Over the past few years, I’ve had the opportunity to tinker with various robots—quadrupeds, bipeds, wheeled robots, and more. This hands-on experience has helped me apply and refine my skills in reinforcement learning. However, despite all this, I’ve always wanted to explore the full Sim2Real pipeline—from designing a robot from scratch (CAD and mechanical design) to fabrication, simulation, training, and ultimately deploying it in the real world.
This blog post captures my journey of attempting this process with the SO-100, a low-cost, easy-to-build robotic arm.
In particular, I will cover the following topics:
- Designing the robot in CAD (SolidWorks) and generating a URDF
- Importing the URDF into Isaac Sim/USD Composer
- Adding sensors and actuators in Isaac Sim
- Performing system identification
- Training a basic policy
- Sim2Real transfer and deployment
More details about the project to be added soon…