Skip to content

qortjdbs/Self-Driving_Algoirthm_with_RL

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 

Repository files navigation

KAIST Pre-URP: Reinfocement Learning-based Self Driving Algirthm in Unity Virtual Environment

This project focuses on developing an advanced autonomous driving system that combines multiple sensing technologies and artificial intelligence algorithms to enable self-driving vehicles to navigate complex environments safely and efficiently.

Key Features:

  • Sensor Fusion: Combines data from LiDAR, cameras, and radar to provide a comprehensive understanding of the vehicle’s surroundings.
  • Object Detection: Uses deep learning algorithms for real-time detection of objects, including vehicles, pedestrians, and obstacles.
  • Path Planning: Implements advanced path planning algorithms to ensure safe and efficient navigation, even in dynamic environments.
  • Reinforcement Learning: Integrates reinforcement learning for adaptive decision-making in complex driving scenarios.
  • Simulation Testing: Leverages a simulation environment to test and validate the autonomous driving model under various traffic and environmental conditions.

Methodology:

  • Sensor Data Integration: Collects and processes real-time data from multiple sensors to generate a high-fidelity map of the environment.
  • Deep Learning for Perception: Uses convolutional neural networks (CNNs) to detect and classify objects in the vehicle's path.
  • Decision Making: Employs reinforcement learning to make decisions about lane changes, obstacle avoidance, and speed adjustments.
  • Path Optimization: Implements algorithms to continuously optimize the vehicle's path for safety and efficiency.

Results:

  • Enhanced Perception: Achieved high accuracy in object detection, even in complex scenarios with multiple objects and occlusions.
  • Safe Navigation: Demonstrated effective navigation in simulation environments, avoiding collisions and optimizing travel time.
  • Adaptive Learning: The reinforcement learning-based decision-making model adapts to new driving conditions over time, improving safety and efficiency.

Conclusion:

This autonomous driving system integrates state-of-the-art sensor fusion, deep learning, and reinforcement learning techniques to provide a robust and adaptable solution for self-driving vehicles. Future work will focus on improving real-world deployment and handling of edge cases in challenging environments.

About

RL-based Self-Driving Algorithm

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages