Values used to calculate reward is also the observation of the agent

[sunchaesk]

Not an error of the program, but one thing I noticed is that the obs: dict contains both the variables and the meters, and the function _compute_reward(), uses the values of the obs dictionary.

So the agent uses the meter to compute the reward but also the same meters are passed into the agent as observation.

I'm not sure but I was thinking this could potentially lead to some problems, maybe biased learning?