# TRPO

Constrains each policy update with a KL-divergence trust region, giving monotonic-improvement guarantees in theory.

## What it is used for

Situations where very conservative, stable policy updates matter; PPO is its simpler descendant.