Use saved reinforcement learning DDPG agent

9 views (last 30 days)
Sayak Mukherjee
Sayak Mukherjee on 26 Sep 2020
I have saved DDPG agent using the optiopn
rlTrainingOptions.SaveAgentValue = 3000
During the simulations number of agents are saved that have episode value greater than 3000. However when I am trying to use the exact same agent for simulation using the command:
simOptions = rlSimulationOptions('MaxSteps',maxSteps);
experience = sim(env,saved_agent,simOptions);
But i an not getting the exact same response as I got during the training. My variance is 0.5 and my variance decay rate is 1e-4. How to replicate the behavior that I got during training using the same agent

Answers (1)

Emmanouil Tzorakoleftherakis
Please see my response here. In short, the behavior you see during training and after training are not expexted to match 100%.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by