The LQG cost becomes gradient dominated under a history-based controller parameterization, yielding global convergence guarantees for policy gradient methods in model-based and model-free settings.
Policy optimization for H2 linear control with H∞ robustness guarantee: Implicit regularization and global convergence,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.OC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
On the Gradient Domination of the LQG Problem
The LQG cost becomes gradient dominated under a history-based controller parameterization, yielding global convergence guarantees for policy gradient methods in model-based and model-free settings.