Reproduce “Separated Proportional-Integral Lagrangian for Chance Constrained Reinforcement Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Baiyu Peng, Yao Mu, Jingliang Duan, Yang Guan, Shengbo Eben Li, Jianyu Chen
SPIL adjusts safety pressure in simulated car-following by combining an immediate response to safety misses with accumulated error, while limiting accumulation when errors are large. The task rewards speed but requires the gap above 2 m over 40 steps. At both tested thresholds, SPIL met safety requirements without reported oscillations and had the highest reward among methods meeting them.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.