Why does this site exist and why should you care right now?
Reinforcement Learning in robotics is a total mess for beginners: fragmented tutorials, heavy theory, zero practical guidance.
Everyone talks about PPO and SAC, but almost no one shows you how to put an agent on a real robot.
Most guides are just theory or code collages that don’t work in real projects.
Here you find a clear path from theory –> code –> simulation –> real robot(follow soon), explained step by step.
My promise: everything you see here is tested in practice and works on affordable hardware.
Here you learn applied RL, not just formulas.
Who is the person behind ReinforcementLearningPath.com?
I’m Dragos Calin. Since 2011, I have gone through many stages in the fields of Robotics and Artificial Intelligence. I have designed and built autonomous systems for agriculture and artificial intelligence applications.
Although I was born in Romania, I like to say that I was raised in the European Union. The values of innovation and collaboration have shaped and inspire me in everything I do.
I am passionate about removing technical barriers. I want to turn complex robotics and artificial intelligence concepts into easy-to-understand solutions. These concepts are tested and validated in real-world projects to ensure their effectiveness. By reading the tutorials, you will gain both the knowledge and detailed step-by-step instructions needed to build your own application.
I’m a practitioner, I’m constantly learning, I don’t claim perfection.
The topics presented on this blog are based on my experience with the Robo-Fermier (eng.: Robo Farmer) project and beyond. Robo-Fermier is an autonomous robot. It navigates between crop rows and performs tasks like mowing grass, scanning for diseased leaves, and targeted spraying.
Here are some images of me and the Robo-Fermier. They show the project’s real-world applications.
The robot is positioned between rows of trees, highlighting its robust design and modular structure.
What you actually find on the site
Types of tutorials:
- RL Fundamentals
- Deep RL Algorithms
- Robotics & Edge AI
- Experiment & Tools
- Sim2Real
- Low-cost hardware (Jetson, Pi, IMU, motors, etc.)
How the tutorials are built
My methodology is:
problem –> minimal theory –> code –> simulation –> results –> debugging –> extensions
How I write and test the tutorials
The code is run by me.
Agents are trained in Gymnasium/ MuJoCo / Unity.
Everything possible is also tested on real hardware.
I display graphs, results, real debugging.
Transparency and limitations
I tell you from the start what NOT to expect from me.
I don’t promise “overnight” results. RL is not magic, it’s not “training in 5 minutes“, it’s not “how to make money with AI“.
I don’t fake results. If an agent doesn’t converge, I say so. If the reward is unstable, I show the real graph, not a customized one.
I don’t do clickbait and I don’t hide technical issues.
If a tutorial has a bug in a specific version of SB3, I say it clearly in the tutorial.
How can you contribute
How can you help me improve this RL & Robotics ecosystem?
- You can send feedback, bug reports, tutorial ideas, share useful articles.
How can you contribute with your own experiments, results or ideas?
- You can also write guest posts, pull requests, examples of how to apply the tutorials.
What do you gain if you actively engage in the community, not just passively read?
- You will have feedback on your projects, visibility, more conceptual clarity.
Let’s connect! If you have any questions or thoughts to share, don’t hesitate to write me at: calinrobotics@gmail.com.
If you have 5 more minutes …
If you want to:
- build real RL agents (not theoretical ones),
- understand problems before they break your robot,
- learn Sim2Real without guessing,
- follow complete end-to-end tutorials from theory –> code –> training –> deployment,
then this is the place.
Join the newsletter and get the insights I don’t post anywhere else.
Short, clear, practical – written for people who actually build things, not just read about them.