Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Recent Trends in Model Predictive Control

Srikshana Sasidaran, Adithya R., Haris Raja V., M.P.Anbarasi
Department of Robotics and Automation Engineering
PSG College of Technology,
Coimbatore, India.

Abstract:- In this paper we are going to present the meeting the restrictions at the present step. The anticipated
recent trends of model predictive control (MPC) and its route with the J provides the best solution and selects the best
techniques are used in modern world. MPC forecasts steering wheel angle sequence to bring the automobile as
plant output behavior using a plant model. The MPC near the reference as feasible. MPC only applies the initial
controller solves the optimization problem across the step of this optimum sequence to the automobile at this time
prediction horizon while adhering to the constraints at step and ignores the rest. The automobile moves a certain
the current phase. This can be used in non-linear distance based on the steering wheel angle used.
problems and it is more precise when compare to the
linear controller such as PID. The controller receives a new measurement of the car's
lateral position at the following time step. It may differ
Keywords:- Model Predictive Control, Feedback Control somewhat from what the MPC controller projected
Strategy, Dynamic Control Algorithms. previously. This might result from an unmeasured
disturbance operating on the vehicle. For example, it may be
I. INTRODUCTION the wind or a slick road surface. The prediction horizon is
moved ahead by a single step, and the controller performs the
The controller’s purpose in a control problem is to same cycle of computations to determine the ideal steering
compute the input to the plant so that the plant output follows wheel angle for the next step. MPC is also known as receding
the desired reference. The technique used by a Model horizon control due to the forward-moving aspect of the
Predictive Controller to compute this input is to forecast the prediction horizon. For example, it may be the wind or a slick
future.MPC predicts plant output behavior using a model of road surface.
the plant. It also employs an optimizer to verify that the
expected future plant production corresponds to the targeted
reference. The following example demonstrates these stages.
Assume that there is an autonomous automobile is kept true
to the path by means of an MPC controller.MPC can handle
MIMO systems like this one, but for the sake of simplicity, it
is assumed that the accelerator is fixed, the automobile is
moving at a constant speed, and only the steering wheel angle
is adjustable to regulate the lateral position of the car.MPC is
known as the prediction horizon since it looks into the future.
It is frequently expressed as the number of future time steps
or the length of time into the future. The MPC controller
must locate the best-predicted path that is closest to the
reference. As a result, it models a variety of future
possibilities. However, it does not perform these simulations
in a random sequence. Instead, it does so in a methodical
manner.

Furthermore, this is where the optimizer comes into Fig 1: Model Predicitve Controller
play. The MPC controller reduces the error between the car's The prediction horizon is moved ahead by a single step,
reference and forecasted paths by solving an optimization and the controller performs the same cycle of computations
problem. It also strives to keep the steering wheel angle to determine the ideal steering wheel angle for the next step.
change from one step to the next to a minimum. Because the MPC is also known as receding horizon control due to the
steering wheel is twisted sharply, the passengers may find the forward-moving aspect of the prediction horizon.MPC is
ride unpleasant. This optimization problem's cost function J referred to in textbooks using general terms. The signals
incorporates both terms and is written as a weighted squared generated by the controller and supplied to the plantare
sum of standard errors and steering wheel angle increments. referred to as manipulated variables, while the plant outputs
MPC ensures that the steering wheel angle and vehicle are referred to as output variables. In the feedback diagram,
position remain within prescribed limits while reducing this State estimator can be observed. In the automobile example,
cost function. These are referred to as constraints. There is a it is assumed that the lateral position of the car could be
limit to how far the steering wheel maybe maneuvered. measured. However, if the states of a system cannot be
Another constraint is the placement of the automobile. To directly measured, they can be inferred by a state estimator
avoid accidents, the vehicle is designed to stay inside its lane and supplied back to the MPC controller.
and not wander off the road.The MPC controller solves the
optimization issue across the prediction horizon while

IJISRT22FEB004 www.ijisrt.com 249

Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
II. NEED FOR MODEL PREDICTIVE CONTROL
Since the 1980s, Model Predictive Controllers have
MPC, which is maybe the most common control among been employed in the process sector. Microprocessors'
specific engineers. However, it stands for model predictive application has extended to different sectors as their
control. MPC is a feedback control method that makes computational capacity has increased.MPC provides all of
predictions about a process's future outputs using a these advantages, but it is essential to remember that it
model.Initially simulations are run in the brain using the demands a robust and fast CPU with enough memory. The
resulting model. These forecast the future trajectory based on reason for this is that MPC solves an online optimization
the control actionsselected. Thenthe best action to bring the issue at every time step.MPC uses a system's model to
anticipated trajectory as near the desired trajectory as feasible anticipate its future behavior and then solves an online
is chosen.These are some of the reasons why model optimization algorithm to choose the optimal control action
predictive control is employed.MPC is capable of dealing that drives the expected output to the reference. It can operate
with multi-input and multi-output systems that may have Multi-Input Multi-Output systems that may have input-output
interactions between their inputs and outputs[Fig 2]. In this interactions.MPC can deal with restrictions and has preview
MIMO system, adjusting the second output influences the capabilities. It is also frequently employed in a variety of
first output. IfPID controllers are used, constructing them sectors.
would be difficult since the two control loops would work
independently as if they had no relationships. Furthermore, III. RELATED WORKS
constructing more effective systems would be significantly
more complex, necessitating the tweaking of far too many A. Development of a Model Predictive Controller for an
controllers’ gains Unstable Heavy Self-balancing Robot
This Model Predictive Controller based approach is
applied to a Two- wheeled Self-Balancing robot which is of
heavy weight. Two wheeled Self balancing robot is similar to
Segway. It is also similar to Inverted Pendulum [2].

Many Non-Linear equations are present in a Two-

Wheeled Self Balancing Robot. The robot is first controlled
using a classic PID controller. Closed loop PD controller is
used to drive the robot. Manually tuned Kp and Kd values are
fed into the system. Then Model tuning for the system is
done to reduce the error. The obtained model is used to
calculate Open loop step response which can be used in
Model Predictive controller (MPC).The manipulated variable
in MPC is the Motor Speed and the controlled variable is the
Tilt Angle. The Parameters of the algorithms are tuned and
final parameters are obtained.MPU algorithm works
Fig 2: MIMO System with MPC efficiently compared to the output of PD controller. The
linear model gives only approximate results whereas MPC
MPC has the advantage of being a multivariable stabilizes the robot for even a strong disturbance in a short
controller that regulates the outputs simultaneously while span of time.
accounting for all system variable interactions. Another
feature of MPC is its ability to deal with constraints. B. Design and Development of Model Predictive Controller
Constraints are required because breaching them may result for Binary Distillation Column
in unforeseen consequences. One of the driving safety The concept proposed here with employs a Model
constraints, for example, is that autos must adhere to speed Predictive Controller in the Industrial Binary Distillation
regulations and maintain a safe distance from other vehicles. Column [23]. The purification of final products in Petroleum
Other controls, such as acceleration limits, are imposed by and Chemical Industries is done by a separation process
the car's physical limitations. If this were a self-driving called as Distillation. Mass transfer or heat energy transfer is
automobile driven by MPC, the controller would follow a enhanced using Distillation process.The control inputs of this
desired trajectory while adhering to these limits. MPC also configuration are the vapor flow rate and liquid flow rate.
has a preview function, which is akin to feedforward control. The objective of this model is to maintain the output product
Assume the autonomous vehicle is traveling on a winding concentration due to disturbance in feed rate and feed
road. If the controller is unaware that a corner is approaching, concentration. The basic idea is to reduce the error and to
it will only apply brakes while entering the corner. However, predict the future response of the distillation system. The
if the automobile includes a front-facing camera that provides system is tested with general PID controller as well as with
information about the vehicle's future trajectory, the Model Predictive Controller using SISO and MIMO systems.
controller will be aware of the forthcoming corner in MPC takes into account of the input variables and the
advance. As a result, it may break sooner in order to stay in manipulated output variables and the control variable output
the lane safely. To increase controller performance, MPC is obtained by minimizing the cost function [Fig 3].Wood
may readily include future reference information into the and Beery Binary distillation system is taken and PID and
control issue. MPC controller is applied to the system. The MATLAB and

IJISRT22FEB004 www.ijisrt.com 250

Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Model Predictive Tool Box is used for the simulation of the it will become impossible for the traditional methods to
system. comprehend and process the data. This is where the Model
Predictive Control algorithm comes into play.This control
algorithm is one of the most or the most versatile and
compatible algorithms to ever exist. This is because; this
control strategy can be combined with any one algorithm or
piece of code to form a completely new type of control
algorithm. These combined algorithms are called or referred
to as hybrids[Fig 4]. To provide a comprehendible
illustration, the predictive control algorithm can be combined
with engineering economics to form the economic non-linear
model predictive control strategy.Neural Networks can be
combined with the model predictive control to form an
Fig. 3: Structure of MPC optimal non-interfering administering algorithm.
The performance of the MPC controller is compared
with that of the traditional PID controller. Based on the
proceedings of the paper, it can be presumed that MPC
controller shows better results compared to PID controller.
MPC controller provides smooth reference tracking, peak
overshoot is reduced significantly and better closed loop
performance.

C. Model Predictive control and Optimization for

Papermaking process
Large scale 2-Dimensional process is involved in
papermaking process. Machine direction control and Cross Fig 4: MPC hybrid algorithms and their applications
direction control are the two challenges to overcome in the
process of papermaking [6].MPC offers a variety of uses in V. BENEFITS OF THE MUTATIONS OF MODEL
paper machine control. MPC handles basic MD control and PREDICTIVE CONTROL STRATEGY
advanced MD-MPC control, which includes economic
optimization and orchestrates transitions across paper The Model Predictive Control Strategy can be
grades.Cross direction can be controlled using MPC, using a surprisingly rewarding, despite its predicaments due it its
properly determined solution approach.The papermaking CD infant stage. One of the most common drawbacks present in
method is a two-dimensional system with a vast size. It has a the prevailing control strategies is that it cannot sustain
high level of input-output coupling. MPC is a common commanding multiple input and output devices at a time. But
approach for regulating multivariable systems, and it has when it comes to the Model Predictive Control algorithm and
evolved into a common advanced control strategy in its hybrids can handle multiple input and multiple output
papermaking systems. However, mill staff have various devices with such ease. These algorithms have the capacity to
challenges in accepting CD-MPC, including the unique make constrained processing of the data to seem like child’s
multivariable control concept and the non-trivial tuning play.
approach.CD-MPC has been successfully implemented in
over 70 paper mills and has been applied to virtually all types
of existing CD processes, including fine paper, board,
newsprints, tissues, and so on. Over the next decade, CD-
MPC will undoubtedly have a substantial influence on
papermaking CD control applications.The ability to combine
numerous CD actuator arrays and multiple CD measurement
arrays into a single CD controller is a key advantage of CD-
MPC.Non-standard CD measurements, such as fiber
orientation, gloss, web formation, and web porosity, are
anticipated to be included into the existing CD-MPC
framework in the next generation of CD-MPC applications.

IV. TYPES OF MODEL PREDICTIVE CONTROL AND

ITS APPLICATIONS

The Model Predictive Control method became a

futuristic necessity because almost all of the industrial
processes are developing and becoming more and more
intricate in leaps and bounds. At one point, these industrial
processes are bound to attain such a level of complexity that
Fig 5: Types of MPC fields

IJISRT22FEB004 www.ijisrt.com 251

Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VII. PROSPECTS OF MODEL PREDICTIVE CONTROL
Initially, the predictive control strategies are divided STRATEGY
into three categories [Fig 5], namely
 Non-linear The Model Predictive Control method when proved
 New up and coming fields practically and economically viable can provide to be useful
 Random fraternity. in a diverse range of applications be it either large scale
intricate systems or non-linear systems.Though this
The non-linear category can further be divided into prospective is quite a challenge to achieve, this control
three sub-categories namely, linearized applications, multi- algorithm can be used to administer the large-scale
model practices and hybrid systems. The multi-model applications like sewage disposal systems and management
approach is used to facilitate the approximation of non- systems of urban traffic. This method can prove to be
prismatic operation variables. And the hybrid systems can advantageous due to its disturbance soundness, global
help with constantly varying dynamic processes with the help stability and confluence capabilities.This method can also be
of logical switching techniques.When it comes to the random utilized in rapid moving applications be it the aerospace
applications, it exercises the concept of probabilistic fraternity or electro-mechanical applications. The Model
uncertainty. This can be classified into two categories as Predictive Control algorithm has been successfully proven in
 Uncertainty and comparatively slower dynamic processes, but has been pretty
 Disturbance. difficult to achieve sustainable efficacy in procedures that
require a massive number of computations to process the
VI. IMPEDIMENTS IN THE APPROACH data.But when it comes down to the non-linearized
applications, this method needs to utilize a titanic amount of
Since the Model Predictive Control strategy is an up- data, calculations and processing iterations to operate
and-coming field in the fraternity of manipulator control and efficiently. Due to these hazards, this has not been
trajectory planning, there are a variety of complications and implemented in more well aged practical operations.
hardships that has to be worked around. The drawbacks and
problems posed by this method can be distinguished into The Model Predictive Control strategy can also be
many diverse points of views including from the algorithm, utilized for controlling and monitoring non-domestic
object application and computing necessities. Some of the buildings as it highly brings down the emissions of Carbon-
most prominent ones are listed below: di-oxide when compared with the pertaining restrained
 In its primitive stage, the MPC method can be applied only control systems and strategies. In addition to that, this
in scale-based applications in production. algorithm also proves to be an economically sustainable
 This control algorithm requires computers that have high alternative to the existing monitoring systems present in full
processing and performance capabilities and also scale buildings and skyscrapers. The energy efficiency
considerably depreciates the dynamic operations. offered by this method is one of the best alternatives to have
 Due to the rising requirements of the MPC algorithm, it been developed.
substantially hinders the possibility of development in not
When it comes to power plants and power stations that
only applications but also in space consumption.
use gas turbines fueled by biomass and natural gases to
 It is very difficult for the normal, orthodox computers and
generate electricity, the Model Predictive Control algorithm
similar devices that is used on a day-to-day basis to can be availed to increase the performance metrics during the
calculate the real time data obtained from this method.
startup of the plant. This can be achieved by minimizing the
 Since this control method is still in its development stage, time it takes to attain the desired set point and by avoiding
this is not one of the most accurate systems to exist. the unforeseen catastrophic and hazardous situations that can
 When the Model Predictive Control strategy was applied in happen when starting up the plant.
real time applications in the industries, it could only satisfy
the initial industrial standard requirements. VIII. CONCLUSION
 When it comes to optimizing the non-linear problem
statements, this approach lacks the procession of proper Model Predictive Control is one of the most promising
effective algorithms. feedback control strategies to be employed given its ability to
 For this method to be successful, it has to compute and prospectively facilitate not only online optimization, but also
calculate a huge amount of data. Thus, this leads to the enables the industrial processes to be controlled with way
computing requirements of this algorithm to be fairly more efficacy than the present prevailing systems. MPC can
massive. be implemented in MIMO systems in contrast to other
 This control strategy utilizes a large amount of online data conventional controllers which can handle only SISO
and applications. So, this method might not be as systems. In spite of the absence of any visible connects
economically feasible like other alternate methods. between this feedback control strategy and the level of its
 The Model Predictive Control strategy is still in incubation. practicality in the day-to-day industrial operations, this
Hence, there is a huge disconnect between its theoretical algorithmic approach has the better aspect of being well
possibilities and its practical applications. suited for a multitude of applications in a variety of industries
 The remunerative and gregarious enlargement and like healthcare, shipping, transport, energy, urban
evolution rate of this approach remains unmet. development, agriculture and aerospace.

IJISRT22FEB004 www.ijisrt.com 252

Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
REFERENCES Control algorithm[J]. Control Engineering Practice,
2014, 28(1):26-33.
[1.] Parisio A, Rikos E, Glielmo L. Stochastic model [18.] Chen Hong-Li, Lan Hai, Shen Yi. Study on generalized
predictive control for economic/environmental predictive control for ¼ type rudder pitching
[2.] operation management of microgrids: An experimental stabilization. Journal of System Simulation, 2006,
case study [J]. Journal of Process Control, 2016, 43:24- 18(12): 3508-3511.
37. [19.] Sun Guang, Huo Wei. Direct-adaptive fuzzy predictive
[3.] Grosso J M, Ocampo-Martínez C, Puig V, et al. control of satellite attitude. Acta Automatica Sinica,
Chance-constrained model predictive control for 2010, 36(8): 1151-1159.
drinking water networks, Journal of Process Control, [20.] Kouvaritakis B, Cannon M, Rakovillc S V, Cheng Q F.
2014, 24(5):504-516. Explicit use of probabilistic distributions in linear
[4.] Percival M W, Wang Y, Grosman B, et al. predictive control. Automatica, 2010, 46(10): 1719-
Development of a multi-parametric model predictive 1724.
control algorithm for insulin delivery in type 1 diabetes [21.] Wang C, Ong C J, Sim M. Convergence properties of
mellitus using clinical parameters, Journal of Process constrained linear system under MPC control law using
Control, 2011, 21(3):391-404. affine disturbance feedback. Automatica, 2009, 45(7):
[5.] ODwyer E, Tommasi L D, Kouramas K, et al. 1715-1720.
Prioritised objectives for model predictive control of [22.] Zhang J. Design of a new PID controller using
building heating systems[J]. Control Engineering predictive functional control optimization for chamber
Practice, 2017, 63. pressure in a coke furnace[J]. Isa Transactions, 2016,
[6.] Chu, Danlei & Forbes, Michael & Backstrom, Johan & 67:208.
Gheorghe, P.Eng, Cristian & Chu, Stephen. (2011). [23.] Killian M, Mayer B, Kozek M. Cooperative fuzzy
Model Predictive Control and Optimization for model predictive control for heating and cooling of
Papermaking Processes. 10.5772/18535. buildings[J]. Energy & Buildings, 2016, 112:130-140.
[7.] Formentin S, Guanetti J, Savaresi S M. Least costly [24.] Rajendran, Sivakumar & Mathew, Shennes. (2013).
energy management for series hybrid electric Design and Development of Model Predictive
vehicles[J]. Control Engineering Practice, 2016, 48:37- Controller for Binary Distillation Column. International
51. Journal of Science Education. 2. 445-451.
[8.] Hartley E N, Trodden P A, Richards A G, et al. Model [25.] Chikasha P N, Dube C. Adaptive Model Predictive
predictive control system design and implementation for Control of a Quadrotor[J]. IFACPapersOnLine, 2017,
spacecraft rendezvous[J]. Control Engineering Practice, 50(2):157–162.
2012, 20(7):695-713. [26.] Vasičkaninova A, Bakosova M, Mészáros A, et al.
[9.] Faulwasser T, Grüne L, Müller M A. Economic Neural network predictive control of a heat
Nonlinear Model Predictive Control[J]. 2017, 5(1):224- exchanger[J]. Applied Thermal Engineering, 2011,
409. 31(13):2094-2100.
[10.] Li J, Wu C, Li S, et al. Optimal disturbance rejection [27.] [25] D. Mayne and S. Rakovic, “Optimal control of
control approach based on a compound neural network constrained piecewise affine discrete time systems using
prediction method[J]. Journal of Process Control, 2014, reverse transformation,” in Proc. 41th IEEE Conf. on
24(10):1516-1526. Decision and Control, Las Vegas, Nevada, USA, Dec.
[11.] Feng X, Patton R. Active fault tolerant control of a wind 2002, pp. 1546–1551.
turbine via fuzzy MPC and moving horizon [28.] M. Okulski and M. Ławryńczuk, "Development of a
estimation[J]. IFAC Proceedings Volumes, 2014, Model Predictive Controller for an Unstable Heavy
47(3):3633-3638. Self-Balancing Robot," 2018 23rd International
[12.] Gorges D. Relations between Model Predictive Control Conference on Methods & Models in Automation &
and Reinforcement Learning[J]. 2017, 50(1):4920-4928. Robotics (MMAR), 2018, pp. 503-508.
[13.] Henson M A. Nonlinear model predictive control: [29.] ] A. Alessio and A. Bemporad, “Feasible mode
current status and future directions.Computers and enumeration and cost comparison for explicit quadratic
Chemical Engineering, 1998, 23(2): 187-202. model predictive control of hybrid systems,” in 2nd
[14.] Kennedy J L, Travis R C. Nonlinear model reduction IFAC Conference on Analysis and Design of Hybrid
and model predictive control of residential buildings Systems, Alghero, Italy, 2006, pp. 302–308.
with energy recovery [J]. Journal of Process Control, [30.] S. J. Qin and T. A. Badgwell, “A survey of industrial
2014, 24(6):723-739. model predictive control technology,” Control
[15.] Fele F, Maestre J M, Hashemy M, et al. Coalitional Engineering Practice, vol. 11, pp. 733–764, 2003.
model predictive control of an irrigation canal[J]. [31.] F. Borrelli, A. Bemporad, M. Fodor, and D. Hrovat,
Journal of Process Control, 2014, 24(4):314-325. “An MPC/hybrid system approach to traction control,”
[16.] Wu X, Shen J, Li Y, et al. Fuzzy modeling and stable IEEE Trans. Contr. Systems Technology, vol. 14, no. 3,
model predictive tracking control of largescale power pp. 541–552, May 2006.
plants[J]. Journal of Process Control, 2014, [32.] N. Giorgetti, G. Ripaccioli, A. Bemporad, I.
24(10):1609-1626. Kolmanovsky, and D. Hrovat, “Hybrid model predictive
[17.] Mythreyi K, Subramanian S C, Kumar R K. Nonlinear control of direct injection stratified charge engines,”
glucose–insulin control considering delays—Part II:

IJISRT22FEB004 www.ijisrt.com 253

Volume 7, Issue 2, February – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
IEEE/ASME Transactions on Mechatronics, vol. 11, no.
5, Aug. 2006
[33.] R. M¨obus, M. Baoti´c, and M. Morari, “Multi-
objective adaptive cruise control,” in Hybrid Systems:
Computation and Control, ser. Lecture Notes in
Computer Science, O. Maler and A. Pnueli, Eds., no.
2623. Springer-Verlag, 2003, pp. 359–374.
[34.] N. Giorgetti, A. Bemporad, H. E. Tseng, and D. Hrovat,
“Hybrid model predictive control application towards
optimal semi-active suspension,” International Journal
of Control, vol. 79, no. 5, pp. 521–533, 2006.
[35.] P. Ortner, P. Langthaler, J. G. Ortiz, and L. del Re,
“MPC for a dieselengine air path using an explicit
approach for constrained systems,” in Proc. 2006 IEEE
Conf. on Control Applications, Munich, Germany, Sept.
2006.

IJISRT22FEB004 www.ijisrt.com 254