International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

Real Time Implementation and Design of

Predictive Controller for Fast Dynamic System
Sumit G. Vyas1, Vinodkumar P. Patel2

.
Abstract—Generalized Predictive Controller is a one type of In [1] Generalized Predictive controller results for DC motor
Model Predictive Controller. Speed Control of DC motor is is compared to Discrete time Predictive Controller. More
taken as a Case study for validation of GPC controller detail and survey on Model Predictive Controller is given in
algorithm implementation on Real Time Hardware. There is a
much gap between theoretical solution and real time [2]. In [3] MPC is implemented for PMSM motor,
implementation of MPC. Basically MPC works better for slow Generalized Predictive Controller is applied to control the
dynamic system, so here efforts are made to Implement MPC Position of Induction Motor in [4]. Robust Model Predictive
for Fast Dynamic System. System Identification of DC motor is Controller is implemented for Fast Dynamic Vehicle System
done using a LABVIEW and system identification tool of in [5]. Above all MPC is designed particularly for specific
MATLAB. First of all ARIX model based GPC controller is application.
designed in MATLAB for Model Obtained from System
Identification of DC motor. Then the control law of GPC In Section II GPC control Strategy is given, In Section III
controller is implemented on ATMEGA328P and Results shows description of DC motor is given, Section IV Describes the
that GPC controller gives better results than discrete time PID System Identification of the System. In Section V Simulation
controller for Set Point Tracking as well as for Disturbance result and effect of tuning parameter of GPC on System
Rejection. Output is shown. Section VI describes the real time
implementation of GPC and PID controller for DC motor
Index Terms—MPC (Model Predictive Controller), GPC
Speed Control is given.
(Generalized Predictive Controller), ARIX Model (Auto
Regressive Integrated Exogenous Model, FDS (Fast Dynamic
System). II. GENERALIZED PREDICTIVE CONTROLLER
GPC is based on minimizing a weighted sum of the Set
I. INTRODUCTION Point error and the control effort and it allows plant models to
Model Predictive Controller is basically suitable only for be updated frequently.GPC makes use of the j-step ahead
Slow Dynamic Systems like Chemical Process, Oil prediction error model.
Refineries, etc. Design of MPC for Fast Dynamic system is
difficult because MPC takes long time to calculate the value ARIX model based Generalized Predictive Control law is
of control action and if the sampling time of control action is given by the following equation,
larger than the FDS like DC motor will not reach at the
Desired Set Point. So MPC with small sampling time and less 𝑢 = 𝑘𝑟 − 𝑘𝐻2 𝑦𝑜𝑙𝑑 − 𝑘𝐻1 𝑢𝑜𝑙𝑑
computation time will be used for FDS.
𝑔𝑘,0 0 …0
There are many types of MPC control algorithms used.
Some of them are listed here: 1) GPC (Generalized Predictive 𝑔𝑘+1,1 𝑔𝑘+1,0 ⋯0
Where G = ,
Controller), 2) DMC (Dynamic Matrix Controller), 3) Steady ⋮ ⋮ ⋮
State Weighted Generalized Predictive Control, etc. 𝑔𝑘+𝑁,𝑁 𝑔𝑘 +𝑁,𝑁−1 ⋯ 𝑔𝑘 +𝑁,0
Generally GPC controller is used for Fast Dynamic System.
Here GPC controller results are compared with PID 𝑔𝑘 ,1 ⋯ 𝑔𝑘,𝑑𝐺 𝑘
controller for DC motor Speed Controller. PID and PI 𝑔𝑘 +1,2 ⋯ 𝑔𝑘+1,𝑑𝐺 𝑘 +1
controller are also used for Fast Dynamic System. 𝐻1 = ,
⋮ ⋮ ⋮
There is much gap between real time implementation and 𝑔𝑘+𝑁,𝑁+1 ⋯ 𝑔𝑘+𝑁,𝑑𝐺 𝑘 +𝑁
theoretical solution of MPC. So here, effort is made for real
time implementation of GPC for DC motor Speed Control. 𝑓𝑘 ,0 ⋯ 𝑓𝑘,𝑑𝐴
More detail on MPC and GPC control algorithm is given in
𝑓 ⋯ 𝑓𝑘+1,𝑑𝐴
[2] and [6]. Now a day’s many researchers tries to reduce the 𝐻2 = 𝑘 +1,0 ,
computation time of MPC. ⋮ ⋮ ⋮
𝑓𝑘 +𝑁,0 ⋯ 𝑓𝑘+𝑁,𝑑𝐴
Manuscript received June, 2015.
Sumit G. Vyas1Student of M.E., Instrumentation and Control Engineering,
L.D. College of Engineering, Ahmedabad,Gujarat, India
Vinodkumar P. Patel2 Associate Professor, Instrumentation and Control 𝑦𝑜𝑙𝑑 = [𝑦 𝑛 … … … … 𝑦(𝑛 − 𝑑𝐴)]𝑇 ,
Engineering, L.D. College of Engineering, Gujarat, India

2401
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

𝑇
𝑢𝑜𝑙𝑑 = ∆𝑢 𝑛 − 1 … … … ∆𝑢 𝑛 − 𝑘 + 1 − 𝑑𝐵 ,

𝑇
𝑢 = ∆𝑢 𝑛 … … … … … … . ∆𝑢 𝑛 + 𝑁 ,

where𝑟 is a trajectory of reference signals.

𝑟 = [𝑟 𝑛 + 𝑘 … … … … … 𝑟(𝑛 + 𝑘 + 𝑁)]𝑇 ,
𝑘 = (𝐺 𝑇 𝐺)−1 𝐺 𝑇 ,

In the above Generalized Predictive Control law, the error

signal and the control effort are weighted over the same
length of time.
Now, generalizing this situation by minimizing the error
from n+k+N1 to n+k+N2, N2≥N1 and the control effort from
n to n + Nu.
The performance index is given by
𝐽𝐺𝑃𝐶 = [𝑦 𝑛 + 𝑘 + 𝑁1 − 𝑟(𝑛 + 𝑘 + 𝑁1 )]2 +
… … …+[𝑦 𝑛 + 𝑘 + 𝑁2 − 𝑟(𝑛 + 𝑘 + 𝑁2 )]2 Fig- 1: DC motor
+𝜌[∆𝑢(𝑛)]2 + … … … + 𝜌[∆𝑢(𝑛 + 𝑁𝑢 )]2
IV. SYSTEM IDENTIFICATION
As a result of this 𝑦 and 𝑢 given by
System Identification uses the different statistical methods
𝑇
𝑦 = [𝑦 𝑛 + 𝑘 + 𝑁1 … … … 𝑦(𝑛 + 𝑘 + 𝑁2 ) ] to generate the model of the System using real time data of
input and output of the system with respect to time measured
𝑇
𝑢 = ∆𝑢 𝑛 … … … … … … . ∆𝑢 𝑛 + 𝑁𝑢 in LABVIEW. LABVIEW stores the input and output
measured data of the system with respect to time at 0.001
The performance of the GPC depends on the parameters sampling time. This data are used in the system identification
N1, N2, Nu and ρ. So the proper selection of these parameters tool of the MATLAB to generate the model of the system.
gives better results. Tuning of these parameters is necessary Figure (2) shows the Interfacing of DC Motor with
which is given in the [6], [7] and [8]. More detail on GPC LABVIEW for the System Identification of the Motor.
controller is given in [6].

III. DC MOTOR
To validate the proposed MPC algorithm, Speed Control of
DC Motor is taken as a Case Study for Fast Dynamic System.
DC Motor Specification is given below. Generally the MPC
gives better results for Slow Dynamics System so here efforts
are made to Implement GPC control algorithm for DC motor
Speed Control.
A. DC Motor Specification
Fig (1) show the DC motor used to validate the GPC
algorithm. Rotating Magnet mechanism on the pulley is put
on motor to give the load on the DC motor to checks whether
the GPC controller can rejects the disturbance or not.

 Specifications:

1. Maximum speed: 1500 Rpm

2. Maximum input Voltage: 10 V DC
3. Maximum input Current: 0.9Amp
4. F to V converter output: 0 to 2 volt for 0 to 1500 rpm. Fig-2: Interfacing of DC motor with LABVIEW to store
the value of input and output values in a Measurement File

As shown in figure DC motor is interfaced with

LABVIEW to obtain the model of the DC Motor and stored
data of input and output of the Motor is used in the system

2402
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

identification tool of the MATLAB. Figure (3) show the V. SIMULATION RESULTS
output and input waveform of the motor stored in the File This section shows the simulation Results of the DC motor
using LABVIEW. Figure (4) shows the estimated output of model with GPC controller and its results are compared with
the obtained model of the DC motor. discrete time PID controller. Sampling time Ts = 0.059s
The Estimated model of the DC motor is obtained as below, taken to implement the PID and GPC controller. PID tuning
is done using Ziegler Nichols closed loop Method.
0.5957 𝑆 + 0.0683 Obtained value of PID tuning parameter are Kp = 6.02, Ti =
𝐺 𝑆 = 0.05 and Td = 0.019. Then this parameter are used to
𝑆 2 + 3.248 𝑆 + 0.339 implement the discrete time PID controller using following
With one zero and two pole and the estimated transfer Equation,
𝑇𝑠 𝑇𝑑
function fits the data by 89.54%. 𝑆0 = 𝐾𝑝 1 + +
Transfer function of DC motor in Z domain is obtained as 2𝑇𝑖 𝑇𝑠
below,
𝑇𝑠 2𝑇𝑑
0.0321 z − 0.03187 𝑆1 = 𝐾𝑝 −1 + −
G z = 2 2𝑇𝑖 𝑇𝑠
z − 1.825 z + 0.8256
𝑇𝑑
𝑆2 = 𝐾𝑝
𝑇𝑠

Discrete Time PID Controller is given by,

1 − 𝑍 −1 𝑢 𝑛 = 𝑆0 + 𝑆1 𝑍 −1 + 𝑆2 𝑍 −2 𝑒(𝑛)

Tuning parameter of GPC are N1, N2, Nu and rho. Values of

tuning parameter of GPC are taken are N1 = 0, N2 = 10, Nu =
5 and rho = 0.5.
N1 = Starting point of Prediction Horizon,
N2 = End point of Prediction Horizon,
Nu = Control Horizon,
rho = Control Weighting parameter.
For all case Set Point of DC Motor taken is 700 rpm
A. DC Motor’s Response without Disturbance
Motor’s closed loop response with PID controller and GPC
controller are show in figure (5). In this response, no
disturbance is given.

Fig- 3: Input and Output response of measured data of DC

motor

Fig-5: System Response without Disturbance

Fig- 4: Response of Actual System and Simulated Model of
DC motor Figure (5) shows the both GPC and PID controller output
response without Disturbance. From the simulation results of

2403
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

the GPC and PID Controller, it is concluded that settling time Figure (8) shows the Control action Response of PID and
of GPC controller is 4 seconds while PID Controller takes GPC Controller. Because of Disturbance, Oscillation occurs
3seconds. So from settling time perspective PID is better than in both controllers at time of 10 seconds. It shows that both
GPC controller but PID controller have large Oscillation then controllers reject the Disturbance effectively. Time taken to
GPC controller. GPC controller gives smooth tracking of the remove the disturbance is almost same for both controller and
Set Point as compared to PID controller. it is approximately 2 to 2.5 seconds.

Fig – 6: Control Action Responsewithout Disturbance Fig –8: Control action Response with Disturbance
C. Set Point Tracking Results of DC Motor
Figure (6) shows the Control action response of the PID and
GPC Controller. In this case variation in the Set Point is given. For 25
seconds Set Point value is 700 rpm and for another 25
B. DC Motor’s Response with Disturbance seconds Set Point value is taken be 0. Figure (9) shows Set
In this case step disturbance is applied after 10 seconds. point tracking of GPC controller is very smooth while PID
Both the Controller takes almost same time to rejects the controller oscillated Set Point Tracking.
disturbance. System Response with Disturbance is shown in
figure (7).

Fig – 9: System Response for Set Point Tracking

Fig –7: System Response with Disturbance Figure (10) shows the control action response of GPC and
PID controller for Set Point tracking case study.

2404
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

Fig – 10: Control Action Response for Set Point Tracking Fig – 12: Effect of Starting Point of Prediction Horizon (N1)
on System Output
Figure (13) shows the effect of End Point of Prediction
D. Effect of GPC Tuning Parameter on the System
Horizon. For all cases N2 = 10 is taken. For N2 = 6 system
Output
output is very slow and it takes more time to reach at set
In this section effect of GPC tuning parameters on the point. For N2 = 13 continuous oscillation occurs in the
System Output is shown. Figure (11) shows the effect of system output. So Small value of N2 makes system very slow
control horizon on System Output. For in all case Control and large value of N2 makes system oscillated. So for N2 =
Horizon Nu = 5 is taken. Decreasing the value of Nu gives 10 it gives system output have neither slow response nor
oscillated output while increasing the value of Nu makes oscillated response.
system slow. For Nu = 5 it gives better results for system
output.

Fig – 13: Effect of End Point of Prediction Horizon (N2) on

System Output
Fig – 11: Effect of Control Horizon (Nu) on System Output Remaining tuning parameter of GPC is Control Weighting
Parameter (rho). In figure (14) effect of the value of rho on
Figure (12) shows the effect of Starting Point of Prediction system output is shown. So it effects almost same as control
Horizon on System Output. For N1 = 0 it gives good results horizon parameter on system output. For rho = 0.3 it gives
but when increasing the value of N1 it makes system very oscillated response on system output and for rho = 0.8 system
slow. output have slow response. So for small value of rho system
output have oscillated response and for rho near to 1 have

2405
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

slow response on system output. For rho = 0.5 it system

output have neither oscillated response nor slow response.

Fig – 14: Effect of Control Weighting Parameter on System

Output Fig – 16: Motor Response for Disturbance Rejection Using
Motor Response fGPC Controller
VI. REAL TIME IMPLEMENTATION OF GPC AND PID
CONTROLLER Motor Response for Set Point Tracking using GPC
Controller is shown in figure (17). There are two values of
In Previous section all are simulation results of GPC and Set Point taken, first one is 900 Rpm and another is 500 Rpm.
PID controller, here results are of real time data of GPC and Each Set Point is given to Motor for 20 seconds. From the
PID controller is shown. Fig (15) shows the interfacing of DC results it is concluded that using GPC Controller DC Motor
motor with Arduino Uno board. In Arduino Uno board tracks the Set Point Smoothly. Blue line in the Results
recursive function of eq (1) is implemented in C++ language. shows the Reference Trajectory. It follows the Set Point
smoothly after 1 to 2 seconds. Set Point tracking and
Disturbance Rejection of the DC Motor using GPC
Controller give better results than the PID Controller.

Fig – 15: Interfacing of DC Motor with Arduino Uno Board

A. GPC Control Algorithm
Figure (16) shows the real implementation of GPC
controller to check the whether GPC rejects the Disturbance
or not peak in circle shows the disturbance given on Motor
using Magnet. It is show that GPC rejects the disturbance
very well. Blue line shows the reference Trajectory. From
the results it is concluded that using GPC Controller Motor
takes approximately 3 to 4 second to reach at the desired Fig – 17: Motor Response for Set Point Tracking Using GPC
Speed and it rejects the disturbance in approximately 1 to 2 Controller
seconds.

2406
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

B. Discrete Time PID Control Algorithm There is large oscillation in PID Controller for Set Point
DC Motor’s Response for Disturbance Rejection Using Tracking. It takes 15 to 16 seconds to settle at a desired set
Discrete time PID Controller is shown in figure (18). Set point which is larger than GPC controller. Overshoot is also
Point is 800 RPM is taken. From the results it is concluded larger in Discrete time PID Controller than GPC controller.
that GPC Controller have better results than Discrete time There is also a difference between Simulation Results and
PID Controller. There is large oscillation occurs in Discrete Real Time Implementation Results of GPC and PID
time PID Controller. DC Motor takes 15 seconds initially to Controller for DC Motor Speed Control.
reach at the desired Speed using PID Controller. Disturbance
on DC motor generates large oscillation and it takes more VII. CONCLUSION
time to rejects the Disturbance than GPC Controller. Settling From the Results of Real Time Performance and
time and Overshoot both are large in PID Controller than Simulation of the DC Motor, it is concluded that GPC
GPC Controller. controller gives better results for both the Set Point Tracking
and Disturbance Rejection than Discrete Time PID
controller. DC Motor using GPC Control takes only 3 to 4
seconds to reach at the Desired Set Point while Discrete Time
PID Controller takes approximately 15 seconds to reach at
the Desired Set Point which is larger than the GPC Control.
In Disturbance Rejection case study, Discrete Time PID
Controller have more effects of Disturbance and it takes
more time to reject that disturbance while GPC control rejects
Disturbance and Continuous Load effects in small time than
PID controller. GPC gives Smooth and Stable Performance
than Discrete Time PID Control Algorithm. GPC control
algorithm performance depends on the proper selection of
tuning parameters which are Control Horizon (Nu), Starting
and End point of Prediction Horizon, Control Weighting
parameter (rho). As Shown in simulation results, for taking
Nu = 2 System have oscillated output while taking Nu = 7
settlingtime of System is increases. For Nu = 5, System has
smooth response. neither it has slow response nor has
oscillated Response. For End point of Prediction Horizon N2
Fig – 18: Motor Response for Disturbance Rejection Using = 13, system have continuous oscillation and for N2 = 6,
PID Controller settling time of system is increases. So for N2 = 10, system
has smooth response neither it is slow nor oscillated
DC Motor’s Output Results for Set Point Tracking using response. Generally Starting Point of Prediction Horizon N1
PID Controller is shown in figure (19). Same as GPC = 0 is taken. After proper selection of tuning parameters,
controller, here also two Set Point values are given to DC GPC control algorithm gives smooth and stable response for
Motor for Set Point Tracking. Results show that for Set Point DC Motor Speed Control.
Tracking GPC controller gives better performance than PID
controller. REFERENCES

[1] Sumit G. Vyas, Vinodkumar P. Patel, “Design of

Predictive Controller For Smooth Set Point Tracking of
Fast Dynamic Controller” , Volume 04, Issue 04,
IJRET April -2015.
[2] Yu-Geng Xi, De-Wei, Shu LIN, “Predictive Control
-Status and Challenges” Vol.39, No.3,
ActaAutomaticaSinica, ScienceDirect, March -2013.
[3] GionataCimini, Valentino Fossi, GianlucaIppoliti,
Stefano Mencarelli, Giuseppe Orlando and
MatteoPirro, “Model Predictive Control Solution for
Permanent Magnet Synchronous Motors”, Industrial
Electronics Society, IECON 2013, IEEE.
[4] Antonio B.S.Junior, Francisco G. Sena, Bismark C.
Torrico, Luiz H.S.C. Barreto,”Generalized Predictive
Control Applied to the Position Control of a Induction
Motor”, Industry Application, IEEE, 2012.
[5] T. Peni, J. Bokor, “Robust Model Predictive Control for
Controlling Fast Vehicle Dynamics”, Control and
Automation 2006, IEEE.
[6] “Digital Control” by KannanMoudgalya.pp.438-443&
312-320.
Fig – 19: Motor Response for Set Point Tracking Using PID
Controller

2407
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR
 International Journal of Science, Engineering and Technology Research (IJSETR), Volume 4, Issue 7, July 2015

[7] “Model Based Predictive Control of Electrical Drives”

by Arne Linder, Rahul Kanchan, Ralph Kennel, Peter
Stolze.
[8] Junxia Mu and David Rees, “Approximate Model
Predictive Control for Gas Turbine Engines”,
Proceeding of the 2004 American Control Conference,
June-July 2004.
[9] “The Good Gain Method for PI and PID Controller
Tuning” by Finn Haugen.
[10] “Model Predictive Control” by Eduardo F. Camacho
and Carlos Bordons.
[11] Yang Su, KokKiong Tan, Tong Heng Lee,
“Computation Delay Compensation for real time
implementation of robust model Predictive Control”,
Journal of Process Control, ELSEVIER,2013.
[12] Vihangkumar V. Naik, D.N. Sonawane, Deepak D.
Ingole and DivyeshGinoya, “Model Predictive Control
of DC servomotor using Active Set Method”, IEEE
Multi-Conference on Systems and Control, IEEE,
2013.
[13] S. Joe Qin, Thomas A. Badgwell, “A Survey of Model
Predictive Control Technology”, Control Engineering
Practice, Science Direct, 2003.
[14] Mark L. Darby, Michael Nikolaou, “MPC: Current
Practices and Challenges”, Chemical Engineering
Practices, 2012.
[15] “PID Controller: Theory, Design and Tuning” by K.
Astrom and T. Hugglund. Pp.134-138.
.

2408
ISSN: 2278 – 7798 All Rights Reserved © 2015 IJSETR