ROS-Enabled Autonomous Vehicle Architecture within CARLA: A Comprehensive Overview

摘要

This paper provides an in-depth exploration of perception, path planning, and control strategies for autonomous vehicles within the CARLA simulation environment. CARLA, known for its accessibility and open-source nature, serves as the backdrop for this comprehensive study. The research investigates LOAM and Lego-LOAM algorithms using LiDAR technology for Simultaneous Localization and Mapping (SLAM) with the goal of improving relocalization efficiency in pre-mapped environments. Moreover, the study integrates global and local path-planning techniques to navigate obstacles and select optimal routes. The conventional PID control and the Stanley Method are scrutinized for managing longitudinal and lateral vehicle control, emphasizing their efficacy. These methodologies are seamlessly integrated and rigorously tested within the simulation environment using the Robot Operating System (ROS). The study concludes by highlighting the ease of transitioning proposed perception and control strategies to real-world implementation through ROS. Ultimately, CARLA emerges as a reliable platform for evaluating innovative solutions that drive advancements in the autonomous vehicle industry.