Date of Award

1-8-2020

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Computer Science

First Advisor

Xiaobu Yuan

Keywords

Autonomous, Gap, Reality, Simulation, Vehicles

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

In order to facilitate acceptance and ensure safety, autonomous vehicles must be tested not only in typical and relatively safe scenarios but also in dangerous and less frequent scenarios. Recent pedestrian fatalities caused by test vehicles of the front-running giants like Google and Tesla suffice the fact that Autonomous Vehicle technology is not yet mature enough and still needs rigorous exposure to a wide range of traffic, landscape, and natural conditions on which the Autonomous Vehicles can be trained on to perform as expected in real traffic conditions. Simulation Environments have been considered as an efficient, safe, flexible and cost-effective option for the training, testing, and validation of Autonomous Vehicle technology. While ad-hoc task-specific use of simulation in Autonomous Driving research is widespread, simulation platforms that bridge the gap between simulation and reality are limited. This research proposes to set up a highly realistic simulation environment (using CARLA driving simulator) to generate realistic data to be used for Autonomous Driving research. Our system is able to recreate the original traffic scenarios based on prior information about the traffic scene. Furthermore, the system will allow to make changes to the original scenarios and create various desired testing scenarios by varying the parameters of traffic actors, such as location, trajectory, speed, motion states, etc. and hence collect more data with ease.

Share

COinS