2018 International Multi-Conference on Industrial Engineering and Modern Technologies (FarEastCon)

The Selecting of Artificial Intelligence Technology

for Control of Mobile Robots

Kafiev Irshat Romanov Petr

Federal State Budgetary Educational Institution of Higher Kolomna Institute (branch) of Federal State Budgetary
Education «Bashkir State Agrarian University» Educational Institution of Higher Education «Moscow
Ufa, Russia Polytechnical University»
e-mail: Kafiev02@mail.ru Kolomna, Russia
e-mail: romanov_p_s@mail.ru

Romanova Irina
Federal State Budgetary Educational Institution of Higher Education «K.G. Razumovsky Moscow State University of
technologies and management (the First Cossack University)»
Moscow, Russia
e-mail: i-p-romanova@yandex.ru

Abstract — The article analyzes the application of artificial wide range of overall sizes [1, 2, 7-11]. A lot of artificial
intelligence methods for control of mobile robots. The proposed intelligence technologies, such as fuzzy logic, artificial neural
method of selecting of the optimal artificial intelligence and neuro-fuzzy networks, artificial immune systems, genetic
technology by the set of available technologies for control of algorithms, swarm intelligence, are actively used during
mobile robots allows to improve the functional characteristics of creating of mobile intelligent robots. The main intelligent
the robot control system as a whole and its individual subsystems functions of mobile robot are processing of sensory
by the optimal coverage of a set of tasks and procedures solved information; assessment of the external situation; making
by the robot control system. The method is based on the adequate decisions on the selecting of behavior goals,
description of the initial information in the form of preference
planning ways to achieve them; implementation of these plans
relations on a set of alternatives. As a solution to the problem of
selecting an optimal information technology is adopted the most
by controlling the movement of the robot [1, 2].
nondominant alternative. The proposed method was applied to At the same time, to control a mobile robot, it is necessary
the selection of the optimal artificial intelligence technology for to select the most effective artificial intelligence technology or
modelling traffic control and technical operation field irrigation a combination of them, since the combination of several
in developing control system of mobile robotic sprinkler system technologies within a single concept of situational control
"Fregat". This method can be implemented in the design of leads to the possibility of creating high-quality systems both in
software not only for mobile robots, but robots for other
terms of speed of operation and in terms of efficiency and
purposes.
reliability in solving applied problems.
Keywords — artificial intelligence; artificial intelligence
technology; mobile robot; intelligent control system; performance II. EXPERIMENTAL SECTION
indicators of artificial intelligence technology; alternative; fuzzy The aim of the study was to develop a method of selection
preference relation.
of artificial intelligence technology for control of mobile
robot.
I. INTRODUCTION
Problem statement. In view of the complexity or
Mobile robotics is an important part of the industry, with impossibility of formalizing the control problems, their
applications in diverse industries, including manufacturing, solution is supposed to be carried out with the involvement of
agriculture, aerospace, mining, and medicine. Mobile robotics artificial intelligence methods.
is currently an effective tool for solving problems linked to a
threat to human life and health, economic and technological When building intelligent control systems for dynamic
limitations. Examples of such tasks are: military intelligence objects, which include intelligent control systems of mobile
operations, monitoring and data collection of environmental robots, it is necessary to take into account the following
trends and conditions, mapping the terrain, the study of other conceptual provisions: the development of the situational
planets, etc. [1-6]. control model (thus the class of possible situations is put in
line a characteristic set of acceptable solutions); the need for a
Today’s scientists are actively working on creating of hierarchical organization of the intelligent control systems of
autonomous land-based, air-based, water-based robots in a mobile robots; the validity of the selection of functional

978-1-5386-9535-7/18/$31.00 ©2018 IEEE

 2018 International Multi-Conference on Industrial Engineering and Modern Technologies (FarEastCon)
information technologies used to solve management problems. indicators that take into account the cost of developing
The movement of the robot is carried out by controlling the appropriate tasks and procedures by means of different
drives of its manipulative and transport systems that provide artificial intelligence technologies [1,19].
the required movement [2].
The system of performance indicators of artificial
The main reasons for using of applications of intellectual intelligence technology designed to implement a certain
technologies on the executive (drive) level of intelligent robot procedure is constructed by the following algorithm [1,19]:
control systems [2] are: the possibility to improve such
characteristics of the drive system as precision, speed, power 1. To define the set of alternative artificial intelligence
consumption, through the use of intelligent technologies; technologies – possible candidates for implementation of this
realization of the optimal mode of the enforcement system as a procedure: X = {xi}, i = 1,I .
whole due to commands in the top-level of control system; the
2. To determine the set of performance indicators (criteria)
drive level of the intelligent robot can have degree of
based on the above considerations and taking into account the
intelligence, enabling to solve the problems of the upper levels
specific features of the procedure: Q = {qk}, k = 1,K .
in cases of failure.
In different sources it is proposed to use different artificial 3. To calculate the rating of relative efficiency of artificial
intelligence technologies for control the robot as a whole or its intelligence technologies for each criterion on the basis of the
subsystems, including drives. In [12-15] it is proposed to use ratios: φ: X × X × Q → [0,1], where φ(x i , x j ,q k ) is adopted as
the methods of expert systems, associative memory, fuzzy the degree of preference of alternative xi to the alternative xj
logic, neural networks to create control systems, in [2] the on the criterion qk.
most promising are the hybrid of two technologies of artificial
In order to make the optimal decision when solving a real
intelligence expert systems and neural networks, in [16] the
task, it is natural to begin with the identification of a set of all
robot uses a learning algorithm with Q - learning. In [17] it is
permissible alternatives. Depending on the available
proposed to solve the control tasks with the help of various
information, this set can be described with varying degrees of
decision-making information technologies. Since electric
clarity. Let X, for example, is a set of artificial intelligence
drives are mainly used at the executive level of modern robot ~
control systems, fuzzy control is increasingly used to improve technologies available for use in the control system, A is a
the efficiency of their control [18]. fuzzy set of X with membership function μ A~ ( x). μ A~ ( x)
The tasks, solved by control systems of mobile robots, are should be understood as fuzzy description of a subset of
connected with various transformations and processing of options based on the characteristic "quality of the resulting
information. They differ significantly, both in terms of the solution". The μ ~ ( x) value determines the degree of validity of
A
content processing of sensor data, and in terms of the amount each specific option for that characteristic. If to take into
of information and computing work performed. The method account only this characteristic, any alternative from the set
for evaluating the degree of intelligence allows to classify the X= {x | x ∈ X , μ ~ ( x) = sup μ ~ ( y )}, (height of fuzzy set, any
A A
tasks, solved at different levels of the robot control system, y∈ X
and to select the theoretical apparatus for their solution [1]. At alternative that has the maximum degree of admissibility) can
the same time, the same task can often be solved by different be considered as the rational choice, because there is no reason
methods, using different approaches related to different to prefer the other. When receiving additional information
artificial intelligence technologies. Thus, it is necessary to (taking into account the values of other characteristics of
select the optimal technology, the most suitable for content artificial intelligence technologies), another choice may be
processing of information on individual procedures and for optimal. Moreover, this additional information may prove to
solving assigned tasks of the mobile robot control system. be such that on its basis it will be possible to identify the only,
It is safe to say that due to the optimal coverage of a set of the best of all alternatives. Information on which one
tasks and procedures solved by control systems of mobile alternative might be preferred to another could be provided in
robots, available artificial intelligence technologies and/or a variety of ways. In particular, if the strict quantitative
their combination can improve the functional characteristics of evaluation criteria are available, this information can be
the mobile robot control system as a whole, as well as its represented in the form of usefulness functions.
individual subsystems. However, this approach to the implementation of
Three groups of indicators are proposed to be used to information processing cannot be used for the assessment of
evaluate the efficiency of artificial intelligence technologies suitability of the artificial intelligence technologies, if the
implemented the tasks and procedures of mobile robot control quality evaluation criteria are available. For example, the
system: developing control system of mobile robotic sprinkler system
[20], when performing the irrigation operation, should take
indicators related to the estimation of time for computer into account the type of plants planted on the field, the shape
implementation of various tasks and procedures by means of of the field relief, the speed of movement of the mobile
appropriate artificial intelligence technologies; robotic sprinkler system across the field, soil moisture, soil
type, weather conditions, etc. Thus, the information coming
indicators that determine the qualitative characteristics of
into the control system of the robot is both quantitative and
the resulting solution;
qualitative, that’s why it is necessary to apply not only

978-1-5386-9535-7/18/$31.00 ©2018 IEEE

 2018 International Multi-Conference on Industrial Engineering and Modern Technologies (FarEastCon)
quantitative, but quality criteria to assess the suitability of If there are a few alternatives which have a most degree of
artificial intelligence technologies for information processing nondominat, it is possible to select any of them coming from
procedures. some additional considerations.
The description of the original information in the form of
preference relations on the set of alternatives is more suitable III. RESULTS SECTION
for the considered application domain [19]. The proposed method was applied to the selection of the
It is suggested the following algorithm for estimation of optimal artificial intelligence technology for modelling traffic
efficiency and selection of artificial intelligence technology control and technical operation field irrigation in developing
for solving task of mobile robot control: control system of mobile robotic sprinkler system "Fregat"
[20].
A. To set the initial data
A. The initial data
1) To define the set of alternative artificial intelligence
technologies – possible candidates for implementation of this 1) Let the set of artificial intelligence technologies
available for use in the control system is X={x1,x2,x3,x4},
task (procedure): X = {xi}, i = 1,I .
where x1 is a method based on genetic algorithm; x2 -
2) To define the set of critical indicators (criteria) for application of neural network; x3 - expert system; x4 -
evaluating of artificial intelligence technologies Q = algorithm based on the fuzzy control.
{qk}, k = 1,K . 2) Let the set of critical indicators (criteria) for evaluating
3) To define fuzzy relation of importance of the indicators of artificial intelligence technologies is Q ={q1,q2,q3,q4},
where q1 is precision; q2 – rate of problem solving; q3 -
μ: Q × Q → [0,1]. The value of μ (q k ,q l ) is understood as the
resource inputs; q4 - validity (correctness) of the problem
degree to which the indicator q k is no less important than the solution.
indicator q l , where qk , ql ∈ Q.
3) Let the fuzzy relation of importance of the indicators is
4) To set for each fixed indicator the fuzzy preference specified by the membership function μ (qi, qj):
relation on the set of alternative artificial intelligence
technology X, i.e. to set a function: φ: X × X × Q → [0,1], the q1 q2 q3 q4
value φ(x i ,x j ,q k ) is interpreted as the degree of preference of q1 1 0.4 0.6 0
alternative xi to the alternative xj by the indicator qk. q2 1 1 0.8 1
B. To carry out calculations in the following sequence q3 0.2 1 1 1
1) To define a fuzzy subset of nondominated alternatives q4 0.8 0 1 1
for the fixed indicator qk by all i,j (i ≠j) on the basis of the 4) Let the fuzzy preference relations on the set of
following relationship: alternative artificial intelligence technology X are set for each
indicator (TABLE I):
ϕ nd ( xi , q k ) = 1 − sup[ϕ ( x j , x i , q k ) − ϕ ( xi , x j , q k )] (1)
x j ∈X TABLE I. Fuzzy Preference Relations
Valuation indicators
2) To calculate the fuzzy preference relation induced on
the set of X by functions ϕ nd ( xi , qk ), μ (qk , ql ) , using the formula: Alternatives Precision q1 Rate of problem solving q2

x1 x2 x3 x4 x1 x2 x3 x4
η ( xi , x j ) = sup min{ϕ nd ( xi , qk ),ϕ nd ( x j , ql ), μ (qk , ql )}. (2)
x1 1 0.8 1 0 1 0.1 0.5 0.3
q k , q l ∈Q
x2 0 1 0.2 1 0.8 1 0.8 0.8
3) To find the fuzzy subset of nondominated alternatives x3 0 0.8 1 0 0.5 0.3 1 0
on the set (X,η) by all i,j (i ≠j) on the basis of correlation: x4 0 0 1 1 0,8 0 0 1

η ( xi ) = 1 − sup[η ( x j , xi ) − η ( xi , x j )].
nd (3)
x j ∈X TABLE I CONTINUATION
4) To correct the subset determined by correlation (3) Valuation indicators
ϕ nd ( xi , qk ), μ (qk , ql ) , using the formula: Alternatives Resource inputs q3 Validity q4
nd x1 x2 x3 x4 x1 x2 x3 x4
η nd ( xi ) = min{η ( xi ),η ( xi , xi )}. (4)
x1 1 0 0.8 0 1 1 0.9 0
5) To accept as a decision of task of selecting an optimal x2 0 1 0 0 0 1 1 1
information technology the alternative with the most degree of x3 0.1 0 1 0.4 0.4 0 1 0
nondominat, i.e. the alternative giving a maximum of function x4 0 1 1 1 0 0 0 1
(4): η nd ( xi ) → max .

B. The calculations

978-1-5386-9535-7/18/$31.00 ©2018 IEEE

 2018 International Multi-Conference on Industrial Engineering and Modern Technologies (FarEastCon)
1) The fuzzy subsets of nondominated alternatives for all References
indicators qk by all i,j (i ≠ j) were set on the basis of the
relationship (1): [1] P.S. Romanov, I.P. Romanova, “Artificial intelligence Systems”,
Kolomna: GSGU; KI (f) MPU, 2017, 244 p.
x1 x2 x3 x4 [2] R. Siegwart, I. R. Nourbakhsh, D. Scaramuzza, “Introduction to
Autonomous Mobile Robots (Intelligent Robotics and Autonomous
q1 1 0.2 0 0 Agents series)”, Publisher: The MIT Press, 2011, 472 p.
ϕ ( xi , qk ) :
nd
q2 0.3 1 0.5 0.2 [3] B. Siciliano, O. Khatib, “Handbooks of robotics”, Springer, 2008, 1628
p.
q3 0 0 0.3 1 [4] R. Robin, “Murphy Introduction to AI Robotics”, Publisher A Bradford
Book, 2000, 400 p.
q4 1 0 0 0
[5] A.V. Linenko, M.F. Tuktarov, Sh. F. Siraev and V.G. Baynazarov,
2) The fuzzy preference relation induced on the set of X “Fuzzy controller in the control system of the linear electric drive of the
were calculated by the formula (2): grain cleaning machine”, vol.12, issue 4, Electrotechnical and
information complexes and systems. Ufa: Ufa State University of
x1 x2 x3 x4 Economics and Service, 2016, pp. 34-40.
q1 1 0.4 0.4 1 [6] R.S. Aipov, A.V. Linenko, M.F. Tuktarov and V.G. Baynazarov,
“Analysis of pulsed operating mode of linear induction drive of grain
η ( xi , x j ) : q2 1 1 0.5 0.8 cleaning machine”, vol.133, Actual Issues of Mechanical Engineering
(AIME 2017). Tomsk: Atlantis Press, 2017, pp. 420-424.
q3 0.5 0.5 0.5 0.5 [7] R. H. Abiyev, I. S. Günsel, S. Abizada, “Fuzzy control of
omnidirectional robot”, Procedia Computer Science, vol. 120, 2017, pp.
q4 1 1 0.5 1 608-616.
3) The fuzzy subset of nondominated alternatives on the [8] B. Thomas, “Embedded Robotics. Mobile Robot Design and
set (X,η) by all i,j (i ≠j) was found by the formula (3): Applications with Embedded Systems”, Third Edition, Springer-Verlag
Berlin Heidelberg, 2008, 556 p.
x1 x2 x3 x4 [9] N. Mendes, P. Neto, “Indirect adaptive fuzzy control for
industrialrobots: A solution for contact applications”, Expert Systems
η ( xi )
nd
0.4 0.8 1 1 with Applications, vol. 42, Issue 22, 1 December 2015, pp. 8929-8935.
4) The subset determined in previous stage was corrected [10] J. W. Chen, H. Zhu, L. Zhang, Y. Sun, “Research on fuzzy control of
by the formula (4): path tracking for underwater vehicle based on genetic algorithm
optimization”, Ocean Engineering, vol. 156, 15 May 2018, pp. 217-223.
x1 x2 x3 x4 [11] X. Wang, M. Fu, H. Ma, Y. Yang, “Lateral control of autonomous
vehicles based on fuzzy logic”, Control Engineering Practice, vol.34,
η ( xi )
nd
0.4 0.8 0.5 1 2015, pp. 1–17. doi: 10.1016/j.conengprac, 2014.09.015.
5) As the decision of task of selecting an optimal [12] P. Chiacchio, S.Chiaverini, “Complex Robotic Systems”, 2006. 209 p.
information technology was accepted the alternative with the [13] D. Janglová, “Neural networks in mobile robot motion”, Cut. Edge
most degree of nondominat, i.e. x4 - the algorithm based on Robot, 2005, vol. 1, № 1, pp. 243–254.
the fuzzy control. [14] A. Zubizarreta, M. Larrea, E. Irigoyen, I. Cabanes, E. Portillo, “Real
time direct kinematic problem computation of the 3PRS robot using
neuralnetworks”, Neurocomputing, vol. 271, 3 January 2018, pp. 104-
IV. DISCUSSION SECTION AND CONCLUSIONS 114.
[15] N. H. Singh, K. Thongam, “Mobile Robot Navigation Using Fuzzy
For the specified control system of mobile robotic Logic in Static Environments”, Procedia Computer Science, vol. 125,
sprinkler system "Fregat" has been developed a mathematical 2018, p.11-17.
model describing the technological process of irrigation the [16] C. Touzet, “Q-learning for robots”, The Handbook of Brain Theory and
field with the use of analog control of the water flow. Neural Networks: second edition, M.A. Arbib ed., MIT Press, 2003, pp.
Described model takes into account the incompleteness and 934-937.
inaccuracy of initial data, in particular, the shape of the [17] P.S. Romanov, I.P. Romanova, “Approaches to the development of the
contour, the speed of movement of the machine across the intelligent control system of mobile robot”, Inženernyj vestnik Dona
(Rus), 2018, № 1. URL: ivdon.ru/ru/magazine/archive/n1y2018/4692.
field, soil moisture. Possible artificial intelligence
technologies, which can be used in the mathematical model in [18] O.S. Amosov, L.N. Amosova, S.N. Ivanov, “Synthesis of optimal
control systems of electromechanical heat generating complex using
these terms (genetic algorithms, application of neural network, fuzzy systems”, Informatics and control systems, 2009, №. 1 (19), pp.
expert system, algorithm based on the fuzzy control) were 73-83.
identified. Also the membership functions of the fuzzy [19] I. R. Kafiev, P.S. Romanov, I. P. Romanova,”Methodical apparatus of
relations of critical indicators for evaluating of artificial choice in information technology for management of agricultural
intelligence technologies and preference of artificial robots”, Journal of the Bashkir State Agrarian University 4(44), Ufa:
intelligence technologies for each indicator were defined. As a Publishing House of Bashkir State Agrarian University, 2017, pp.62-68.
result of the selection of the optimal artificial intelligence [20] P.S. Romanov, A.I. Ryazantsev, A.O. Antipov, I.P. Romanova, “Fuzzy
control of a robotic sprinkler type machine «Fregat»”, /Inženernyj
technologie was obtained the algorithm based on fuzzy vestnik Dona (Rus), 2018, №2. URL: ivdon.ru/ru/magazine/archive
control. This mathematical model can be implemented in the /n2y2018/4819.
software design not only for mobile robots but also robots for
other purposes (for example, intelligent robots), used in
various industries and human activities.

978-1-5386-9535-7/18/$31.00 ©2018 IEEE