized in the following lemma and stability analysis
tξ ≤ max{η1−1 |ξ1 (0)| 2 , η2−1 |ξ2 (0)| 2 }
1 1
(28) is provided.

123
1442 H. Xie et al.


Lemma 1 Given the NWMR system in (4) and (11), = −sT H G Ĉ − M0−1 δ
and the control law (39), ĉi has an upper bound and 
4
there exists a positive value ci in (12) such that cˆi ≤ + c̄i c˙ˆi + sT H GC − sT H GC
ci (i = 1, 2, 3, 4) always holds. i=1

= −sT H G C − M0−1 δ − sT H GC − Ĉ
Proof Supposing the sliding surfaces s1 and s2 have

4
not arrived at zero and ĉi is increasing and there exists + c̄i c˙ˆi .
time instances t1 and t2 such that: i=1

ĉ1 (t1 ) + ĉ3 (t1 )|v| > |m −1 μ

0 δ1 + k1 s1 + k3 sig(s1 ) | Following ci > ĉi from Lemma 1, we have
(40)

ĉ2 (t2 ) + ĉ4 (t2 )|ω| > |I0−1 δ2 + k2 s2 + k4 sig(s2 )μ |. V̇ ≤ −sT H G C − M0−1 δ − sT H GC − Ĉ
(41) 
4
− |c̄i |c˙ˆi
i=1
Based on (12), when t3 = max{t1 , t2 }, ĉi will be large
= −s Q − ζ1 |s1 ||c̄1 | − ζ2 |s2 ||c̄2 | − ζ3 |v||s1 ||c̄3 |
T
enough to make the sliding surfaces s1 and s2 reach to
zero in a finite time ts . Then, the value of ĉi (t) will hold − ζ4 |w||s2 ||c̄4 | (44)
at ĉi (t3 + ts ) finally.
Under Assumption 1 and the continuity property of where
ĉi , it is clear that t3 + ts is finite and for all t, ĉi (t) has  T
an upper bound. Therefore, there exists such a positive C = c1 + c3 |v|, c2 + c4 |ω|
value ci in (12) satisfying ĉi ≤ ci . 
Q = [q1 , q2 ]T = H G C − M0−1 δ + C − Ĉ .
This completes the proof of Lemma1. 

Lemma 2 Consider the NWMR system in (4) and (11), To make the expression compact, the following sym-
then under the AFNTSMDC law (39) the tracking error bols are defined:
vector pe converges to zero in a finite time.
1 = ζ1 |s1 |, 2 = ζ2 |s2 |
Proof Firstly, by substituting the control law (39) into
3 = ζ3 |v||s1 |, 4 = ζ4 |w||s2 |.
(30), one can obtain

ṡ = −K 1 s − K 2 sig(s)μ − H G L Ĉ − M0−1 δ . (42) It is obvious that i > 0 (i = 1, 2, 3, 4), thus (44)
Define the adaptive estimation error c̄i = cˆi − ci (i = can be rewritten as
1, 2, 3, 4) and choose the Lyapunov function as 
√ |s1 | |s2 | |c̄1 | |c̄2 |
V̇ ≤ − 2 q1 √ + q2 √ + 3 √ + 4 √
1 2
4
1 T 2 2 2 2
V = s s+ c̄i (43) 
2 2 |c̄3 | |c̄4 |
i=1 + 5 √ +  6 √
2 2
Using (42) and evaluating the derivative of V along  
√ |s1 | |s2 | |c̄1 | |c̄3 | |c̄3 | |c̄4 |
this system trajectory with the proposed AFNTSMDC ≤ − 2η3 √ + √ + √ + √ + √ + √
input yields 2 2 2 2 2 2
√ 1
≤ − 2η3 V 2 (45)

4
V̇ = s T ṡ + c̄i c˙ˆi
i=1 and

= −K 1 s − K 2 s μ+1
− s H G L Ĉ −
T
M0−1 δ
η3 = min{q1 , q2 , 1 , 2 , 3 , 4 }. (46)

4
+ c̄i c˙ˆi Therefore, according to the result in Appendix A, the
i=1
inequality (45) satisfies the finite-time stability crite-
 
4
rion. More specific, V will converge from any initial
≤−sT H G Ĉ − M0−1 δ + c̄i c˙ˆi
i=1
condition V (0) to zero in a finite time ts shown in the

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1443

following equation: Remark 1 The controller (39) requires the inverse of

√ 1 the matrix H G M0−1 B. Similar issue has appeared in
2V 2 (0) [13,15] while examining the singularity of the con-
ts ≤ . (47)
η3 trol laws. In our method, the singularity issue can be
This implies that the sliding variable s and the esti- resolved by the proposed tracking-error function (22).
mation error c¯i will converge to zero in a finite time This can be straightly justified by the determinant of
ts . the matrix H G M0−1 B as follows:
When sliding surface is first arrived (i.e., s = 0  
and ξ̇ = −αξ − βsig(ξ )γ ), the finite-time converging 2b ρ
|H G M0−1 B| =− xe + 1 .
condition of ξ is met. After a finite time tξ when ξ = 0, r m 0 I0 (|xe | + ρ)(1 + ye2 )
according to the definition of ξ in (22), we have (52)

xe = 0 It can be seen that for 0 < ρ < 1, the func-
(48) ρ
φe = −tan−1 ye . tion (|x |+ρ)(1+y 2 ) x e = −1 because of the functions
e e
ρxe
|xe |+ρ ∈ (−1, 1) and 1
1+ye2
∈ (0, 1]. Therefore, the
Substituting the condition (48) into (18), the dynamics
of ye can be written as inverse of H G M0−1 B is finite.

Remark 2 In the adaptation law (37)–(38) , the sliding

y˙e = −vr sin(tan−1 ye ) variable s is generally chattering around zero due to
vr ye
= − . (49) system uncertainties and measurement noises, resulting
1 + ye2 in overly large estimations of ci causing control input
saturation. Hence, the dead zone technique [24] can
To investigate the stability of the dynamics of ye , we be employed in practice to moderate this issue. More
choose a Lyapunov candidate Vye = 21 ye2 . Thus, V̇ye specifically, the following rules are implemented:
can be expressed as

c˙ˆ1 = ζ1 |s1 |, c˙ˆ3 = ζ3 |v||s1 |, for |s1 | > ε
vr ye2 (53)
V̇ye = −  cˆ˙1 = c˙ˆ3 = 0, for |s1 | ≤ ε
1 + ye2 
c˙ˆ2 = ζ2 |s2 |, c˙ˆ4 = ζ4 |ω||s2 |, for |s2 | > ε
vr |ye | √ |ye | (54)
≤ − 2√ c˙ˆ2 = c˙ˆ4 = 0, for |s2 | ≤ ε
1 + ye2 2
√ 1
= − 2η4 Vye2 (50) where ε is a small positive threshold value selected
as 0.05 in our case. It is clear that when s is within
where η4 = √vr |ye |2 and vr > 0. According to the region ε, cˆ1 and cˆ2 will not increase but retain the
1+ye present values. One can easily verify that when both
Appendix A, ye converges to zero in the finite time |s1 | < ε and |s2 | < ε, (45) still holds. Therefore, the
satisfying finite-time stability property is still retained in practice.
√ 1
2V 2 (0) Remark 3 The boundary layer technique can be used
t ye ≤ − . (51) to compromise between control accuracy and the chat-
η4
tering induced by the reaching control. This can be
Recall the condition in (48). When ye reaches zero, accomplished by replacing sgn(·) in the reaching law
φe reaches zero at the same time. Therefore, under the (32) with a saturation function given by
proposed AFNTSMDC law, the x-axis tracking error 
sgn(ϑ), for |ϑ| > 
xe converges from any initial condition to zero in a sat(ϑ) = (55)
finite time of txe = ts + tξ and it takes additional finite ϑ−1 , for |ϑ| ≤ 
time t ye for ye and φe to reach zero. where  denotes the boundary layer thickness and
The proof is thus completed. 
  = diag([0.04, 0.04]) is chosen in our case to reach

123
1444 H. Xie et al.

Fig. 2 Block diagram of

the proposed unified control
structure for NWMR
systems

a balance between chattering reduction and acceptable (2) Selection of ρ: Before xe reaches zero, a smaller
tracking errors. value of ρ in the tracking-error function (22) can
increase the convergence rate of ye and φe . How-
Remark 4 To emphasize the advantage of the proposed ever, it may invoke oscillations in transient response
control scheme which is shown in Fig. 2, a comparison of φe . Thus, ρ is set to be 0.8.
is made between the AFNTSMDC method and a clas- (3) Selection of K 1 , K 2 , μ: The control parameters K 1 ,
sic control scheme that is presented in Fig. 3. In Fig. 3, K 2 , and μ in the reaching law (32) dominate the sys-
the inner velocity controller and an outer loop controller tem robustness. Increasing the values of K 1 and K 2
are designed based on the dynamic model and the kine- strengthens the system robustness at the cost of con-
matic model separately, whereas the AFNTSMDC is trol signal smoothness. Meanwhile, μ can balance
the controller that integrates the control requirements the control signal chattering and the robustness. In
for both the kinematic and dynamic model. Thus, the our case, K 1 = diag([9, 4]), K 2 = diag([15, 5]),
proposed control scheme simplifies the system struc- and μ = 0.8 are chosen in the implementation.
ture which potentially leads to more reliable perfor- (4) Selection of ci (0) and ζi (i = 1, 2, 3, 4): ci (0)
mances, a reduction in the number of tuning parame- denote the initial values of ci . Proper selected of
ters and tuning processes, and less computational and values of ci (0) will reduce the adaptation time. The
hardware cost. Further comparisons and statistical val- adaptive gains ζi in (37)–(38) determine the conver-
idations regarding the practical performance will be gence rate of the adaptive estimation error, whereas
presented in Sect. 4. a large value of them may cause control input sat-
uration and overshoots. Thus, we find c1 (0) = 14,
c2 (0) = 14, c3 (0) = 8, c2 (0) = 6, ζ1 = 0.3,
3.3 Control parameters selection ζ2 = 0.4, ζ3 = 0.4, ζ4 = 0.5 are sufficient for the
experiments.
During the implementation, trade-offs between the
desired tracking performance and other factors such as
control input saturation, control command smoothness,
and measurement noises are expected. In the follow-
4 Experimental results
ing, we will discuss the controller parameters selection
guideline for the proposed AFNTSMDC law and give
To demonstrate the effectiveness of the proposed
their values for the NWMR under study.
AFNTSMDC method in the presence of external dis-
(1) Selection of α, β, γ : The parameters in α, β, and γ turbances and load variations, experiments are con-
affect the dynamic behaviors of the sliding surface ducted on the NWMR shown in Fig. 4. In addition,
s in (27). An increment of these parameters can experimental comparisons are made with other exist-
lead to faster convergence of ξ towards zero but ing control methods, i.e., a classic kinematic control
may incur a rising in the amplitude of the tracking- (CKC) method [19], a recently proposed robust sliding
error overshoots. For the NWMR constructed for mode kinematic control (RSMKC) method [15], and a
the experiment, α = diag([7, 7]), β = diag([7, 6]) conventional sliding mode dynamic control (CSMDC)
and γ = 1.67 are chosen. method.

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1445

Fig. 3 Block diagram of an

inner-outer control structure
for the NWMR system [12]

4.1 Control methods for comparison

Classic kinematic control (CKC) As shown in Fig. 3,

an outer loop CKC law is given in the following form
[19]:
z d = K E pe + K D (56)
with

vd k 0 0
zd = , K E = e1 ,
ωd 0 vr ke2 0
vr cosφe
KD = (57)
vr ke3 sinφe + ωr
Fig. 4 Experimental platform of a nonholonomic wheeled
where ke1 , ke2 , and ke3 are all positive constants which mobile robot
are set to be 1.7, 1.7, and 1.5 in this experiment.
Robust Sliding Mode Kinematic Control (RSMKC):
where u = [u 1 , u 2 ]T denotes the output of the PI
Similarly, an RSMKC method is proposed in [15]
⎧ velocity controller; z e = z d − z which is the velocity
−1
⎨ sr = kr 1 φe + vr tan ye
⎪ tracking error; k P and k I are tuning parameters and
xe (58) they are set to 30 and 250 in our case, respectively.
⎪
⎩ −1
 Conventional Sliding Mode Dynamic Control
z d = −Br Ar Fr + Dr + K r sgn(sr )
(CSMDC): For comparison, a CSMDC method is listed
where below whose control scheme is the same as the one
  discussed in Sect. 2. However, the sliding surface and
vr vr xe
0 1+ye2
kr 1 0 −(kr 1 + 1+ye2
)
Ar = , Br = reaching law are designed based on the conventional
1 0 0 −1 ye
⎡ ⎤
SMC method which is given as
vr cosφe ⎧
v˙ tan−1 ye k 0 ⎪ scsm = λc ξ + ξ̇
Fr = ⎣ vr sinφe ⎦ , Dr = r , Kr = r 2 ⎪
⎪
⎪ 
0 0 kr 3 ⎨τ −1 −1
ωr eq2 = −(H G M0 B) λc ξ̇ + Ḣ ṗe + H ( Ḟ + ĠV )
⎪
⎪ −1 −1
τr 2 = −(H G M0 B) K c sgn(scsm )
⎪
⎪
where kr 1 , kr 2 , and kr 3 are tuning parameters which are ⎩
τ2 = τeq2 + τr 2
set to 0.40, 0.02, and 0.02, respectively.
(60)
It is worth noting that both kinematic controllers pre-
sented above are coupled with an inner velocity con- with
troller which is designed for the actual velocity to fol-
low the desired velocity generated by the kinematic kc1 0 λ 0
Kc = , λC = c1
controller. A PI velocity controller is used in experi- 0 kc2 0 λc2
ments, and it can be described as
where ξ is with (22); kc1 , kc2 , λc1 , and λc2 are positive
 T control parameters which are chosen as 21, 16, 3 and
u = k P z e (T ) + k I z e (t) dt (59) 4, respectively.
0

123
1446 H. Xie et al.

Table 1 Model parameters of the NWMR In addition, the AFNTSMDC obtains the smallest RMS
Parameter Symbol Nominal value with a value of 1.918 cm in X -axis tracking error. There
is no significant lead for the AFNTSM regarding RMS
Mass of the NWMR m0 4.500 kg and MAX, the transition in Y -axis is the smoothest
Moment of inertia of the NWMR I0 0.560 kg m2 shown in Figure 5d. Figure 6 presents the control inputs
Wheel radius r 0.042 m of all controllers, where the AFNTSMDC exhibits per-
Distance between the driving wheels 2b 0.372 m sistently smooth control signals without singularity
Distance between Om and Ol h 0.110 m issue that matches the theoretical design. Nevertheless,
due to the impact of measurement noise, there is still
a small level of chattering in the control input, which
4.2 Experimental platform is inevitable in the experiments but has not caused any
implementation issue in our case.
Our experimental setup for the NWMR is shown in
Fig. 4 with its nominal physical parameters listed in
Table 1. It consists of two dc motors (Maxon) which are 4.4 Circular path tracking performance with load
integrated with encoders to measure the rotary angles. uncertainty and disturbance
Two motor drivers (Maxon ESCON-36/2) are also used
to control the motor power. Two caster wheels are To further verify the performance robustness in the
located at the rear of the robot. The data acquisition and presence of modeling uncertainties, we place a 3.60-
controller are implemented with the real-time micro- kg cylindrical payload on the mass center of the mobile
controller (NI myRIO). The sampling period is set to robot, i.e., making M = diag([3.60, 0.01]). Mean-
0.01 seconds for all controllers implemented. The feed- while, the circular path reference is still used on the
back posture signals of the NWMR are obtained by control system with an initial posture offset set on
using a posture estimator with the motor encoder sig- the robot. Compared with the previous experiment,
nals as the system input. various performance degradation can be observed in
Fig. 7c–e. However, the AFNTSMDC maintains its
advantages with the shortest settling time with a value
4.3 Circular path tracking performance with initial of 1.128 s and the smallest RMS of 2.806 cm in the
posture offset X -axis tracking error listed in Table 3 because the pay-
load uncertainties have been explicitly considered dur-
A circular path is typically used for testing the track- ing the design process as shown in (11). Meanwhile,
following performance of mobile robots. In the exper- the fast convergence feature and the smoothness of
iments conducted, the desired circular path is config- the AFNTSM are observed in Fig. 7d, whereas the
ured with an angular speed of 0.70 rads/s and a radius kinematic controllers (i.e., CKC and RSMKC) suffer
of 0.60 m. In addition, the initial posture offset is set to from severe performance degradation due to their weak
be (−0.10, −0.10, 0). robustness against system uncertainties. From the con-
Experimental results on the trajectory and tracking trol input in Fig. 8, it can be observed that there is a
error profiles are shown in Fig. 5. We can see that the significant increment in the amplitude and the number
AFNTSMDC method achieves the fastest convergence of spikes in control input signals except those under
rate and the least oscillations. Meanwhile, the track- CSMDC and AFNTSMDC. The increased amplitudes
ing results in Fig. 5 are also summarized in Table 2, in Fig. 8b are 39.54%, 40.83%, 35.41%, and 24.05%
where the root-mean-square (RMS), the maximum for the CKC, RSMKC, CSMDC, and AFNTSMDC,
value (MAX), and the settling time of tracking errors respectively. Hence, this experiment validates that the
are listed. Figure 5a, b shows that all controllers guaran- proposed controller is more robust against modeling
tee to converge to the desired trajectory. In particular, uncertainties in comparison with other control meth-
Fig. 5c shows that the convergence of X -axis track- ods.
ing error under AFNTSMDC is with a settling time of Later the disturbances d1 and d2 as shown in Figs. 2
0.684 s which is 788%, 603%, and 249% faster than and 3 are set as shock disturbances acting on the wheels
that of the CKC, RSMKC, and CSMDC, respectively. of the NWMR. The disturbance behaves as a half-sine

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1447

(a) (c)
0.2
CKC

X-axis error (m)

1 RSMKC
CSMDC
0.1
AFNTSMDC
Y-axis (m)

0.5
0
CKC
0 RSMKC 0 5 10 15 20
CSMDC
AFNTSMDC (d)
Reference
-0.5 0.1 CKC

Y-axis error (m)

-0.5 0 0.5 1 1.5 RSMKC
X-axis (m) 0.05 CSMDC
AFNTSMDC
0
(b)
15
-0.05
1
Orientation angle (rad/s)

0 5 10 15 20

(e)
Orientation angle error (rad)
0.5
10

0.2 CKC
0 RSMKC
0 0.5 1 1.5
CKC CSMDC
0.1
5 RSMKC AFNTSMDC
CSMDC
AFNTSMDC 0
Reference
0 -0.1
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

Fig. 5 Circular tracking with initial posture offset. a Trajectory profiles. b Orientation angles. c Tracking errors of X -axis displacement.
d Tracking errors of Y -axis displacement. e Tracking errors of orientation angle

Table 2 Tracking performance summary of circular trajectory with initial offset

Tracking errors Performance metrics CKC RSMKC CSMDC AFNTSMDC

xe RMS (cm) 3.298 3.288 3.304 1.918

MAX (cm) 11.604 12.115 17.524 13.019
Settling time (s) 6.070 4.810 2.380 0.680
ye RMS (cm) 2.436 2.528 3.754 3.159
MAX (cm) 10.021 10.014 10.684 10.210
Settling time (s) 7.520 7.590 9.360 8.870
φe RMS (rad) 0.015 0.023 0.047 0.030
MAX (rad) 0.032 0.064 0.215 0.172
Settling time (s) 5.930 7.390 3.980 4.460

123
1448 H. Xie et al.

Fig. 6 Control input with (a)

initial posture offset. a
Right motor voltage input. b 5

Right motor input (V)

Left motor voltage input
0

-5 CKC
RSMKC
CSMDC
-10
AFNTSMDC

0 2 4 6 8 10 12 14 16 18 20

(b)
5
Left motor input (V)

-5
CKC
RSMKC
-10 CSMDC
AFNTSMDC

0 2 4 6 8 10 12 14 16 18 20
Time (s)

waveform with a duration of 0.5 s and an amplitude of In summary, the proposed AFNTSMDC controller
6.0
 V. The disturbance can be modeled as performs the most robustly and fastest in the trajectory
d1 = d2 = 6|sin(2π t)| 12 s ≤ t ≤ 12.5 s tracking control for the NWMR. It is also worth noting
(61) that the tracking errors under all controllers still contain
d1 = d2 = 0 Other time.
a small amount of steady-state error in practice due
The disturbances occur at the 12th second when all the
to the coarse sensor resolution and the trade-off for
controllers reach steady state and last for another 0.5 s.
reducing the control chattering, which will be improved
Thus, the data collected from the 11th to 18th are inter-
in our future work.
cepted from a 20-second experiment and it is shown in
Fig. 9 for better readability. One of the most notable
features of the AFNTSMDC is the fast convergence
rate and robustness in response to an external distur- 5 Conclusion
bance. It can be seen from Fig. 9a that the tracking
error under the AFNTSMDC has a MAX peak value of In this paper, we developed an AFNTSMDC method
1.624 cm, which is 74%, 66%, and 25% smaller than for the NWMR system to accomplish trajectory track-
the MAX peaks values under the CKC, RSMKC, and ing control in a finite time. The AFNTSMDC method
CSMDC. In addition, the settling time of the X -axis resolves the singularity issue presented in previous
tracking error is 0.525 s under AFNTSMDC in com- research and possesses strong robustness against exter-
parison with 0.520 s, 0.870 s, and 0.670 s under the nal disturbances and modeling uncertainties. Compar-
CKC, RSMKC, and CSMDC listed in Table 4. While isons are made with existing techniques such as the
obtaining stronger robustness, only the same level of CKC, RSMKC, and CSMDC methods under a series
control inputs is required under AFNTSMDC, which of circular trajectory tracking experiments. It has been
can be seen from Fig. 10. Thus, the results verify that validated that the proposed control method possesses
the AFNTSMDC is more robust and faster in terms of stronger robustness and a faster convergence rate when
disturbance rejection capability and settling-time con- compensating for initial posture offset, load uncer-
vergence. tainty, and external disturbances. In addition, the pro-

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1449

(a) (c)
0.3

X-axis error (m)

CKC
1 0.2 RSMKC
CSMDC
AFNTSMDC
0.1
Y-axis (m)

0.5

0
CKC
0 RSMKC 0 5 10 15 20
CSMDC
AFNTSMDC (d)
Reference
-0.5 0.1 CKC
RSMKC

Y-axis error (m)

-0.5 0 0.5 1 1.5
0.05 CSMDC
X-axis (m)
AFNTSMDC
0
(b)
14
-0.05
12 1
Orientation angle (rad/s)

0 5 10 15 20

(e)
Orientation angle error (rad)
10 0.5

8 0.3 CKC
0 RSMKC
6 0 0.5 1 1.5
CKC 0.2 CSMDC
RSMKC AFNTSMDC
4 0.1
CSMDC
AFNTSMDC
2 0
Reference
0 -0.1
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)

Fig. 7 Circular tracking with the load. a Trajectory profiles. b Orientation angles. c Tracking errors of X -axis displacement. d Tracking
errors of Y -axis displacement. e Tracking errors of orientation angle

Table 3 Tracking performance summary of circular trajectory with load uncertainty

Tracking errors Performance metrics CKC RSMKC CSMDC AFNTSMDC

xe RMS (cm) 3.224 3.883 4.740 2.806

MAX (cm) 12.165 12.248 21.436 15.229
Settling time (s) 6.490 6.280 1.810 1.120
ye RMS (cm) 2.148 3.408 3.933 2.587
MAX (cm) 10.046 10.015 10.727 10.015
Settling time (s) 8.380 9.270 6.330 7.180
φe RMS (rad) 0.017 0.035 0.065 0.024
MAX (rad) 0.041 0.075 0.285 0.125
Settling time (s) 6.120 9.640 5.350 3.860

123
1450 H. Xie et al.

Fig. 8 Control input with (a)

the load. a Right motor 5
voltage input. b Left motor
voltage input

Right motor input (V)

0

-5
CKC
RSMKC
-10 CSMDC
AFNTSMDC

-15
0 2 4 6 8 10 12 14 16 18 20

(b)
5
Left motor input (V)

-5
CKC
RSMKC
-10 CSMDC
AFNTSMDC

-15
0 2 4 6 8 10 12 14 16 18 20
Time (s)

Fig. 9 Circular tracking (a)

under external disturbances. 0.02
X-axis error (m)

a Tracking errors of X -axis

displacement. b Tracking 0 CKC

errors of Y -axis RSMKC

displacement. c Tracking -0.02 CSMDC

AFNTSMDC
errors of orientation angle
-0.04
11 12 13 14 15 16 17 18

(b)
0.02
Y-axis error (m)

0 CKC
RSMKC
-0.02 CSMDC
AFNTSMDC
-0.04
11 12 13 14 15 16 17 18

(c)
Orientation angle

0.02
error (rad)

0 CKC
RSMKC
-0.02
CSMDC
-0.04 AFNTSMDC

11 12 13 14 15 16 17 18
Time (s)

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1451

Table 4 Tracking performance summary of circular trajectory with disturbance

Tracking errors Performance metrics CKC RSMKC CSMDC AFNTSMDC

xe RMS (cm) 3.227 3.887 4.748 2.805

MAX (cm) 2.827 2.695 2.035 1.624
Settling time (s) 1.830 0.870 0.670 0.520
ye RMS (cm) 2.144 3.407 3.937 2.682
MAX (cm) 0.785 0.901 1.234 1.004
Settling time (s) 2.370 2.660 2.280 2.090
φe RMS (rad) 0.004 0.006 0.008 0.003
MAX (rad) 0.013 0.025 0.026 0.011
Settling time (s) 1.070 1.310 1.080 1.120

Fig. 10 Control input (a)

under external disturbances. 5
a Right motor voltage input.
Right motor input (V)

b Left motor voltage input

0

-5 CKC
RSMKC
CSMDC
-10
AFNTSMDC

11 12 13 14 15 16 17 18

(b)
5
Left motor input (V)

-5 CKC
RSMKC
CSMDC
-10
AFNTSMDC

11 12 13 14 15 16 17 18
Time (s)

posed control scheme simplifies the control structure In our future work, we will investigate the chattering-
and unifies the design process in comparison with the free reaching law to inherently eliminate the chattering
classic cascaded control scheme. Thus, the design pro- phenomenon without sacrificing tracking accuracy.
cess has been simplified since there is no need to
Funding There is no any funding associated with this paper and
design two individual controllers for different objec- the authors declare that there is no known competing financial
tives, which also potentially leads to more reliable per- interests or personal relationships that could have appeared to
formances, reduction in control parameters, tuning pro- influence the work reported in this paper.
cess, and computational and hardware requirements.

123
1452 H. Xie et al.

Data availability The data collected in this study are available where η1 = α1 |ξ1 | − β1 |ξ1 |μ . According to Appendix
from the corresponding author on reasonable request. A, ξ converges to zero in the finite time satisfying
Declarations √ 1
2Vξ21 (0)
tξ 1 ≤ (B4)
Conflict of interest The authors declare that they have no con- η1
flict of interest.
√ 1
Similarly, we can have tξ 2 ≤ η2−1 2Vξ22 (0) by defining
Open Access This article is licensed under a Creative Com-
mons Attribution 4.0 International License, which permits use, Vξ2 = 21 ξ22 for the system ξ̇2 = −α2 ξ2 − β2 sig(ξ2 )μ .
sharing, adaptation, distribution and reproduction in any medium Thus, we can conclude that the tracking-error function
or format, as long as you give appropriate credit to the original ξ can converge to zero in a finite time tξ given as
author(s) and the source, provide a link to the Creative Com-
mons licence, and indicate if changes were made. The images or tξ = max{tξ 1 , tξ 2 }. (B5)
other third party material in this article are included in the article’s
Creative Commons licence, unless indicated otherwise in a credit
line to the material. If material is not included in the article’s Cre-
ative Commons licence and your intended use is not permitted by References
statutory regulation or exceeds the permitted use, you will need
to obtain permission directly from the copyright holder. To view
a copy of this licence, visit http://creativecommons.org/licenses/ 1. Fabregas, E., Farias, G., Aranda-Escolástico, E., Garcia, G.,
by/4.0/. Chaos, D., Dormido-Canto, S., Bencomo, S.D.: Simulation
and experimental results of a new control strategy for point
stabilization of nonholonomic mobile robots. IEEE Trans.
Ind. Electron. 67(8), 6679–6687 (2019)
Appendix A 2. Brockett, R.W.: Asymptotic stability and feedback stabiliza-
tion. Differ. Geom. Control Theory 27(1), 181–191 (1983)
3. Fierro, R., Lewis, F.L.: Control of a nonholonomic mobile
Given the following first-order nonlinear differential
robot: backstepping kinematics into dynamics. J. Robot.
inequality: Syst. 14(3), 149–163 (1997)
4. Fukao, T., Nakagawa, H., Adachi, N.: Adaptive tracking
V̇ (x) + ιV ϕ (x) ≤ 0 (A1) control of a nonholonomic mobile robot. IEEE Trans. Robot.
where V (x) represents a positive Lyapunov function Autom. 16(5), 609–615 (2000)
5. Fierro, R., Lewis, F.L.: Control of a nonholonomic mobile
with respect to the state x ∈ R, ι > 0, 0 < ϕ < 1, robot using neural networks. IEEE Trans. Neural Netw. 9(4),
then for any given initial condition V (x(0)) = V (0), 589–600 (1998)
the function V (x) converges to the origin in the finite 6. Zeng-Guang, H., An-Min, Z., Long, C., Min, T.: Adaptive
time given by control of an electrically driven nonholonomic mobile robot
via backstepping and fuzzy approach. IEEE Trans. Control
V 1−ϕ (0) Syst. Technol. 17(4), 803–815 (2009)
tv ≤ . (A2) 7. Huang, J., Wen, C., Wang, W., Jiang, Z.P.: Adaptive output
ι(1 − ϕ) feedback tracking control of a nonholonomic mobile robot.
The derivation is referred to [25,26] and references Automatica 50(3), 821–831 (2014)
8. Zheng, J., Wang, H., Man, Z., Jin, J., Fu, M.: Robust motion
therein.
control of a linear motor positioner using fast nonsingu-
lar terminal sliding mode. IEEE ASME Trans. Mechatron.
20(4), 1743–1752 (2015)
Appendix B 9. Sun, Z., Xie, H., Zheng, J., Man, Z., He, D.: Path-following
control of Mecanum-wheels omnidirectional mobile robots
using nonsingular terminal sliding mode. Mech. Syst. Signal
For the system ξ̇1 = −α1 ξ1 − β1 sig(ξ1 )μ , define the Process. 147, 107128 (2021)
Lyapunov function Vξ1 = 21 ξ12 10. Yang, J.M., Kim, J.H.: Sliding mode control for trajectory
tracking of nonholonomic wheeled mobile robots. IEEE
V̇ξ1 = ξ1 ξ̇1 Trans. Robot. 15(3), 578–587 (1999)
11. Chwa, D.: Sliding-mode tracking control of nonholonomic
= − α1 ξ12 − β1 |ξ1 |μ+1 wheeled mobile robots in polar coordinates. IEEE Trans.
√ |ξ1 | Control Syst. Technol. 12(4), 637–644 (2004)
≤ − (α1 |ξ1 | − β1 |ξ1 |μ ) 2 √ 12. Xie, H., Zheng, J., Chai, R., Nguyen, H.T.: Robust tracking
2 control of a differential drive wheeled mobile robot using fast
√ 1 nonsingular terminal sliding mode. Comput. Electr. Eng. 96,
= − 2η1 V (t) 2 (B3) 107488 (2021)

123
Finite-time tracking control for nonholonomic wheeled mobile robot 1453

13. Youssef, E., Martins, N.A., Pieri, D., Moreno U.F.: PD- 20. Zhang, Y., Liu, G., Luo, B.: Finite-time cascaded tracking
super-twisting second order sliding mode tracking control control approach for mobile robots. Inf. Sci. 284, 31–43
for a nonholonomic wheeled mobile robot. In: Proceedings (2014)
of the 19th World Congress the International Federation of 21. Shi, S., Yu, X., Khoo, S.: Robust finite-time tracking control
Automatic Control, Cape Town, South Africa. August 24– of nonholonomic mobile robots without velocity measure-
29, pp. 3827–3832 (2014) ments. Int. J. Control 89(2), 411–423 (2016)
14. Begnini, M., Bertol, D.W., Martins, N.A.: A robust adap- 22. Zhai, J.Y., Song, Z.B.: Adaptive sliding mode trajectory
tive fuzzy variable structure tracking control for the wheeled tracking control for wheeled mobile robots. Int. J. Control
mobile robot: simulation and experimental results. Control 92(10), 2255–2262 (2019)
Eng. Pract. 64, 27–43 (2017) 23. Shao, K., Zheng, J., Huang, K., Wang, H., Man, Z., Fu,
15. Nath, K., Yesmin, A., Nanda, A., Bera, M.K.: Event- M.: Finite-time control of a linear motor positioner using
triggered sliding-mode control of two wheeled mobile robot: adaptive recursive terminal sliding mode. IEEE Trans. Ind.
an experimental validation. IEEE Trans. Emerg. Sel. Topics Electron. 67(8), 6659–6668 (2020)
Power Electron. 2(3), 218–226 (2021) 24. Utkin, V.I.: Sliding modes in control and optimization.
16. Mobayen, S.: Finite-time tracking control of chained-form Springer, Berlin (2013)
nonholonomic systems with external disturbances based on 25. Moulay, E., Perruquetti, W.: Finite time stability and sta-
recursive terminal sliding mode method. Nonlinear Dyn. bilization of a class of continuous systems. J. Math. Anal.
80(1–2), 669–683 (2015) 323(2), 1430–1443 (2006)
17. Cui, M., Liu, W., Liu, H., Jiang, H., Wang, Z.: Extended state 26. Plestan, F., Shtessel, Y., Brégeault, V., Poznyak, A.: New
observer-based adaptive sliding mode control of differential- methodologies for adaptive sliding mode control. Int. J. Con-
driving mobile robot with uncertainties. Nonlinear Dyn. trol 83(9), 1907–1919 (2010)
83(1–2), 667–683 (2016)
18. Ou, M., Sun, H., Zhang, Z., Gu, S.: Fixed-time trajectory
tracking control for nonholonomic mobile robot based on Publisher’s Note Springer Nature remains neutral with regard
visual servoing. Nonlinear Dyn. (2022) to jurisdictional claims in published maps and institutional affil-
19. Kanayama, Y., Kimura, Y., Miyazaki, F., Noguchi, T.: A iations.
stable tracking control method for an autonomous mobile
robot. In: Proceedings of the IEEE international conference
on robotics and automation, Cincinnati, OH, USA, pp. 384–
389 (1990)

123