TIMELINE
14 Weeks
To Working Testbed
Autonomous Driving & AI
TIMELINE
To Working Testbed
DETECTION
Traffic Sign & Obstacle
RL MODEL
Adaptive Decision-Making
SCENARIOS
Urban & Highway Tested
SUCCESS
Lane-Keep Completion Rate
When the risks of reality stall the speed of innovation
Testing autonomous systems on physical roads is prohibitively dangerous and expensive. Every real-world failure carries a cost that slows the essential validation needed for safe, widespread deployment.
The Intellema Design Challenge
Developing autonomous vehicles requires rigorous validation that is often too risky or costly to perform in real-world environments. This project utilized the CARLA simulator to create a high-fidelity testbed for evaluating computer vision and reinforcement learning models safely.
The framework integrated YOLOv8 and LaneNet to handle real-time perception tasks like obstacle avoidance and lane tracking. The system optimized decision-making for complex maneuvers, ensuring robust performance across diverse urban and highway scenarios.
Realistic, interactive driving environments generated via the CARLA simulator to provide a safe testbed for training and validation.
Perception suite integrating YOLOv8 for traffic sign and obstacle detection with LaneNet for precision lane tracking.
OpenCV-based pipeline facilitating image cleaning, segmentation, and feature enhancement to improve model robustness.
Deep Q-Network (DQN) implementation enabling adaptive decision-making for complex tasks like merging and dynamic obstacle avoidance.
