Simulation-Environments-AI-based-Autonomous-Driving-Solutions--a-Comparative-Evaluation

This thesis presents a comprehensive approach to evaluate and enhance autonomous vehicle simulators. The research comprises two primary works. Firstly, a systematic evaluation methodology is introduced, establishing a scoring system based on 73 discerning parameters to assist users in efficiently selecting a suitable simulator. This method streamlines the process, enabling users to assess simulator capability aligned with their specific needs. Secondly, the thesis explores the development of a prototype simulator driven by a deep generative model. This model aims to simulate the transition of 2D lidar data and generate diverse scenarios crucial for training machine learning models in autonomous driving. The study assesses the performance of the system of comparison by evaluating existing simulators and user preferences. Furthermore, the results of the proposed prototype simulator are discussed.