Regarding the Goal orientation

I see that the trifinger_simulation.tasks.move_cube_on_trajectory module provides the get_active_goal function. While working with the function I realised that it returns only the goal positions and not the orientation. In my understanding there needs to be a goal position and orientation of the cube (orientation only relevant in difficulty 4). Am I missing something here?

In the “move cube on trajectory” task, the orientation is not considered, that’s why the goal only contains the positions.

To compute the reward, we are internally using the function from move_cube (which is the task from last year’s challenge) but we use the “levels 1-3”-case, so the orientation is ignored.
You should not need to call this function directly, rather use move_cube_on_trajectory.evaluate_state().

I just added the description of move_cube.evaluate_state to the documentation as it was the easiest way to include the exact equation of the reward function. However, as mentioned above, only the “levels 1-3”-case is relevant for this year’s challenge, so the description for level 4 can be ignored.

I can see now that this is not very clear, the way I wrote it in the documentation. Sorry for the confusion!

1 Like

I just updated the documentation (added the relevant description to move_cube_on_trajectory.evaluate_state and removed move_cube.evaluate_state). I hope that makes it clearer.

Ah, I see. So is difficulty 4 irrelevant for the pre-qual and stage 1 of the competition, or for the whole competition as such?

For the whole competition. The different difficulty levels were used last year but are not relevant for this year’s tasks.

1 Like