All news

Lab news

Model-informed safe reinforcement learning accepted to IROS 2026

Victor Paredes and Ayonga Hereid deposited Model-Informed Safe Reinforcement Learning for Bipedal Locomotion via Step-to-Step Prediction on arXiv on September 28, 2026. Their record reports acceptance to IROS 2026.

The authors’ arXiv announcement provides the acceptance status. This news item is dated to that announcement.

The study evaluates the approach in full-order Digit simulation using MuJoCo. Its safety certificate is derived for the reduced-order model; the results do not establish a general hardware safety guarantee.

Read the paper and citation.