RL Environment: get obs from last episode

Hi,

I defined a Reinforcement Learning Environment based on the rlCreateEnvTemplate.

How can I limit the change of the actions, which are choosen by the agent while having a predefined action range (-50V<action<150V) ? (in my case I have voltages as actions.)

I think about something like this:

abs(action(i-1)-action(i)) < 10

for step i. But I don't know how to access the action from the previous step (which would be action(i-1)).

Another approach would be to use the change in voltage as action and then add this change to the voltage from the previous step. Again I would have to access a value from the previous step and I don't know how to get this value.

Thank you for any advice :)

Answer 1

[Katharina Schmidt]

I solved it by using LoggedSignals to store the action values from the last step.