Collision avoidance with machine learning

Abstract

The top four causes of deadly car accidents are speeding, drinking while operating a vehicle, driver weariness, and distracted driving. Speeding has the highest percentage among all factors that cause deadly car accidents. As cars become more automated, there is hope that a reduction in road fatalities will be observed. The goal of the study is to research and develop machine-learning algorithms to decrease car accidents and improve human safety on roads. The recent systems have achieved good results in which they were able to provide safer rides than human-controlled cars. Such systems used sensors and cameras that helped to eliminate human error. However, the behavior of these systems under various circumstances remains understudied until now. Additionally, these systems are not advanced enough to detect and moderate every possible impact. Therefore, the purpose of this paper is to apply a Machine learning backpropagation algorithm based on six different features (speed of the car, speed of the neighbor car, level of alcohol in blood, weather condition, road condition, and whether the brakes are applied). These features are used to train the system to predict the car accident and thus avoid a car crash.