• Title: "Exploring the Feasibility of Sim2Real Transfer in Reinforcement Learning"
• Author: Lucas Driessens
• Date: June 1, 2024 (expected completion date)
• Institution: Howest (Bachelor Multimedia & Communication Technology)

Want to read more? : go look at the thesis.pdf

Exploring the Feasibility of Sim2Real Transfer in Reinforcement Learning

This repository contains the research and implementation files for the bachelor thesis by Lucas Driessens, presented at Howest, focused on sim-to-real transfer in reinforcement learning, particularly navigating a maze using a remote-controlled (RC) car.

https://github.com/driessenslucas/researchproject/assets/91117911/972541d6-5010-4f73-a56f-5b60bb8b4f3e

Project Overview

This thesis investigates the transition of reinforcement learning algorithms from simulated environments to real-world applications. The core of this research revolves around the practical deployment of a Double Deep Q-Network (DDQN) for controlling an RC car in a physical maze setting, highlighting both the potential and challenges of such technological transfers.

Objectives

• Validate the practical application of simulated RL algorithms in real-world scenarios.
• Assess the efficacy of DDQN in real-world maze navigation.
• Explore the interplay between theoretical concepts and their real-world implementations.

Key Insights

• Sim-to-Real Transition Challenges: Detailed exploration of adjustments needed to translate simulation training into real-world effectiveness.
• Algorithmic Efficacy: Analysis on the performance of DDQN in navigating complex real-world environments.
• Practical Implementation Considerations: Discussion on overcoming real-world implementation challenges, including sensor integration and environmental variability.

Contributions

This project welcomes academic collaborations and further research enhancements. Contributions that improve the algorithms, extend the simulations, or refine the real-world testing components are highly encouraged.

Citation

Please cite this thesis if it assists or influences your research: Lucas Driessens, "Exploring the Feasibility of Sim2Real Transfer in Reinforcement Learning", Bachelor Thesis, Howest, June 2024.

License

This project is released under the MIT License - see the LICENSE.md file for details.

Contact

For inquiries or further information, please contact: Lucas Driessens - lucas.driessens@hotmail.com

Acknowledgments

• Special thanks to guest speakers and faculty advisors who provided invaluable insights and guidance throughout this research journey.
• Gratitude towards Howest for the support and resources provided during this academic pursuit.