question during training & its application #92
Description
Great work!
I'm planning to apply the model to a real robot, and I have some question regarding.
When training, I notice that the result wasn't converge fast. I'm wondering if it's right to change some of the reward function to encourage it to converge fast, I didn't get good result by doing so. And it's there any effective way to get a good result?
I intend to apply it to a small Biped Robot. I could imagine that could be a sim2real gap concerning lidar msg stability and intense Oscillation of output msg. May I ask if there is any tips about this?