Open grizzi opened 4 years ago
For the time it took, good job! I was wondering why it is so shaky.
thanks :). I think that this is related to the cost function. This is not penalizing the input (force on the cart), neither the velocity of the pendulum and the cart, just the position of the pendulum and the one of the cart. The code is here
Very cool, is it time to try the double inverted pendulum? https://www.youtube.com/watch?v=ki9BZeugVxc
@JenJenChung I know that work, it is called PILCO and I have done my semester thesis on something similar. I do not know if it would work out of the box :thinking: On monday I have a meeting with Farbod. He has apparently already something implemented that we could use. Lets see, I will keep you updated!
To get an idea of the algorithm I have implemented it and tried on the pole-cart problem. This is the result: https://youtu.be/TQnbzJ7scFs
@marco-tognon-robotics @JenJenChung