Enhancement of Advanced Driver Assistance System (Adas) Using Machine Learning

Abstract

Most of the road accidents can be attributed to human errors. Advanced driver assistance system (ADAS) is an electronic system that guides a vehicle driver while driving. It is designed with a safe human-machine interface that is intended to increase vehicle safety and road safety. ADAS is developed to automate, adapt and enhance vehicle systems for safety and better driving. An increasing number of modern vehicles have ADAS such as collision avoidance, lane departure warning, automotive night vision, driver monitoring system, anti-lock braking system and automatic parking system. ADAS relies on input from multiple data sources like lidar, radar, and camera. This paper describes the implementation of ADAS using machine and deep learning algorithms. We implement a model which has a 360-degree camera (lens on two sides of 170 degrees each), lidar, ultrasonic sensor, and radar that provide the input for ADAS. We implement the ADAS by training this whole model using deep learning (advanced machine learning) by designing a neural network using Python in TensorFlow. Generative adversarial networks (GANs) are used in object detection when a hazed image (foggy, rainy, etc.) is detected. This reduces the sensor complexity and area in the vehicle. Results gained from the study and their implications are presented.