Fuzzy Interface Systems

IMAT3406 Fuzzy Logic and KBS

Abstract

This paper will be a complete report of my workings to produce

a fuzzy logic system which requires three inputs and one output.
The fuzzy logic system will be designed for automatic braking systems on modern
vehicles.

Archie Khuman
Introduction
The task that I have been set for this assignment is to design, build, test and evaluate a fuzzy logic system. The
system itself can be anything relevant to fuzzy logic systems, however, there must be a minimum of 3 input
variables, and at least 1 output variable.

I have decided to make a fuzzy logic system for vehicles, the system will be an automatic braking
system which would in theory cause the vehicle to brake automatically depending on speed and
distance of obstacles. I believe this will be great for this task as this system would have been
vigorously tested before it is available to be put on current vehicles. Currently there are some road
safe vehicles do have this system in place.

Literature review
Introduction
This literature review will be on how fuzzy logic systems are being used for modern vehicles to enhance safety
and performance, these papers go into detail on how vehicle auto breaking systems, cruise control as well as
general performance as well as some traffic congestion can be improved with fuzzy logic systems, like a fuzzy
controller. I will only be looking at a small spectrum of land vehicle; trains, cars, vans and trucks. I will not be
exploring other vehicles for this may reduce the focus of the research as each different vehicle made operates
differently and to cover a fuzzy logic-based system for all types of vehicles would be too much for a small
literature review.

The reason why fuzzy logic would be very useful for vehicles is because fuzzy logic does not use a
standard binary system of 1 and 0, as seen in figure 1. This would mean that if a car is not close to any Figure 1
obstruction then the value would return as 0, if
there is one close then it would return the value 0 1
of 1, and this would cause the car to apply full
brake pressure to the vehicle as the value as jumped. This would cause great discomfort to the passengers of
said car, also causing some issues for the cars general condition and possibly causing more traffic issues if this
example occurred in a congested road with multiple cars in.

Fuzzy logic uses probability so a way to imagine a scale would be a gradient of 1 to 0 which would have
the values 0.0 - 0.9 within it. You can see an example of this in figure 2. The reason why this is preferred is Figure 2
because instead of applying full break pressure
when the value returns as not 0.0 the breaks 0.0 0.5 1.0
would slowly apply break pressure as the value
slowly increased from 0.0 to 1.0. A good example is of a person trying to park their car, they would slowly
approach the obstructions to flag the auto break system and this would cause the returned values to slowly
climb up, as this is happening the brakes are being applied simultaneously as the value climbs causing the car
to calmly stop. No discomfort would be felt by the passengers of the vehicle as well as giving other drivers the
necessary reaction time to deal with a slowing car instead of a dead stopped car.

I have decided to research this because I want to develop a fuzzy logic system for modern day vehicles to
improve their performance as well as the safety and comfort of the passengers of said vehicle.

Page 1 of 11
Literature Research
Paper 1 Table 1
The first source I went through was one that peaked my interest as I would like to see
examples and tests for these vehicles in local environments, but to do find this would
be incredibly hard to find. To help with this I found a source which uses the UK road
laws, as Australia uses the same road laws as the United Kingdom. This is a small issue
with me as some road laws may not allow pedestrians to walk on non-zebra crossed
roads i.e. American road laws. The first paper is appropriately titled “Using FUZZY
Logic to Control Vehicles' Velocity in Traffic System.”

Within this paper the author has implemented a fuzzy controller into vehicles control
systems and began testing their acceleration and deceleration with multiple factors to
get a good estimate on how effective fuzzy logic systems are to improve the
performance of vehicle. This is a small simulation, but the results should be enough to
validate opinions on how affective fuzzy logic can be.

If you look at table 1 you will find that each factor has been listed as well as each minor factor that the main
ones are influenced by. These will be used for other tests within the paper as these would be expected factors
for the environment of the car being driven.

Figure 3 “first experiment performed for this paper. It

Figure 3
illustrates the step response of a dynamic vehicle by using
FLC. There are three purposes of this experiment:
1) make sure the desired velocity can be approached;
2) understand the impacts of inputs of controller on the
vehicle velocity;
3) test the traffic dynamic interaction between the vehicles,
especially starting from station to desired velocity.”

All vehicles used in this simulation had a fuzzy controller

implemented into them. Figure 3 shows that for three
different vehicles the results where different, in detail we
can see that at the 18 second mark vehicle 1, and 2 had a dip
in velocity but vehicle 3 did not. This may have been
influenced by the cars being in manual as this can lead to slower velocity upon gear changes, this is one
weakness for this paper, it’s very jargon in its terminology, as there are multiple figures and tables which give
evidence but are presented with pre-explained formula that could confuse some people when researching
this. I do prefer papers that have factual proven evidence when presenting the authors ideas as I can assess
the information to their ideas and say whether or not they do support those ideas or disprove them

The conclusion to this paper referees to figure 3 and how the experiment was carried out, but I believe that
the conclusion could have went into more detail about how each vehicle did compared to the other in terms of
speed, as vehicle 1 outperformed all others during that test. Though the author did reference how the vehicle
beforehand did need a fuzzy controller to reach their desired velocity and that the traffic situation for any
vehicle in practice would have varying performance.

Paper 2
The second paper I went to read was titled “Fuzzy Logic Controller on Automated Car Braking System” and it
went into detail about simulating a fuzzy logic controller or FLC in an auto break car stopping before it collides
with an obstacle. This is similar to the fist paper in which both ran a simulation of a car using the same system,
but the first paper used many cars as well as different types of simulations as the author said that they tested
the cars in traffic environments.

The paper differs from the first one by also having it compare FLC to PI and PD fuzzy logic controllers. This
paper uses the software MATLAB which is the same software I will be using so the information from this paper

Page 2 of 11
Figure 4
is incredibly useful to me and my work, she is also concerned about passenger safety like I am with my
assignment, this is what pushed me to read further into the paper as she shares mostly the same ideas as me.
The author goes through in detail with graphs given on how effective each type of fuzzy logic controller is,
making the paper non-bias as there is comparable evidence within it to back the ideas of the author.

While reading this paper I found it very educational for my assignment as the author when to put down how
the simulation will be modelled as well as the differences between both PD-FLC and PI-FLC, you can see what
she put down for the paper in figure 4. I really enjoyed this as it provided enough context for many people to
read and understand when looking at the results of her tests.

At the end of the paper the author concludes that; “PD-Fuzzy Logic Controller give best performance in braking
system compared to the other type of Fuzzy Controller. PD-FLC is able to brake the car faster than PI-FLC and
FLC. In addition, safety is guaranteed and able to avoid collision due to faster braking action” This is very well
spoken as it clearly gives a good ending explanation as well as further information on what her tests have
shown. I believe that I may use this paper when working on my project as it has enhanced my undertstanding
on how I can use MATLAB to work on my project.

Paper 3
The third paper is quite different from the rest as it takes cases into account when going into the same topic of
fuzzy logic, when the past two have been using tests and results to back the authors ideas. The paper is titled
“Different control applications on a vehicle using fuzzy logic control” and the paper does not look directly at
auto braking systems like the others, it looks at seat suspension which focuses more on passenger comfort
which is something that I am interested in. Through this review I will be looking at each case the authors use
Figure 5
for the paper and I will be comparing them to each other.

Firstly, before looking at the first case I will mention that each case is tested for output
of passenger seat displacement, this may include bouncing and lateral movement when
the car is in motion. This paper has multiple graphs showing the results for each case
and what each one means. Now the first case the authors said that “the model used
was a vehicle model with passenger seat having a controller under it”, this is shown in
their tests as the first figures of each result are of this first case and they do show that
they have the worst effect on passenger seat stabilisation. You can see an example of
how this is presented in figure 5.

As you can already see from figure 5, the second case has more influence on passenger
seat stability, this is because the second case uses a “passive passenger seat but
suspensions having controllers parallel to them” this is explained to be better than the
first case as the passenger is controlling the seat which makes it easier for the passenger
to manage how well their seat is for comfort. The third and final case used in this paper is one which
incorporates both of the previous cases into one, the author says it is; “with passenger seat having a controller
under it together with suspensions having controllers parallel to them, respectively”. This case seemed to have
the best results out of the three, which means that for passenger comfort the way to improve that is to have
both passenger control and suspensions under the seat.

Page 3 of 11
This paper is presented differently from the last two as it has multiple methods for testing which is similar to
the second paper, but this one uses three instead of two and the main point of this paper is quite different to
the second and first one as this one is mainly about passenger safety and comfort. This was the main reason
why I chose to review this paper as it has more focused ideas on an aspect that I would like to include in my
assignment.

Paper 4
The fourth paper I looked at was titled “A review of fuzzy methods in automotive engineering applications”
and it was a review of all the information from fuzzy system tests within a multitude of vehicles and whether
or not the evidence backs up the argument of; does fuzzy logic systems help improve any aspect of driving. i.e.
performance of the car, stability of the vehicle while it is in motion, and what types of vehicles get the most
benefit from fuzzy logic systems.

I believe it is not only necessary to have another opinion on this topic, but to have a well-informed
Figure 6
opinion. This paper goes through almost all of the past
experiments of fuzzy logic on vehicle systems and
provides an idea on what the results are to the
experiments, as well as research on the manufacturing
of vehicles with fuzzy logic systems implemented
within the vehicle systems, an example of one is
Figure 6 “displays statistics of publications related to
industrial or ready-to-installation specimens of road
vehicle systems with fuzzy controllers or fuzzy models
embedded into the processing unit” This shows that
there have been an invonsistant but varying amount
of different vehicles with fuzzy controllers within the
systems.

To have someone already go through the research that I myself will have to go through is very useful as it
condenses some of the information into helpful sizes for learning and reviewing.

This paper is completely different to the other papers I have reviewed as the past papers have their authors
going through their own experiments with fuzzy controllers and systems whereas this papers author has done
no experiments for their paper but they have been through many other tests and gone through the findings to
come to a conclusion that it is advantageous to implement the fuzzy controllers onto vehicles to assist them
with auto breaking as well as passenger comfort.

Paper 5
The fifth and final paper I went to look at was titled “Fuzzy Logic Approach to Autonomous Car Parking Using
MATLAB”. I went to look at this paper as it offers the most direct help when it comes to this topic as it uses the
software that I have to use, having a very similar topic involving fuzzy logic systems and auto systems on
vehicles, as well as going into detail on how the experiments were carried out within this paper.

This paper is quite similar to the second one I went through as they both use the same software for the
experiments used for each paper. I do believe that it is very useful to possess useful information on this topic
as well as having information that was made with the same format that I myself will have to try to create.

The paper goes into how they set up the experiment as the author provides screenshots of the software while
they were making the experiments, I will be using these as inspiration for my work as I believe I may need help
for some of this assignment.

Page 4 of 11
The authors conclusion is quite educated as it goes into both the negatives and positives of their findings, this
differs from most of the papers I previously covered as they go into detail on the positives of the author’s
experiments as well as reiterating what they have previously said in their papers. I find this quite useful as the
authors have thought about what their tests came back with and they are acting upon it by calling out what
went wrong as well as giving examples of it; “But, also, any extraordinary event that involves with any of the
procedures could make this system prone to failure, like a long distance to the sidewalk, a long or ample car,
ample obstacles, etc.”

Conclusion
Coming to the end of this I believe that the papers I looked at helped me with deciding on how I want to work
on this assignment, each paper has presented useful methods for making an auto braking system with fuzzy
logic as well as how much of a positive affect this may have on actual live vehicles.

I would have to say that the most useful papers I reviewed would have to be the second and fifth paper as they
put forward my issue first when compared to the other three papers, though they each approach the challenge
at a different angle which in turn shows more positives for the use of fuzzy controllers/systems within current
day vehicles.

Though it must be said that paper four goes into a lot of depth that I myself will not have to go into. Paper two
follows this issue but due to the use of MATLAB within the paper I must conclude that it has a higher priority
on my list of papers which I may use as I progress with this assignment.

The system
The system that I ended up making was one that can be implemented as an auto braking system or ABS. To
make this work within MATLAB and using fuzzy logic I had to go and evaluate what inputs would be needed for
the final output of the ABS which is ‘Applied brake pressure’.

Using the information, I gathered from the papers within the literature review I decided that the necessary
inputs for the ABS would be the distance of the vehicle to any objects that would be in the way. A way this
information could be read on a real vehicle would be from the sensors that are placed on the front and back
bumper as well as cameras on the front and back of the car. With these devices on vehicles it should be quite
easy to know how close and far objects are around the car.

Another input for the ABS would be the current speed of the vehicle as this is vital information when braking is
concerned as different speeds lead to different brake distances, this is how far the car would move when the
brakes are applied. This would be easy to read for vehicles as they have to display the current speed of the
vehicle to the driver, so the information can be given to the systems within vehicles which can allow the
vehicle to slow itself down when it reaches certain speeds.
Figure 7
The third and final input that I went with was ‘current brake
pressure’ I believe that this is needed as the driver or user of
the vehicle may have already applied the brakes to the
vehicle, therefore I wanted to not have the ABS and the user
stack brake pressure for the vehicle as this could cause the
vehicle to stop abruptly and this could cause great discomfort
to the user as well as cause some hazards when trying to park
or stop on a busy road.

These three inputs will help to give the final input of ‘Applied
brake pressure’. You can see these inputs and output in figure
7.

Page 5 of 11
Input variable: speed
Figure 8
The first input is speed and this variable need three
membership functions, as there are mainly three speeds
that I have identified; slow, intermediate and fast.

These speeds are not static, so I used the gaussmf for the
type, this is because there can be a gradual change to the
speed of the vehicle. I have had this idea in mind as I have
made each membership function intersect around the half
way mark on each other so that when a vehicle goes from
fast to slow, they will gradually go down into intermediate
and then they go into slow speeds.

You can see these changes in figure 8. I have also made the
range of the input go from zero to ten. I decided to do this
because I wanted to have a wide range for speeds which
can allow me to make the membership functions to cross into each other.

Input variable: distance

The second input I used was distance, this input is absolutely necessary when thinking of inputs for an ABS.
Much like the first input there are three membership functions needed, they are; close, medium and far. Figure 9
These membership functions are different though as I
made these become a trapmf type as I believe that
distance can be kept constant and that means that there
needs to be a point on each membership function that has
a constant value. I have made this for all three of my
membership functions, but each one is a bit different from
the other as the distance maximum value for each distance
is different. I have kept the range for this variable at zero
to ten, this is just like the first variable as I want to keep
the range consistent between all inputs as this will make it
easier for the data to be given to the output as well as
being read. You can see the values I made for this variable
in figure 9.

I made the values for ‘close’ and ‘medium’ not intersect as

I believe that these two don’t share similar values when it comes to distance, as well as making it clear on what
constitutes as close and medium distances, as there may not be enough time for the vehicle to stop and not
collide with the object that is close to the vehicle. This is not the same for ‘medium’ and ‘far’ as there can be
some discretion on how far the object may be as the user and the system has enough time to react and apply
the brakes on the vehicle.

Input variable: current brake pressure

The third and final input variable I used was ‘current brake pressure’, it took me some time to think of a third
input for this assignment as the last two input variables are a necessity when making an ABS, but when
thinking more about how this may affect the user while the vehicle is in motion it came to me that the user has
an input for the vehicles brake pressure. This meant that I would have to make sure that bots the ABS and the
users braking does not stack and cause the brakes to wear down as well as causing discomfort for the user.

Page 6 of 11
When it came to the membership functions for this input, I decided to only add two, this is because when it Figure 10
comes to brake pressure, I see that there are only two different
states, these are; small and large. These two membership
functions can gradually change from one to the other, that is why I
made it a gaussmf type like the first input variable. This meant that
the system can gradually go from a small brake pressure to a large
one with relative ease and smoothness, meaning that the vehicle
will not go to a complete stop and can slowly decrease its speed
allowing for the most comfortable stop to the user. You can see
this input variable from figure 10.

Input variable: applied brake pressure

The output variable for the ABS would be called ‘applied brake
pressure’, this output was done similarly to the third input as they
both are for the brake pressure of the vehicle. This means that
Figure 11
the output has two membership variables which are called;
‘small’ and ‘large’. As I explained before there are only really two
membership functions as they gradually change into one
another. You can see that they intersect around the half way
mark on figure 11.

The output is different to the third input via range as I believe

that the more accurate the final output is, the more improved
the braking performance of the vehicle will be and that is a large
factor for this task as I want to include user comfort as much as
possible, so that if this does get implemented to modern
vehicles it can pass regular safety checks as this would need to
cause little to no discomfort to the user.

Rules
The rules for my fuzzy system where set
Figure 12
with and statements as the majority of the
inputs would be needed together to get
the correct final output, you can see from
figure 12 that there are quite a lot of rules
to be set, 16 to be accurate.

This is because some of them are of the

first two variables and not the third one as
I did keep the user in mind. This is needed
because the user may not have applied the
brakes already therefore a new rule would
need to be created in which the third
variables value is none.

When the user has input a small amount of

brake pressure the system should put an
equal amount of brake pressure given the
distance and speed to the object in the
way, on the opposite side of that rule, the
system may have to place a large amount of pressure to stop the vehicle faster, if the distance is ‘close’ and
the speed is either ‘medium’ or ‘fast’.

Page 7 of 11
Experimental Design and Evaluation Figure 13
When it came to the testing the performance of my fuzzy system,
I first began to test each of the rules that I set for my ABS.

You can see the results of this test from figure 13 and figure 14.
On these figures you can see a small red line, these lines are the
inputs for each of the variables. You can see that when it has no
affect for the output or the inputs the charts are not filled in as
well as having no lines for some inputs that have a value of none
for their respected input during these tests.

A good example of how well my ABS would work is to look at rule

6 within figure 13. You can see that the line is on the right side for
these inputs meaning that they are large inputs. This means that
for the first variable the vehicle is moving at a fast speed, the Figure 14
second variable which is distance is set to ‘medium’, this is because
that membership function has the most value for the given input
whereas the ‘far’ membership function has less. The third and final
variable is ‘current brake pressure’ is set to zero or none, meaning
that the input would be ‘small’ if not none. This leads to the output
which shows that a large amount of brake pressure is applied, but
the applied brake pressure is not one hundred percent large as you
can see that only half of the chart is coloured in, so we can assume
that a reasonably large amount of pressure is gradually applied as
these inputs are given. This was the main reason why I decided to
make the output type a gaussmf to make sure that the user has as
much comfort as possible when the vehicle is stopping.
Figure 15
The tests that I conducted do not only contain these charts, I also went
to see the results in many other formats to make it easier for reading,
you can see in figure 15. This is a 3D graph which shows the main two
variables that have to be constant to produce an output. You can see
that this graph shows all the possible outputs with all the possible
inputs from the two input variable. A good example of this is when the
distance is low, and the speed is high the applied break pressure is at its
maximum. This scenario would most likely not happen as the user
would be able to manage their speed when they approach an object
that is in their way or when they are parking said vehicle.

This 3D graph is useful to see how my rules will affect the output, but
because this does not include the third input, I cannot use this as my final result for my tests. This is because I
have set rules in which all three variables are needed for the output and this graph does not show the third
variable and how it impacts the output. This is why I have used the results from figure 13 and figure 14 for my
testing, as these figures contain the possible inputs and outputs for each of my rules that I have set making
them very reliable and useful information when testing my fuzzy systems.

If this system was used for modern ABS on vehicles, then I would refer to the two figures for the testing as
well as possibly test them more if not change the range on the inputs to get a more accurate output.

Page 8 of 11
Critical reflection
When it comes to how I performed for this assignment and the quality of the work, I would say that they are
both good, but I do have some changes in my mind that I would like to do if I was set the same task.

I would ideally like to compare my work to the work of the papers, these improvements would be similar in
nature to paper 2 as I believe that going through the systems and testing it with multiple vehicles as well as
different types of fuzzy systems would create a safer ABS for the user as well as bring back better results so
that I may be able to compare them to each other. I was only able to do this for my testing because I had made
so many rules for my system to follow allowing me to look at the possible outputs from all angles.

Another part of this assignment that I feel that I could improve on was making more inputs for this fuzzy
system as this would make it much easier to get a greatly accurate result. I believe the three inputs I made for
my system were quite good as they take everything about the vehicle and the user into consideration when it
came to completing this assignment. Though I will say that given more time I could maybe think of more inputs
for my system so that it affects the users comfort a bit more. A good example of how I could do this would be
with paper 3 in which the authors went into detail about how they managed seat stability between tests. This
is just a thought on how I could improve my work as I could think of ways to manage user comfort, this could
possibly lead to me making a second output for this work.

The testing phase of my assignment was completed in an exceptional way as I was able to go and change the
possible inputs for my variables which would lead to vastly different outputs. This was made even better by
having all of my rules present for the testing as I could change the values for all of my rules and see what the
output for each one could be. I will say that the 3D graph was a good idea though due to how many inputs I
needed for this assignment I couldn’t reliably use that for solid testing evidence, though some of my rules did
state that the third inputs value would be zero. This meant that I could use the graph for my testing evidence
though I could not rely on it completely hence why I used the first method as it covered all the points that I
made for this fuzzy system.

Conclusion
If I were to complete this assignment again with more time, I believe I could improve upon the work I already
had, for instance I would like to spend more time comparing my work to the papers I reviewed as well as going
to read more papers for my literature review.

Looking back to my literature review, I believe that it was handled quite well, as I went above the necessary
page recommendation. This is due to my papers that I read and reviewed, I needed decent size of reviews as
well as multiple papers that cover different parts of the topic I covered. This allowed me to have ideas of how I
want the information to be read as well as given. A good example is of the ‘brake pressure’ variables. Paper 3
gave me the ideas of trying to make my system safe as well as comfortable for the user in the vehicle.

Thinking about how I handled the development side of this assignment, I think that it can be improved some
what as I said before, if I had more time, I would go into more detail on what I want my fuzzy system to be as
well as trying to broaden out on what my fuzzy system can do. You can see this already in my work as I wanted
to cover the comfort of the user a great deal for my system. The literature review helped me think of ways to
cover this as well. If I spent more time looking at other papers as well as comparing them and the authors test,
I know that I can make a great fuzzy system with many inputs as well as possibly having multiple outputs for
my fuzzy system.

Spending more time on this assignment could also improve my testing as I could go into more detail on how
my work compares to the work of others who have tried the same task as me. I have already covered this
some bit so I will say that with my thoughts on how I could improvement my system, the testing would also
benefit from having a better system as there would be more data to test as well as reflect on.

Page 9 of 11
Bibliography
 (Table 1, Figure 3) AlGhamdi, A. (2013). Using FUZZY Logic to Control Vehicles' Velocity in Traffic System. 1st
ed. [ebook] Gold Coast, Australia: The International Federation of Automatic Control, pp.54 - 59. Available
at: https://ac.els-cdn.com/S1474667015349090/1-s2.0-S1474667015349090-main.pdf?_tid=55f95757-
ef93-4e0a-aa2d-996bce693123&acdnat=1542558786_d55153d40ce73a06b3dfb17c146bf579 [Accessed 17
Nov. 2018].
 (Figure 4) Mamat, M. and Ghani, N. (2009). Fuzzy Logic Controller on Automated Car Braking System. 4th
ed. [ebook] Christchurch, New Zealand: IEEE International Conference on Control and Automation, pp.2371
- 2375. Available at: https://ieeexplore.ieee.org/document/5410301 [Accessed 17 Nov. 2018].
 (Figure 5) YAGIZ, N., SAKMAN, L. and GUCLU, R. (2008). Different control applications on a vehicle using
fuzzy logic control. 33rd ed. [ebook] N/A, India: N/A, pp.15 - 24. Available at:
https://www.ias.ac.in/article/fulltext/sadh/033/01/0015-0025 [Accessed 18 Nov. 2018].
 (Figure 6) Ivanov, V. (2015). A review of fuzzy methods in automotive engineering applications. 29th ed.
[ebook] N/A: Published online at SpringerLink.com, pp.1 - 10. Available at:
https://link.springer.com/content/pdf/10.1007%2Fs12544-015-0179-z.pdf [Accessed 20 Nov. 2018].
 Ángel, M., Gutiérrez, H., Palomares, R. and Flores, C. (2018). Fuzzy Logic Approach to Autonomous Car
Parking Using MATLAB. 15th ed. [ebook] Puebla: CONIELECOMP, pp.1 - 6. Available at:
https://www.researchgate.net/publication/4162403_Fuzzy_Logic_Approach_to_Autonomous_Car_Parking
_Using_MATLAB [Accessed 19 Nov. 2018].

Page 10 of 11