A Study on Autonomous Driving Simulation Using a Deep Learning Process Model

Abstract

Along with artificial intelligence technologies, deep learning technology, which has recently received a great deal of attention, has been studied on the basis of developed artificial neural networks. This thesis deals with the detection, recognition, judgment, and control that are included in the basic technologies of the autonomous driving subsystems to achieve fully autonomous driving. And this work solves many problems in this area. The use of the CARLA simulation in this project is the development of a deep learning intelligent autonomous driving system in the road environment. Autonomous driving recognizes the situation by processing the data collected through images from multiple sensors or lidars and cameras in real-time. In the cloud server process using real data, explore various deep learning models for traffic flow prediction, return the model trained onboard, perform the prediction and solve the problem of fully autonomous driving, including a module of control, which is a CARLA simulation.