Case Study
Category:
Autonomous Driving & AI
Impact:
14 Weeks | Safe R&D Testbed
Developing and validating autonomous vehicles in the real world is costly, risky, and often unsafe. To address this, I built a simulation framework that integrates computer vision (CV) and reinforcement learning (RL), enabling realistic and safe testing of autonomous driving systems. Using the CARLA simulator, the system replicates real-world driving conditions to support lane detection, obstacle avoidance, and traffic sign recognition in diverse environments.
Implemented CARLA simulator to generate realistic, interactive driving environments for training and validation.
Integrated YOLOv8 for traffic sign recognition and obstacle detection, and LaneNet for accurate lane detection and tracking.
Created a preprocessing pipeline using OpenCV for image cleaning, segmentation, and feature enhancement.
Applied a Deep Q-Network (DQN) to enable adaptive decision-making for dynamic driving tasks, including merging and obstacle avoidance.
