Communications on Applied Electronics (CAE) – ISSN : 2394-4714

Foundation of Computer Science FCS, New York, USA

Volume 6 – No.7, February 2017 – www.caeaccess.org

Model Predictive Control for Positioning and Navigation

of Mobile Robot with Cooperation of UAV
Moustafa M. Kurdi Imad A. Elzein Alex K. Dadykin
Belarusian National Technical Belarusian National Technical Belarusian National Technical
University University University
Minsk, Belarus Minsk, Belarus Minsk, Belarus

ABSTRACT nonlinear and MIMO (Multiple Inputs Multiple outputs)

The purpose of navigation system is to help mobile robot in developed a multivariable control.
order to select an optimal and short path to reach the target. In A major difficulty in a mobile robot design is the controller
most of these systems, GPS are used to determine the robot design for trajectory tracking. The trajectory control may be
position. There are errors in positioning using GPS. This accomplished through numerous control techniques. When
paper considers the problem of navigating a Mobile Robot in considering a path that faithfully represents the desired
an unknown environment while maintaining visibility with a trajectory, it is necessary to introduce an approach that
(movable or non-movable) target by means of Fuzzy Model incorporates a data feed related to the actual location of the
Predictive Control (FMPC). The approach combines input mobile robot. This perspective is fundamental to running a
variables from different resources such as: GPS, RVS (Robot control to ensure the displacement of the robot in a desired
Vision System), and QVS (Quad-copter Vision System). In trajectory.
this paper, a new approach based on Fuzzy Model Predictive
Control (FMPC) is proposed to solve the positioning and The main objective of this work is to introduce and develop a
navigation problems for mobile robot. multivariable controller for mapping 4-wheel mobile robot
path using a model in state space that allows the robots to
General Terms move smoothly with obstacles avoidance. Robot’s position
Navigation; Model Predictive Control; Mobile Robot plays a key role in path tracking. The objective of the study of
path planning in mobile robotics is to give robots the ability to
Keywords plan their own movements, without the need for direct
Navigation; Fuzzy; MPC; Mobile Robot; UAV interference of humans. As a result, our approach is to use
model predictive control for robot trajectory generation
1. INTRODUCTION (navigation) and positioning.
Mobile robotics is the area of robotics that groups the robots
that have no fixed base, or they can get around within a Path trajectory for a four-wheel mobile robot (Belarus 132N)
limited or no space. Robot’s locomotion should be recognized is in fact a feedback control problem to which model-based
as a moving object, save its position within this space predictive control (MPC) has been an effective mechanism
understand the physical limits of space and adapt providing [1,2]. Most MPC control applications are based on linear
ways and fixed objects that can be obstacles to be diverted. models of dynamic systems to predict outputs over a certain
horizon. Future sequences of control signals are evaluated by
The robots have potential application in areas such as (i) minimizing a cost index that takes account of the future output
Support to medical services (Transportation of food, prediction errors over a reference trajectory, as well as control
medication, medical exams); (ii) Automatic cleaning of areas efforts. However, when the systems are non-linear and are
(Supermarkets, airports, industrial sites); (iii) Agricultural and operated over a large dynamic range, the use of linear models
Forests (fertilization, planting, fire preventing, tree cutting); becomes impractical, and the identification of non-linear
(iv) Hazard Environments (catastrophic areas, volcanoes, models for control becomes absolutely necessary.
nuclear power plants, inspection of gas or oil pipes); (v)
Military (Surveillance vehicles, Monitoring vehicles); (vi) Fuzzy Model Predictive Control (FMPC) [3], [4] combines
Safety ( Surveillance of large areas, buildings, airports, car input variables from different resources such as: GPS, robot
parking); and (vii) Civil Transportation (Inspection of image system, and quad-copter video system in order to
airplanes, trains). achieve localization i.e. Cartesian coordinates (x,y) and the
orientation θ.
A 4-wheel mobile robot is a robot that can move in any
direction in the horizontal plane depending on angle of This paper is organized as follows: Section II presents the four
rotation Ɵ. Thus, for locomotion, it is necessary that it has a wheeled mobile robot design, and Section III presents the
computer embedded system responsible for controlling all its fuzzy model predictive control. The proposed control system
functions. This computer system can send and receive is demonstrated in Section IV. Our conclusions and thoughts
information so that the robot can perform its functions. Thus, on future extensions are summarized in Section V.
this is called a control system. This system can be divided into
simpler control modules. 2. FOUR WHEELED MOBILE ROBOT
The Mobile Robot is a strongly coupled system which DESIGN
influence the rotation of front wheels directly influences the The wheel has been by far the most popular locomotion (how
speed of the other two wheels, as well as the resulting velocity the robot moves through its environment) mechanism in
of the center of mass of the robot. For this coupled system, mobile robotics and in man-made vehicles in general. It can

17
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

achieve very good efficiencies, and it does so with a relatively 2.2.1 Kinematic model of mobile robot
simple mechanical implementation. Almost all wheeled robot The kinematic model [5] of a mobile robot at point p(xp , yp,
research tends to focus on the problems of traction and θp) as shown in Fig. 3 is given by:
stability, maneuverability, and control: can the robot wheels
provide sufficient traction and stability for the robot to cover (1)
all of the desired terrain, and does the robot's wheeled
configuration enable sufficient control over the velocity of the (2)
robot?
Where x and y represent the coordinates of the centre of the
2.1 Design of Belarus-132N Mobile Robot axis of the actuated wheels on the plane X-Y and θ is the
The Mobile robot used in the experiment use chassis tractors angle formed by the longitudinal axis of the robot and the X
"Belarus-132N" platform with the petrol engine (HONDA axis. V represents the linear velocity, ω represents the angular
GX390). Its weight is about 400 kg with dimensions of velocity, V1 and V2 - longitudinal velocity of the wheel while
120×120×180 cm3. General view of the current mobile robot K - drive transmission ratio. The kinematic model of a four
system is shown in Fig. 1. wheeled mobile robot is shown in Fig. 4.

Fig 1: General view of the Belarus 132N Robotic system Fig 4: kinematic Model of a Wheeled Mobile Robot

2.2 Discrete Time Model of Belarus-132N 2.2.2 Discrete Time Model of mobile robot
Considering an example break Ts and a zero-appeal hold, eq.
Wheeled Mobile Robot (1) movements in discrete-time of a mobile robot at point p(xp
The robot is assumed to be moved in a flat plane where the , yp, θp) to:
motion P can be described in Ixp and Iyp direction using four
wheels. The direction θ determines the rotation of mobile
robot in counter clockwise. The robot motion can be described
in two coordinate systems: global coordinate system Ix ,Iy axis
and Local coordinate system Ex , Ey as shown in Fig. 2.

Consequently a virtual robot, the heading q d(t)=[ xd(t) yd(t)

θd(t) ] is viewed as coming about the accompanying kinematic
model, Mathematical statement of the virtual robot can be
communicated as:
(4)
Where pd=(xd , yd, θd) represents the desired pose, Vd the
desired linear speed, wd the desired angular.
Fig 2: Location of Wheeled Mobile Robot with respect to
global and local coordinate system Considering an example break Ts and a zero-appeal hold, eq.
(4) movements in discrete-time to:

(5)

Fig 3: Kinematic Position of a Wheeled Mobile Robot

18
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

3. OVERVIEW OF
MODELPREDICTIVE CONTROL
Model Predictive Control (MPC), is playing a vital propelled
control technique for troublesome multivariable control
problems and is widely implemented within industry and
getting more popular to those who are interested in control to
fulfill their duties. In fact predictive control would describe an
approach to the control design rather than a specific
algorithm. And thus an interpretation of the approach is
required to define a suitable algorithm that fits the needs. If a
clear and defined dynamic model is available for a process,
then the model and present measurements could be utilized to
reach forthcoming/future output values. At that point the
proper fluctuations in the input variables can be computed in Fig 5: Block diagram for model predictive control
accordance of the measurements and predictions.
Fundamentally, any changes in the input variables are 4. GENERALIZED NONLINEAR
managed with respect to the I/P-I/O relationships that are PREDICTIVECONTROL
introduced by the process model. The output variables are Figure 6 demonstrates the basic principle of MPC. At a
referred as controlled variables (CVs), and on the other hand denoted time (t) certain measurements are provided. These
input variables are referred to as manipulated variables measurements will trigger the controller and kick it to
(MVs). In addition, the measured disturbance variables are commence predicting the forthcoming (future) dynamic act of
known as (DVs) and in other terms may be called feed a system over a predicted horizon called Tp and furthermore
forward variables. determines the input at which optimization of predefined
Model predictive control is well known to provide many open loop performance goals are met.
favorable leverages; first as we begin by the process model it As long as neither disturbances exist nor mismatching model
would start in collecting the static as well as the dynamic plant presence is evident, in this case if there was a way for
interaction between the input, output, and DVs; secondly, the the infinite horizon optimization issue to take place in solving
constraints if any are existing on I/P and O/P which are taken it, hence at this point and at time of t=0 the input function can
into effect in the sense of systematic aspect; thirdly, the be incorporated in regards of the system in a condition where
control computations are possible to interact with finding the all times “t” is greater or equal to zero (t ≥ 0). However and as
ultimate set points, and the fourthly; building a model of a matter of fact, this is not feasible generally. Because of the
accurate predication can detect, trace and flag any upcoming factors of disturbances and the stressing mismatching model
issues. Realizing that the main objective that MPC has been plant, the real and predicted behaviors of the system will not
illustrated by Qin and Badgwell as: i) to avoid infringement of be the same. The integration of the feedback would depend on
constraints on I/O-O/P; ii) push part or few of the O/P the acquired I/P function of the open loop and shall be ready
variables to the set point (SP), while keeping up other O/Ps in for implementation whenever a next measurement is around.
defined set of ranges; iii) to set a static status on the input
variables; iv) and in the absence of a sensor to control the A difference in the time among the re-calculation incident and
most possible process variables. measurement can change but regularly is expected being fixed
however often it is assumed to be fixed, i.e. a measurement is
MPC is a comprehensive control system that consists of going to occur throughout each δ sampling time units. During
accompanying modules [6], [7], [8], [9]. MPC system block a time (t + δ) a new measurement is used, and later due to
diagram is shown in Fig. 5. In order to predict the output these measurements the process of both the optimization and
variables values a process model is going to effectively used. prediction will re-run to figure out and get another recent
Normally any variations between real and predicted outputs input function (new) in conjunction with both the
are going to be acting as a feedback signal to a prediction control/prediction horizons advancing ahead forward.
module. At every sampling instance those predictions are
utilized as a part of the following MPC calculations which
starts with the set point and control computations. Note that
inequality constraints whether occurring on the input and
output variables, can be involved in any of the above
calculations.
Targets which are known as set points and serve for control
calculations would be evaluated and computed from the
process steady state model (linear model). When the process
settings are changing the set points values change as well due
to these variations in the process circumstance particularly in
the changes of the inequality constraints. The changes and
variations in the constraint are mainly due to fluctuations in
process surroundings, apparatus, devices, etc.

Fig 6: Principle of Model Predictive Control

19
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

Nevertheless, the input can be assumed as constant during a Additionally, assume that the function is set for any
sampling time δ. The computations of the input that will be permissible control vector functions and
applied permits adding constraints on the states/inputs and , meets the system expectations (6) and (11) respectively;
more could go for optimizing certain cost function. and being nearby at a pace of .
Technically speaking the behavior of predicted system is
going to be different from a closed loop; therefore the stability The identified systematic model of differential equations (11)
of the closed loop will require more analysis to achieve it. has formed a close associative relation between the predictive
and mathematical models (6) of the control object.
According to the literature a nonlinear set of differential
equations as outlined and shown below will touch base the
issue of stabilization as we move through a set of systems [6],
[7]:
(6)

Which is represented to input and state constraints as set out

below,
(7)
Where x(t) denotes the states
vectors along with the inputs. The set of feasible input values
is denoted by X. Note that μ and X shown as constraints
according to the below form:
(8) Fig 7: Predictive plant movement
A delayed time , at the x-axis would be and during an initial
Note that umin, umaxas well as xmin, xmaxare assumed as constant state, let us consider =T. So far, the control object with an
vectors.We presume that the aim and main objective of a unidentified precise model of the form (6) is forward moving
control object (6) is to assure: through the influence of certain control executed
feedback scheme, and through the time it was in a state.
Let us consider an explicit control that would work as a
; (9) function of time interval and being having a full integration
with the system (11) at a defined interval with the initial state.
The resultant incomplete solution will be viewed as the
Where the vector functions; and identify a predicted behavior of the control object however through the
preferred passage of object. zone of prediction horizon as shown in Fig. 7.
The control of objects’ moves can be set or regulated by a We need to stress out that because of the differences that falls
specific functional; (1). among the dynamics of a real object and a predictive model of
(10) transit on the sector. Such a segment will be assumed an
entire difference, knowing that any sort of concurrence is
Optimal control problem would definitely need to look for a valid at the initial point only.
control action of a specified class, which assure the
accomplishment of setup goals (9) without forgetting the At this point we can formulate a mathematical problem of
choosing the optimal control based on the forecast. Further we
constraints and hence the
will suppose that the purpose of management is to provide a
functional minimizes (10).
predetermined pattern of behavior (11), it is determined as in
To overcome the problems of optimal control various (9), and vector functions and where - the
techniques are used and applied in the multiple forms and state vector; - the vector control,
approaches where we can take as a sample the theoretical
applications in model predictive control that was considered The structure of all controllers of the MPC family follows a
as the backbone of feedback based on the structure of control strategy as shown in Fig. 8 as follows: (i) Future output values
action that rely on and utilize the measured information on the are calculated every time t; (ii) Future control signals using an
object’s state. optimization method, process, and the reference trajectory is
calculated so as to converge as much as possible; (iii)
To illustrate and go through a predictive control theory, we Typically, this optimization method is quadratic as a function
will need to consider a mathematical model of the control of the difference between the predicted output and the
object (6) as well as we will need a set of differential predicted trajectory; (iv) The control signals measured only
equations that can be such as: the first sample found.

, (11)

20
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

Fig 8: Model Predictive Control Scheme

By introducing model predictive control approach a fuzzy model [12].
formulation of a constrained optimal control algorithm is and
incorporated throughout the receding horizon scheme [10]. The first two categories above are known as Mamdani fuzzy
models. Accordingly, TS fuzzy modeling structure can be
It is obvious that any sort of a model is constructed as a result defined as follows: (i) The input space separated into sub
of the physical plant characteristics as well as the control spaces; (ii) Then each sub-space system model converges to
problem normally and initially would be formulated by taking more simple linear models; (iii) Finally, as required by fuzzy
into account the objectives and constraints of the system. subsystem models of the rule base on the weighted average,
the overall structure is created fuzzy model [4].
4.1 Fuzzy Modeling
A highly nonlinear system controlled by a linear Model The path above in reference [10] Fuzzy Bend (Convolution) is
Predictive Controller may not give a satisfactory control expressed as the model. Accordingly, a non-linear system is
solution. To solve this problem, Nonlinear Model Predictive decomposed into p sub system and the behavior of each sub-
Control methods are proposed. But it is not easy to apply system to be linear or linear-like. TSA's modeling
these methods practically since the nonlinearity increases the requirement, a fuzzy linear-like pattern must be defined for
computational effort. Nonlinear process is defined by a set of each sub-claims [13]. These definitions are fuzzy inference
linear sub-models and algorithms decreasing the (fuzzy implication) is called and in the sub-region (Ri , i = 1,2,
computational effort to design controllers are proposed. In this .., p) is defined. In this way, the cause and effect relationship
study, an alternative approach to these methods, Fuzzy Model model, the control signal u, output y, and it is represented by
Predictive Control approach is proposed. discrete time difference equation in terms of the sampling
instant n. Each region is represented by a fuzzy Cartesian
According to the Fuzzy Model Predictive Control approach, a product as follows [10].
nonlinear process is described by a fuzzy convolution model
that consists of a number of linear or quasi-linear subsystems. Ri = y(n)× y(n − )1 ×...× y(n − m + )1 ×u(n)×
For each subsystem, a Model Predictive Controller is u(n −1 ) ×...× u(n −l + 1) (12)
designed and the control effort is fuzzily merged by using the
same rules used to merge subsystems. y(n-j) in the above expression, n-j instantly measured output
value u (n-j) , n-j is instantly measured input values.
Fuzzy logic, linguistic variables are expressed based on an
expert knowledge, it is a method used in order to quantify the Fuzzy inference rules are consisting of symbolic premise of
value of linguistic and linguistic rules [11]. Fuzzy logic is "if" part consisting of the numerical results and the phrase
frequently used in the modeling of complex systems. "then" consists of. In other words, the lower bending fuzzy
Depending on the nature of inference rules, there are three model has a structure as follows [10]:
different models of fuzzy systems: (i) Linguistic fuzzy Ri : if
models; (ii) Fuzzy related models; (iii) Takagi - Sugeno (TS)

21
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

y(n)Ai0 , y(n-1)Ai1, ….. , y(n-m+1)Aim-1 Accordingly, y (n + 1) is transformed to equation with the

following form [13]:
and (13)
(16)
u(n)Bi0, u(n- 1)Bi1, ….. , u(n-l+1)Bil-1
Then; Equation (13) for each fuzzy sub-models as seen when
calculating yj (n+1) and yj (n) is used instead of using y(n). n
In the above-mentioned common: is immediately used such an expression in order to minimize
the estimation error is known y(n) [10].
• Aij ,i’th fuzzy inference Amedaki y (n-j) of the output fuzzy
sets. After the output values calculated FMPC structure will be
introduced. However, it is necessary to explain the first classic
• Bij ,i’th fuzzy inference Amedaki (n-j) of the input fuzzy
MPC structure to explain the first classic MPC structure for it.
sets.
• hij ,i’th fuzzy inference Amedaki impulse response 5. PROPOSED SYSTEM
coefficients are. Differential GPS navigation system which complements the
fixed network of terrestrial base stations, improves position
• T is the model horizon. accuracy of one meter. But this accuracy is not quite enough
• Δ (n), u (n) and u (n-1) is the difference between [10]. to ensure the normal operation of the control systems of
vehicles, robots and unmanned aerial vehicles.
In this study, instead of impulse response of the subsystem,
differential equations were used. This change did not cause A new method to obtain accuracy is a combination of work
any difference in the rule base, just brought out a calculation standard GPS, RVS (Robot Vision System), and QVS (Quad-
of the difference in the form of sub-systems. copter Vision System)as shown in Fig.10. The GPS system
provides in this case, the exact locations of the binding data,
All pieces are pin fuzzy inference in the fuzzy bending model. RVS, and QVS are used to ensure the continuity of location
System output y (n + 1), the weighted average of the output determination process. Combining the data from the GPS,
value calculated by the fuzzy inferences can be found by [10]. RVS, and QVS allow obtaining the exact value of the
coordinates.
(14)

Wj term in the above equation, j th fuzzy inference degree of

reality, that is the multiplier ignition. Ignition factor fuzzy sets
"if" part is calculated on the basis of the following form [10]:

(15)
We use Triangular membership functions for each fuzzy
linguistic value as shown in Fig. 9 of the total equal to 1.

Fig 10: General scheme of QMRS

Fig 9: Membership functions used for the fuzzy model

22
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

Fig 11: Finding the Optimal Path of Mobile Robot

Using this method allows us to achieve a very high level of projects each found object with a vertex on a RVS-smart-
accuracy, but requires computing resources, which do not map-graph. Each found object has the following properties:
have the most powerful microprocessors and
microcontrollers. This will create a highly specialized a) Identifier - a unique number assigned to the new object
navigation systems (Fuzzy Logic) designed for each specific detection. Number assigned to this object as long as there is
area, for car-robots, agricultural robots and other automated no decision on the loss of the object;
equipment for civilian and military purposes. Moreover, high- b) Coordinate and orientation of object (xi , yi, θi);
precision positioning modules may become available in
smart-phones and other portable electronics without c) Width of new object;
increasing their ultimate cost. d) Length of new object;
5.1 Detection and Object tracking e) Height of new object;
Automatic detection and tracking of objects observed in a
sequence of images is an actual problem, often occurring in f) Life Time - the number of frames of a video sequence in
the development of navigation systems manned and which the object was present since the first detection;
unmanned aircraft. At the same time one of the most g) Object type.
important challenges is to ensure detection and tracking under
uncertainty relative spatial arrangement of the camera and 5.1.2 Quad-copter Vision System (QVS)
lens. There detection and tracking task airfield on board the QVS use image-down-view processing system to extract
aircraft in conditions when the initial relative positioning of elements or objects by processing each image in the video
the airfield and the aircraft is unknown. An essential sequence. QVS projects would each found an object with a
requirement for the developed algorithm is the ability to vertex on a QVS-smart-map-graph. Each found object has the
implement it in the onboard data processing systems. following properties:

5.1.1 Robot Vision System (RVS) a) Identifier - a unique number assigned to the new object
RVS use image processing system to extract elements or detection. Number assigned to this object as long as there is
objects by processing each image in the video sequence. RVS no decision on the loss of the object;

23
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

b) Coordinate and orientation of object (xr , yr, θr); regions found on the frame. The algorithm will measure the
distance between each pairs of vertices located at Smart-Map-
c) Width of new object; Graph.
d) Length of new object; The algorithm use information of object identifier from both
e) Height of new object; RVS and QVS, if the objects appear twice (coincide) on
smart-graph-map i.e. the object has been detected by both
f) Life Time - the number of frames of a video sequence in systems so the algorithm neglect one of them.
which the object was present since the first detection;
A recursive procedure is repeated for each received sub-
g) Object type. graph. The feasibility of using this algorithm is caused by the
presence of such parameter of the detected object, as the time
5.1.3 Trajectory Analysis of RQVS of life. At the stage of preliminary detection of the desired
Trajectory analysis algorithm is based on the decomposition
object is taken with the greatest lifetime and thus reduces the
of a smart-map-graph. At the entrance of the algorithm are the
probability of false detection.
measurement data - parameters of an image extracted by
processing each image in the video sequence. Downstream - a With the trajectory analysis algorithm analyzes a list of
list of found objects and their properties, in which each objects found in the previous step. The result of the algorithm
element of the list is marked with a unique number. for tracking the object coordinates of the object are taken,
having previously memorized identifier. The coordinates of
The first step of the algorithm is to construct a smart-map-
the found object can then be used to adjust the rate of moving
graph; the first group of vertices is the set of objects known on
of Mobile Robotas shown in Fig.11.
the current frame, and the second - the set of connected

Fig 12: General scheme of the proposed system of FMPC

24
 Communications on Applied Electronics (CAE) – ISSN : 2394-4714
Foundation of Computer Science FCS, New York, USA
Volume 6 – No.7, February 2017 – www.caeaccess.org

5.2 Calculation of new position (x, y, Ɵ) Predictive Control-Part I. The Basic Algorithm,
Fuzzy Model Predictive Control (FMPC) process input Automatica, Vol. 23, No. 2, pages 137-148, 1987.
variables from different resources such as: GPS, robot vision [2] J. B. Rawlings, Tutorial Overview of Model Predictive
system (RVS), and quad-copter vision system (QVS) in order Control, IEEE Control Systems Magazine, Vol. 20, No.
to achieve localization i.e. Cartesian coordinates (X ,Y) and 3, June 2000, pages 38-52.
the orientation θ.
[3] A. Jadbabaie, J. Yu, and J. Hauser, “Stabilizing receding
The algorithm of Fuzzy Model Predictive Control can be horizon control of nonlinear systems: A control
described as follows. Lyapunov function approach,” in Proceedings of the
1) Use the coordinate of GPS for Mobile robot. American Control Conference, San Diego, CA, USA,
1999, pp. 1535–1539.
2) Calculate the optical flow for the frame for both
RVS and QVS of smart-graph-map. [4] G. DE Nicolao, L. Magni, and R. Scattolini, “Stabilizing
receding horizon control of nonlinear time-varying
3) The total optical flow into two parts: the "terrestrial" systems,” IEEE Transactions on Automatic Control, vol.
and "un-terrestrial ". 43, no. 7, pp. 1030–1036, 1998.
4) In the "un-terrestrial " of the move to a cylindrical [5] R. Siegwart, ‎I. R. Nourbakhsh and D. Scaramuzza,”
coordinate system and calculate the angle of rotation Introduction to Autonomous Mobile Robots “,2nd
θ with respect to the last two frames, thereby Edition,2004, ISBN: 978-0-262-01535-6.
determining the angle of rotation.
[6] I. A. Elzein, Y. N. Petrenko, “An Integration of A
5) The "terrestrial" side of the highlight optical flow Predictive Control Model and MPPT for PV Station”.
vectors and to calculate displacement in the plane International Conference on Smart Systems and
XOY, received vector (x, y). Technologies, Osijek, Croatia. IEEE Trans. ISBN:978-1-
5090-3718-6. pp. 275-280, 2016.
6) Add (x, y) to the original position of the mobile
robot (Xm ,Ym ) in order to get a new position of [7] Sotnikova MV. Development toolkit for solving control
(X, Y). problems with non-linear predictive models in MATLAB
Tr. V Int. Conf. "Design of engineering and scientific
7) Go to step 1 for the next frame. Periodically update
applications in the environment MATLAB”, 2011, p.
the key points. In general, the process of calculating
299-319.
the optical flow is shown in Fig.12.
[8] Veremey EI, MV Sotnikova, Plasma stabilization on the
6. CONCLUSION AND FUTURE WORK basis of the forecast of a steady linear approximation, St.
In this paper, the motion control system and algorithm for a Petersburg. Univ. Ser. 10. 2011, Vol. 1, p. 116-133.
Mobile Robot was presented. The robot positioning and
localization was calculated using Fuzzy Model Predictive [9] Sotnikova MV, Questions of stability of motion in
Control. control systems with predictive models, Bulletin of
Voronezh State Technical University, 2012, p. 72-79.
The algorithm of FMPC use data from different inputs such
as: GPS, RVS, and QVS. The program is able to take into [10] Huang, Y. L. and Lou, H. H. and Gong, J. P. and Edgar,
consideration the occurrence of obstacles along the path of T. F., 2000. Fuzzy model predictive control, IEE
travel. The successful generation and simulation of two Transactions on Fuzzy Systems, 8(6).
different kinds of static obstacles from RVS and QVS are [11] Passino, K. M. and Yurkovich, S., 1998. Fuzzy Control,
projected on Smart-Graph-Map. Addison-Wesley Longman, California, 52.
The program has been developed in such a way that it [12] Nikolaou, M., Model Predictive Controllers: “A Critical
increases efficiency of moving and localization of mobile Synthesis of Theory and Industrial Needs”, Houston, TX
robot by the help of quad-copter which can be introduced into 77204-4792.
other related applications for unmanned navigation designs.
[13] Takagi, T. and Sugeno, M., 1985. Fuzzy identification of
7. REFERENCES systems and its application to modeling and control,
[1] D. Clarke, C. Mohtadi, and P. Tuffs, Generalised IEEE Trans. Syst., Man, Cybern.,15, 116–132.

25