Page 1 of 17

Transactions on Engineering and Computing Sciences - Vol. 11, No. 1

Publication Date: February 25, 2023

DOI:10.14738/tecs.111.14012. Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences,

11(1). 45-61.

Services for Science and Education – United Kingdom

Artificial Intelligence Techniques Application in UAS

Nickolay Zosimovych

School of Robotics, Xi’an Jiaotong-Liverpool University, China

ABSTRACT

This paper depicts the progress of an appliance of Artificial Intelligence (AI) for

Unmanned Aerial System (UAS) control. This job was done as part of the conditions

for lectures and labs in AI at the School of Robotics, and a beginning project platform

for research projects at Xi’an Jiaotong-Liverpool University for a flexible, healthy,

and intelligent unmanned aerial vehicle (UAV) mission control system. A method is

outlined which allows a base level application for applying an artificial intelligence

method, fuzzy logic, to aspects of control logic for UAV flight. The benefits of using

fuzzy logic for a movement planner were shown in this research: a small number of

failures, near-optimal pathways, and low control attempts. It is these benefits that

make it a nice tool for additional improvement. Some elements of UAV flight,

automated direction, altitude is held, and prevention of dynamic obstacles, have

been implemented and results analysis displayed. Certain potential clear future

works that were not investigated in this study are motion-planning in three

dimensions, prevention of dynamic obstacles, and more robust analysis.

Keywords: Artificial Intelligence (AI); Unmanned Aerial System (UAS); Unmanned Aerial

Vehicle (UAV); Fuzzy Logic; controller; inaccuracy.

INTRODUCTION

Unmanned Aerial Systems (UAS), additionally stated to as drones or unmanned aerial vehicles

(UAVs), carry out an extensive not only archeological, urban, environmental, and agricultural

monitoring, military application that also reveals their potential in supporting warfare efforts

[1]. UAS consists of aircraft components, sensor payloads, and a ground control base. Later,

operated by operator in adding up to a committed human “pilot” (supplemented in some

situations by another “spotter” to confirm safety), differs commonly in its structure depending

on the base and mission. Faithful control systems may be committed to large UAVs and installed

onboard vehicles or in automobile trailers to allow proximity to UAVs restricted by variety or

transfer abilities. The smallest categories of UAVs are often escorted by ground-control places

consisting of laptops and other components small enough to be carried easily with the aircraft

in small vehicles, or in backpacks.

In recent years, the use of UAVs that can run flight routes independently and develop geo- referenced sensor data has grown dramatically for environmental and wildlife monitoring

purposes [2]. Certain questions confining the broader use of UAVs for environment

management and exploration involve UAV regulations, effective costs, and public insight. One

of the most crucial limitations, nevertheless, is the demand to create or apply improved

computerized image detection algorithms created for this mission.

Page 2 of 17

46

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Last decade universities in many countries all over the world keep on to study UAS and continue

to involve the applications for artificial intelligence and traditional control methods, which still

guide as military as civil aircrafts [3].

Current study efforts are looking at developing the UAS to full up self-governing mission. Even

though there appear to be several descriptions of what the autonomy includes, each discipline

has emphasized its sense and gains. Some publications, like [4] talk about a good summary of

the problems engaged in achieving control of the development methods now. Authors are

outlining the heterogeneous modelling and design of an advanced control system approach.

One of the key questions faced is the fact that the systems and sensors concerned are

heterogeneous in nature. These systems must be integrated and determined into an operating

system. For the greatest part, this will be correct for the most complicated UAV systems, which

effort full autonomy [3].

BACKGROUND

Innovative purposes of artificial intelligence (AI), which consist of intelligent agents, are

offering new fields of study [5]. Numerous investigations have increased throughout the last

decade within the applications of AI techniques put on to the ideas of exploring for the UAS. So

often modern UAS can be defined as any platform that is operated without humans onboard.

UAS, which exists today, involves quadrotors, helicopters, airplanes, balloons, and sometimes

satellites. Their independence differs from human interface and a remote console to fully

autonomous take-off and landing.

Prior investigation, performed in Ref. [6], applied AI techniques such as expert systems and

schedulers, which used rule-based systems. These systems tried to create the art of flying an

aircraft into a rational sequence of actions to maintain control of specific functions of the

aircraft. This idea will still be a vital element with systems designed for future purposes.

The primary idea of employing an expert system to control an aircraft looks straightforward at

first but shows how hard it is to use. Certain problems that still happen today contain the fact

that an expert pilot’s decision-making procedure is tricky to imitate with laptops. The dynamic

nature and dynamic circumstances influencing aircraft or drones are fields that must be fitted

to such an expert system.

Next study attempts started to separate the many tasks engaged in the UAV flight control into

adaptable stages. Focused attempts of using logic methods related to fuzzy logic and neural

networks, are being used to enhance the mathematical results for flight control. Dr. Anthony J.

Calise from Georgia Tech (USA) in Ref. [7] said that “several areas of control for UAV’s can be

adapted to the use of neural networks”. The writer provides some descriptions of the

advantages of employing neural networks in addition to the UAS typical control methods

because of the integration of these systems.

THE PROBLEM STATEMENT

Last investigation attempts are being made to contain an integration of more AI techniques.

These recent research projects are the curious innovations taking place. How does a robot fly

Page 3 of 17

47

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

an aircraft like a human can? What does a human guess about in the practice of responding to

the environment? Can a system “think” sufficient to fly a UAV fully autonomously? These are

problems forcing UAS to examine.

Investigation of the uses of AI comes to a place where the methodologies, algorithms and

models need to be used for additional interpretation of their usefulness and pertinency to any

certain project for a UAS [8], [9]. Most of the investigation papers reviewed gave excellent

theory but did not give adequate evidence on the mechanics or measures for employing the

techniques drawn up to set up a research endeavor.

The aim of this study has been to draw some of the Al methods applied for UAV flight control

and examine several of the instruments used to apply AI methods. The aim is to achieve with

the application of employing AI methods for really managing various characteristics of UAV

flight. These kits and procedures provide as the project’s essential turn.

Artificial Intelligence of Things

Artificial Intelligence of Things (AIoT) is the mixture of AI and IoT (Internet of Things) [10]. AI

has constantly relied on large data sets to build effective algorithms. IoT might consume alive

data to AI to create further intricate algorithms and use “sense” to real-time data. AI could be

used to convert IoT data into valuable knowledge for enhanced decision-making procedures.

Significance, AIoT is equally valuable for both tools.

Throughout a mission, if a drone operates machine learning processes to all the data gathered

by its various sensors, it might make moment-to-moment judgments around how to react to its

situation. For example, in reply to wildfire, drones could be forwarded to give up fire retardant.

Centered on the situations and datasets trapped by their IoT-enabled sensors, they might

decide separately on the greatest place to drop the material. These terms would involve the

wind direction and speed; existing temperature; and the percentage of fire enclosed [11].

Merging the information from the UAV sensors and the data point from the IoT, UAS could

obtain useful situational understanding for autonomous decision making. A potential

development of a UAS missions could include UAVs that find out of houses, verify how many

people are in a relatively failed construction, and with other records inputs, such as data from

social media, construction types, and well-being data on inhabitants, the UAV could assess if a

trapped individual is like diabetic and needs insulin or so [10].

UAVs Swarm

The highly advanced utilizes of AI in UAS missions to-date are in the uploading and treating of

taken pictures in the cloud after the mission. Thanks to enhancing communication technology

and small-scale supercomputers, UAVs can now handle and deliver only appropriate data to the

cloud. The following stage of UAS application that specialists predict is a swarm around 100 or

even much more of intelligent UAVs that work simultaneously, communicate with each other,

and use AI to get real-time decisions at the same time.

Page 4 of 17

48

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Generally, there are two main elements for UAVs swarm to use: communication amongst its

representatives and AI. Being able to connect with the other UAS creates a feedback system that

is intended for band support. AI facilitates each UAV to fly on its own and to play a role in its

environment, as well as the swarm, through pre-programmed options or choices created by AI.

The major question for an efficient application of swarm UAVs is networking the devices to

guarantee that they are informed of their environment and have got the capacity to implement

a particular mission profile [10]. Synchronized evaluation is vital to efficiently swarm UAVs

missions. It means to the outcome of UAV to go through with an environment as one in sort to

effectively deal with that environment. Each UAV employed determines where to go based on

its environments, they communicate at standard periods to distribute information on their

world and get to additional conclusions concerning their managed investigation centered on a

group pattern.

Due to engineering developments in data transmission expertise and in onboard

supercomputer capacity, the use of swarm UASs is technologically reasonable. To date they

have primarily applied to perform regular duties for entertainment functions such as

mentioned at Ref. [12] at the 2017 Super Bowl halftime show, where Intel flew hundreds of

drones.

Synchronized UAS mission could be employed to get significantly more situational perception.

For example, first respondents could release a rescue raid of hundreds of UAVs to search for a

destroyed region, draw it, and use AI to recognize prospective targets in a brief quantity of time.

Even though these kinds of processes remained respected science fiction in the past, numerous

effective strategies show that they are scientifically and engineering achievable. Nevertheless,

such tasks can only be tested on a bigger level when policies are set [10].

THE APPROACH

There is an expansion in study and articles on achievements and suggested exploring below the

issue of UAS. The area was pointed along to consist of AI techniques that could be used easily

in a lab setting [3].

Throughout the achievement of the stages in Figure 1, the subsequent design was done. The

choice to use Python programming occurred due to a condition for the AI group. By the way,

this programming language turns out to be extremely valuable for this type of project. The

subsequent synopsis is given for the common methodology:

1. The initial attempt to solve the problem was to examine the past and current endeavors

of UAV flight control. This study was performed by literature analysis.

2. Investigation was made as soon as separating the option for elevator control to keep

cruise altitude. An effort was made to get a fuzzy logic program or applet written in

Python.

Page 5 of 17

49

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

3. The methodology for the problem was to learn basic Python programming techniques

for doing virtual UAV flight files, transforming these files if needed, and running or

modifying a Fuzzy Logic application.

The methodologies realized in most of the inquiry included Expert Systems, Neural Networks,

Fuzzy Logic, and Hybrid patterns of these and PID control systems. The AI methods preferred

for use were Fuzzy Logic, Expert System, and later a sequence of Neural / Fuzzy for adaptive

capability.

The first AI method used during the problem was Fuzzy Logic. This procedure was applied to

control the altitude carry of a modeled UAV. Two products were executed. The first result

concerned writing Python code to imitate a flight by means of calibrated steps of altitude climbs.

The second purpose concerned learning a demo copy of a commercial software for creating a

Fuzzy Control System (FuzzyTech 5.x). The output data after the first treatment was applied for

input into the demo program. The output data file from FuzzyTech was brought into an Excel

spread sheet for evaluation next to the input data.

An experimental model was needed to accept a confined application of a simple Fuzzy control

system. The desired altitude hold was applied with an additional input for acceleration control.

The acceleration control occurred for upcoming work including degrees of power required

during different mission scenarios (take-off, level flight, disturbed flight, and approach).

Fuzzy Logic Controllers have been very effectively employed in numerous engineering

products over the past few years. But this sort of control system is not extensively established

between control experts [13].

Page 6 of 17

50

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Fig. 1. Design Approach

Fuzzy Logic Technique

Currently, numerous defense organizations and police forces are employing drones and

different types of unmanned vehicles in fuzzy logic like the human being’s thinking and

interpretation process. Contrasting conventional control approach, which is usually point-to- point control, fuzzy control is a range-to-range control. Technically, the input and outputs of

the controller are the same as the traditional methods, so the input is the error in the controlled

Page 7 of 17

51

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

variable, and the output is the control amount. Nevertheless, the output of a fuzzy control is

obtained from fuzzifications of equal inputs and outputs applying the combined affiliation

purposes. A sharp input will be transferred to the various participants of the related

participation functions established on their existing value. From this point of view, the output

is built on its affiliation of the various affiliation purposes, which can be believed as a series of

inputs [13].

To execute fuzzy logic method to an actual product entails the next three steps:

1. Fuzzification – convert classical data into membership functions.

2. Inference – combine membership functions with control rules to derive the fuzzy output.

3. Defuzzification – use a method to calculate each associated output and pick up the final

output from a look-up table.

All devices can handle crispy or traditional data such as either “true” or “false”. To enable

devices to carry out hazy linguistic input such as “somehow satisfied”, the sharp input and

output ought to be transformed to linguistic variables with fuzzy parts, for example: “High”,

“Medium” and “Low”, or “Fast”, “Medium” and “Slow”. Contrasted with traditional control

methods, like a static PID, a fuzzy controller is able of adjusting itself to the dynamics of the

plant. If the transfer function of the plant varies over time, it is required to get arrangement

with a PID controller, even though fuzzy logic factors stay constant.

Fig. 2. Fuzzy controller principle in Veronte Autopilot [13]

The chosen experimental project was defined (Figure 3) but the purpose selected for the

simulator graphical user interface (GUI) FZightGearproved too complicated for the given

growth moment. This will be a key purpose for any investigation since it delivers a visual

interpretation of the UAS control impacts. The software is free delivery and very good

maintained.

Page 8 of 17

52

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Fig. 3. AI Technique Application

Hardware Improvements

Drone distribution is an increasing area due to fast high-tech developments and to curiosity

from appropriate firms in this application. The fast, consistent, and cost-effective delivery of

vigorous provisions to publics precious by main disasters is a central component of

humanitarian release. But the carriage of loads can be vulnerable to several issues such as

injured infrastructure, roadblocks, and floods. So, UAVs can play a significant role in the “last

mile” delivery. They have been used to passage small payloads over brief distances and at a high

amount. The humanitarian drone boxes may contain food, water, shelter kits or some medical

supplies [10].

Innovations in battery equipment, like hydrogen fuel cells, and the capability to carry

progressively more heavy payloads will push the UAVs delivery business in the future years.

Alongside artificial intelligence software, the drone's capacity to transfer payloads could

increase humanitarian reactions to disasters. In some missions, specialists and AI envision that

UAS will not only be important for hunting impacted people but then also for rescuing them.

Following freely recognizing people to be rescued, UAS could give up goods such as food, two- way radio communication, cell phones or medicine to the sufferers.

Page 10 of 17

54

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Lastly, the error among the moving angle and the target angle (Figure 8) is defined by next

seven association functions: Negative Big, Negative Medium, Negative Small, Zero, Positive

Small, Positive Medium, and Positive Big.Thfjgghm.

Fig. 5. Input: distance between obstacle and agent (very far is for an obstacle out of sensing

range)

Fig. 6.

Decision-Making Rule Base

For this question frame, four inputs are used for the fuzzification interface, and two outputs are

done after defuzzification. Inputs into the structure are as follows: space from the UAV to the

obstacle, angle between the UAV and the obstacle, the distance to the target, and the inaccuracy

between the recent heading angle of the UAV and the angle of the target in relation to the

inertial reference frame. The obstacle responses were used by Dong et Ref. [14] for fuzzy path

tracing and demonstrated nice outcomes for obstacle avoidance. The target keys were selected

built on the notion that there is only a negligible amount of data regarding the object; that is,

GPS coordinates. The crops for the structure were also used by Dong et Ref. [14] and based on

considerable information that employs these as the control inputs to a UAVs.

Page 11 of 17

55

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

The space to the obstacle (Figure 5) is depicted by four association functions [14]: Close,

Medium, Far, and Very Far (Out of Sensing Range). The point among the barrier and the UAV

(Figure 5) is described by six association functions: Negative Big, Negative Medium, Negative

Small, Positive Small, Positive Medium, and Positive Big. Related to the obstacle space, the

distance to the target (Figure 6) is described by three association functions: On Top, Medium,

and Far. Ultimately, the inaccuracy between the heading angle and the target angle (Figure 7)

is defined by seven association functions [14]: Negative Big, Negative Medium, Negative Small,

Zero, Positive Small, Positive Medium, and Positive Big.

Fig. 7. Input: distance between target and agent

Fig. 8. Input: angle between target and agent heading angle

With these responses and a law basis, the control input for the structure is found. That is, the

outputs of the fuzzy inference structure, the percent of the maximum velocity and the heading

angle change, are employed as the control into the system. Consequently, the outputs of the

fuzzy logic controller (Figure 4) are the percent of maximum velocity and the heading angle

vary.

Page 12 of 17

56

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

The output velocity (Figure 8) can be split into four association functions: Very Slow, Slow, Fast,

and Very Fast. The output angle change over (Figure 8) is parallel to the target angle. Hence,

there are seven association functions described by [14]: Negative Big, Negative Medium,

Negative Small, Zero, Positive Small, Positive Medium, and Positive Big.

Decision-Making Rule Base

Rules concerning the inputs and outputs for the fuzzy logic controller (Figure 4) are set up in

the form of if-then statements and are based on heuristics and human knowledge with piloting

through an environment. The rules for the fuzzy inference structure can be added up in some

simple decision-making logic. There are 40 rules for this setup, and the rules can be cracked up

into two conditions: if there is an obstacle within the sensing scale or not. As soon as the rules

were made, the association functions were set based on the space moved when matched to

optimal solutions up to the value of the solutions achieved a stage.

The primary aim of the controller, when there are no obstacles inside its sensing range, turns

out to be too proposal an immediate path to the agent’s main target. In this situation, the space

to the obstacles is set “very far” away (considerably bigger than the sensing radius). Route

design is later made by modifying the UAV’s heading angle to match that of the angle of the

target in the inertial reference frame or by running the fault among the two angles to zero. For

the reasons in which there are no obstacles in scale and the target is very far away, the UAV

leans to its highest running speed. When the vehicle goes to its object place, the vehicle corrects

its velocity to slow down and achieve goal purposes as still driving the angle error to zero. Then

again, these rules cut off from the rules used to drive a car, for example, and can be summarized

as follows [14]:

(1) If Do = Very Far and Dt = Far, then V = Very Fast.

(2) If Do = Very Far and Dt = Med., then V = Slow.

(3) If Do = Very Far and Dt = On Top, then V = Very Slow.

(4) If Do = Very Far and Θt = Neg. Big, then Θc = Neg. Big.

(5) If Do = Very Far and Θt = Neg. Med., then Θc = Neg. Med.

(6) If Do = Very Far and Θt = Neg. Small, then Θc = Neg. Small.

(7) If Do = Very Far and Θt = Zero, then Θc = Zero.

(8) If Do = Very Far and Θt = Pos. Small, then Θc = Pos. Small.

(9) If Do = Very Far and Θt = Pos. Med, then Θc = Pos. Med.

(10) If Do = Very Far and Θt = Pos. Big, then Θc = Pos. Big.

When barriers are discovered inside the detecting scope, the vehicle corrects its velocity and

moving angle handling info about obstacle gaps, obstacle angles, and target spot to escape, and

then retrieve, from the impediment. The agent should slow down and adjust the way to prevent

it. This means running the moving angle error to around 90◦. When it is obvious of this

impediment, it can remain on its path to the target. The set of rules that explain the shift in

moving speed and angle when a problem is discovered can be added up below [14]:

(11) If Do = Far and Θo = Pos. Med, then V = Very Fast and Θc = Zero.

(12) If Do = Far and Θo = Pos Small, then V = Fast and Θc = Neg. Small.

Page 14 of 17

58

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

happen at regular or big, dipped hurdle. Nevertheless, the benefit to using fuzzy logic is that it

is flexible to integrating supplementary laws to avoid these problems. In this situation, it

employs only one point as the input to compute the space to the obstacle and constantly utilizes

the nearest point to the obstacle. Expected to this, further logic must be involved.

To prevent this question, plain logic was applied and validated to redefine barriers that happen

inside the perception scale. If the vehicle correctly estimates the obstacle inside the sensing

scale, it can affect these problems. That is, if the vehicle surrounds the regions inside a certain

lowest amount like “safe” radius it can route out of the settings that affect these jots.

The mathematical interpretation is as respects: if the �! macro plan is discretized, O = {O1, O2, .

. . , On}, ∈ �! is the set of polygonal shape (as dipped as curved) barriers where EO does up the

�"barrier borders. In the discretized space, the EO barrier borders can be defined by a set of

points � = {�#, �!, ... , �$} that get up the borders.

Let (xV , yV ) imply the place of the UAV at a particular moment. Next, the set of places that are

inside the perception scale, �%, of the UAV is defined by �" = {�"#, �"!, ... , �"&}, where PI ⊂ P for

which the subsequent is correct:

+,�"&,( − �)/

!

+ ,�"&,* − �)/

!

≤ �%. (1)

For this set of obstacle spots withing the UAV’s detecting variety, the agent also redefines the

place as respects: for 1 to l, where �,� ∈ � and � ≠ �, if

+,�"&,( − �"+,(/

!

+ ,�"&,* − �"+,*/

!

≤ �,, (2)

where �, is the “secure” space from the UAV to an obstacle, but then �- = {�-#, �-!, ... , �-.}

turn into a set of goals that produces a new hurdle boundary among goals �/ and �+, where the

set of points �- is defined by

� = 01!",$21!%,$3

01!",&21!%,&3 , � = �"/,* − �"/,( ∙ �, (3)

�. = � ∙ �. + � for �. = �"/,( to �"+,(.

Though this is simple sense, it takes lower some expectations and terms. This has as slow as the

lowest “safe” radius is big plenty in contrast to the drone minimum turn radius. Thus, if the

perception limit is big, the vehicle will have plenty of time to turn to avoid the obstacle.

This reasoning stands for not simply dished obstacles, but also for convex obstacles. If a dished

obstacle is too small for the UAV to turn out of, the process would describe the entire area as an

obstacle. This would make sense since the drone would not be capable of navigate in and out of

Page 15 of 17

59

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

the barrier, and, hence, the UAV will not way do its way into the barrier in the initial area. For

convex obstacles, the vehicle would just be redefining an area inside the obstacle.

If the dished barrier is big enough for the UAV to endeavor into, the drone will be capable to

redefine the barrier and steer out, even with a slightest turn radius. This is indicated in Figure

9, where the blue color is the sensing range and red one is the obstacle [15].

Fig. 9. Drone Redefining a Large, Concave Obstacle [15]

CONCLUSION

The AI methods can be significantly increased with additional effort in this progress. Efforts can

be made to develop the work of AI methods so an evaluation can be made versus extra

traditional flying controllers. Offered here is the design and confirmation of a fuzzy logic

controller (FLC) for movement designing in real-time in a two-dimensional, unidentified

environment. As a result, the combination of code plus the ability to visually examine the

execution and evaluate the outcomes real time will require a robust, feasible, and low-cost

platform for research with AI methods which are used to UAV flight control.

The fuzzy reasoning structure was tested on numerous, static barriers and for moving objects.

The Fuzzy intelligence system was matched to an optimum methodology and a procedure built

on possible UAS fields that as well action plans in real time.

The resemblance to another path preparing techniques showed that the FLC has good quality

outcomes, and they can be gained online with any kind of environment. This was displayed to

be incredibly consistent, with only nearly a 3% error rate across all issues. The inaccuracies

that made happen were when the UAV navigated into a region that did not have an opening,

and it could not properly maneuver out. But this technique is not perfect, the proposed method

demonstrated a considerably better failure rate at about 18% overall (and 34% for complex

environments). Furthermore, the artificial potential field solution process applied considerably

additional control attempt (about 10 times) than the FLC.

It was also indicated that in most cases when the environment is relatively straightforward

(only contains polygons or “closed” obstacles) the Fuzzy inference systems (FIS) makes near- optimal results steadily. That is, the potential fields method was able to reach within 5.7% of

the perfect cause in straightforward conditions, in 10.7% for normal conditions, and in 6.5%

for challenging situations for separate ran (7.7% overall). This defeats the entire process for

the potential fields’ method, which mentioned above, at 9.7%.

Page 16 of 17

60

Transactions on Engineering and Computing Sciences (TECS) Vol 11, Issue 1, February- 2023

Services for Science and Education – United Kingdom

Although the operation is fewer consistent for obstacle improved conditions, the outcomes are

even capable to be achieved in actual period and with sufficient proximity to optimality. As

conditions get gradually denser, the ideal result will be trickier to get in a sufficient time. This

will make it impractical to depend on securing an optimum explanation still if the situation is

totally established earlier to mission.

The outcomes shown in this work showing that the proposed algorithm surpasses the potential

fields method in dependability of achievement of a result, in close-optimality, and quantity of

control attempt applied. Anything is further, it was able to do so in actual time with limited data,

which the optimum result was not.

But quantitatively it is clear that the FLC surpasses the artificial possible field result and optimal

methods, there are some qualitative benefits to the fuzzy controller. The decreased failure ratio

is a distinctive characteristic of fuzzy logic. Fuzzy Logic lets the customer to obtain a lot more

data about the situation in a more effective way. Furthermore, the existing tools permit the user

to simply create and operate FISs. This becomes it an especially effective tool that permits the

operator to look at the impacts of integrating further info with minimal attempt.

The benefits of using Fuzzy Logic for a movement planner were shown in this research: a small

number of failures, near-optimal pathways, and low control attempt. It is these benefits that

make it a nice tool for additional improvement. Certain potential clear future works that were

not investigated in this study are motion-planning in three dimensions, prevention of dynamic

obstacles, and more robust analysis.

References

[1]. Watts ,A.C., Ambrosia, V.G., Hinkley E.E., Unmanned Aircraft Systems in Remote Sensing and Scientific

Research: Classification and Considerations of Use. Remote Sensing, 2012. 4: p. 1671-1692.

[2]. Liang, Z., et al., The detection and quantification of retinopathy using digital angiograms. Medical Imaging,

IEEE Transactions on, 1994. 13(4): p. 619-626. Gonzalez, L.F., Montes, G.A., Puig, R., Johnson, S., Mengersen

L., Gaston, K.J., Unmanned Aerial Vehicles (UAVs) and Artificial Intelligence Revolutionizing Wildlife

Monitoring and Concervation, MDPI. Sensors, Vol. 16, Issue 1, Basel, Switzerland, 2016. Available online:

https://www.mdpi.com/1424-8220/16/1/97 .

[3]. Dufrene, W.R. Application of Artificial Intelligence Techniques in Uninhabited Aerial Vehicle Flight.

Proceedings. Digital Avionics Systems Conference, 2003. DASC '03. The 22nd Vol. 2. Available online:

https://www.researchgate.net/publication/4043886_Application_of_artificial_intelligence_techniques_in_

uninhabited_aerial_vehicle_flight .

[4]. Liu, X., J. Liu, J. Eker, Lee, E.A., Heterogeneous Modeling and Design of Control Systems, to appear in

Software-Enabled Control: Information Technology for Dynamical Systems, 2002, New York City: IEEE

Press.

[5]. Cannady, J., DCIS 760 Artificial Intelligence: (Class Lecture Viewgraphs), 2002, Nova Southeastern

University, Graduate School of Information Sciences. Available online: http://scis.nova.edukannad .

[6]. Handelman D., Stengel R., Perspectives on the Use of Rule-Based Control, Artificial Intelligence in Real-Time

Control, 1989. Pergamon Press, New York, p. 27-32.

Page 17 of 17

61

Zosimovych, N. (2023). Artificial Intelligence Techniques Application in UAS. Transactions on Engineering and Computing Sciences, 11(1). 45-61.

URL: http://dx.doi.org/10.14738/tecs.111.14012

[7]. Control Using an Artificial Neural Network. Proceedings. Australian Conference on Robotics and Automation,

2000. (ACRA 2000), Aug. 30 – Sept. 1, Melbourne.

[8]. Elmore, W., Control of Intelligent Unmanned Vehicles in Dynamic Environments, 2000. Available online:

http://www.asinc.codpdfs/unmannedVehiclesMay2002.pdf .

[9]. Sehgal, B., Deters, R. W., Selig, M.S., Icing Encounter Flight Simulator. AIAA 2002-0817, Reno, NV, 2002, Jan.

[10]. Oren, C., Verity, A., Artifcial Intelligence (AI) Applied to Unmanned Aerial Vehicles (UAVs) and its Impact on

Humanitarian Action, 2020, Digital Humanitarian Network.

[11]. Kranz M., Inside the “hive-mind”:how AI-powered drone “swarms” can benefit society. Information Age.

February 15, 2019. Available online: https://www.information-age.com/ai-powered-drone-swarms- 123479202/# .

[12]. Feist J., Intel flies hundreds of drones during Super Bowl Halfime with Lady Gaga, Drone Rush. February 6,

2017. Available online: http://dronerush.com/intel-shooting-star-drones-super-bowl-halfime-lady-gaga- 5984/ .

[13]. Fuzzy Logic Controller: A new control technique for drones, EMBERTION, October 8, 2021.

Sabo C., Cohen K., Fuzzy logic UAV (Unmanned Aerial Vehicles) motion planning. Advances in Fuzzy

Systems, Jan., 2012. Available online:

https://www.researchgate.net/publication/258331480_Fuzzy_Logic_UAV_Unmanned_Aerial_Vehicle_Mot

ion_Planning