Addressing the Sim2Real Gap in Robotic Scenarios

Simulation has been well-established for robotics education and integrated robot software testing for years, there is an ongoing debate in the research community about the ability to transfer robotics skills learned in simulation to reality, a concept termed Sim2Real transfer [1]. Testing learning methods in real robots is often dangerous, since those methods often attempt dangerous behaviours that can damage the robot. For this reason, most successful learning methods for robotics applications are mainly demonstrated in simulation environments. A simulated environment though can never replicate the real world, and methods attempting to transfer learned behaviours from simulation to reality will have to deal with, what is called, the reality gap. The reality gap refers to the discrepancies between the simulated and the real, physical world. The Sim2Real transfer concerns the task of transferring robot skills acquired in a simulated environment to a physical setting [2].

[1] Höfer, S., Bekris, K., Handa, A., Gamboa, J. C., Golemo, F., Mozifian, M., ... & White, M. (2020). Perspectives on sim2real transfer for robotics: A summary of the r: Ss 2020 workshop. arXiv preprint arXiv:2012.03806.

[2] Dimitropoulos, K., Hatzilygeroudis, I., & Chatzilygeroudis, K. (2022). A brief survey of Sim2Real methods for robot learning. Advances in Service and Industrial Robotics: RAAD 2022, 133-140.