Journal of Mechanical Science and Technology 38 (4) 2024
1007/s12206-024-0234-6
Journal of Mechanical Science and Technology 38 (4) 2024
Original Article
DOI 10.1007/s12206-024-0234-6
Keywords:
· Perturbed MIMO system
· Robust generalized dynamic inversion
· Recursive sliding mode control
· Adaptation laws
· Piece-wise Lyapunov function
Correspondence to:
Xiaodong Liu
xdliuros@dlut.edu.cn
Citation:
Abbas, N., Liu, X., Iqbal, J. (2024). Robust
GDI-based adaptive recursive sliding
mode control (RGDI-ARSMC) for a highly
nonlinear MIMO system with varying
dynamics of UAV. Journal of Mechanical
Science and Technology 38 (4)
(2024) ?~?.
http://doi.org/10.1007/s12206-024-0234-6
Received March 19th, 2023
Revised September 7th, 2023
Accepted December 3rd, 2023
† Recommended by Editor
Dongho Oh
© The Korean Society of Mechanical
Engineers and Springer-Verlag GmbH
Germany, part of Springer Nature 2024
DOI 10.
Robust GDI-based adaptive recursive
sliding mode control (RGDI-ARSMC) for
a highly nonlinear MIMO system with
varying dynamics of UAV
Nadir Abbas1, Xiaodong Liu1 and Jamshed Iqbal2
1 2
School of Electrical Engineering, Dalian University of Technology, Dalian 116024, China, School of
Computer Science, Faculty of Science and Engineering, University of Hull, Kingyston upon Hull, UK
Abstract The novelty of the proposed work lies in the control technique, referred to as the
robust generalized dynamic inversion based adaptive recursive sliding mode control (RGDI-
ARSMC), for addressing various challenges to control a highly coupled and perturbed system
called twin rotor MIMO systems (TRMS) UAV. The continuous disturbances, varying parameter
values, actuator failure, and unmodeled states are the challenges related to the proposed con-
troller. The method aims to effectively mitigate unwanted signals, including coupling effects, un-
known states, gyroscopic disturbance torque, parametric uncertainties, and other disturbances.
The control design process is divided into two phases: the first involves estimating the deviation
between the actual and desired output angles and conducting a stability phase analysis. The
confined stability-based Lyapunov stability was verified. While the second phase involves the
addition of a robust term and the use of an adaptive recursive design procedure to determine the
controller parameters for pitch and yaw angles. The proposed control strategy is compared with
other techniques such as classical sliding mode control, backstepping, and RGDI-SMC controls.
The proposed strategy is also implemented in real-time to characterize its performance. On the
basis of obtained results, the considered perturbations were effectively addressed by the aug-
mentation of adaptation laws and recursive control design.
1. Introduction
The control community field has long been interested in variable structure systems (VSS)
control because of their highly nonlinear behavior, varying dynamics, coupling effects, and sen-
sitivity to parametric disturbance during considered controller. The control of unmanned aerial
vehicles (UAVs) is particularly challenging due to matched and mismatched perturbations.
These systems are of interest because of their increasing use in various environments for
common security services and defense operations [1, 2]. Twin rotor MIMO systems (TRMS)
are a type of UAV system that has gained attention due to their ability to tilt their angle of flight,
hover, take-off, and land in irregular locations [3]. TRMS have high coupling and nonlinear dy-
namics, uncertainties, and gyroscopic torque, which make their control a challenging problem
for control researchers. These systems have expanding applications in various fields [4]. The
main challenges in TRMS include propeller rotation, coupling between rotors, changing propel-
ler rotation speed, sensitivity to parametric perturbations, and the time-varying nature of the
system [5, 6]. The controller task must guarantee the following challenges: (i) the varying dy-
namics of TRMS; (ii) the varying dynamics of the system include unmatched disturbances and
varying parametric dynamics; (iii) the unavailability of time-varying unmodeled states of the
system.
To address these issues, researchers have proposed a variety of linear, nonlinear, and intel-
ligent control strategies such as robust observer [7, 8], adaptive SMC strategy [9], learning-
based adaptive model predictive control (MPC), linear MPC [10, 11], nonlinear techniques
1
 Journal of Mechanical Science and Technology 38 (4) 2024
based on online adaptive laws for UAV in Ref. [12], validation
of adaptive RBFNN [13], adaptation laws based hierarchical
SMC approach [7, 14], and terminal SMC [15]. The adaptive
laws-based neural networks (NN) backstepping control strat-
egy in [16], and adaptive nonlinear recursive control based on
fuzzy logic control are discussed respectively in Ref. [17].
Type-2 fuzzy adaptive backstepping control for the nonlinear
coupled systems is also elaborated in Ref. [18], to evaluate the
significance of emerging research in control, the integration of
nonlinear control techniques with traditional control methods is
being studied in Refs. [19, 20]. The goal of these controllers is
to ensure that the TRMS system is stable and able to handle
internal and external disturbances, parametric uncertainties,
and unmodeled states. One of the first steps in designing a
controller for a non-linear system is to ensure its stability.
Higher order coupled non-linear systems must be analyzed
using mathematical tools [21, 22]. Nonlinear dynamic inversion
(NDI), which is a feedback linearization approach, is employed
at UAV to make the mathematical model simpler. However,
this method will ignore the important terms of nonlinearities,
singularities, and matrix inversions. To overcome these limita-
tions, generalized dynamic inversion (GDI) is used to solve
non-square inversions due to inverse problems in matrix [23].
The left inversion approach is used to establish linear differen-
tial equations and is inverted using the Moore-Penrose gener-
alized inverse (MPGI) technique, which is based on the Gre-
ville method [24]. This approach aims to avoid inversion issues
and also avoid ignoring important square or high-degree terms.
Robust generalized dynamic inversion (RGDI) controllers have
been designed for various aerospace and robotic applications
and have been shown to be effective [25]. In previous research,
RGDI-based control theory, mixed-optimization control, and
adaptations laws-based recursive sliding control are applied in
Refs. [26-28], at UAV, and its performance was characterized
through an experimental test of the prototype. Three different
methods (linear and nonlinear control) are elaborated on the
basis of their technical strength and a detailed comparison is
provided in Appendix A.5. However, it is important to address
the chattering effect, which is a problem that can occur in
higher-order complex systems with fast switching of control
inputs, in any control strategy. The RGDI-ARSMC directly de-
pends on the mathematical analysis-based model of the con-
sidered system [29]. In Ref. [30], a robust adaptation laws-
based controller was developed to address parametric uncer-
tainties and loss of thrust anomalies. The controller uses an
adaptive law to follow the reference point of the vehicle in both
vertical and horizontal positions. This control approach offers
finite-time convergence, reduces the problem, and provides a
law called parameter-tuning law to eliminate disturbances. An
adaptive recursive method with a finite-time convergence tech-
nique was employed to create a control law for stabilizing a
nonlinear system. The controller used a hybrid approach for full
system trajectory tracking and ensured stability in closed-loop
through hybrid Lyapunov analysis. The adaptive law was used
to calculate the controller coefficients, and the global stability of
2
DOI 10.1007/s12206-024-0234-6
the closed-loop system was proven through Lyapunov analysis.
When the system experiences perturbations, the relative de-
gree of the higher-order system can be affected, leading to
singularity issues. This may affect the stability and conver-
gence time of the closed-loop system. These perturbations can
be time-varying or state-dependent, and they may render the
system unstable, causing a loss of solution uniqueness. Fur-
ther studies are needed to understand these effects. The con-
tribution outline of the paper is enlisted as:
·Our research illustrates our ability to proficiently mitigate
the consequences of parametric alterations and unac-
counted-for system states by achieving an improved finite-
time convergence. Furthermore, we establish stability cri-
teria for the perturbed system by employing a piecewise
linear Lyapunov function.
·The Euclidean error norm method is utilized to compute
the discrepancy between the control angles and the de-
sired angles, serving as an estimation of the system's
state error.
·The adaptive recursive controller structure, constructed
upon the mathematical model of the analyzed system, ac-
centuates its remarkable resilience in the face of contem-
plated parametric variations and state-dependent unmod-
eled states.
·A control methodology is applied to achieve stabilization of
a nonlinear system through the utilization of an adaptive
recursive approach coupled with a finite-time convergence
technique. This approach not only mitigates the inherent
challenges but also incorporates a parameter-tuning law to
nullify disturbances. Consequently, the outcome is a con-
trol law guaranteeing finite-time convergence.
·The efficacy of the dynamic approach's robustness was
showcased under demanding conditions, including noise,
concurrent parametric fluctuations, and the application of
disturbance torque to both rotors. The effectiveness of this
approach is evident through a comparative analysis con-
ducted via MATLAB simulations.
·Based on experimental validation, suggestions were pro-
vided for control engineers to gain a better understanding
of the control design and system behavior.
The remainder of this paper has the following sections: the
mathematical modeling in Sec. 2, while the inclusion of robust
terms is provided in Sec. 3. The experimental setup with a
detailed description of the simulation response is provided in
Sec. 4. Finally, the conclusion based on validated results is
presented in Sec. 5.
2. Twin rotor MIMO system (TRMS) UAV
Before understanding the mathematical modeling of the
TRMS, it is important to understand the various parameters
and control outputs of the system [30]. The TRMS is a labora-
tory tool used to study the flight control of helicopters [31]. It
has two rotors, as shown in Fig. 1. The design of these rotors is
important as they are influenced by various forces such as
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

Fig. 1. Basic schematic sketch of TRMS [32]. Fig. 2. Block diagram of UAV.

gravity, propulsion, centrifugal force, friction, and disturbance these two systems may affect the overall performance and
torque. To counteract these forces, the control input is provided stability of the aircraft. The NDI control method, although effec-
through motors. To understand and simplify the mathematical tive in linearizing nonlinear systems is explained in detail in Sec.
model, it is important to understand the mathematical assump- 1 of supporting material. Nonlinear dynamic inversion (NDI),
tions that are used in the NDI process. The TRMS has two which is a feedback linearization approach, is employed at
degrees of freedom, allowing for movement in two directions, UAV to make the mathematical model simpler. However, this
the horizontal plane and the azimuthal plane, which are derived method will ignore the important terms of nonlinearities, singu-
in the model. larities, and matrix inversions. To overcome these limitations,
generalized dynamic inversion (GDI) is used to solve non-
dθ a1 2 b1 M 0.0326 square inversions due to inverse problems in matrix. The left
= τ 1 + τ 1 − g sin (θ ) + sin ( 2θ )ϕ 2
dt I1 I1 I1 2 I1 inversion approach is used to establish linear differential equa-
k tions and is inverted using the Moore-Penrose generalized
B1θ
−
I1 I1
(
θ − gy cos (θ ) ϕ a1τ 12 + b1τ 1 ) inverse (MPGI) technique, which is based on the Greville
method. Singularity based issue is being resolved by designing
dϕ a2 2 b2 B k
= τ 2 + τ 2 − 1ϕ ϕ − c 1.75 ( a1τ 12 + b1τ 1 ) . RGDI control provided with details in Sec. 2 of supporting ma-
dt I 2 I2 I2 I2 terial.

The principle of momentum conservation is also applied to

the rotor, resulting in similar momentum equations. Differential 3. Inclusion of adaptive recursive robust
equations for both rotors are derived as follows: term
The RGDI-based ARSMC law is a control method that com-
T10 k
τ1 = τ 1 + 1 u1 . bines the conventional generalized dynamic inversion (GDI)
T11 T11
method with an adaptive RSMC term. This method is specifi-
cally for controlling MIMO systems like TRMS. The implemen-
For tail motor: tation of ARSMC for twin-rotor MIMO systems can be challeng-
ing as it requires knowledge of the system's dynamics and the
T20 k design of a suitable Lyapunov function. Additionally, the per-
τ2 = τ 2 + 2 u2
T21 T21 formance of the ARSMC may be affected by the choice of ad-
aptation laws and the design of the sliding surface. To address
where k1 and k2 are the motor gains, T10 , T11 and T20 , these problems, a hybrid controller is developed to enable the
T22 are the motor parameters, τ 1 , τ 2 are rotors momentum, full system to track a trajectory and maintain stability in a
uθ , uφ are controlled actions of the vertical plane and hori- closed-loop configuration. The stabilizing functions counteract
zontal plane respectively. The nomenclature, parametric values nonlinearities that impact the system's stability. Previous re-
with units are given in Sec. 1 of supporting material. The block search [26-28] has shown that the inclusion of sliding mode
diagram in Fig. 2 represents two output states (pitch angle and control as a robust term, while ARSMC design can ensure
yaw angle), which are likely related to the control of the aircraft. stability of the complex system with a sharp response towards
The coupling effect is evident in a figure which illustrates both convergence.
rotors and angles with labeled blocks. This interaction between The sliding surface of TRMS with recursive backstepping can

3
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

be defined as: ⎧ α ξ1 = −cξ1 eξ1 + x1d

⎪
⎪ αξ2 = −cˆξ1 eξ1 + x1 + x2 − cξ1 ( cˆξ2 eξ2 + αξ1 ) x1d
s = eξ + cξa eξ ( t ) + cξa ∫ eξ ( t ) dt ⎨ (2)
⎪ αξ3 = −cξ4 eξ4 + x1d
⎪α = −e − cˆ e + x − x − c −cˆ e − x 
here cξ , cξ are the gain constants to enforce sliding, eξ is
a b
⎩ ξ4 ξ4 ξ 5 ξ5 4 5 ξ4 ξ4 ξ4 ( 2d x2d )
an error tracking state which will be calculated via adaptive
backstepping. The adaptive backstepping method is chosen for where ĉξ and ĉξ are the estimate of ĉξ and ĉξ respec-
2 5 2 5

its ability to provide stable robustness in desired position track- tively.

ing problems and its capability to control TRMS position in the Theorem 1: If the TRMS position system is governed by Eq.
presence of uncertainties and disturbances. In this section, an (1) while being accompanied by the adaptation law given by Eq.
adaptive backstepping for position trajectory tracking control is (2), the assurance of convergence is established [29]. Addi-
implemented, taking the output vector for TRMS position as [x, tionally, the adaptation laws for parametric stability are pro-
y]. The variables used in the design procedure for the MIMO vided as follows:
system are:
⎧iˆ = n c 2 ,
⎪ ξ2 1 ξ2

⎡θ = x1 and eξθ = x1 − x1d , eξθ ∈ eξ1 , eξ2 , eξ3 ⎤ ⎨

ˆ
⎪⎩iξ5 = n2cξ22 .
⎢φ = x .
⎣⎢ 3 and eξφ = x3 − x2d , eξφ ∈ eξ4 , eξ5 , eξ6 ⎥⎦⎥

Here, n1 and n2 represent positive constants. The applica-

Step 1: The first step of backstepping control design is to de- tion of [Barbalat's Lemma] is employed to assess the signifi-
fine the position tracking errors as: cance of the theorem under consideration [33]. Above lemma
must require as:
Lemma 1: The function f ( t ) is uniform function and
⎡eξ ⎤ ⎡x − x ⎤
⎡⎣ eξ z ⎤⎦ = ⎢ θ ⎥ => ⎢ 1 1d ⎥ t
e
⎣⎢ ξφ ⎦⎥ ⎣ x3 − x2d ⎦ limt →+∞ ∫ f (τ ) dτ validated, then f ( t ) must be converged
0

⎡ Tracking error of pitch position (angle) ⎤ at zero (origin) asymptotically [29].

=> ⎢ ⎥.
⎣ Tracking error of yaw position (angle) ⎦ Proof: To provide confined convergence of the system with
explaining ĉξ , and ĉξ as the parameters of a system, The
2 5

Lyapunov stability analysis is applied. For ĉξ : the candidate

Step 2: Introducing new arbitrary control input 2

function is introduced for the considered subsystem.

⎡ eξ ⎤ ⎡ x1 − x1d ⎤ 1 2
⎢ 1⎥ ⎢ ⎥ vξ2 = vξ1 + cξ (3)
⎣⎡eξθ ⎦⎤ = ⎢ eξ2 ⎥ => ⎢ x2 − α1 ⎥ 2n1 2
⎢ eξ ⎥ ⎢⎣ x5 − α 2 ⎥⎦
⎣ 3⎦
⎡Vξ1 ⎤ ⎡1 / 2eξ21 ⎤ where cξ denotes the error. The time derivative of Eq. (3) is
2

⎢ ⎥ ⎢ ⎥
⎡⎣Vξθ ⎤⎦ = ⎢Vξ2 ⎥ => ⎢1 / 2eξ22 ⎥ .
1 
⎢Vξ ⎥ ⎢1 / 2eξ ⎥
2 vξ2 = −cξ1 eξ21 − cˆξ2 eξ22 + cξ cξ
⎣ 3⎦ ⎣ 3 ⎦ n1 2 2
1 
= −cξ1 eξ21 − ( cξ2 − cξ2 ) eξ22 − cξ cˆξ (4)
Step 3: The required condition for Lyapunov function to fulfill n1 2 2
the asymptotic stability as: ⎛ 1 ⎞
= −cξ1 eξ21 − cξ2 eξ21 + cξ2 ⎜ eξ22 − cˆξ2 ⎟ .
⎝ n1 ⎠
V = −Vξ12 − Vξ22 − Vξ23
⎡ eξ4 ⎤ ⎡ x3 − x2d ⎤ ⎛ 1 ⎞
In the Eq. (4), the mentioned term cξ ⎜ eξ2 − cˆξ ⎟ will be
⎡eξ ⎤ = ⎢⎢ eξ ⎥⎥ => ⎢ x4 − α 3 ⎥ ⎝ n1 ⎠ 2 2 2

⎣ φ⎦ 5 ⎢ ⎥
⎢ ⎥ ⎣⎢ x6 − α 4 ⎦⎥
⎣ eξ6 ⎦ equal to zero. By taking cξ as positive constant, the deriva-
2

⎡Vξ4 ⎤ ⎡ eξ4ξ4 ⎤ ⎡eξ4 ( x3 − x2 d ) ⎤ tive of cξ can be expressed as cξ = 0 − cˆξ . Now the candi-
⎡Vξ ⎤ = ⎢⎢Vξ ⎥⎥ = ⎢⎢ eξ ξ5 ⎥⎥ => ⎢⎢ eξ ( x4 − α 3 ) ⎥⎥ .
2 2 2

 (1)
⎣ φ⎦ 5 5 5 date function will be elaborated in Eq. (4), can be written as:
⎢Vξ ⎥ ⎢ eξ ξξ ⎥ ⎢ eξ ( x6 − α 4 ) ⎥
⎣ 6⎦ ⎣ 6 6⎦ ⎣ 6 ⎦
vξ2 = −cξ1 eξ21 − cξ2 eξ22 ≤ 0 . (5)
The arbitrary control laws for the pitch and yaw position are
formulated as follows: Thus, the stability condition is satisfied through Eq. (5). To

4
 Journal of Mechanical Science and Technology 38 (4) 2024
guarantee the stability of the positioning system, the Lyapunov
candidate function for the system's position is chosen:
1⎛ 1 ⎞
vξ = ⎜ eξ21 + eξ22 + cξ22 + eξ24 2eξ25 + n2 cξ25 ⎟ .
2⎝ n1 ⎠
The time derivative of the Lyapunov position is
vξ = ( −cξ1 eξ21 − cξ2 eξ22 − cξ4 eξ24 − cξ5 eξ25 ) ≤ 0 .
where cξ , cξ and cξ , cξ are the parameters of pitch an-
1 2 5 4
gle as well as yaw angle respectively. Therefore, the system's
stability is ensured by Eqs. (6) and (7), providing the capability
of flight trajectory tracking. In this section, an adaptive recursive
method for the trajectory tracking of a considered system is
developed. Adaptive laws are employed to calculate the pa-
rameter of the proposed controller. By taking a derivative of
sliding surface,
¨
s = eξ + cξ eξ ( t ) + cξ eξ ( t ) .
By using Eq. (8) and rearranging equations to get ensure the
negative definite provided in Sec. 3 of supporting material.
Thus, it is not possible to ensure the finite-time closed-loop
stability of sliding mode dynamics. Nevertheless, it is possible
to attain semi-global practical stability of the ARSMC through
appropriate design of the SMC gain.
Theorem 2: The real integers C * > 0 for all real integers
ρ ∈ (0, 1) which will provide the negative definite of V with
sliding dynamics elaborated by in Sec. 3 of supporting material
for all values of ρ ( eξ ; eξ ; v; t ) > ρ * and C > C * .
z r
Proof: Let ρ * be a real scalar number as a constant entry
in the coverage of ρ ( eξ ; eξ ; v; t ) , ρ * , i.e., ρ ∈ (0, 1). Also,
define C ( eξ ; eξ ; t ) as:
z r
z r
ρ* −1
C ( eξ z ; eξr ; t ) = − B ( eξ , eξ , t ) .
ρ* z r
It follows that C ( eξ ; eξ ; t ) > C ( eξ ; eξ ; v; t ) whenever ρ ( eξ ;
z r z r
eξr ; v; t ) > ρ . Accordingly, let D be a neighborhood of ( eξ ; eξ
*
= ( 02 ; 02 ) , and choose a sliding gain constant C
C * > maxC ( eξ z ; eξr ; t ) Then the negative definite value of V <
D
0 ensures along any closed-loop track that starts within D
whenever ρ ( eξ ; eξ ; v; t ) ≥ ρ * and C > C * . The finite number
z r
C *
is ensured for any range D because of B ( eξ , eξ , t ) ,
which is globally bounded by virtue of implementing the DSGI
A* , which results in globally bounded trajectories. The control-
ler design steps with its variable representation via synoptic
scheme is described in Fig. 3, to control the highly coupled
nonlinear system.
DOI 10.1007/s12206-024-0234-6
(6)
(7)
Fig. 3. Block diagram of controller for UAV.
Remark 1: According to the statement of theorem 2, the gain
must be positive and gain C increased in a way that the posi-
tive bound ρ * will be obtained to ensure of the situation (con-
dition) is V < 0 will hold D for ρ ( eξ ; eξ ; v; t ) < ρ * . A specific
domain must be defined by ρ ( eξ ; eξ ; v; t ) < ρ * will be fol-
z r
z r
lowed by state trajectory ea which should be in this range.
The driving ρ * is close to the zero for driving ea which is
also close to zero for uniformly bounded. The following condi-
(8)
tion makes the attitude error trajectory eξ = 02x1 stable. Due z
increase in C * will affect the enlargement of D, it must be
followed as semi-global stability. In the SMC method, the sys-
tem's trajectory is guided along a manifold by the use of multi-
ple control structures that follow a specified switching condition.
The system's structure is defined by switching functions, which
can be either scalar or vector. The switching surface, repre-
sented by s(x) = 0, is a line on the phase plane. To verify its
effectiveness, we conducted a comparison study and evalu-
ated the performance of the controller in terms of vertical stabil-
ity (pitch angle) and horizontal stability (yaw angle). Additionally,
the controller was tested in the presence of disturbances, such
as external disturbances, parametric uncertainties, coupling
effects, and noise signals Secs. 4 and 5 of supporting material.
The step input tracking response in the simulation confirms that
the control theory regarding the convergence time of the an-
gles, specifically the pitch angle and yaw angle, is logical and
results in a sharp stability response. While the detailed re-
sponse of different control strategies (SMC, backstepping,
z
RGDI based SMC) in Sec. 4 of supporting material and for the
worth of the proposed strategy is validated by providing differ-
z r
) ent reference inputs in Sec. 5 of supporting material.
*
such that
4. Hardware and system setup
The realization process and key ideas for implementation are
briefly elaborated with system interconnection details. The real-
time implementation of the prototype can be seen in Fig. 4 with
z r
a laboratory setup and other components. It tests the effective-
ness of the simulation results by applying various disturbances
to each rotor of the highly coupled system. The controller is
designed to handle disturbances such as noise, unmodeled
states, uncertainties, and coupling effects. The experimental
response of the pitch angle with a sinusoidal input as a refer-
5
 Journal of Mechanical Science and Technology 38 (4) 2024
Fig. 4. Experimental apparatus (prototype) with real-time workshop.
Fig. 5. Pitch angle-experimental sinusoidal response.
ence input is applied in Figs. 5 and 6 represents the response
of the yaw angle under all applied disturbances (parametric
disturbances, coupling effects, and noise signal). Some impor-
tant points that can be verified from the obtained results and
some observations as a control engineer are also described
here to understand issues related to hardware implementation.
The pitch angle response of the prototype elaborates that the
controller performs very well against all disturbances and the
convergence time also verifies the logical time to converge the
main rotor before the tail rotor (yaw angle) response. A subplot
also shows the attenuation due to the applied noise signal and
6
DOI 10.1007/s12206-024-0234-6
Fig. 6. Yaw angle-experimental sinusoidal response.
Fig. 7. Control actions under sinusoidal input.
also the range of disturbance being tackled by an efficient con-
troller. Other subplots verify how the controller can manage the
considered disturbance over time. The system response be-
comes more stable over time, which can make a remarkable
difference with respect to other applied control methods to date.
The convergence time and effect of attenuation are more in the
yaw angle. The reason behind this change is that the coupling
effect due to disturbance torque is more due to the weighted
rotor as well as the blades of the main rotor. The sudden and
sharp variation in yaw angle is generated by different factors
created by the main rotor. Noise (disturbance) can greatly im-
pact the accuracy of actuators and input control signals, caus-
ing errors. To address this issue, a first-order filter based on the
Butterworth filter is employed to filter out noise and obtain the
actual actuator input. The control actions of both angles are
represented in Fig. 7 under sinusoidal input.
5. Conclusions
The objective of this study was to design an adaptive recur-
sive technique based on generalized dynamic inversion for a
highly nonlinear and cross-coupled multiple input multiple output
(MIMO) system. This technique was tested on a UAV (TRMS)
for flight path tracking and stabilization and used to develop
 Journal of Mechanical Science and Technology 38 (4) 2024
novel robust controllers for controlling the UAV in the presence
of uncertainties. The control strategy consisted of two phases:
understanding the behavior of the system, which is challenging
due to high coupling and disturbance torque, and designing
time-varying dynamic constraints. Nonlinear dynamic inversion
(NDI) was used to provide a simplified model of TRMS. GDI
was used to address the limitations of NDI and singularity is-
sues. In the second phase, output states were tracked by refer-
ence trajectories, and sinusoidal reference tracking of states
ensured robustness and stability validation against the nonlinear
behavior of the coupled system with uncertainties. The inclusion
of a robust term in previous research controllers was developed
to increase robustness against external perturbations and un-
modeled states. The RGDI-ARSMC method is based on the
mathematical model of the TRMS system which addresses
varying parametric uncertainties and loss of thrust anomalies.
The controller uses an adaptive law to track the desired trajec-
tory of the vehicle in both vertical and horizontal angles (posi-
tions). This control approach offers finite-time convergence,
reduces the problem, and provides a parameter-tuning law to
eliminate external perturbations. A novel reaching law based on
an adaptive recursive approach with a finite-time convergence
technique was used to generate a control law for stabilizing the
nonlinear system. The hybrid controller was designed for full
system trajectory tracking and stability in closed-loop and pro-
vided by using hybrid Lyapunov analysis. The adaptive law was
used to estimate the controller coefficients and the global stabil-
ity of the closed-loop system was proved using Lyapunov
analysis. Accurate fast-tracking and error convergence per-
formance in all cases of perturbations (noise matrix, parametric
disturbance) reveal the effectiveness of the applied controller.
Numerical simulation and real-time experiments were con-
ducted to evaluate the performance of the developed control
system. The experimental results also provide some sugges-
tions for control engineers to consider.
·Experimental testing has confirmed that the robust control
system's performance during real-time implementation is
greatly influenced by the particular adaptation law applied
for parameter estimation.
·Substantial uncertainties in the physical parameters can
introduce nonlinearity in the system's behavior, necessitat-
ing the use of a recursive adaptation law to address this
issue accurately. Additionally, the presence of high-
amplitude noise signals can severely disrupt input actua-
tors and the high-voltage range.
5.1 Insights of future guidance
There are some general insights into future offers and guide-
lines regarding proposed research.
·Future research is likely to focus on developing advanced
control algorithms that integrate the principles of GDI-
based adaptive control and recursive sliding mode control.
These algorithms should be tailored to address the specific
challenges posed by highly nonlinear MIMO UAV systems
DOI 10.1007/s12206-024-0234-6
with varying dynamics.
·Guidelines will emphasize the importance of robust pa-
rameter estimation techniques that can adapt to the
changing dynamics of UAVs in real-time. This might in-
clude machine learning-based approaches for online pa-
rameter identification.
·Future work may involve the development of more accu-
rate nonlinear dynamic’s models for UAVs. This could in-
clude accounting for aerodynamic effects, wind distur-
bances, and variations in vehicle configurations.
·Researchers and practitioners will likely be encouraged to
conduct extensive experimental validations to demonstrate
the effectiveness of RGDI-ARSMC in real-world scenarios.
This could involve test flights with actual UAV platforms to
showcase the control system's robustness.
·Guidelines may stress the importance of designing control
systems that are not only robust but also fault-tolerant and
resilient. This is crucial for UAVs operating in challenging
environments where failures or disturbances can occur.
·Future offers might include the development of open-
source software and resources for researchers and engi-
neers working on RGDI-ARSMC for UAVs. This can pro-
mote collaboration and accelerate advancements in the
field.
Acknowledgments
This research work is funded by the National Key R&D Pro-
gram of China under Grant No. 2018YFB1702200.
References
[1] M. Bucolo, A. Buscarino, L. Fortuna and S. Gagliano, Bifurca-
tion scenarios for pilot induced oscillations, Aerospace Science
and Technology, 106 (2020) 106194.
[2] B. Geranmehr, E. Khanmirza and S. Kazemi, Trajectory control
of aggressive maneuver by agile autonomous helicopter, Pro-
ceedings of the Institution of Mechanical Engineers, Part G:
Journal of Aerospace Engineering, 233 (4) (2019) 15261536.
[3] S. Ahmad, A. A. Uppal, M. R. Azam and J. Iqbal, Chattering
free sliding mode control and state dependent kalman filter de-
sign for underground gasification energy conversion process,
Electronics, 12 (4) (2023) 876.
[4] D. Y. Dube and H. G. Patel, Suppressing the noise in meas-
ured signals for the control of helicopters, Fluctuation and
Noise Letters, 18 (1) (2019) 1950002.
[5] H. Liu, J. Xi and Y. Zhong, Robust attitude stabilization for
nonlinear quadrotor systems with uncertainties and delays,
IEEE Transactions on Industrial Electronics, 64 (7) (2017) 5585-
5594.
[6] O. Mofid and S. Mobayen, Adaptive sliding mode control for
finite-time stability of quad-rotor uavs with parametric uncer-
tainties, ISA Transactions, 72 (2018) 1-14.
[7] K. Wulff, T. Posielek and J. Reger, Compensation of unmatched
disturbances via sliding-mode control: A comparison of classical
7
 Journal of Mechanical Science and Technology 38 (4) 2024
results and recent methods using integral and higher-order slid-
ing-mode, Variable-Structure Systems and SlidingMode Con-
trol: From Theory to Practice, 53 (2) (2020) 237-272.
[8] F. R. López-Estrada, J.-C. Ponsart, D. Theilliol, Y. Zhang and
C.-M. Astorga-Zaragoza, LPV model-based tracking control
and robust sensor fault diagnosis for a quadrotor UAV, Journal
of Intelligent and Robotic Systems, 84 (1) (2016) 163-177.
[9] H. Castañeda and J. L. Gordillo, Spatial modeling and robust
flight control based on adaptive sliding mode approach for a
quadrotor uav, Journal of Intelligent and Robotic Systems, 93
(1) (2019) 101-111.
[10] G. Cao, E. M.-K. Lai and F. Alam, Gaussian process model
predictive control of an unmanned quadrotor, Journal of Intelli-
gent and Robotic Systems, 88 (1) (2017) 147-162.
[11] M. Mehndiratta, E. Kayacan, S. Patel, E. Kayacan and G.
Chowdhary, Learning-based fast nonlinear model predictive
control for custom-made 3 d printed ground and aerial robots,
Handbook of Model Predictive Control, Birkhäuser, Cham
(2019) 581605.
[12] R. C. Avram, X. Zhang and J. Muse, Quadrotor actuator fault
diagnosis and accommodation using nonlinear adaptive esti-
mators, IEEE Transactions on Control Systems Technology,
25 (6) (2017) 2219-2226.
[13] M. Z. Ghellab, S. Zeghlache, A. Djerioui and L. Benyettou,
Experimental validation of adaptive rbfnn global fast dynamic
terminal sliding mode control for twin rotor mimo system
against wind effects, Measurement, 168 (2021) 108472.
[14] J. Moreno-Valenzuela, R. Pérez-Alcocer, M. Guerrero-Medina
and A. Dzul, Nonlinear pid-type controller for quadrotor trajec-
tory tracking, IEEE/ASME Transactions on Mechatronics, 23
(5) (2018) 2436-2447.
[15] J.-J. Xiong and G.-B. Zhang, Global fast dynamic terminal
sliding mode control for a quadrotor UAV, ISA Transactions, 66
(2017) 233-240.
[16] H. Liu, Y. Pan, J. Cao, H. Wang and Y. Zhou, Adaptive neural
network backstepping control of fractional-order nonlinear sys-
tems with actuator faults, IEEE Transactions on Neural Net-
works and Learning Systems, 31 (12) (2020) 5166-5177.
[17] H. Liu, Y. Pan, S. Li and Y. Chen, Adaptive fuzzy backstep-
ping control of fractional-order nonlinear systems, IEEE Trans-
actions on Systems, Man, and Cybernetics: Systems, 47 (8)
(2017) 2209-2217.
[18] A. A. Jafari, S. M. A. Mohammadi and M. H. Naseriyeh, Adap-
tive type-2 fuzzy backstepping control of uncertain fractionalor-
der nonlinear systems with unknown dead-zone, Applied
Mathematical Modelling, 69 (2019) 506-532.
[19] J. Iqbal, M. Ullah, S. G. Khan, B. Khelifa and S. Cuković,
Nonlinear control systems-a brief overview of historical and re-
cent advances, Nonlinear Engineering, 6 (4) (2017) 301-312.
[20] S. H. Shah, S. Khan, I. ul Haq, K. Shah and A. Abid, Compli-
ance control of robotic walk assist device via integral sliding
mode control, 2019 16th International Bhurban Conference on
Applied Sciences and Technology (IBCAST), Islamabad, Paki-
stan (2019) 515-520.
[21] A. N. Panteli, D. I. Manolas, V. A. Riziotis and K. V. Spiliopou-
8
DOI 10.1007/s12206-024-0234-6
los, Comparative study of two geometrically non-linear beam
approaches for the coupled wind turbine system, Journal of
Wind Engineering and Industrial Aerodynamics, 231 (2022)
105231.
[22] S. Murtaza, Z. Ahmad, I. E. Ali, Z. Akhtar, F. Tchier, H.
Ahmad and S.-W. Yao, Analysis and numerical simulation of
fractal-fractional order non-linear couple stress nanofluid with
cadmium telluride nanoparticles, Journal of King Saud Univer-
sity-Science, 35 (4) (2023) 102618.
[23] A. H. Bajodah, H. Mibar and U. Ansari, Aircraft motion de-
coupling of roll and yaw dynamics using generalized dynamic
inversion control, 2018 26th Mediterranean Conference on
Control and Automation (MED), Zadar, Croatia (2018) 1-9.
[24] U. Ansari, A. H. Bajodah and M. T. Hamayun, Quadrotor
control via robust generalized dynamic inversion and adaptive
non-singular terminal sliding mode, Asian Journal of Control,
21 (3) (2019) 1237-1249.
[25] U. Ansari and A. H. Bajodah, Robust generalized dynamic
inversion based control of autonomous underwater vehicles,
Proceedings of the Institution of Mechanical Engineers, Part M:
Journal of Engineering for the Maritime Environment, 232 (4)
(2018) 434-447.
[26] N. Abbas, X. Pan, A. Raheem, R. Shakoor, Z. A. Arfeen, M.
Rashid, F. Umer, N. Safdar and X. Liu, Real-time robust gen-
eralized dynamic inversion based optimization control for cou-
pled twin rotor mimo system, Scientific Reports, 12 (1) (2022)
17852.
[27] N. Abbas and X. Liu, A mixed dynamic optimization with μ-
synthesis (dk iterations) via gain scheduling for varying dynam-
ics of decoupled twin-rotor MIMO system based on the method
of inequality, Journal of Control Engineering and Applied In-
formatics, 24 (4) (2022).
[28] N. Abbas, Z. Abbas and X. Liu, Adaptive recursive sliding
mode control (ARSMC)-based UAV control for future smart cit-
ies, Applied Sciences, 13 (11) (2023) 6790.
[29] S. B. F. Asl and S. S. Moosapour, Adaptive backstepping fast
terminal sliding mode controller design for ducted fan engine of
thrust-vectored aircraft, Aerospace Science and Technology,
71 (2017) 521-529.
[30] Z. T. Dydek, A. M. Annaswamy and E. Lavretsky, Adaptive
control of quadrotor uavs: A design trade study with flight
evaluations, IEEE Transactions on Control Systems Technol-
ogy, 21 (4) (2012) 1400-1406.
[31] A. Rahideh and M. Shaheed, Mathematical dynamic modelling
of a twin-rotor multiple input-multiple output system, Proceed-
ings of the Institution of Mechanical Engineers, Part I: Journal of
Systems and Control Engineering, 221 (1) (2007) 89-101.
[32] S. Khan, S. Bendoukha, W. Naeem and J. Iqbal, Experimen-
tal validation of an integral sliding mode-based lqg for the pitch
control of a uav-mimicking platform, Advances in Electrical and
Electronic Engineering, 17 (3) (2019) 275-284.
[33] M. Ran, Q. Wang, D. Hou and C. Dong, Backstepping design
of missile guidance and control based on adaptive fuzzy sliding
mode control, Chinese Journal of Aeronautics, 27 (3) (2014)
634-642.
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

Appendix f1 = x1

A.1 Feedback linearization inversion f 2 = x2

f 3 = x3
The control process regarding linearization, also known as f 4 = x4
NDI control, is a strategy that linearizes nonlinear systems by
f 5 = x5
canceling out nonlinear terms and simplifying the model. The
system modeling is based on state vectors to represent the f 6 = x6 .
system's dynamics for the control algorithm. This method pro-
vides enough knowledge of nonlinear systems to design a We know that to linearize the system take Jacobean and
sophisticated controller and is widely used in aerospace and then check stability at origin.
robotics applications:
⎡ ∂f1 "
∂f1 ⎤
x ( t ) = Ax ( t ) + Bu ( t ) ⎢ ∂x1 ∂x n ⎥
⎢ ⎥
y ( t ) = Cx ( t ) J =⎢ # % # ⎥.
⎢ ∂f n ∂f n ⎥
⎢ ∂x " ∂x n ⎥⎦
⎣ 1
while u ∈ R shows the signal at the input signal, y ∈ R
expresses the output of the system and u ∈ R represents the
states of the system in the form of a state vector. The system's After taking Jacobean and putting point (0, 0) then resulting
vectors can be explained as: system matrices are given below.

⎡ 0 0 ⎤
T
x = ⎡⎣θ θ ϕ ϕ ⎤⎦ 1 0 0 0
⎢ M ⎥
⎢− g − B1Ψ b1
y = [θ ϕ ] 0 ⎥
T
0 0
⎢ I1 I1 I1 ⎥
⎢ ⎥
where θ is elevation plane angle, φ is azimuth angle, τ 1 ⎢ 0 0 0 1 0 0 ⎥
⎢ B1ϕ k b2 ⎥
shows main rotor momentum and τ 2 shows tail rotor momen- A=⎢ 0 0 0 − − c 1.75 ⎥
tum. ⎢ I2 I2 I2 ⎥
⎢ ⎥
⎢ 0 T
0 0 0 − 10 0 ⎥
⎢ T11 ⎥
A.1.1 Linearization of TRMS ⎢ ⎥
⎢ 0 T20 ⎥
0 0 0 0 −
To linearize the system first we write its nonlinear equations ⎢⎣ T21 ⎥⎦
in state space equations and all parameters are shown in
⎡0 0⎤
equations. State space equations of non-linear system are ⎢0
⎢ 0 ⎥⎥
given below.
⎢0 0⎥
⎢ ⎥
dΨ M 0 0⎥
a b
B=⎢
0.0326
= 1 τ 12 + 1 τ 1 − g sin ( Ψ ) + sin ( 2Ψ ) ϕ 2 ⎢
dt I1 I1 I1 2 I1 k1 ⎥
⎢0 ⎥
B1Ψ  k gy ⎢ T11 ⎥
− Ψ− cos ( Ψ ) ϕ ( a1τ 12 + b1τ 1 ) ⎢k ⎥
I1 I1 ⎢ 2 0⎥
d ϕ a2 2 b2 B k ⎣⎢ T21 ⎦⎥
= τ 2 + τ 2 − 1ϕ ϕ − c 1.75 ( a1τ 12 + b1τ 1 )
dt I2 I2 I2 I2 ⎡1 0 0 0 0 0 ⎤
C=⎢ ⎥.
d τ1 T k ⎣0 0 1 0 0 0 ⎦
= − 10 τ 1 + 1 u1
dt T11 T11
d τ2 T k By putting all constant values of TRMS from table we get fi-
= − 20 τ 2 + 2 u2 .
dt T21 T21 nal matrices as

To linearize the system, Let the plant can be represented as ⎡ 0 1 0 0 0 0 ⎤

⎢ −4.7059 −0.0882 0 0 1.3588 0 ⎥⎥
⎢
x = Ax + Bu (A.1) ⎢ 0 0 0 1 0 0 ⎥
y = Cx . (A.2) A=⎢ ⎥
⎢ 0 0 0 −5 1.617 4.5 ⎥
⎢ 0 0 0 0 −0.9091 0 ⎥
Where x ∈ R as states, u ∈ R as the control input and ⎢ ⎥
y ∈ R as the measured output. Consider ⎣⎢ 0 0 0 0 0 −1 ⎦⎥

9
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

Table A.1. TRMS system parameters. Table A.2. Previous research vs proposed strategy.
-2 2
I1 = Moment of inertia of vertical rotor 6.8×10 kgm Settling time pitch Settling time yaw Robust against
Control strategies
I2 = Moment of inertia of horizontal rotor 2×10 kgm
-2 2 angle (rad/s) angle (rad/s) perturbations

a1 = Static characteristic parameter 0.0135 Mixed optimization

3.4 3.6 Good
with μ-synthesis
b1 = Static characteristic parameter 0.0924
RGDI based
a2 = Static characteristic parameter 0.02 4 4.3 Good
optimi-zation
a2 = Static characteristic parameter 0.09 ARSMC 3.2 3.4 Good
Mg = Gravity momentum 0.32 Nm Proposed strategy 0.8 1 Good
-3
B1Ψ = Friction momentum function parameter 6×10 Nm·s/rad
-3 2
B2 Ψ = Friction momentum function parameter 1×10 Nms /rad
-1 (pitch angle and yaw angle), which are likely related to the
B1ϕ = Friction momentum function parameter 1×10 Nm·s/rad
-2 2
control of the aircraft. The coupling effect is evident in a figure
B2ϕ = Friction momentum function parameter 1×10 Nms /rad
which illustrates both rotors and angles with labeled blocks.
kgy = Gyroscopic momentum parameter 0.05 s/rad
This interaction between these two systems may affect the
k1 = Motor 1 gain 1 2 3.5 0.2 − 1.1 overall performance and stability of the aircraft. The NDI control
K2 = Motor 2 gain 0.8 method, although effective in linearizing nonlinear systems,
T11 = Motor 1 denominator parameter 1.1 does come with some challenges during implementation.
T10 = Motor 1 denominator parameter 1 These issues are related to singularity (where control becomes
T21 = Motor 2 denominator 1 difficult due to near-zero input), cancellation of important
T20 = Motor 1 denominator parameter 1 nonlinear terms that may affect performance, and the need for
large control inputs. Table A.2 provides the specific values with
kc = Cross reaction momentum gain 2
units of the system's parameters. These are crucial in under-
standing the system's behavior and implementing the NDI con-
⎡0 0 ⎤
⎢0 0 ⎥ trol strategy.
⎢ ⎥
⎢0 0 ⎥
B=⎢ ⎥ A.2 RGDI control design
⎢0 0 ⎥
⎢1 0 ⎥ To design control law for VSS, equations of TRMS are rear-
⎢ ⎥
⎣⎢0 0.8⎥⎦ ranged as:
⎡1 0 0 0 0 0 ⎤
C=⎢ ⎥ x zξ = xrξ
⎣0 0 1 0 0 0 ⎦
(A.3)
( )
xrξ = A xrξ , t + Buξ
where values of constants are given in Table A.1.

A.1.2 State space equation’s of linearized model here xzξ = [ x1 , x3 ] and xrξ = [ x2 , x4 ] , which are dynamic

Nonlinear equations of TRMS are linearized and state space states having pitch angle and yaw angle. uξ = ⎡uθξ wθξ ,
⎣ ( )
equations are obtained from linearized model of TRMS.
( )
uϕξ wϕξ ⎤
⎦
⎡ ⎤
and w = ⎣ wθξ wϕξ ⎦ are the angular position (rota-
x1 = x2 tional angle) of both rotors (main rotor and tail rotor). The
x2 = −4.7059 x1 − 0.0882 x2 + 1.3588 x5 Euclidean norm of a linear time-invariant (LTI) system is a
x3 = x4 measure of the distance of a vector from the origin, also known
x4 = −5 x4 + 1.617 x5 + 4.5 x6 as the Euclidean distance. This can be thought of as the mag-
x5 = −0.9091x5 + u1 nitude of the difference between the true value of a system and
its approximate value. In the present research, we present a
x6 = − x6 + 0.8u 2 .
method for calculating the deviation of the system state from its
desired state, referred to as the Euclidean error norm, by com-
The state variables of pitch position and yaw position are
paring control angles to desired angles. The state deviation of
also simplified as; the attitude deviation function ρ can be defined mathemati-
⎡θ ⎤ ⎡ x1 ⎤
cally using the Euclidean error norm as shown in the following
⎢φ ⎥ = ⎢ x ⎥ . expression:
⎣ ⎦ ⎣ 3⎦
ρ = ezξ 2w = r1e12ξ + r2e22ξ = eTzξ D ( r1 , r2 ) ezξ (A.4)
The block diagram in Fig. 2 represents two output states

10
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

here e1 = x1 − x1 , e2 = x3 − x2d and ezξ = ⎡⎣e1ξ , e2ξ ⎤⎦ . Remem- will be perturbed by a particular issue (singularity) during dy-
d
namic inversion. The matrix rank may be changed and discon-
ber that the value of r1 & r2 are positive constants with defined
tinuity will come into the MPGI. The scaling factor augmenta-
integers and D represents the matrix with diagonal values r1 , tion can resolve this kind of issues and leads towards a con-
and r2 as diagonal position in the matrix as a diagonal mem- fined solution as discussed in Ref. [31] with demonstration
ber. The equations obtained, will follow the order of differential provided as:
Eq. (A.4) as well as the same attitude deviation function:
γ
v ( t ) = −v ( t ) + , v ( 0) > 0 (A.11)
ρ + c1 ( t ) ρ + c2 ( t ) ρ = 0 ezξ ( t )
2
(A.5)

here c1 , and c2 are parameters that must be asymptotically
converged at defined dynamics of the focused system [33].
The computation of varying dynamics can be obtained by tak-
ing the time derivative as:
where ezξ ( t ) = ⎡⎣e1ξ ( t ) , e2ξ ( t ) ⎤⎦ . The mathematical expression
Eq. (A.11) verify that how the system can be stable asymptoti-
cally. The modified equations of the system are also expressed
below as a function:

( ) (
A xzξ , xrξ , t U = B xzξ , xrξ , t . ) (A.6)
( ) (
A* xzξ , xrξ , v, t = AT xzξ , xrξ , t )
(A.12)
{ A( x ) ( ) }
−1
xd = [ 
here  x2 d ] . The obtained values can be used in the
x1d ,  zξ , xrξ , t AT xzξ , xrξ , t + v ( t ) .
Eq. (A.3), and the transformed system can be expressed here
as:
The extended condition of controller input is as:

( ) (
A xzξ , xrξ , t U = B xzξ , xrξ , t ) (A.7)
( ) ( ) (
U * = A xzξ , xrξ , v, t B xzξ , xrξ , t + P xzξ , xrξ , t Y . ) (A.13)

The control function is given as:

The modified equation structure of dynamically scaled gen-
eralized inverse (DSGI) is given as:
B = −2eTzξ D ( r1 , r2 ) ezξ − 2c1eTzξ D ( r1 , r2 ) ezξ
− c2 eTzξ D ( r1 , r2 ) ezξ − 2ezTξ D ( r1 , r2 ) A+ (A.8)
( ) ( ) ( ) (
xrξ = A* xrξ , t + BA xzξ , xrξ , v, t B xzξ , xrξ , t + P xzξ , xrξ , t Y . (A.14) )
− 2e D ( r1 , r2 ) 
T
zξ xd .

With the reference to Eq. (A.5), the infinite system must be
parameterized via a known method called the Goreville ap-
proach. This method recommends that:
( ) ( ) (
U = A+ xzξ , xrξ , t B xzξ , xzξ , t + P xzξ , xzξ , t Y
here the mathematical notation Y defines the expression for
the control vector for the system and the projection matrix for
the considered states of the system represented by P . The
notation A+ is known as the MPGI and can be explained via
expression as:
( ){ A ( x ) (
A+ = AT xzξ , xrξ , t zξ , xrξ , t AT xzξ , xzξ , t
(
The elements in expression A xzξ , xrξ , v, t will be defined )
(bounded) for each t ≥ 0 explained in Ref. [32]. The system
stability of TRMS (tail rotor moment system) was previously
verified in our research using the Controllability and Ob-
servability criteria via the full rank property [26] for the system
) (A.9) matrix. This stability analysis provides a solid foundation for
designing a suitable controller for the system. The following
section of the research will focus on the controller constraints
based on the system dynamics.
A.3 Stability analysis of RGDI-based controller
Using Eqs. (A.5) and (A.8), the new equation can be ob-
tained from Eq. (A.14):
)}
−1
. (A.10)
s = A ( eξ z , eξr , t ) u − B ( eξ z , eξr , t ) . (A.15)

Singularity can occur during the dynamic inversion of a sys-

tem matrix if the matrix becomes singular. This type of issue The stability convergence of the attitude error needs to vali-
can arise during the dynamic inversion of a system matrix and date and for this we use Eqs. (A.5) and (A.8) for u to place
is known as a singularity issue. Changes in the rank of the them in Eq. (A.14), which will provide the following yields:
matrix can cause this issue, resulting in a discontinuous order
in the matrix function. This discontinuous behavior can lead to ⎛ s ⎞
s = A ⎜ A* B + PY − CA* ⎟u − B . (A.16)
unlimited values of the generalized matrix. The RGDI controller ⎝ & s &⎠

11
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

Additionally, by utilizing the equation for the null projection

matrix P from Eq. (A.10) and applying the property of the
pseudo-inverse, Eq. (A.15) can be rephrased as:

{ }
s = ρ ( eξ z , eξr , v, t ) − 1 B − Cρ ( eξ z , eξr , v, t )
s
&s&
.

The positive unbounded function as a Lyapunov candidate

function is elaborated as:

1 2
V= s
2
Fig. A.1. Block diagram of RGDI-ARSMC regarding state variables for
TRMS.
will helps to get the sliding gain of C. By taking the time deriva-
tive of the function Eq. (15), we will get the following expression: eξr ; v; t ) ≥ ρ * and C > C * . The finite number C * is ensured
( )
for any range D because of B eξ z , eξr , t , which is globally
s bounded by virtue of implementing the DSGI A* given by Eq.
V = s {ρ − 1} B − Cρ . (A.17) (A.13), which results in globally bounded trajectories. The con-
&s&
troller design steps with its variable representation via synoptic
scheme is described in Fig. 3, to control the highly coupled
The function C ( eξ , eξ , v, t ) that will definitely affirms
z r
nonlinear system.
Remark 1: According to the statement of theorem 2, the gain
ρ −1
C ( eξ z , eξr , v, t ) > B must be positive and gain C increased in a way that the posi-
ρ
tive bound ρ * will be obtained to ensure of the situation (con-

that must ensure the negative definite of V , which shows that

( )
dition) is V < 0 will hold D for ρ eξ z ; eξr ; v; t < ρ * . A specific

the stability of the system in the finite time of s = 0, can be ob-

tained from the Lyapunov's direct method [10]. It can also be
ensured that the finite-time stability of eξ = 02×1 is upheld, as
is clear from the definition of s. Thus, it is not possible to ensure
the finite-time closed-loop stability of sliding mode dynamics.
( )
domain must be defined by ρ eξ z ; eξr ; v; t < ρ * will be fol-
lowed by state trajectory ea which should be in this range. The
driving ρ * is close to the zero for driving ea which is also close
n to zero for uniformly bounded. The following condition makes
the attitude error trajectory eξ = 02 x1 stable. Due increase in
z

C * will affect the enlargement of D, it must be followed as

Nevertheless, it is possible to attain semi-global practical stabil-
semi-global stability. In the SMC method, the system's trajec-
ity of the ARSMC through appropriate design of the SMC gain.
tory is guided along a manifold by the use of multiple control
Theorem 2: The real integers C * > 0 for all real integers
structures that follow a specified switching condition. The sys-
ρ ∈ (0, 1) which will provide the negative definite of V with
tem's structure is defined by switching functions, which can be
sliding dynamics elaborated by Eq. (A.16) for all values of
either scalar or vector. The switching surface, represented by
ρ ( eξ ; eξ ; v; t ) > ρ * and C > C * .
z r s(x) = 0, is a line on the phase plane. To better understand the
Proof: Let ρ * be a real scalar number as a constant entry in design steps and variables used in the TRMS controller, a flow
the coverage of ρ ( eξ ; eξ ; v; t ) , i.e., ρ ∈ (0, 1). Also, define
z r
chart with all its variables representation and design flow. The
effectiveness of the TRMS controller is evaluated through
(
C eξ z ; eξr ; t ) as:
simulation and experimental validation, using three different
cases that examine its performance in the presence of various
ρ* −1 perturbations.
(
C eξ z ; eξr ; t = − ) ρ*
B ( eξ , eξ , t ) .
z r

A.4 Case 1
It follows that C ( eξ ; eξ ; t ) > C ( eξ ; eξ ; v; t ) whenever
ρ ( eξ ; eξ ; v; t ) > ρ * . Accordingly, let D be a neighborhood of
z r z r
In our novel methodology, we have compared the design
z r

( eξ ; eξ ) = ( 02 ; 02 ) , and choose a sliding gain constant C *

z r
pattern with our previous lab work, specifically the nonlinear
such that paper and RGDI paper. This comparison demonstrates how
the RGDI-ARSMC method is superior to other techniques such
as SMC, Backstepping, and the RGDI controller. The novel
C * > maxC eξ z ; eξr ; t .
D
( ) RGDI-ARSMC strategy effectively addresses key issues such
as finite-time convergence, robustness for parametric perturba-
Then the negative definite value of V < 0 ensures along any tions, and singularity. To verify its effectiveness, we conducted
closed-loop track that starts within D whenever ρ ( eξ ; z
a comparison study and evaluated the performance of the

12
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

Fig. A.4. Pitch angle- ramp response of the proposed RGDI-ARSMC strat-
Fig. A.2. Step response of pitch angle. egy Input.

Fig. A.5. Yaw angle- ramp response of the proposed RGDI-ARSMC strat-
Fig. A.3. Step response of yaw angle. egy input.

controller in terms of vertical stability (pitch angle) and horizon-

tal stability (yaw angle). Additionally, the controller was tested
in the presence of disturbances, such as external disturbances,
parametric uncertainties, coupling effects, and noise signals.
Figs. A.2 and A.3 represent the brief response with their con-
vergence time in subplots. The figure that shows the step input
for both the pitch angle and yaw angle of the TRMS illustrates
that the pitch angle has a faster convergence time than the
yaw angle. This delay is caused by the fact that the main rotor
must be stabilized first in order to counteract the disturbance
Fig. A.6. Pitch angle-sinusoidal response of the proposed RGDI-ARSMC
generated by the tail rotor, such as the gyroscopic torque effect strategy input.
and coupling effect. As a result, the convergence time of the
tail rotor cannot be faster than that of the main rotor. The step
input tracking response in the simulation confirms that the con-
trol theory regarding the convergence time of the angles, spe-
cifically the pitch angle and yaw angle, is logical and results in
a sharp stability response.

A.5 Case 2
The first case study is aimed at understanding the behavior
of the controller by applying three different reference inputs,
including ramp and sinusoidal inputs. The simulation response
of the highly coupled system, with both matched and mis-
matched perturbations, can be seen in the figures. To further
demonstrate the effectiveness of the controller, a different input,
specifically a ramp input and a sinusoidal input are, is applied
as shown in Figs. A.4-A.7, respectively. The subplot in the
simulation response of the angles is provided to understand
the exact time of convergence as well as the confined stability
Fig. A.7. Yaw angle-sinusoidal response of the proposed RGDI-ARSMC
strategy input.
tackled by RGDI-ARSMC. It is worth noting that the remarkable
difference in the convergence time of both outputs is due to the
recursive structure of the controller, which is designed based
on the arbitrary controller for each state of the system.
A detailed comparison regarding previous research [26-28] is

13
 Journal of Mechanical Science and Technology 38 (4) 2024 DOI 10.1007/s12206-024-0234-6

comprehensively provided in Table 2 and we can notice the
efficient convergence time of proposed controller [31].
Nadir Abbas is doing Ph.D. from 2019 in
School of Electronic Information and Elec-
trical Engineering with major of Control
Theory from Dalian University of Technol-
ogy, Dalian 116024, China. He is working
as a Ph.D. researcher in the Control The-
ory Lab of Prof. Dr Xiaodong Liu with ma-
jor in Control Theory. He received Master
and bachelor degree from Pakistan in telecommunication engi-
neering and automation control respectively in 2012 & 2016 from
top ranked universities. with good research background.
Jamshed Iqbal is Senior Lecturer at
University of Hull. He received double
M.Sc. M.Sc. (University of Engineering
and Technology, Taxila). M.Sc. (Aalto
University School of Science and Tech-
nology). MSc (Lulea University of Tech-
nology) and Ph.D. (University of Genoa ).
Currently, he is looking after Mechatron-
ics and Robotics programme. His research interests include
robotics, mechatronics and control systems.

Xiaodong Liu received the B.S. degree

from Northeastern Normal University,
Changchun, China, in 1986, and the M.S.
degree from Jilin University, Jilin, China,
in 1989, both in mathematics, and the
Ph.D. degree in control theory and con-
trol engineering from Northeastern Uni-
versity, Shenyang, China, in 2003. He is
currently a Professor with the Research Center of Information
and Control, Dalian University of Technology, Dalian, China.

14