question on pendulum reward function

Open Haichao-Zhang opened this issue 5 years ago • 0 comments

Can you explain why this reward function (-cos(theta)-0.1*sin(theta) ... ) is used for pendulum? https://github.com/WilsonWangTHU/POPLIN/blob/edd8dba50f9049c6164eda774602bef0c299cb51/dmbrl/config/gym_pendulum.py#L104 And why does it need to be different from the original reward function from openai-gym?

Feb 11 '20 01:02 Haichao-Zhang