Domain Randomization for Sim2Real Transfer
Lilian Weng
Domain randomization addresses the reality gap in robotics by training models in simulators with randomized physical parameters so they generalize to real-world conditions. The technique varies friction, damping, mass, and other physical properties during simulation training to reduce sensitivity to parameter mismatches between simulator and real robots. This approach enables robots trained entirely in simulation to function on physical systems without additional real-world fine-tuning.
Why it matters
In Robotics, one of the hardest problems is how to make your model transfer to the real world. Due to the sample inefficiency of deep RL algorithms and the cost of data collection on real robots, we often need to train models in a simulator which theoretically provides an infinite amount of data. However, the reality gap between the simulator and the physical world often leads to failure when working with physical robots. The gap is triggered by an inconsistency between physical parameters (i.e. friction, kp, damping, mass, density) and, more fatally, the incorrect physical modeling (i.e. collision between soft surfaces).