Maps and Scenario Framework in CARLA
Maithili Karande
Vishwakarma University, Pune, India
Abstrct:
In this research, we explore the use of CARLA, a high-fidelity simulation environment, to train and evaluate machine learning models, particularly reinforcement learning agents, for autonomous driving. CARLA offers a range of realistic 3D maps that replicate urban environments, enabling the simulation of diverse driving scenarios. These maps include detailed roads, intersections, buildings, pedestrians, vehicles, and dynamic elements, creating a comprehensive framework for testing autonomous driving algorithms. The primary objective is to develop and refine algorithms that enable vehicles to autonomously navigate these environments safely and efficiently, exhibiting human-like driving behaviors. This involves tasks such as perception, decision-making, and control, all tested under various challenging conditions including adverse weather, pedestrian interactions, and complex traffic scenarios. Specific testing scenarios include intersection navigation, lane changes, pedestrian crossings, adverse weather conditions, emergency braking, construction zone navigation, and highway merging. These scenarios allow for the systematic assessment of algorithm performance, ensuring robustness and adaptability in various conditions. Ultimately, CARLA maps provide a structured and risk-mitigated environment for iterative experimentation, fostering innovation and collaboration within the autonomous driving community. By enabling comprehensive and realistic testing, CARLA significantly contributes to the advancement of autonomous driving technology, helping to develop robust algorithms capable of handling the complexities of realworld driving. This approach not only accelerates the development process but also ensures that the algorithms are thoroughly tested in diverse and challenging conditions, which is crucial for their deployment in real-world scenarios. CARLA thus serves as an essential tool in the pursuit of safe, efficient, and reliable autonomous vehicles.
Keywords:
CARLA, autonomous driving, maps, scenarios, perception
Published on: 05-2024
Journal Name: Science Management Design Journal
Volume: 02
Issue: 01
Pages: 16-21
Month: May
Year: 2024