International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 05 | May 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 7354
Self Driving Car Using Machine Learning
Tej Kurani
1
, Nidhip Kathiriya
2
, Uday Mistry
3
, Prof. Lukesh Kadu
4
, Prof. Harish Motekar
5
1,2,3
Student
4,5
Assistant Professor
1,2,3,4,5
Dept. of Information Technology, Shah and Anchor Kutchhi Engineering College,
Mumbai, Maharashtra, India.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The fundamental idea behind the project is to
develop an automated car that can sense its environment
and move without human input. This paper proposes Car
automation, which is accomplished by recognizing the road,
signals, obstacles, stop signs, responding and making
decisions, such as changing the course of the vehicle,
stopping red signals, stopping signs, and moving on green
signals using Neural Network. Self-driving car processes
input, tracks a track and sends instructions to the actuators
that control acceleration, braking, and steering. The
software tracks traffic by means of hard-coded rules,
preventive algorithms, predictive modeling, and "smart"
discrimination on objects, helping the software to follow
rules on transport.
Key Words: Self-driving, Neural Network, Actuators,
changing course, predictive modeling, preventive
algorithms, smart discrimination.
1. INTRODUCTION
Human-driven Cars use protection systems to Identify
Barriers and stop-offs in some high-end cars, but none of
them is completely driverless. The automation feature of
existing cars is insufficient to allow cars to drive itself.
There is a constant need for drivers, without it the car is
inaccessible. But with self-driving cars, we can make the
presence of cars on the road constantly. The Driver
constantly needs to monitor signals, road safety signs,
barriers, and lanes for traditional cars and make decisions
accordingly. Self- driving is no longer a futuristic dream,
but it is becoming a reality. Companies proclaim their
dedication to developing and launching autonomous cars
and many of them talk about the level of autonomy being
developed. Certainly, autonomous driving can be
dangerous to some but it also has its advantages. This
would result in reduced traffic congestion, reduced
emissions, lower travel costs for all, and a reduction in the
cost of new roads and services. It would also immensely
improve the mobility of people with old and physical
disabilities.
(a) (b)
Fig -1: Traditional cars(a) and Autonomous cars(b)
The fields of expertise are self-driving model cars in a
model area with few sensors[6] such as Tesla, and such
cars use numerous sensors such as lidar and radar. The
way we seek to achieve the autonomy of cars is to model it
on RC cars on the 1/10 scale. With the aid of the pi camera
and the ultrasonic sensor, the car can sense its
environment, and data collected from the two are
transmitted on the server through the Raspberry Pi, where
we are running the neural network, where the images are
being processed to detect lane markings. The car
automatically drives on its own according to the lane
marking, once it is trained. In actual vehicles, the same
algorithm and techniques can be used for automation.
2. RELATED WORK
We have studied many papers so far and we have come
across the different techniques and technologies which are
used to develop the system
In literature[1], an autonomous platform for cars, using
the softmax function, is presented which squashes the
outputs of each unit between 0 to 1. The softmax function
acts as a sigmoid function by ranging output, while that is
not done by an actual softmax function. The use of a neural
network helps to give output in real-time. Before
implementing it, they have tested the model on the
MATLAB simulator. The system only uses a single camera