Anda di halaman 1dari 10

1

DHANALAKSHMI COLLEGE OF ENGINEERING

ARTIFICIAL NEURAL NETWORKS AND FUZZY LOGIC IN UNMANNED


AIRCRAFT

SUBMITTED BY:

SANDEEP.R (E.E.E-IV YEAR)


HARI PRABHU.R (E.E.E-IV
YEAR)

EMAIL ID:

san_deep05@yahoo.com
2
hariprabu24@gmail.com

CONTACT NO:
9952055137
1

Abstract: A.BASIC BLOCK DIAGRAM OF NEURAL NETWORK:


The aim of this paper is to automate the aircraft for the purpose of
aviation and defense using Neural Network and Fuzzy Logic. STIMULI
Proponents suggest that the peration cost of Unmanned Combat Air
Craft (UCAV) is lesser than the manned aircraft. Fuzzy establishes, RESPONSE RECEPTORS NEURAL EFFECTORS
manipulates and solves numerous probabilities that exist in between NETWORK
two-element set. As the probability increases the need for
sophisticated problem solving network arises which is
comprehensively met by Neural Network. The techniques
used are fuzzy control system, obstacle spotting, fuel controlling
system, temperature and pressure controlling system, neuro-fuzzy EFFECTORS
control, navigation system. Also the inclusion of Ultrasonic
Propulsion in firing the Missiles is an added feature. Thus the need
for sending Pilots into warfare is much reduced. Unmanned
systems are the center of the next big step in military aviation.
One of the most awaited applications is the unmanned aircraft Fig. 1.BASIC BLOCK DIAGRAM OF NEURAL NETWORK
combat implemented using neural and fuzzy logic. It has many
The fundamental part of a network is called a neuron or a
unique applications today as voiced earlier. Lot of manpower can be
saved and Ultra high precision navigation can be got through this. nerve cell. The fundamental part of a network is called a
The performance of this system is being studied worldwide by some neuron or a nerve cell.
important U.S. and U.K. utilities. The wide acceptance of this is an The neuron has soma where the cell nucleus is located. Tree
added advantage that unmanned aircrafts are going to create a big like nerve fibers associated with soma are called dendrites
turn over in the entire world in the near future. which receive signals from other neurons. A single long fiber
extending from the soma is the axon which branch to strands
I.INTRODUCTION: and sub strands connecting to many other neurons at the
synapses. The dendrites serve as receptors for signals and
This paper presents the complete automation of the aircraft axon serves for
from the point of takeoff to landing and control in defense
purpose. Here we discuss about all the possible situations that transmission of the generated neural activity to other nerve
might emerge in and around the aircraft in air, and the cells. The size of a cell body or soma is approximately 10-80
methodology used by the machine through fuzzy logic and Um and the gap at synaptic junction is 200 nm wide.
neural network systems to meet the complexities. Dynamic
changes in the temperature and pressure at each and every In the state of inactivity in the interior of the neuron, the
second are controlled. It has an efficient Radar, air current protoplasm is negatively charged against the surrounding
controller, auto refueling system, trajectory optimization and neural liquid containing positive sodium (Na+) ions. The
anti-firing missiles, to have a complete aviation. Before going resulting resting potential is –70mv and the (Na+) ions are
into the paper in detail, let us have a glimpse of fuzzy logic impenetrable in the cell membrane. When signals arrive at
and neural networks. synaptic junctions they depolarize the resulting potential to –
60mv and hence the (Na+) ions penetrate into protoplasm and
II.NEURAL NETWORK: hence reduce the potential difference. Within a few milli
seconds the membrane regains its properties and also the
Due the required need for advancements in speech, vision, neuron regains its resting potential.
language, processing, etc., which was not fulfilled by the
earlier technologies and also due to the inspiration of the III.BIOLOGICAL NEURON
biological neural network (BNN) the scientist created the
Artificial neural network (ANN) which stimulates the from other neurons
operation of biological neural network (BNN).The neural synapse
network continuously receives information, perceives it and
makes approximate decisions. The receptors convert the dendrites
stimuli from the human system and convert it into electrical
impulse that conveys information to the neural network. The
effectors convert the electrical impulses into discernible
nucleus
response as system outputs.

Soma

axon
2

The soma of the neuron acts as a kind of summing


device due to the net depolarizing effect of its input signals. A network with two states of a neuron (0 or 1, and –1 or 1) is
Every neuron is estimated to receive a converging input on an called “discrete”, and the same with a continuous output is
average of about 10^4 synapses. The density of neuron in the called “analog”. If, in a discrete network at a particular
cortical tissue is about 15 x 10^4 neurons /mm^2, the total time‘t’, the state of every neuron is updated. The network is
number of neurons is approximately 10^11 and hence there is said to be synchronous. A network is feed forward, if there is
about 10^15 synaptic connections in human brain. The neural no closed chain of dependence among neural states. The same
networks store information in the strengths of the inter network is feed backward, if there is such a closed chain.
connections and perform massively parallel operations. When the output of the network depends upon the current
input, the network is static (no memory). If the output of the
IV.ARTIFICIAL NEURAL NETWORK (ANN): network depends upon past inputs or outputs, the network is
dynamic (recurrent). If the interconnection among neurons
An artificial neural network is an abstract simulation of a
change with time, the network is dynamic; it is called non-
nervous system. Due to this self organizing and adaptive
adaptive. The synaptic weight updating of the networks can be
nature the model offers potentially a new parallel processing
carried out by supervised methods or by unsupervised
model. This model could be more robust and user friendly
methods or by fixed weight association networks methods.
than the traditional approaches. ANN can be viewed as
Some of the supervised learning algorithms are the
computing elements, simulating the structure and function of
perceptions, decision-based neural network, adaptive linear
the biological neural network. These networks are expected to
element and hidden Markov analysis. Various unsupervised
solve the problems in a manner which is different from
algorithm are self-organizing feature map, competitive
conventional mapping. Neural networks are used to mimic the
learning, adaptive resonance theory and the principal
operational details of the human brain in a computer. Neural
component analysis the fixed weight networks are hamming
networks are made of artificial ‘neurons’, which is actually
net, hopfield net and the combinational optimization.
simplified versions of the natural neurons that occur in the
human brain. It is hoped that it would be possible to replicate
some of the desirable features of the human brain by V.FUZZY LOGIC:
constructing networks that consist of a large number of Fuzzy logic refers to a logical system that generalizes classical
neurons. two-valued logic for reasoning under uncertainty. In abroad
ANN are computing elements, which are based on the sense, fuzzy logic refers to all of the theories and technologies
structure and function of the biological neurons. These that employ fuzzy sets, which are classes with unsharp
networks have nodes or neurons, which are described by boundaries. The idea of fuzzy sets was born in July 1964.
difference or differential equations. The nodes are Lofti a. Zadeh is a well-respected professor in the department
interconnected layer-wise or interconnected among of electrical engineering and computer science at University
themselves. Each node in the successive layer receives the of California, Berkeley. Fuzzy logic generalizes notion of
inner product of synaptic weights with the outputs of the truth values in classical logic into a matter of degree. A
nodes in the previous layer. The inner product is called the statement in fuzzy logic, thus, can be particularly true. A truth
activation value. The activation value is passed through a non- value is a number between 0 and 1. An important goal of
linear function.When the vectors are binary or bipolar, hard fuzzy logic is to be able to make reasonable inference even
limiting non-linearity is used. When the vectors are analog, a when the condition of an implication rule is partially satisfied.
squashed function is used. Some of the squashed functions are This capability is some times referred to as approximate
sigmoid (0 to1), tan h (-1 to +1), Gaussian, logarithmic and reasoning. This is achieved in fuzzy logic by two related
exponential. techniques.
1. Representing the meaning of a fuzzy implication
X rule using a fuzzy relation
2. Obtaining an inferred conclusion by applying the
compositional rule of inference to the fuzzy
X Wij implication relation.
Fuzzy mapping rule can be represented as a fuzzy relation
X
between antecedent variables and consequent variables.
NEURAL MODEL:

Fig. 3.FUNCTION OF A NEURON


3
Neural models have found many applications in control. In the actual outputs, it creates new classification and input-
general, a network may be trained to respond to variations of output association and it generates new rules.
the input such that the desired output is maintained. If the
input is not given clear then it is not always sure that the
possibility of getting the output is correct.
CLOSED –LOOP FANN CONTROLLER

Ouy

INPUT OUTPUT
Input

Fig. 4.NEURAL MODEL


OPEN LOOP ANN CONTROLLER
Output

Fig. 6 .CLOSED-LOOP FANN CONROLLER


A.FUZZY MODEL:
Fuzzy systems are also used in the control. The input output
relationships are known in a “fuzzy way”. The rules of the VI.OPERATION OF AIRCRAFT USING NEURO FUZZY SYSTEMS:
control plant, the variables and their membership function are The most important factors that are to considered for the
known. Here no learning takes place as in neural networks. purpose of aviation are represented in the block diagram it
Based on this, the sensors of the plant provide inputs to the consists of the following factors such as.,
fuzzy system and, based on the rules, the fuzzy systems • Linear dynamics: To maintain the trajectory
generates an output that controls the plant. In this case, the path in linear direction.
behavior of the plant should be known very well so that a • Angular dynamics: Takes care of the turning
complete set of rules is used. If the set is not complete, then operations.
the plant may become unstable. • Speed controller: Controls the speed of the
air craft.
Input • Altitude controller: Controls of the height of the
flight in air.
A.NAVIGATION SYSTEM:
The navigation is the important part for any flight. The path of
the flight has to be specified in a clear form. A well-planned
trajectory for the flight is stored in a database called as
trajectory database. It consists of number of probable
trajectories it can travel. It takes preferably the shortest and
easiest path. A monitoring base on the ground monitors the
path of the trajectory.
The sophisticated high level trajectory optimization program
gives the optimum commands for steering control to maintain
its trajectory. The infra red sensor,temperature and pressure
output sensors present in the flight senses the surroundings
immediately and also regularly with a very short time period.
The possibilities of the obstacle,enemy crafts ,temperature and
pressure changes are checked and the program activates the
Fig. 5.OPEN LOOP FUZZY CONTROLLER neural network during flight. The neural network overcomes
the difficulties with the help of different controls.
B.NEURO-FUZZY CONTROL: It is from this point the real fuzzy theory blossoms in the form
The neural network combines both these worlds. They of neural and fuzzy logic in unmanned aircraft combat theory.
consider fuzzy inputs and at the same time they are capable of The trajectories are combination of lines and curves to
learning. Here the forward network is fuzzy and the feed back overcome the complexities. The horizontal and vertical
is neural network. The neural network receives the inputs and movements given by fuzzy depends on the speed controller
4
which follows the following fuzzy depends on the angle of
turns.

Fig. 7 .GENERAL BLOCK


5
Here we consider the maximum turn angle to be 180 degere,if
the turn angle is more than 180 it can be reached in the other
direction.

TABLE I

ANGLE FRACTION OF TURN (X)


0 0
45 0.25
90 0.50
135 0.75
180 STORAGE 1.00
TANK
CONTROL SYSTEM
ENGINE
The following FUEL
coding does the above – stimulated results.
SUPPLIER
Fraction (x)=(angle (x))/180
FUEL
The angles in terms of fractions obtained
SENSOR
as0,0.25,0.50,0.75,1.00 are given nominations. These
nominations are CODE
given to the code converter,which converts it
into digitalCONVERTER
form as shown below. Depending upon the angle
the speed is controlled
PROCESS to say.
X = 010 minimum
LOGIC turn angle so reduction in speed is less
PROCESS LOGIC
CONTROL SYSTEM TABLE II
ANGLE THROTTLECONDITIONS DIGITAL
CONTROLLER OUTPUT
0 DATABASENIL 001
45
VACCUMISER
MINIMUM 010
CODE CONVERTOR
90 FUZZIFIERMEDIUM 011
135 TEMPERATURE LARGE 100
180 SENSOR MAXIMUM 101
PRESSURE
SENSOR
SENSOR
Fuzzy System
So for minimum turn angle the speed is reduced slightly. Thus
following a perfect trajectory nature.

Fig. 8 .BLOCK DIAGRAM OF NAVIGATION

NEURO FUZZY
SYSTEMS
ANGULAR
DYNAMICS
LINEAR .ACCESSORIES:
DYNAMICS • Temperature and Pressure Controller: To maintain a
ALTITUDE constant temperature and pressure inside the aircraft.
CONTROLLER
SPEED
CONTROLLER

NAVIGATION
FUZZY
NETWORK
OR
NETWORK
Fuzzifier
RULES

DEF
6
• Air Currents Controller: Facilitates steady navigation
in spite of change in air currents. The data given below have a range, considering the normal

UZZI
• Fuel Control System: Helps in automation of fuel traveling parameters. The fuzzified output in the degree oh
system. hotness is given for easy understanding purpose.
• Obstacle Spotting: Helps in detecting the enemy The nominations are given and the corresponding degrees are

FIER
crew crafts and prevents from other unavoidable sent to the code converter, which provides the following.
collision. According to the nominations the conditions are given from
cold to hot the temperature limit can also be varied according
to our convenience. The code converter then produces similar
Formal
VII.TEMPERATURE AND PRESSURE CONTROL: such values.
Control
The main aim of this design of this system is to take care of TABLE IV
design and pressure at each and every second. There
the temperature
is an imbalance inside and outside the plane. When we travel TEMPERATURE CONDITION DIGITAL
higher altitude the temperature increases, pressure increases
Plant OUTPUT
and vice-versa. Being unable to handle these extreme
model
conditions the plane collapses. 30 COLD 001
Output
In our case, the temperature and pressure sensors detect any 40 COOL 010
changes in conditions provided. The output is given to the 50 GENTLE 011
fuzzier which activates the code converter. The code converter 60 NORMAL 100
does the job of defuzzier. The defuzzification is an important
step based on which the output may or may not be successful. 70 WARM 101
Several techniques are involved in this method. The three 80 HOT 110
important methods used are given as follows.,
 Maximizer : By which maximum output is
selected
 Weighted average: This averages weighted The output of the database is given to process logic. The
process logic manipulates and finds solutions. These solutions
possible outputs.
are fed into the control system. The control system then
 Centroid : Which finds output’s center of activates the vacummiser (or) the throttle controller
mass. accordingly. Thus the temperature and pressure are
maintained at the normal conditions.
Temperature pressure vaccumiser pumps air out
VIII. WORKING OF FUZZIER:
The fuzzier maintains a boundary value of hotness, say 30 to
80. All the values between this have a corresponding digital Temperature pressure throttle controller pumps air
value based on which the fuzzier works. So a code is designed
to perform the operation. IX.AIR CURRENT CONTROLLER:
Hot) (x) = 0 if temp (x) <30 The air current is also an important factor to be considered for
(temp (x) – 30)/ 50 if 30< temp(x) <80 the safe navigation of the flight. The effect is more
1 if temp(x) >80 pronounced in high velocities. It leads to fatal mishap in the
navigation.
Data’s are manipulated as per the following:
TABLE V

TABLE III Angle Safety factor Digital output


TEMPERATURE DEGREE OF 10 0.0 001
HOTNESS 15 0.2 010
30 0.0 20 0.4 011
40 0.2 25 0.6 100
50 0.4 30 0.8 101
60 0.6 35 1.0 110
70 0.8
80 1.0
7
40 0.6 Medium 100
50 0.8 High 101
60 1.0 Severe 110

C.CODING ANGLE:
Say for a plane if it turns away of 10 degree from trajectory to
60 degree the values are simulated as angle. Accordingly the
safety factors are decided. The angle is maintained such that
the aircraft does not deviate away from the trajectory.

Angle Safety factor Digital output


10 0.0 001
15 0.2 010
20 0.4 011
25 0.6 100
30 0.8 101
35 1.0 110
TABLE VII

Fig. 9.TEMPERATURE AND PRESSURE CONTROL SYSTEM

The resultant output of the angle coding and the speed adding
is fed to the direction sensor were the altitude and the speed
The two parameters to be considered are: controller are activated to face the input.
X.FUEL CONTROL SYSTEM:
 Speed with which the flight travels.
The automated machine needs fuel checking and must be kept
 The angle with which the air current strikes
constant always above a fixed limit. The auto fuel system
the flight.
handles the problem effectively. It uses the principle of weight
Let us assume that for a well-built plane if the air balance.
current speed is below 10km/h it has no effect on the
flight. On the other hand if the speed is above 80
km/h the flight collapses.

A.CODING:
F(x) = 0 if speed (x) <10
Speed ((x)-10)/80; speed (x) =10<x<80
1 if speed (x)>80

Fig. 10.AUTO FUELLING SYSTEM


B.SPEED VALUES
TABLE VI
The sensor works on the above mentioned principle. The
Speed of air Degree of Conditions Digital engine fuel supplier is rested on one arm of the balance. The
risk output other end is directly connected to the moving contact of the
10 0.0 Mild 001 variable rheostat, which has current flowing through them.
20 0.2 Gentle 010 The voltage drop registered.
30 0.4 Slow 011 Let ‘x’ be the fixed value of the voltage, which corresponds to
the cutoff weight of the engine supplier below which fuel has
8
to be supplied. The obtained value is transferred to the code personnel.Indian nuclear energy is still in a gambling
converter that gives digital equivalent of ‘x’ the valve opens state.So,we have to shift for safer sides for better tomorrow.
and supplies the required fuel when the
input is equal to the value ‘x’ the valve closes engine. Thus
the auto fuel system is done effectively.

A.OBSTACLE SPOTTING:
REFERENCES:
The infra red radiation which surrounds the flights spots the
[1].Artificial neural networks in real life applications-By Juan
obstacle. The neural system senses the exact dimension of the
Ramon Rabunal, Julian Dorrado.
obstacle. If the obstacle is in X-axis the neural system which
activates the altitude controller which in turn increases the Y- [2].Simulation and modeling:current technologies and
coordinate so that the altitude is increased If the object is in Y- applications- By asim El sheikh,asim abdel rahman el
axis the neural system activates the speed controller to sheikh,abid al ajeeli.
increase the speed and hence the velocity in the x-axis [3].www.ai-junkie.com/ann/evolved/nnt1.html
increases. Thus the speed and altitude is effectively controlled [4].www.wiley.com/WileyCDA/WileyTitle.html
by the neural systems to prevent a collision with the obstacle. [5].www.fuzzy-logic.com/fuzzy_logic_120.html
[6].www.amazon.com/Understanding-Neural-Networks-
FuzzyLogic/dp/0780311280
[7]. www.grc.nasa.gov/WWW/K-12/airplane/atmosi.html
XI.CONCLUSION:
One of the most awaited applications is the unmanned aircraft
combat implemented using neural and fuzzy logic. It has
many unique applications today as voiced earlier. Lot of
manpower can be saved and Ultra high precision navigation
can be got through this. The performance of this system is
being studied worldwide by some important U.S. and U.K.
utilities. The wide acceptance of this is an added advantage
that unmanned aircrafts are going to create a big turn over in
the entire world in the near future. Latest unmanned aircrafts
that are under constructions are…

TABLE VIII

Specification Saab Urav Saab Uav Saab Ucav


s
Length (m) 8-Mar 20-Jul 12-Jul
Span (m) 6-Feb Oct-40 8-May
Weight (Kg) 500-2000 1000-6000 1000-6000
Max. Speed MO.6-0.9 MO.5-M0.8 MO.8 plus
Altitude (ft) 46,000 65,500 46,000

These are unmanned aircrafts that are designed to serve the


purpose.

XII.RECOMMENDATIONS WE SUGGEST:
The operation and research to launch unmanned aircrafts is
taking place in an rigorous manner in U.K and U.S
agencies.India as we all know at present is facing a tough
situation.India at present has launched chandrayaan-I and we
as the citizens of India urge the government to turn its view to
develop safe and secure war which doesn’t affect any army

Anda mungkin juga menyukai