I am working on training a robot limb with 4 degrees of freedom with LearningAgents.
The limb is trained to reach 3d goal positions with it’s tip. To speed up training i would gradually expand the randomized target goal positions to be close first to finally extend to the maximum reach of the robot limb.
But i fail to implement this, because i cannot find a way to get the actual iteration number during training. I fail to find a class in the plugin that uses or stores this iteration number.
Am i missing something ?
Is it possible to get the number out of the typical PPOTrainer set up ?
Something similar like: PPOTrainer→GetEpisodeNum(); ?
i am using C++ and Blueprints in my project, so i could derive custom classes if needed.
Thank you in advantage.