Would it make sense to restrict the action to what's possible?

[shamoons]

If the cartpole is already all the way at the right, we can't really select that action. So would it make sense to disallow that from either the random case (by sampling again) or the network case (by choosing the next highest Q value that the network predicts)?

[daysofthunder98]

The episode itself terminates if the cartpole deviates from either side by more than 15 degrees, so the experience is recorded and (hopefully) the agent learns from it.

[WorksWellWithOthers]

The action shouldn't be restricted since it's the goal of the agent to learn what action to take for the most reward.