ft[sensors]: Gives FullSensor access to all attributes of environment.

[maxspahn]

Adresses #159. Replaces sensing from pybullet through indices by accessing environment items directly. Splits unittest for sensors into separate files.

[c-salmi]

Also was thinking maybe it would be better to change the way mpscenes updates the position here and use the function from pybullet resetBaseVelocity to reset to a non-zero velocity, so we could just read from pybullet directly and still support moveable obstacles. What do you think?

[maxspahn]

Also obstacles with flag movable=True will give wrong sensor data when interacted with I think, right?

That is very much true! Good point. For movable obstacle, we cannot rely on the obstacles itself and we should catch this case.

Also was thinking maybe it would be better to change the way mpscenes updates the position here and use the function from pybullet resetBaseVelocity to reset to a non-zero velocity, so we could just read from pybullet directly and still support moveable obstacles. What do you think?

I agree with that, especially in combination with the movable obstacles issue. After this change in the mpscenes using resetBaseVelocity, position and velocity request should directly use pybullet. However, I would like to keep the access to the other goal and obstacle information through the getattr function.

For now, I suggest to ignore the movable obstacles and update after https://github.com/maxspahn/motion_planning_scenes/issues/56 is solved.

[c-salmi]

Okay, yes that sounds good :+1:

[maxspahn]

I actually realized the changes in mpscenes are so minor (litereally one line) that we can wait for that.

[maxspahn]

That should do it @c-salmi