Reinforcement Learning
hard
~50 hours
Sim-to-Real Robot Control with Domain Randomization
Train an RL agent in simulation (MuJoCo/Isaac Gym) with domain randomization, then demonstrate transfer to a real or more realistic environment. Track sim-to-real gap metrics.
Skills Demonstrated
Sim-to-real transfer
Domain randomization
SAC/PPO implementation
Robotics control
Implementation Steps
- Set up MuJoCo environment for a manipulation task
- Implement SAC agent with entropy regularization
- Add domain randomization: vary mass, friction, sensor noise
- Train ensemble of policies across randomized environments
- Evaluate transfer to held-out environment configurations
- Build visualization dashboard showing sim vs real performance gap
Interview Relevance
Why this project matters for interviews
Robotics RL is a growing field at Tesla, Google DeepMind, and Boston Dynamics. Sim-to-real transfer is the critical unsolved challenge — showing you've tackled it demonstrates real research capability.