Intelligent transportation is gaining much attention about autonomous driving systems. This research will present a fully developed comprehensive framework to implement the navigation of autonomous vehicles using CARLA. Deep learning-based object detection by the use of YOLOv5 is used in the current project, which identifies both real-time vehicle traffic, traffic lights, and obstacles. In parallel, lane detection is achieved through image processing techniques, such as edge detection and Hough Line Transform. The proposed system integrates these capabilities with decision-making algorithms to facilitate responsive vehicle control in dynamic environments. In addition, the validation of autonomous behaviour from a holistic point of view uses third-party camera views. Multiple vehicle traffic simulations add realism to the environment. A pilot function simulates the actual condition, with a synchronized simulation approach for the correct timing in processing and analysing data. The outputs include saved images, processed frames, and videos demonstrating the efficacy of the system under varying scenarios. This research integrates modern computer vision techniques with simulated environments to create a robust testing platform for autonomous navigation. Future research can extend this framework to real-world applications or incorporate advanced reinforcement learning algorithms for enhanced decision-making capabilities.
Navigating the Future: Deep Reinforcement Learning for Smart Autonomous Vehicles
2025-03-19
551257 byte
Conference paper
Electronic Resource
English
NAVIGATING AERIAL VEHICLES USING DEEP REINFORCEMENT LEARNING
European Patent Office | 2021
|Navigating autonomous vehicles in uncertain environments with distributional reinforcement learning
SAGE Publications | 2024
|Systems and Methods for Navigating Aerial Vehicles Using Deep Reinforcement Learning
European Patent Office | 2021
|A Deep Reinforcement Learning Driving Policy for Autonomous Road Vehicles
ArXiv | 2019
|