Traversing the Reality Gap via Simulator Tuning
This addresses the problem of data scarcity and transfer learning for roboticists, though it is incremental as it builds on existing simulation tuning methods.
The paper tackles the reality gap in robotics by tuning simulation parameters to improve transfer of learned behaviors to real-world tasks, showing that optimized physics parameters can narrow the gap and identifying key factors like friction and actuator velocity.
The large demand for simulated data has made the reality gap a problem on the forefront of robotics. We propose a method to traverse the gap by tuning available simulation parameters. Through the optimisation of physics engine parameters, we show that we are able to narrow the gap between simulated solutions and a real world dataset, and thus allow more ready transfer of leaned behaviours between the two. We subsequently gain understanding as to the importance of specific simulator parameters, which is of broad interest to the robotic machine learning community. We find that even optimised for different tasks that different physics engine perform better in certain scenarios and that friction and maximum actuator velocity are tightly bounded parameters that greatly impact the transference of simulated solutions.