Hello everybody,
now that the team code is published on GitHub (https://github.com/RoboCupJuniorTC/rcj-soccer-sim-2021-robot-code/) we would like to encourage all teams to show how they approached the challenges of the simulated environment which is limited in some ways while at the same time providing cleaner data than sensors would.
I’ll link to two videos TFA-312 have done about theirs to start with and would encourage everybody to ask questions and show their own approaches.
Regards,
David
@Leo I have taken the liberty to link your two YouTube-Videos in this post to kick this off, thank you for sharing such a detailed description of your approach.
Ball handling video:
@Leo my best guess as to the behaviour of the ball being reflected of the walls (upper video, ~2min45s) would be that the rotation of the ball plays into that as the walls are flat in the model.
I could also imagine some degree of impact from numerical error as the tick rate of the simulation can have the ball more or less far “inside the wall” when collision is detected (movement is calculated first for a time step, then the updated position is checked for collisions, that’s how the ball was able to go through the wall: enough speed that it was not in collision in the simulation timestep before or after passing through the wall).
I think that the impact of ball rotation is much more significant but have not validated any of these hypotheses.
Regards,
David