You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi @anand-bala, I have been playing around with STL specs in cartpole example from this repo, however I could not reproduce the results shown in paper (https://jdeshmukh.github.io/Papers/iros19.pdf). The issue seems to be with spec definition in \bin\safe_rl\specs : stl.G(stl.F(x_dot< abs(0.01)) & (abs(theta) < 5) & (abs(x) < 0.5) & stl.F(abs(theta) < 1)). Isn't theta in radians?
FYI i have used stl.G(stl.F(abs(x_dot) < 0.01) & (abs(theta) < np.radians(2)) & (abs(x) < 0.5)) and alternatively stl.G(stl.F((abs(x_dot) < 0.01) & (abs(theta) < np.radians(12)) & (abs(x) < 0.5))) - based on textual definition in paper but also standard cart pole spec: stl.G(stl.F(abs(x_dot) < 0.01) & (abs(theta) < np.radians(12)) & (abs(x) < 2.4))
Am I missing something??
Cheers for the interesting work,
The text was updated successfully, but these errors were encountered:
Hi @anand-bala, I have been playing around with STL specs in cartpole example from this repo, however I could not reproduce the results shown in paper (https://jdeshmukh.github.io/Papers/iros19.pdf). The issue seems to be with spec definition in \bin\safe_rl\specs : stl.G(stl.F(x_dot< abs(0.01)) & (abs(theta) < 5) & (abs(x) < 0.5) & stl.F(abs(theta) < 1)). Isn't theta in radians?
FYI i have used stl.G(stl.F(abs(x_dot) < 0.01) & (abs(theta) < np.radians(2)) & (abs(x) < 0.5)) and alternatively stl.G(stl.F((abs(x_dot) < 0.01) & (abs(theta) < np.radians(12)) & (abs(x) < 0.5))) - based on textual definition in paper but also standard cart pole spec: stl.G(stl.F(abs(x_dot) < 0.01) & (abs(theta) < np.radians(12)) & (abs(x) < 2.4))
Am I missing something??
Cheers for the interesting work,
The text was updated successfully, but these errors were encountered: