Closed Zilch123 closed 4 years ago
Solve by changing the reward structure and by tuning the hyperparameters. ~80% accuracy
This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.
Mouse as input rather the keyboard keys. Took velocity of the mouse as current mouse position - Prev Mouse position. VectorActions are the velocity of the mouse.
When it's mouse position as input it learns. While using the mouse velocity doesn't work, "No episode was completed since last summary." For player mode, the velocity is calculated in heuristic and action is given as velocity.
The Velocity of the mouse controls the ball velocity. Deltax, Deltay Usually ranges from -1.0 to 1.0.
Kinematics is enabled and scripting controls the whole game.