Deep Reinforcement Learning for Adverse Garage Scenario Generation

Abstract--Autonomous vehicles need to travel over 11 billion miles to ensure their safety. Therefore, the importance of simulation testing before real-world testing is self-evident. In recent years, the release of 3D simulators for autonomous driving, represented by Carla and CarSim, marks the transition of autonomous driving simulation testing environments from simple 2D overhead views to complex 3D models. During simulation testing, experimenters need to build static scenes and dynamic traffic flows, pedestrian flows, and other experimental elements to construct experimental scenarios. When building static scenes in 3D simulators, experimenters often need to manually construct 3D models, set parameters and attributes, which is time-consuming and labor-intensive. This thesis proposes an automated program generation framework. The generated 3D ground scenes are displayed in the Carla simulator, where experimenters can use this scene for navigation algorithm simulation testing. However, experiments have shown that autonomous vehicles need to travel over 11 billion miles to ensure their safety [2]. In practical use and testing, traffic accidents caused by autonomous A. Background The Self-Driving System, also known as the Autonomous As one of the most critical quality assurance technologies, Driving System (ADS), is a comprehensive integration of ADS testing has garnered attention from both academia and hardware and software designed to autonomously manage industry [3]. Nonetheless, due to the numerous components motion control based on its perception and understanding of and high complexity of ADS, testing faces many challenges. Naturalistic Field Operational Testing (N-FOT) to simulationbased Perception, decision-making, and control constitute the three testing [4], also known as simulation testing.