EuRoC Forum

[AM-Robotics]

Hello,
I have two questions regarding task 5:
1) In the provided scene-files the puzzle parts in task 5 are always "two-dimensional". Can we assume that also for the competition?
2) Is it possible, that the puzzle parts have to be flipped over by the robot to fit the puzzle? What i mean is: are there cases where a part has to be rotated by 180° around an axis in the table plane?
Thanks

[fschmidt]

1) In the provided scene-files the puzzle parts in task 5 are always "two-dimensional". Can we assume that also for the competition?

yes, you can assume this.

2) Is it possible, that the puzzle parts have to be flipped over by the robot to fit the puzzle?

yes, this is possible and your solution should be prepared to deal with it.

thanks for this clarification.

[red]

I try to sum up all the information provided in the forum concerning the puzzle. Just to verify if i got all the informations correct.

1) all parts are 2 dimensional and composed of equally sized cubes (boxes)
2) the color of the parts is chosen from one of the following: 0000ff - blue, 00ff00 - green, 00ffff - cyan, ff0000 - red, ff00ff - magenta, ffff00 - yellow
3) objects might have the same color but will then be clearly distinguishable because of their shape
4) no touching parts have the same color
5) minimum size of the boxes is 60% (~ 0.03 m) of the current v1 size for task 5: The boxes will not be smaller than the width of the gripper jaws
6) the order of placement is not important
7) the origin of the first box will be in the origin of the part
8) Parts will always lie flat on the table. Nevertheless they will still have a z position offset depending on the size of the object.
9) rotation of parts is possible around all axis (but only in steps of 90 degree) Later on it was stated that only rotations around z-axis are possible. --> i assume here we are talking about the relative_puzzle_part_target_poses, see code below
10) maybe the parts have to be flipped over (rotated by 180 around an inplane axis) --> this is due to the initial position (start_pose) of the parts
11) the puzzle fixture is fixed to the table and cannot be moved
12) the fixed sides of the fixture are always on top of the x-axis and the y-axis in the local frame of the puzzle fixture.

Especially concerning (9) it would be very helpful to get a file for testing.
Up to now both provided puzzles were defined without any rotations:

Code: Select all

relative_puzzle_part_target_poses:
    part...: [..., ..., 0.5*boxsize , 0.0, 0.0, 0.0]

Somewhere (unfortunatelly i can't find it anymore) it was written, that the provided scenes will give us a range of to be expected variations of the tasks. So up to now NO rotations in the relative_puzzle_part_target_poses.

Thanks for your replies!

Edit on 30.10.2014: clarification of my point concerning (9) and (10)

[Robotica]

Hello,

I would like to know whether the objects can have any shape possible (while still composed of small boxes). I am asking this because for instance if one of the puzzle pieces is a cuboid with a square face made of four basic box shapes, it will definitely be too big for the gripper to grasp and to lift it (or turn it if it has been flipped).

Furthermore I would like to ask how many pieces there are in the puzzle. Reading the technical annex, I got the impression there would only be four pieces but task5_2 shows six pieces lying on the table...

Thank you for your help.

[fschmidt]

it will definitely be too big for the gripper to grasp and to lift it

parts won't be too big to grasp them. and they also won't be to heavy to lift them. (compare the "object variety" thread http://euroc.iphpbb3.com/forum/viewtopic.php?nxu=20112706nx57588&f=4&p=33#p33

beside the feasibility to grasp them there is no restriction on the
initial object pose.

)

how many pieces there are in the puzzle.

you are right that the techincal annex suggests in section 3.5.2 that there are only 4 parts. this happened by accident and is also not specified in the "Annex III - Scoring System for Stage I" (EuRoC_Call4Challengers_Annex3.pdf page 12).

the number of parts is limited by the fact that it is a 4x4=16 blocks field, the number of valid colors 6 and the statement given in the "object variety" thread that

objects might have the same color but will then be clearly
distinguishable because of their shape.

[fschmidt]

i agree with all your points with the following additions:

8) Parts will always lie flat on the table. Nevertheless they will still have a z position offset depending on the size of the object.

the initial pose of the puzzle parts will be "anywhere on the table" - the physics engine will force the parts to come to rest on one of their sides. compare part4 in task5_v1 (the origin of the part is 1.5 times the boxsize above the table plane) - if you consider this case as "lying flat on the table" - then yes.
i'd suggest to take no assumptions on the initial orientation of the puzzle parts other than what is given by physics and the fact that they will rest on the table - i.e. the won't be stacked ontop of each other.

10) maybe the parts have to be flipped over (rotated by 180 around an inplane axis)

again: i'd suggest to take no assumptions on the initial orientation of the puzzle parts. -> yes, they might have to be flipped.

9) rotation of parts is possible around all axis (but only in steps of 90 degree) Later on it was stated that only rotations around z-axis are possible. --> i assume here we are talking about the relative_puzzle_part_target_poses

yes, that was about the relative_puzzle_part_target_poses.

Especially concerning (9) it would be very helpful to get a file for testing.
...
Somewhere (unfortunatelly i can't find it anymore) it was written, that the provided scenes will give us a range of to be expected variations of the tasks. So up to now NO rotations in the relative_puzzle_part_target_poses.

we can hardly provide you with enough example scenes to excite all parameters within their allowed ranges. the simulator is setup in a way that challengers can create and test their own scenarios simply be copying & editing the human readable (yaml) text files below /opt/euroc_c2s1/scenes/ within the simservVM or your native setup.

it was not yet specified anywhere but we didn't plan to use the rotational components in the relative_puzzle_part_target_poses as in our opinion those are not of interest to test challengers robotic assembly skills.

so here is a new limitation: relative_puzzle_part_target_poses in task5 will always have all rotational component set to 0.

[red]

Thank you very much for the clarification.
Now it is time for fine tuning

[AM-Robotics]

I have two more questions concerning the puzzle fixture:
1) is the size of the fixture always the same?
2) might there be cases where the puzzle pieces have there initial position on the fixture?
Thanks!

[fschmidt]

in short:

1) is the size of the fixture always the same?

yes.

2) might there be cases where the puzzle pieces have there initial position on the fixture?

no.