This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
SPiDR: A Simple Approach for Zero-Shot Safety in Sim-to-Real Transfer
Created by
Haebom
Author
Yarden As, Chengrui Qu, Benjamin Unger, Dongho Kang, Max van der Hart, Laixi Shi, Stelian Coros, Adam Wierman, Andreas Krause
Outline
SPiDR (Sim-to-real via Pessimistic Domain Randomization) is an algorithm designed to address the sim-to-real gap problem encountered in real-world policies trained in simulators. SPiDR uses domain randomization to incorporate uncertainty about the sim-to-real gap into safety constraints, enabling safe sim-to-real transfer. SPiDR is compatible with existing training pipelines and has demonstrated its safety and performance through sim-to-sim benchmarks and experiments on real robot platforms.
Takeaways, Limitations
•
Takeaways:
◦
Despite the sim-to-real gap, it ensures safety and maintains strong performance.
◦
Compatible with existing training pipelines.
◦
Provides a scalable algorithm for secure sim-to-real transfer.
•
Limitations:
◦
The specific Limitations was not presented in the paper.
