Driverless Car - Design of a Parallel and Self-Organizing System

EasyChair Preprint no. 1248

Abstract

For autonomous vehicles, several real-time systems must work tightly together. These real-time systems, include environment mapping and understanding, localization, route planning and movement control. All these real-time systems work simultaneously and use artificial neural networks which are self-organizing systems. The self-driving car itself needs to be equipped with the appropriate computational hardware such as parallel computing power of modern graphics processors and software infrastructure for supporting implementation of DNN & CNNs. There are two approaches for applying deep learning in self-driving cars. The first one is semantic abstraction and the second is end-to-end learning system. Our chosen approach is semantic abstraction where the problem of autonomous driving is broken down into several components and at the end, these components are glued together into master network that makes the driving decisions. Also implementation of image classification in traffic signs dataset using Deep Neural Network with TensorFlow is presented.

Keyphrases: Artificial Neural Networks, self-organizing system, Semantic Abstraction

@Booklet{EasyChair:1248,
  author = {Poondru Prithvinath Reddy},
  title = {Driverless Car - Design of a Parallel and Self-Organizing System},
  howpublished = {EasyChair Preprint no. 1248},

  year = {EasyChair, 2019}}