Fuzzy Alarm System based on Human-centered Approach

Elena Magán

, Agapito Ledezma

, Paz Sesmero

and Araceli Sanchis

Computer Science and Engineering Deparment, Universidad Carlos III de Madrid, Av. de la Universidad 30,

Leganés (Madrid), Spain

Keywords: ADAS, Advanced Driver Assistance System, Alarm System, Fuzzy Logic, Multi-agent Systems, Ontology.

Abstract: This paper presents an Advanced Driver Assistance System (ADAS), based on a fuzzy logic decision support

system and developed by using a multi-agent system. The ADAS is designed so that it can detect dangerous

situations on urban environments and alert the driver about them if necessary. For that, it collects data from

the car, the car’s surroundings and the driver, and represents the information as an OWL ontology. Then, a

fuzzy logic inference system uses this information to evaluate whether there is danger or not. The system can

detect 9 dangerous situations by using a repository of 14 fuzzy rules, based on a previous work and expanded

on this one. Although with limitations, the results show that the ADAS can alert the driver when the driver is

in a dangerous situation.

1 INTRODUCTION

According to a status report launched by the World

Health Organization (WHO) in 2018, road traffic

crashes are one of the leading causes of death in the

world, being also the leading killer of people aged 5-

29 years (WHO, 2018). Moreover, distractions are

one of the main causes of road traffic accidents –

recent studies estimate that almost 70% of crashes are

caused by driver's distractions (Dingus et al., 2016).

For these reasons, the development of Advanced

Driving Assistance Systems (ADAS) can make a

huge impact on the issue of preventing road traffic

accidents. These systems are active components

installed on vehicles and are designed to assist the

driver continuously to prevent dangerous situations

(Bengler et al., 2014), so they could help minimize

the consequences of human error and, thus, to reduce

the number of road accidents.

This paper presents an ADAS alarm system based

on fuzzy logic that detects potentially dangerous

situations and acts as a co-driver, warning the real

driver by using visual and sounding stimuli. This work

continues the ADAS developed in (Zamora, Sipele,

Ledezma Espino and Sanchis de Miguel, 2017), where

https://orcid.org/0000-0001-9797-6146

https://orcid.org/0000-0002-0041-6829

https://orcid.org/0000-0001-9473-6809

https://orcid.org/0000-0002-1429-4092

the authors used classical logic to build the decision-

making system. In this work, this system is expanded

both by using fuzzy logic and by increasing the number

of dangerous situations detected.

Whereas classical logic represents information in

a binary way –that is, a clause can either be true or be

false–, fuzzy logic works with statements that can be

partially true or false (Zadeh, 1988). With fuzzy logic,

one can handle imprecise or rough information,

which, in relation to the development of ADAS, is

extremely convenient to determine potentially

dangerous situations. Moreover, since fuzzy logic’s

output is also a grade of truth, the intensity of the

alarm can be variable according to the activation

value of the rules that detect the dangerous situations.

The development environment of the ADAS

consists of a driving simulator based on the STISIM

Drive software (Intelligent Systems Technology,

2019). With this simulator, it is possible to reproduce

realistically the driving environment of a vehicle in

real-time.

The driving environment interacts with a multi-

agent system, where there are multiple intelligent

agents with different functions. Some agents are in

charge of the data collection, their mission being

448

Magán, E., Ledezma, A., Sesmero, P. and Sanchis, A.

Fuzzy Alarm System based on Human-centered Approach.

DOI: 10.5220/0009348704480455

In Proceedings of the 6th International Conference on Vehicle Technology and Intelligent Transport Systems (VEHITS 2020), pages 448-455

ISBN: 978-989-758-419-0

obtaining information both from the simulation –

about the cars, pedestrians, light traffics, etc.– and

from the driving environment –such as the steering

wheel angle–, and there are also agents that receive

and process that data, so that they can decide if it is

necessary to alert the driver.

The paper is organized as follows: Section 2

provides an overview of the background and related

work of ADAS, fuzzy logic and driving simulators.

The design of the proposed ADAS is described on

Section 3. Section 4 explains the experimental

settings and the results obtained, Finally, Section 5

presents the conclusions and proposes some related

future works.

2 BACKGROUND AND RELATED

WORK

Since road safety is one important issue in today’s

society, many ADAS –active components installed on

vehicles that assist the driver in dangerous situations–

are being investigated and developed nowadays, both

by business and academic researches (Pérez,

Gonzalez Bautista and Milanes, 2015).

In fact, there are some ADAS that are already

being commercialized. On one hand, many car brands

develop their own ADAS, like Ford’s Co-Pilot360™

Technology (Ford Motor Company, 2019) or Volvo’s

IntelliSafe Technology (Volvo Car Corporation,

2019). On the other hand, there are also companies

that sell ADAS systems that can be retrofitted on any

vehicle, as in the case of Mobileye Systems

(Mobileye, 2019). These products usually include

multiple systems to avoid risks while driving, like

lane departure warnings, collision avoidance or

pedestrian protection, among others.

However, these companies do not share their

researches nor the complete results obtained by their

products, so in this case it is more relevant to evaluate

the more recent academic works about ADAS.

On this matter, an interesting research is the

NAVIEYES project (Duguleana, Florin and Gheorghe,

2015). NAVIEYES is a smartphone-based ADAS that,

by using the dual camera of a modern smartphone, is

able to monitor both the outside and the inside of a

vehicle, and can alert the driver if necessary. Although

this is a practical and cheap approach, it requires a

calibration process beforehand that could discourage

the driver from using it regularly.

However, there are also extensive research about

vehicle-integrated ADAS. Several of them focus on the

development of individual ADAS systems, prepared to

detect exclusively one kind of dangerous situation.

Among these works, there are systems to supervise the

vehicle safety distance (Attia, Ismail and Ali, 2016), to

protect pedestrians from danger (Sanatkumar, Gandhe

and Dhulekar, 2015), or even to detect speed bumps

(Wilson, Babu and Tharumar, 2015).

This paper is based on a previous work, where a

rule-based system is developed to detect and prevent

multiple dangerous situations (Zamora et al., 2017).

To continue with the research, the system has been

modified to use fuzzy logic on the decision-making

system, and new dangerous situations have been

studied. As a starting point, some tests have been

performed on the previous system, which revealed

some limitations. The most relevant one was induced

by an alarm that detected that a parked car was going

to join the road, since this alarm generated false

positives continuously whenever there were vehicles

parked on the right. This wouldn’t be useful on an

urban environment, where there are usually parked

cars, so it has been decided to replace this alarm with

another one that can detect similar dangerous

situations without those false positives.

In this work, it has been decided not to use the

pedal activity of the vehicle as an input of the ADAS,

since preliminary tests revealed that these values are

driver dependent and are not easily generalized. This

is because the driver’s behavior determines both the

pedal’s usage and the perception-reaction time (Lee

and Yeo, 2016), which would involve a more

complex approach that it has been decided not to deal

with in this work.

3 SYSTEM DESCRIPTION

This section describes the main elements of the

ADAS system: the ontology that gathers the

information about the environment, the dangerous

situations the ADAS must detect, the fuzzy inference

system that evaluates if the driver is involved on one

of those situations, and the Human-Computer

Interface that warns the driver if necessary.

3.1 Ontology Data

With the information retrieved by simulated sensors

(LiDAR sensor, frontal/rear cameras), and the data

collected from the driver’s vehicle, the ADAS has all

the necessary information for the decision-making

system. This information is represented on an

ontology, that is based on the previous work from

(Zamora et al., 2017) and has been extended. Figure

1 shows the ontology used for this work.

Fuzzy Alarm System based on Human-centered Approach

449

Figure 1: Ontology diagram.

The main changes made to the previous ontology

are: extracting numeric values of attributes previously

categorized, and adding some information about the

surrounding vehicles –such as their speed, trajectory,

and angle with respect to the driver’s vehicle.

This ontology represents the information of the

whole simulation environment, but not every piece of

data is used on the decision-making system. The

relevant classes and attributes for this project, with

their possible values, are described below.

 Class MyCar: Data related to the vehicle of

the driver, like the steering wheel angle or the

level of stepping of the pedals.

Relevant Attributes: realSteeringWheelAngle

([-450,450]), realSpeed ([0,200]), realClutch

([0,1]), realBrake ([0,1]), realThrottle ([0,1]),

gear (-1, 1, 2, 3, 4, 5, 6).

 Class CarContext: Information about the

surroundings of the driver. That includes both

the vehicles and the pedestrians detected.

 Relevant Attributes: hasCar<zone> (instance

of class Car that represents the car on the

<zone> position), hasPedestrian (0 to 5

instances of class Pedestrian that represent the

closest pedestrians detected).

 Class Car: Data that defines a vehicle of

CarContext, like the distance to the driver, the

speed, etc.

Relevant Attributes: state (moving, stopped),

realDistance ([0, ∞ ]), angle ([-180,180]),

realSpeed ([0,200]), trajectory (north, south,

east, west).

 Class Pedestrian: Information about a

detected pedestrian: the distance to the driver,

the trajectory it follows...

Relevant Attributes: angle ([-180,180]), distance

([0, ∞]), trajectory (north, south, east, west).

3.2 Dangerous Situations Detected

With this information, the dangerous situations that

this ADAS can detect are:

 Situation 1. Risk of Frontal Collision. While

driving normally, there is a vehicle in front of

the car and the distance to said vehicle is too

short, so if the front car stops suddenly it could

cause a collision.

 Situation 2. Risk of Running over (Frontal).

While driving normally, a pedestrian crosses

the road in front of the driver, but the distance

to the vehicle is too short and if the driver

doesn’t stop the car it will cause an accident.

 Situation 3. Risk of Running over (Frontal,

Not Visualized). While driving normally, a

pedestrian is crossing the road in front of the

driver, but there are vehicles obstructing the

driver’s vision and the pedestrian is not visible.

 Situation 4. Risk of Rear Collision (Rear

Vehicle Approaching). While driving

normally, there is a vehicle on the rear of the

car and the distance to said vehicle is too short,

so if the driver stops suddenly it could cause a

collision.

 Situation 5. Risk of Rear Collision (Reverse).

While driving on reverse, there is a vehicle on

the rear of the car and the distance to said

vehicle is too short, so if the driver doesn’t stop

the car it could cause a collision.

 Situation 6. Risk of Lateral Collision (Turn).

While driving normally, the driver tries to turn

the car, but there is already a vehicle on that

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

450

position, so if the driver turns that way it could

cause a collision.

 Situation 7. Risk of Running over (Rear).

While driving on reverse, a pedestrian crosses

the road behind the car, but the distance to the

vehicle is too short and if the driver doesn’t

stop the car it will cause an accident.

 Situation 8. Risk of Lateral Collision

(Intersection). While arriving at an intersection,

there is a car approaching from any of the sides

of the intersection, so if the driver doesn’t stop

the car it could cause a collision.

 Situation 9. Risk of Overtaking on the Right

Lane. While driving normally, a vehicle on the

right lane is traveling faster than the car,

overtaking irresponsibly. If the driver doesn’t

notice and tries to change lanes, or if the other

vehicle tries to change lanes while being too

close to the driver, it could cause an accident.

3.3 Fuzzy Inference System

Since the decision-making system of the ADAS is

based on fuzzy logic, fuzzy rules have been defined

to detect the potentially dangerous situation. For this

project, the decision-making system is designed as a

Mamdani fuzzy inference system (Mamdani and

Assilian, 1975). To design a fuzzy logic system of this

kind, it’s necessary to define the following aspects:

both input and output fuzzy sets, and fuzzy rules.

3.3.1 Input Fuzzy Sets

Input fuzzy sets represent the values that the variables

can take. For example, the “speed” variable could be

classified with 3 fuzzy sets: low, medium, or high.

The fuzzy system would receive a numeric value for

speed and assert the grade of truth of that variable for

each fuzzy set.

To be able to do that, it is necessary that every

fuzzy set is defined by a function µA(x) –membership

function of the fuzzy set A for an input value of x–,

that will assign a value in the 0 to 1 range depending

on the input x. 0 would mean that it is completely

false that x belongs to the fuzzy set A, while 1

represents that it is completely true.

To describe the variables and their possible fuzzy

sets, each fuzzy set will be designed as a continuous

function, represented by ordered pairs that must be

connected linearly. To understand this, an ordered

pair of (60,1) represents that the variable input of 60

has a grade of truth of 1. Additionally, a fuzzy set that

is represented by just one ordered pair is known as a

singleton.

For this ADAS, fuzzy sets have been designed to

deal with the information extracted from the

ontology: distance to the vehicles, speed, etc. The

variables and their corresponding fuzzy sets –as

defined by their ordered pairs– are:

a) Pedestrians position: “angle” variable:

- behind_right: (125,0) (140,1) (165,1)

(180,0)

- behind: (-180,1) (-175,1) (-165,0) (165,0)

(175,1) (180,1)

- behind_left: (-180,0) (-165,1) (-140,1)

(-125,0)

- front_left: (-55,0) (-40,1) (-15,1) (0,0)

- front: (-15,0) (-5,1) (5,1) (15,0)

- front_right: (0,0) (15,1) (40,1) (55,0)

b) Surrounding cars position: “carPosition”

variable:

- behind_right: (125,0) (140,1) (160,1)

(175,0)

- behind: (-180,1) (-175,1) (-170,0) (170,0)

(175,1) (180,1)

- behind_left: (-175,0) (-160,1) (-140,1)

(-125,0)

- left: (-135,0) (-115,1) (-65,1) (-45,0)

- front_left: (-55,0) (-40,1) (-20,1) (-5,0)

- front: (-10,0) (-5,1) (5,1) (10,0)

- front_right: (5,0) (20,1) (40,1) (55,0)

- right: (45,0) (65,1) (115,1) (135,0)

c) Surrounding cars state: “carState” variable

- moving: (1,1)

d) Distance to pedestrians and cars: “distance”,

“carDistance” and “carReactionTime”

- very_close: (0,1) (1,1) (1.5,0)

- close: (1,0) (1.5,1) (3,1) (3.5,0)

- normal: (3,0) (3.5,1) (4.5,1) (5,0)

- far: (4.5,0) (5,1) (20,1)

e) Surrounding cars speed: “carSpeed” variable

- high: (-100,1) (-15,1) (-10,0)

- medium: (-15,0) (0,1) (15,0)

- low: (10,0) (15,1) (100,1)

f) Driver’s steering wheel angle:

“steeringWheelAngle” variable

- left: (-450,1) (-90,1) (-40,0)

- center_left: (-45,0) (-40,1) (-20,1) (-15,0)

- center: (-20,0) (0,1) (20,0)

- center_right: (15,0) (20,1) (40,1) (45,0)

- right: (40,0) (90,1) (450,1)

g) Pedestrians and cars trajectory: “trajectory”

and “carTrajectory” variables

- north: (1,1)

- east: (3,1)

Fuzzy Alarm System based on Human-centered Approach

451

- south: (5,1)

- west: (7,1)

h) Variables that represent if there is a car closer

than a pedestrian at a certain position:

“distance_center_right_less_eq_ped” and

“distance_center_left_less_eq_ped”

variables

- no: (0,1)

- yes: (1,1)

i) Driver’s gear: “gear” variable

- reverse: (-1,1)

- first: (1,1)

- second: (2,1)

- third: (3,1)

- fourth: (4,1)

- fifth: (5,1)

- sixth: (6,1)

3.3.2 Output Fuzzy Sets

Output fuzzy sets represent the results that the fuzzy

system provides. The outputs are defined as fuzzy sets

too, so that the inference system can provide which is

the grade of truth of the output.

For this ADAS, an output has been established for

every possible rule that can be triggered. All of these

outputs are defined by linear growth functions from 0

to 1, where the defuzzification method of Center of

Gravity is applied. That is, the only fuzzy set of output

variables will be defined as:

- yes: (0,0) (1,1)

In addition, a threshold of 0.5 has been defined –

that is, if the activation value of a rule is lower than

0.5, the ADAS will not alert the driver of that danger.

3.3.3 Rules

Fuzzy rules combine both the input and the output

fuzzy sets to detect the dangerous situations studied.

Some of these situations are defined by two rules, one

with the conditions for the right and another for the

left. In these cases, the conditions for the right will be

used as an example.

a) Situation 1: Risk of Frontal Collision.

IF carPosition IS front AND carDistance

IS (close OR very_close) AND gear IS NOT

reverse

THEN alarm1 IS yes;

b) Situation 2: Risk of Running over (2 Rules).

IF trajectory IS west AND angle IS (front

OR front_right) AND distance IS (close

OR very_close) AND gear IS NOT reverse

THEN alarm2_1 IS yes;

c) Situation 3: Risk of Running Over (Frontal,

Not Visualized, 2 Rules).

IF carPosition IS front_right AND

carDistance IS (close OR very_close OR

normal) AND carState IS NOT moving

AND trajectory IS west AND angle IS

(front_right OR front) AND distance IS

(normal OR close OR very_close) AND

distance_center_right_less_eq_ped IS

yes AND gear IS NOT reverse

THEN alarm3_1 IS yes;

d) Situation 4: Risk of Rear Collision (Rear

Vehicle Approaching).

IF carPosition IS behind AND carState IS

moving AND carReactionTime IS very_low

AND carTrajectory IS north AND gear IS

NOT reverse

THEN alarm4 IS yes;

e) Situation 5: Risk of Rear Collision (Reverse).

IF carPosition IS behind AND carDistance

IS (close OR very_close) AND gear IS

reverse

THEN alarm5 IS yes;

f) Situation 6: Risk of Lateral Collision (Turn,

2 Rules).

IF (steeringWheelAngle IS (right OR

center_right) AND carPosition IS right

AND carDistance IS very_close OR

(carPosition IS behind_right AND

carReactionTime IS (low OR very_low))

AND gear IS NOT reverse

THEN alarm6_1 IS yes;

g) Situation 7: Risk of Running over (Rear, 2

Rules).

IF gear IS reverse AND angle IS (behind

OR behind_right) AND trajectory IS west

AND distance IS (close OR very_close)

THEN alarm7_1 IS yes;

h) Situation 8: Risk of Lateral Collision

(Intersection, 2 Rules).

IF carPosition IS (right OR front_right)

AND carState IS moving AND

carReactionTime IS (normal OR low OR

very_low) AND carTrajectory IS west AND

gear IS NOT reverse

THEN alarm8_1 IS yes;

i) Situation 9: Risk of Overtaking on the Right

Lane.

IF carPosition IS (right OR

behind_right) AND carSpeed IS high AND

carTrajectory IS north AND

carReactionTime IS very_low

THEN alarm9 IS yes;

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

452

3.4 HCI Messages

To warn the driver of all dangerous situations, seven

alarms have been established, based on the work by

(Zamora et al., 2017). These alarms share something

in common: they must show an image on the

simulator interface so that the driver can recognize

the symbol showed, and thus, the risk situation that is

happening. Optionally, a sound alarm could be played

at the most dangerous situations (alarms 1, 2 and 3).

On the designed system it is possible that multiple

alarms are activated at the same time. Because of that,

it has been implemented a priority system, so that the

ADAS can choose which one of the activated alarms

must be shown depending on its importance. To

define the hierarchy, risks of a car accident have been

analyzed to determine what are the most dangerous

situations for the driver, based on local statistics about

road traffic crashes by type of accident (DGT, 2017).

The proposed hierarchy is shown in Table 1:

Table 1: Alarm hierarchy.

Priority Alarm

Situation(s)

detected

Alarm 3: Running over a NOT

visualized pedestrian

Alarm 2: Running over a

visualized pedestrian

2, 7

3 Alarm 1: Frontal collision 1

Alarm 6: Lateral collision

(intersection)

5 Alarm 5: Lateral collision (turn) 6

6 Alarm 4: Rear collision 4, 5

Alarm 7: Overtaking on the right

lane

4 EXPERIMENTAL RESULTS

4.1 Experimental Setup

To verify the system, the testing will be done on

simulated environments because of the dangerous

situations that are subject to study. As the ADAS is at

an early stage of development, to perform

experiments that jeopardize people without the

guarantee that the system works would be both

irresponsible and unethical.

That way, a scenario has been designed for each

one of the situations. That scenario consists on the

simulation of the dangerous situation that the ADAS

is supposed to detect, so that it can be proven that the

alarm warns the driver in that situation. A description

of the simulator and the scenarios can be found in

(Zamora, Ledezma and Sanchis, 2016) and (Sipele,

Zamora, Ledezma and Sanchis, 2016).

To complement these unit tests, the ADAS has

been used while driving on a random urban

environment –that is, a default driving scenario has

been selected from the STISIM Driver Simulator

repository, so that it can be checked whether the

ADAS is useful or not on a real environment.

4.2 Experimental Results

First, the ADAS has been tested on the prepared

scenarios. The results obtained from these unit tests

have been generally satisfactory, since the ADAS

alerts the driver of all of the potentially dangerous

situations stablished in those scenarios.

However, those are prepared scenarios, and may

not be realistic. To study the effectiveness of the

ADAS on a real situation, it has been tested on a

random urban environment. This environment has

been extracted from the default driving scenarios

provided by the STISIM Driver Simulator repository,

and consists on a urban scenario that lasts about 15

minutes and challenges the driver with some hazards.

Figure 2 shows three different situations where

alarms are activated on the simulated environment.

Figure 2: Alarms activated.

As it can be seen on Figure 2, when an alarm is

activated an image is showed on the left side of the

dashboard. On the first example, the ADAS is alerting

of a risk of rear collision (alarm 4), as proved by the

proximity of the vehicle visible on the rear view

mirror. On the second one, it alerts the driver about

the pedestrian that is crossing the street (alarm 2). On

the last one it can be observed that a vehicle is

approaching from the left, and the ADAS is alerting

about a risk of lateral collision with it (alarm 6).

Fuzzy Alarm System based on Human-centered Approach

453

Table 2 shows the results obtained while testing the

ADAS on this regular, non-prepared, driving scenario.

Table 2: Results.

Alarm True positives False positives

1 11 33

2 5 0

3 5 2

4 3 1

5 7 13

6 2 0

7 1 2

As we can see, even though the ADAS warns the

driver about the dangerous situations, a lot of false

positives have been detected. Figure 3 shows an

example of these false positives detected, where the

ADAS is alerting the driver of a frontal collision

without being any danger ahead.

Figure 3: False positive from alarm 1.

These false positives are mostly from alarms 1

and 5, and their causes have been analyzed in order to

minimize them in a future version:

 Alarm 1 is designed so that it can alert the

driver of risks of frontal collision. Some of

these false positives are caused by the data used

to determine the position of the surrounding

vehicles –that is, the angle with respect to the

driver. Due to this representation, the ADAS

detects some vehicles that are not exactly in

front of the vehicle as such, alerting the driver

on unnecessary situations. One possible

solution to this problem would be to change the

representation of the vehicles’ position to

lateral and longitudinal distances, which would

be more precise.

There have also been detected false positives

on curves, since sometimes the car does indeed

have a vehicle at the front, but there is no

danger because the driver is taking the turn and

is not going to hit a vehicle on another lane. In

this case, it would be convenient to study the

value of the steering wheel angle, so that the

ADAS can analyze if there is going to be a

collision or not.

 Alarm 5 warns the driver when it detects that

the car can hit laterally another vehicle. Since

the steering wheel is very sensitive, it is

considered that the driver has to move it just a

bit to turn the vehicle. This causes the ADAS

to alert the driver at the slightest move of the

steering wheel, even if the driver is just

straightening up the car. To address this

problem, a possible solution would be to

consider the lateral speed of the vehicle instead

of the longitudinal speed that is currently used,

so that the ADAS would know how quickly the

vehicle is approaching to the side and whether

if it is a risky move or not.

A numerical comparison with the activation of the

alarms from the work by (Zamora et al., 2017)

wouldn’t make sense, because both the dangerous

situations detected and the alarms activated have been

changed and would be different. However, given the

variety of dangers avoided and the minimization of

previous false positives, it can be considered that the

new ADAS improves the previous one, even though

it is still necessary to reduce the false positives. As far

as we know, there are no other related works within

this specific research line, so a comparison with a

baseline cannot be showed.

5 CONCLUSIONS AND FUTURE

WORKS

On one hand, the current ADAS can detect the

potentially dangerous situations established on the

previous system, and does so while reducing the false

positives produced by the previous system. On the

other hand, the ability to detect situations has been

improved, because the new ADAS can warn the

driver on more diverse situations. Therefore, we

consider that the new system improves the previous

one, although there are still some false positives that

reduce the effectiveness of the new ADAS.

With that, it has been proven that it is viable to

develop an ADAS based on a decision-making

system by using fuzzy logic. This kind of system

provides great flexibility to represent the environment

information, which makes the process of making the

rules that use that information easy and intuitive.

As for the future works that could come from this

project, first it would be convenient to improve the

current system so that the problems detected while

VEHITS 2020 - 6th International Conference on Vehicle Technology and Intelligent Transport Systems

454

testing can be avoided. That problems include both

the false positives observed and the possible

improvement of the implementation that uses the

pedals' information. The CAOS research group, from

the Carlos III University of Madrid, is currently

working on a driver-monitoring system that could be

added to the system, so that the information about the

driver –like the area they are looking at– helps the

ADAS to detect risks more accurately.

Beyond the improvement of the ADAS developed

on this project, another line of work would be to

extend the system so that it can assist the driver in

more diverse situations. That way, there are numerous

devices that could be implemented, like a lane-

keeping alert system.

Finally, the ADAS could be further developed,

allowing it to take control of the vehicle in extremely

dangerous situations –e.g. if there is a risk of running

over a pedestrian and the driver hasn't started to brake

the car, the ADAS could stop the car by itself.

ACKNOWLEDGEMENTS

This work has been supported by the Spanish

Ministry of Science, Innovation and Universities,

RTI2018-096036B-C22, TRA2015-63708-R and

TRA2016-78886-C3-1-R projects.

REFERENCES

Attia, H. A., Ismail, S., & Ali, H. Y. (2016). Vehicle safety

distance alarming system. 2016 5th International

Conference on Electronic Devices, Systems and

Applications (ICEDSA).

Bengler, K., Dietmayer, K., Färber, B., Maurer, M., Stiller,

C., & Winner, H. (2014). Three Decades of Driver

Assistance Systems - Review and Future Perspectives.

IEEE Intelligent Transportation Systems Magazine,

6(4), 6-22.

DGT. (2017). Las principales cifras de siniestralidad vial.

Edición ampliada. España 2017. Retrieved May 13,

2019, from: http://www.dgt.es/

Dingus, T. A., Guo, F., Lee, S., Antin, J. F., Perez, M.,

Buchanan-King, M., & Hankey, J. (2016). Driver crash

risk factors and prevalence evaluation using naturalistic

driving data. Proceedings of the National Academy of

Sciences 113(10), 2636-2641.

Duguleana, M., Florin, G., & Gheorghe, M. (2015). Using

Dual Camera Smartphones As Advanced Driver

Assistance Systems: NAVIEYES System Architecture.

Proceedings of the 8th ACM International Conference

on PErvasive Technologies Related to Assistive

Environments. Corfu, Greece: ACM.

Ford Motor Company. (2019). Ford Driver Assist

Technologies | Ford Co-Pilot 360. Retrieved October

25, 2019, from: https://www.ford.com/technology/

Intelligent Systems Technology. (2019). Stisim drive: Car

driving simulator & simulation software. Retrieved

November 15, 2019, from: https://stisimdrive.com/

research/m100-system/

Lee, D., & Yeo, H. (2016). Real-Time Rear-End Collision-

Warning System Using a Multilayer Perceptron Neural

Network. IEEE Transactions on Intelligent

Transportation Systems.

Mamdani, E. H., & Assilian, S. (1975). An experiment in

linguistic synthesis with a fuzzy logic controller.

International Journal of Man-Machine Studies, 7(1), 1-

13.

Mobileye. (2019). Collision Avoidance System for

Distracted Driving. Retrieved October 26, 2019, from:

https://www.mobileye.com/uk/fleets/products/

Pérez, J., Gonzalez Bautista, D., & Milanes, V. (2015).

Vehicle Control in ADAS Applications: State of the

Art. In A. Perallos, U. Hernandez-Jayo, E. Onieva, & I.

J. García Zuazola, Intelligent Transport Systems:

Technologies and Applications (pp. 206-219).

Sanatkumar, R., Gandhe, S., & Dhulekar, P. (2015).

Pedestrian Protection System for ADAS using ARM 9.

International Journal of Computer Applications

127(2):19-23.

Sipele, O., Zamora, V., Ledezma, A., & Sanchis, A. (2016).

First Symposium SEGVAUTO-TRIES-CM.

Proceedings of the First Symposium SEGVAUTO-

TRIES-CM. Technologies for a Safe, Accessible and

Sustainable Mobility, (pp. 27-30). Madrid.

Volvo Car Corporation. (2019). Safety | Volvo Cars.

Retrieved October 25, 2019, from:

https://www.volvocars.com/intl/why-volvo/human-

innovation/future-of-driving/safety

WHO. (2018). Global status report on road safety 2018. In

Violence and Injury Prevention and Disability (VIP).

WHO Library Cataloguing-in-Publication Data.

Wilson, D., Kennedy Babu, C. N., & Tharumar, S. (2015).

Advance Driver Assistance System (ADAS) – Speed

Bump Detection. 2015 IEEE International Conference

on Computational Intelligence and Computing

Research (ICCIC).

Zadeh, L. A. (1988). Fuzzy Logic. (IEE, Ed.) Computer,

21(4), 83-93.

Zamora, V. M., Sipele, O., Ledezma Espino, A. I., &

Sanchis de Miguel, A. (2017). Intelligent Agents for

Supporting Driving Tasks: An Ontology-based Alarms

System. 3rd International Conference on Vehicle

Technology and Intelligent Transport Systems. Porto,

Portugal.

Zamora, V., Ledezma, A., & Sanchis, A. (2016). First

Symposium SEGVAUTO-TRIES-CM. Proceedings of

the First Symposium SEGVAUTO-TRIES-CM.

Technologies for a Safe, Accessible and Sustainable

Mobility, (pp. 23-26). Madrid.

Fuzzy Alarm System based on Human-centered Approach

455