Terminal state in inverted_pendulum_env.py about pybullet-gym HOT 3 CLOSED

Hello,
According to the python reference, the precedence of the "not" operator is higher than the precedence of the "and" operator which is higher than the precedence of the "or" operator.
Which means that:
done = (not np.isfinite(ob).all()) or (np.abs(ob[1]) > .2)
is equivalent to:
done = not np.isfinite(ob).all() or np.abs(ob[1]) > .2
ie the line of code in the repo.
Also,

notdone = np.isfinite(ob).all() and (np.abs(ob[1]) <= .2) 
done = not notdone

is equivalent to:
done = not (np.isfinite(ob).all() and (np.abs(ob[1]) <= .2))
not:
done = not np.isfinite(ob).all() and (np.abs(ob[1]) <= .2)
due to the precedence of "not" over "and", the parenthesis are required.

from pybullet-gym.

Hi, thanks for pointing out.

Yes, I agree on both arguments
1.

Which means that:
done = (not np.isfinite(ob).all()) or (np.abs(ob[1]) > .2)
is equivalent to:
done = not np.isfinite(ob).all() or np.abs(ob[1]) > .2

The reason why I retain the parenthesis in (not np.isfinite(ob).all()) and (np.abs(ob[1]) > .2) is to make the logic clear with less confusion.

notdone = np.isfinite(ob).all() and (np.abs(ob[1]) <= .2) 
done = not notdone

is equivalent to:

done = not (np.isfinite(ob).all() and (np.abs(ob[1]) <= .2))

not:

done = not np.isfinite(ob).all() and (np.abs(ob[1]) <= .2)

due to the precedence of "not" over "and", the parenthesis are required.

Thanks for pointing out my previous mistake in step1 but steps from step2 onwards are correct.

done = not (np.isfinite(ob).all() and (np.abs(ob[1]) <= .2))

is the same as my step2.

Thanks!

from pybullet-gym.

To conclude, the logic of

is equivalent to that in the original OpenAI Gym.

from pybullet-gym.