Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
Available online at www.sciencedirect.com

ScienceDirect
Procedia Computer Science 00 (2018) 000–000
Procedia Computer
Procedia Science
Computer 14300
Science (2018) 663–670
(2018) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

8th International Conference on Advances in Computing and Communication (ICACC-2018)

8th International Conference on Advances in Computing and Communication (ICACC-2018)
Model
Model Predictive
Predictive Control:
Control: Simulation
Simulation Studies
Studies for
for the
the
Implementation on Vertical Take-Off and Landing Lab Prototype
Implementation on Vertical Take-Off and Landing Lab Prototype
Anoop S, K Rahul Sharma
Anoop S, K Rahul Sharma
Department of Electrical and Electronics Engineering, Amrita School of Engineering, Coimbatore
Department of Electrical and Electronics
Amrita Vishwa Engineering,
Vidyapeetham,Amrita School
641112, Indiaof Engineering, Coimbatore
Amrita Vishwa Vidyapeetham, 641112, India

Abstract
Abstract
Model Predictive Control (MPC) is one of the advanced control techniques. The ability to act prior to the setpoint change by
Model Predictive
predicting Control
the future (MPC)
control is makes
actions one ofMPC
the advanced controlfor
a better choice techniques. The abilitywith
control applications to act prior to setpoint
predefined the setpoint changeThis
trajectory. by
predicting the future control actions makes MPC a better choice for control applications with predefined setpoint
paper is a preliminary work done for the development of control system for an autonomous air-borne surveillance system. A trajectory. This
paper is a preliminary
brief comparison amongwork done for theunconstrained
the constrained, development MPCs
of control system for an autonomous air-borne
and Proportional-Integral-Derivative (PID) surveillance
controller is system. A
presented.
brief comparison
Simulation among
study of the constrained,ofunconstrained
the implementation MPCs Vertical
MPC on the QNET and Proportional-Integral-Derivative
Take-Off and Landing (VTOL) (PID) controller
prototype is presented.
is investigated in
Simulation study
this paper. The of the implementation
software of this
platforms used in MPC on the
study QNET
include Vertical Take-Off
MATLAB and Landing (VTOL) prototype is investigated in
and Simulink.
this paper. The software platforms used in this study include MATLAB and Simulink.
c 2018

© 2018 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier B.V.
B.V.
c 2018
 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
(https://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an
Selection open
Selection and access article
and peer-review under
peer-review under the CC BY-NC-ND
under responsibility
responsibility of
of the license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
the scientific
scientific committee
committee ofof the
the 8th
8th International
International Conference
Conference on
on Advances
Advances in
in
Selection and
Computing andpeer-review under(ICACC-2018).
Communication responsibility of the scientific committee of the 8th International Conference on Advances in
(ICACC-2018).
Computing and Communication (ICACC-2018).
Keywords: MPC;VTOL;PID;constraints;horizon
Keywords: MPC;VTOL;PID;constraints;horizon

1. Introduction
1. Introduction
Inspite of the development of advanced control strategies, the major portion of the control tasks are being done
by Inspite of the development
the conventional of advanced control strategies,
Proportional-Integral-Derivative the majorsince
(PID) controllers portion
theyofare
thecost
control tasks are
effective, easybeing done
to imple-
by the conventional Proportional-Integral-Derivative (PID) controllers since they are cost effective,
ment and maintain, technically simple, guaranteed performance and flexible. The basic principle behind a PID control easy to imple-
ment
scheme andismaintain, technically
that, whenever simple,from
a deviation guaranteed performance
the set-point andcontroller
occurs, the flexible. The
willbasic principle
initiate behind
the control a PIDtocontrol
actions make
scheme
the deviation minimum. It sounds good for a control application with no knowledge on the future set-point.toFor
is that, whenever a deviation from the set-point occurs, the controller will initiate the control actions makean
the
application where the set-point or trajectory for the future is set and known to the controller, why the controller has an
deviation minimum. It sounds good for a control application with no knowledge on the future set-point. For to
application where in
wait till a change theset-point
set-pointtooroccur?
trajectory for thePredictive
A Model future is set and known
Control (MPC)tostrategy,
the controller, why formerly
which was the controller
used has to
in the
wait till a change in set-point to occur? A Model Predictive Control
process control industries has the potential in the aforementioned scenario [3]. (MPC) strategy, which was formerly used in the
process control industries has the potential in the aforementioned scenario [3].

∗ Corresponding author. Tel.: +91-9645687820.

∗ Corresponding
E-mail address:author. Tel.: +91-9645687820.
s anoop@cb.students.amrita.edu
E-mail address: s anoop@cb.students.amrita.edu

1877-0509 
1877-0509 ©c 2018 The Authors.
2018 The Authors. Published
Published by
by Elsevier
Elsevier B.V.
B.V.
1877-0509
This is c 2018 Thearticle
Authors. Published by Elsevier B.V.
This is an
anopen
openaccess
access articleunder
underthethe
CCCC
BY-NC-ND
BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
license (https://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an and
Selection
Selection openpeer-review
and access article
peer-review under
under
under the CC BY-NC-ND
responsibility
responsibilityofofthe license
thescientific(https://creativecommons.org/licenses/by-nc-nd/4.0/)
committee
scientific committee ofofthethe
8th8th
International Conference
International on Advances
Conference in Computing
on Advances and
in Computing
Selection
and and peer-review
Communication
Communication under responsibility of the scientific committee of the 8th International Conference on Advances in Computing and
(ICACC-2018).
(ICACC-2018).
Communication (ICACC-2018).
10.1016/j.procs.2018.10.450
664
2 AnoopSharma
Anoop S,K Rahul S et al./ /Procedia
ProcediaComputer
ComputerScience
Science00143 (2018)
(2018) 663–670
000–000

l1
an
Lh randF
Moto

l3
l2
θ
m1g

ight m3g
terwe
Coun

m2g

Fig. 1. (a) QNET VTOL system with one degree of freedom; (b) Free body diagram of VTOL.

MPC is comparatively younger in the control engineering field [1][2]. As the name indicates, prediction is the
inherent peculiarity of an MPC. There are various control algorithms for MPC depending on the application such as
nonlinear MPC, robust MPC, explicit MPC so on.The advantage of having a sight on the future makes it an important
member in the advanced controller family. The user has the flexibility to design or tune the MPC structure according
to the concerned application, viz. there is no fixed configuration for the MPC. Other main advantages are constraints
on the parameters can be incorporated into the controller, easy to implement on multivariable systems, systems hav-
ing complex dynamic behaviour can be controlled and more physical realization of the controlled system is possible.
An investigation has been started on the implementation of a position and altitude control of an autonomous air
borne system. As the application is for an aerial surveillance system, the trajectory of the flight is pre-set. The work
carried out in this paper is the feasibility study of MPC. Similar work has been reported on a lab scale helicopter pro-
totype using LQR [6] and MPC [1], where a trajectory tracking control of quadrotor is considered using conventional
linear MPC. Whereas this paper deals with a single DOF lab prototype with incremental MPC. Simulation analysis of
MPC with input and output constraints is done with MATLAB software platform. A Vertical Take-Off and Landing
(VTOL) prototype is considered in this work, which is the QNET VTOL setup (which will be simply called as VTOL
in the rest of the paper) with Control and Instrumentation laboratory at the Amrita Vishwa Vidyapeetham, Coimbatore
campus, as shown in fig. 1 (a). It represents single degree of freedom of VTOL concept [4]. The pitch is generated
using a DC fan, which can be interpreted as a thruster of a fixed wing VTOL flight or the rotor of a helicopter or
quadrotor. An optical encoder is used to measure the pitch angle.
The organization of the paper is as follows; Section 1 deals with the problem identification, state of the art develop-
ments and the literature survey. Section 2 describes the incremental MPC configuration, its cost function, constraints
and the system of equations. Section 3 gives the description and modelling of the VTOL and section 4 showcases the
simulation analysis done and section 5 concludes the work and discusses the future work.

2. Model Predictive Control

An MPC is an optimal controller based on real time numerical optimization. The plant output is predicted using
an estimated system model, which is called the predictor model. The plant input is optimized at each time instance
according to a penalty function and constraints. The main goal of an MPC is to predict a set of future manipulated
variable to get an optimized output. As mentioned in the previous section, the cost function to be minimised to get
an optimal solution for the control problem can be designed by the user. In a broader sense, the parameters to be
considered for the optimization function such as control input, states or outputs can be decided by the user. In this
work the cost function is as given in (1)which considers only the output and the control action,

J = (Y − W)T Q(Y − W) + UT RU (1)

 Anoop S,K Rahul
AnoopSharma / Procedia
S et al. Computer
/ Procedia Science
Computer 00 143
Science (2018) 000–000
(2018) 663–670 3
665

where,
 
 w(k + 1) 
 w(k + 2) 
 
 .. 
 . 
W =    (2)
 w(k + Nc ) 
 .. 

 . 

w(k + N p )
where Y is the predicted output which will be explained later in (12), W is reference trajectory as given in (2) and Q
and R are the weight matrices which act as the tuning parameters for MPC.
Considering a discrete time, single input single output (SISO) system with the following state equation in (3) and
output equation in (4),

x(k + 1) = Ax(k) + Bu(k) (3)

y(k) = Cx(k) (4)

where x(k) ∈ n is the state vector, y(k) ∈ m is the output vector and u(k) ∈  p is the controlled variable. Matrices
A ∈ (n×n) is the state matrix, B ∈ (n×p) is the input matrix and C ∈ (m×n) is the output matrix.
The predictor model of the MPC considers an augmented state vector with control input as a state as given in (5).
This incremental MPC configuration will add the ability to control the change in control action also, which is not
available in a normal MPC.
 
x(k)
(5)
u(k)
where,

u(k) = u(k − 1) + ∆u(k) (6)

The predictor model can be described as,

   
x(k + 1) x(k)
=M + N∆u(k) (7)
u(k) u(k − 1)
 
x(k)
y(k) = O (8)
u(k − 1)
where,
 
A B
M= (9)
0 1
 
B
N= (10)
1
 
O= C 0 (11)

The 1 and 0 in equations (9) - (11) represents identity and zero matrices, whose size depends on the system.
Prediction horizon decides how long the output and states of the system have to be predicted, which is denoted
as N p . Prediction horizon is one of the parameters that we decide before the controller implementation and which
usually remains unaltered throughout the process. Control horizon is the number of moves or steps to be optimized for
an interval, which is denoted as Nc . There is no fixed rules for the selection of prediction and control horizons, but it
is done based on the experience on the system. Usually the control horizon is less than or equal to prediction horizon,
666 Anoop S et al. / Procedia Computer Science 143 (2018) 663–670
4 Anoop S,K Rahul Sharma / Procedia Computer Science 00 (2018) 000–000

i.e., Nc ≤ N p In (1), the error is calculated between the predicted output and the actual output. The predicted output
consists of forced as well as natural response as given in [5],
 
x(k)
Y = GU + F (12)
u(k − 1)
where,  
 ON 0 0 ... 0 
 OMN ON 0 . . . 0 
 2 
G =  OM N OMN ON ... 0  (13)
 .. .. .. .. .. 
 . . . . . 
 
OMN p −1 N OMN p −2 N OMN p −3 N . . . OMN p −Nc N
 
 OM 
 OM2 
 
 3  
F =  OM  (14)
 .. 
 . 
 
OMN p
The constraints on the states, output and control action can be imposed on the problem as in (15) and can be solved
with the help of quadratic programing.

AU ≤ b (15)

where A is a matrix consisting of elements corresponding to the imposed constraints such as maximum and minimum
control input, maximum and minimum output, and so on and b holds the upper and lower bounds for the corre-
sponding parameter of A. Sizes of A and b depend on the constraints imposed on the parameters. U is the vector
consisting of all the predicted change in control action.

3. VTOL Modeling

The free body diagram of the VTOL is given in fig. 1 (b), where m1 , m2 and m3 , are the masses of motor and
fan assembly, counter weight and the link connecting these two respectively. Lh , l1 , l2 and l3 are the lengths of the
represented sections in fig. 1 (b).
The model of the VTOL can be obtained with the help of first principle of modelling. The torque-current equation
can be obtained as (16),

Kt Im + m2 gl2 cos θ(t) − m1 gl1 cos θ(t) − m3 gl3 cos θ(t) = 0 (16)

where Kt is the torque constant, θ is the pitch angle and Im is the motor current.
The equation of motion of the motor is as given in (17),

J θ̈ + Bm θ̇ + K s θ = Kt Im (17)

where J is the moment of inertia, Bm is the damping coefficient and K s is the stiffness.
The current to pitch transfer function model and the state space model of the system can be obtained as (18).
Kt
Gt f (s) = Bm Ks
(18)
J(s2 + J s+ J )
The parameters of VTOL which are used for the study are given in Table 1. The state space representation of the
above transfer function model which is used for the entire simulation is given by (19) and (20). The states of the
system considered are the pitch and the pitch rate. The control input is the voltage applied to the motor.
 Anoop S et al. / Procedia Computer Science 143 (2018) 663–670 667
Anoop S,K Rahul Sharma / Procedia Computer Science 00 (2018) 000–000 5

Table 1. Model Parameters of VTOL

Parameter Value Unit (t)

J 0.00347 Kgm2
Bm 0.002 Nms/rad
Ks 0.0373 Nm/rad
Kt 0.0108 Nm/A

Pitch Current
reference reference Voltage Pitch
PID Controller PI Controller Motor

Current

Fig. 2. Cascade control structure of the VTOL.

      
ẋ1 0 1 x1 0
= + u (19)
ẋ2 −10.7 −0.576 x2 3.112
   x1 
y= 10 (20)
x2
The block diagram representation of the VTOL is shown in fig. 2. The obtained model is represented using the

MPC Setpoint

Predicted
Output
output +
Plant Predictor Model -

Predicted input Error

Optimizer

Cost function Constraints

Fig. 3. Model Predictive Control system for VTOL.

dashed box. This LabVIEW based plant has a cascade control scheme as shown in fig. 2. The inner loop has a PI
controller which will control the current according to a current set point. This current set point is the output of a PID
controller in the outer loop. The PID controller controls the pitch according to a given reference. The control structure
with MPC for VTOL is shown in fig. 3. The measured output from the plant will be given to the predictor model along
with the predicted control action. The predicted output of the model will then be compared to the set point and the
error is given to the optimizer along with the cost function and the constraints, if any.
668 Anoop S et al. / Procedia Computer Science 143 (2018) 663–670
6 Anoop S,K Rahul Sharma / Procedia Computer Science 00 (2018) 000–000

Output
0.6

Set point
0.4 Output

0.2
Pitch (rad)

-0.2

-0.4

-0.6
20 25 30 35 40 45 50 55
Time (s)
Fig. 4. PID controller simulation results - output.

Control Input:
2.5

2 Input

1.5

0.5
Current (A)

-0.5

-1

-1.5

-2

-2.5
20 25 30 35 40 45 50 55
Time (s)
Fig. 5. PID controller simulation results - control input.

4. Simulation Analysis

The software platform used for the simulation studies is MATLAB 2017a. Comparison among the unconstrained
and constrained MPC configuration is done using the simulation results. A PID controller has also been used for
the simulation analysis and comparison, since the VTOL prototype is implemented with a PID controller. The per-
formance of the PID and MPC are not compared here, because PID does not having any prediction capability. The
simulation result for a PID controlled VTOL is given in fig. 4 and 5. Figure 4 shows the output response of the system
with the set point and fig. 5 shows the control action. It is clear from the response that, the PID controller will manip-
ulate the controlled variable only when there occurs a change in the set point. It has been observed that since the step
change has occured at 50 seconds, the output of the system doesn’t tend to change until then.
The simulation results for MPC without imposing any constraints on the output or control action is shown in fig. 6
and 7. Figure 6 shows the plot of output and set point and fig. 7 shows the variations in the control input. It is clearly
seen from the response that, the controller has foreseen the change in set point at 50 seconds and the output has started
to cope up with the future variation quite before the change occurred. The control action has started at 48 seconds,
which is the result of the pre-set control horizon of the MPC. The MPC configuration considered for this simulation
study has prediction horizon and control horizon equal. The prediction of the MPC can be varied by varying the pre-
diction window. Increasing the prediction horizon will overload the controller and response will become slower. On
the other hand a small horizon will make the working of the MPC improper. So there should be a trade-off between the
computation complexity and accuracy for the selection of horizon window. From fig. 6 it can be seen that the output
of the VTOL is going beyond zero and 0.5, which are considered as constraints for the VTOL. Similarly the current
 Anoop
Anoop S,K Rahul S et al.
Sharma / Procedia
/ Procedia Computer
Computer Science
Science 00 143 (2018)
(2018) 663–670
000–000 669
7

Output
0.6
Set point
Output
0.4

Pitch (rad) 0.2

-0.2

-0.4

-0.6
20 25 30 35 40 45 50 55
Time (s)
Fig. 6. Unonstrained MPC simulation results - output.

Control Input
3

Input
2

1
Current (A)

-1

-2

-3
20 25 30 35 40 45 50 55
Time (s)
Fig. 7. Unonstrained MPC simulation results - control input.

Output

y m ax
Output
0.4 Set point

0.2
Pitch (rad)

-0.2

-0.4
y m in

-0.6
20 25 30 35 40 45 50 55
Time (s)
Fig. 8. Constrained MPC simulation results - output.

input to VTOL also goes above 1.8 A, which is the upper constraint on control input.
Figure 8 and 9 shows the simulation response of the VTOL with constraint MPC. Constraints are considered for
the output and control action. The pitch of the VTOL is constrained as 0.5 radians as maximum and 0 radians as
670 Anoop S et al. / Procedia Computer Science 143 (2018) 663–670
8 Anoop S,K Rahul Sharma / Procedia Computer Science 00 (2018) 000–000

Control Input
2
u m ax
Input
1.5

0.5
Current (A)

-0.5

-1

-1.5
u m in
-2
20 25 30 35 40 45 50 55
Time (s)
Fig. 9. Constrained MPC simulation results - control input.

minimum, which can be seen in fig. 8. The control input is constrained to 1.8 A and change in control action is limited
to 0.5 A which is clearly seen in fig. 9.

5. Conclusion and Future Scope

The aim of this work is to study the response of the VTOL with MPC, as a part of the hardware implementation
of an airborne surveillance system. Simulation analysis was done on the VTOL system with PID controller, MPC
without any constraints on the output and input and MPC with constraints on the output and input. The simulation
shows the effectiveness of the MPC with constraints. The control input and output of the VTOL has got limited within
the specified limits in the constrained MPC.
The actual control configuration implemented on the QNET VTOL is of cascade structure. Future prospects of this
work include development of a cascaded MPC configuration for the system, its validation by software and hardware
experiments which will conclude the work.

References

[1] Ganga, G., Dharmana, M.M., 2017. Mpc controller for trajectory tracking control of quadcopter, in: 2017 International Conference on Circuit
,Power and Computing Technologies (ICCPCT), pp. 1–6. doi:10.1109/ICCPCT.2017.8074380.
[2] Jalali, A.A., Nadimi, V., 2006. A survey on robust model predictive control from 1999-2006, in: 2006 International Conference on Computa-
tional Inteligence for Modelling Control and Automation and International Conference on Intelligent Agents Web Technologies and Interna-
tional Commerce (CIMCA’06), pp. 207–207. doi:10.1109/CIMCA.2006.29.
[3] Kozk, ., 2016. From pid to mpc: Control engineering methods development and applications, in: 2016 Cybernetics Informatics (K I), pp. 1–7.
doi:10.1109/CYBERI.2016.7438634.
[4] Mondal, S., Mahanta, C., 2013. Observer based sliding mode control strategy for vertical take-off and landing (vtol) aircraft system, in: 2013
IEEE 8th Conference on Industrial Electronics and Applications (ICIEA), pp. 1–6. doi:10.1109/ICIEA.2013.6566330.
[5] Rahul Sharma, K., Honc, D., Dušek, F., 2015. Model predictive control of trajectory tracking of differentially steered mobile robot, in:
Abraham, A., Jiang, X.H., Snášel, V., Pan, J.S. (Eds.), Intelligent Data Analysis and Applications, Springer International Publishing, Cham.
pp. 85–95.
[6] Sini, S., Vivek, A., Nandagopal, J.L., 2017. Trajectory tracking of 3-dof lab helicopter by robust lqr, in: 2017 International Conference on
Circuit ,Power and Computing Technologies (ICCPCT), pp. 1–7. doi:10.1109/ICCPCT.2017.8074217.