DANA HADDAD
SIMULATION. HARDWARE. TRANSFER.
I study how legged robots adapt to the real world, using system identification and narrower randomization ranges.
Let’s talk researchResearch
Wider randomization
wasn’t more robust.
It was making the policy timid.
The quadruped pipeline started in simulation. Retraining after each hardware change took most of a day, and the policy kept overfitting to the simulator.
Adding system identification and narrowing the randomization ranges brought retraining to about 40 percent of the previous time.
System identificationDomain randomizationSim-to-real transfer
Background
Building a research
practice in robotics.
I’m a master’s student in robotics and a research assistant at the Legged Locomotion Lab, applying for PhD positions in legged robotics.
- Education
- MSc Robotics
2025 to 2027, expected - Publications
- Two workshop papers
ICRA 2026 workshops - Tools
- Python, PyTorch, Isaac Gym,
ROS 2 and C++
Legged robotics research