How to use RNN+DDPG together?

I have a custom RL environment that changes over time. Therefore I am planning to use the timestamp as one of the state information. I am also planning to use RNN as the NN architecture.

My problem has continuous action space. How can I use RNN+DDPG together? (or any other continuous action schemes)

Answer 1

[Emmanouil Tzorakoleftherakis]

The ability to create LSTM policies with DDPG is available starting in R2021a.

Hope that helps

[laha_M]

Thanks, Emmanouil. Is there any exmple of similar kind?

[Emmanouil Tzorakoleftherakis]

See this example in R2021a. Beyond that, see some short short ones that show how to set up the neural nets here.

[laha_M]

Thanks, Emmanouil.