Clipped Utility Function with Upper-Bound Clipping

Consider a reinforcement learning agent being trained with a policy gradient method. For a given state-action pair, the ratio of the new policy's probability to the old policy's probability is 3.0. The estimated advantage for this action is positive. The algorithm incorporates a clipping mechanism defined as min(ratio, 1 + ε), where ε is set to 0.2. What is the primary effect of this mechanism on the policy update for this specific step?

Asymmetric Effect of Upper-Bound Clipping

A policy update mechanism uses a function to adjust the policy probability ratio, defined as min(ratio, 1 + ε). Given ε = 0.2, match each original ratio value on the left with its corresponding adjusted value on the right after the function is applied.