Functionality improvement of driverless cars using traffic data and artificial neural networks

This paper aims to improve the functionality of driverless cars using traffic data based on a follow-up study. Artificial neural networks are used for training a driverless car to react and decide whether to accelerate or decelerate with respect to the speed and acceleration of adjacent vehicles as well as their types and positions to the car. To this end, an Equivalent Vehicle (EV) is introduced and defined for each car in the traffic data and used for training of a neural network. A feed-forward and a time-delay neural network (TDNN) are used to simulate a driver’s reaction. The TDNN tends to show a better performance compared to a feed-forward neural network. This method is illustrated using the NGSIM traffic data.

Keywords:

Driverless car , Artificial neural networks , NGSIM traffic data , Driver’s behavior