Added support for loading the same SDF file multiple times. #2426

liangfok · 2016-05-23T10:36:52Z

Closes #2546.

This PR contains changes necessary to load the same SDF file multiple times.
You can run the demo by executing the following:

$ cd [drake distro]/drake/examples/Cars
$ ../../../build/bin/drake-visualizer &
$ ../../pod-build/bin/multi_car_sim_lcm

Here's what you should see in Drake Visualizer:

The example source code is in [drake distro]/drake/examples/cars/multi_car_sim_lcm.cc.

You can also run the following unit test:

$ cd /drake/drake/pod-build/
$ ctest -VV -R multi_car_sim_lcm

This change is

RussTedrake · 2016-05-23T11:15:00Z

Reviewed 9 of 9 files at r1.
Review status: all files reviewed at latest revision, 5 unresolved discussions.

drake/examples/Cars/CMakeLists.txt, line 38 [r1] (raw file):

  add_dependencies(multi_car_sim drake_lcmtypes lcmtype_agg_hpp)
  target_link_libraries(multi_car_sim drakeCars drakeRBSystem drakeLCMSystem)

add as a test, too?

drake/examples/Cars/multi_car_sim.cpp, line 118 [r1] (raw file):

  rigid_body_sys->penetration_stiffness = 5000.0;
  rigid_body_sys->penetration_damping = rigid_body_sys->penetration_stiffness / 10.0;
  rigid_body_sys->friction_coefficient = 10.0;  // essentially infinite friction

looks like it's already been moved there. obviously zap all of the above before your actual merge.

drake/systems/plants/RigidBodySystem.h, line 181 [r1] (raw file):

  void addRobotFromSDF(const std::string& sdf_filename,
                       const std::string* model_name_postfix,

inserting this in the middle of the arguments will unnecessarily break the api. recommend putting it at the end?

drake/systems/plants/RigidBodyTree.cpp, line 107 [r1] (raw file):

  SortTree();

  std::map<std::string, int> model_name_to_robotnum;

i think that the longer-term fix is probably to purge all occurrences of robotnum which is essentially a UID, in favor of some data interface using pointers (e.g. the RigidBody has a link to the specific model_name string (or object), which no longer has to be unique). We should figure out the best design for the c++ and then I'm happy to help connect it back through to matlab.

this is a good fix for the short term.

drake/systems/plants/RigidBodyTree.h, line 118 [r1] (raw file):

                           DrakeJoint::QUATERNION,
                       std::shared_ptr<RigidBodyFrame> weld_to_frame = nullptr);

again, recommend it as the last argument.

Comments from Reviewable

…e/multi_prius_demo # Conflicts: # drake/examples/Cars/car_simulation.cc # drake/systems/plants/RigidBodySystem.cpp # drake/systems/plants/RigidBodyTreeSDF.cpp

…e/multi_prius_demo # Conflicts: # drake/examples/Cars/car_simulation.cc

…into feature/multi_prius_demo

…pp" to be "multi_car_sim_lcm.cpp".

… to co-exist in the rigid body tree. They are differentiated by the model ID.

liangfok · 2016-06-04T05:53:43Z

I addressed all reviewer comments and also updated this PR based on recent discussions about keeping the model name intact without adding a postfix. A model ID is still being used to differentiate between models within a rigid body tree. A future PR will be issued that removes this ID and switch to using pointers to differentiate between different models.

+@RussTedrake for feature review and +@jwnimmer-tri for platform review

Previously, liangfok (Chien-Liang Fok) wrote…

Added support for loading the same SDF file multiple times.

This is not yet ready to be merged (I quickly hacked in the new feature), but it contains changes necessary to load the same SDF file multiple times.

You can run the demo by executing the following:
$ cd [drake distro]/drake/examples/Cars

$ ../../../build/bin/drake-visualizer &

$ ../../pod-build/bin/multi_car_sim
Here's what you should see in Drake Visualizer:

The example source code is in [drake distro]/drake/examples/cars/multi_car_sim.cpp.

Review status: 0 of 12 files reviewed at latest revision, 5 unresolved discussions.

drake/examples/Cars/CMakeLists.txt, line 38 [r1] (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

add as a test, too?

Done. The new unit test is called "multi_car_sim_lcm". The command for running it is:

$ ctest -VV -R multi_car_sim_lcm

drake/systems/plants/RigidBodySystem.h, line 181 [r1] (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

inserting this in the middle of the arguments will unnecessarily break the api. recommend putting it at the end?

Done. I removed the `model_name_postfix` in r3. Just the model ID is used to differentiate between otherwise identically-named models.

drake/systems/plants/RigidBodyTree.h, line 118 [r1] (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

again, recommend it as the last argument.

Done. I actually removed the `model_name_postfix` in r3. Just the model ID is used to differentiate between otherwise identically-named models.

drake/examples/Cars/multi_car_sim.cpp, line 118 [r1] (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

looks like it's already been moved there. obviously zap all of the above before your actual merge.

Done. I moved it to a new method called `SetRigidBodySystemParameters()`

Comments from Reviewable

liangfok · 2016-06-04T15:34:43Z

Two unit tests are failing: rigid_body_tree_test and load_model_test. I am able to replicate these failures locally and will debug them soon.

Previously, liangfok (Chien-Liang Fok) wrote…

I addressed all reviewer comments and also updated this PR based on recent discussions about keeping the model name intact without adding a postfix. A model ID is still being used to differentiate between models within a rigid body tree. A future PR will be issued that removes this ID and switch to using pointers to differentiate between different models.

+@RussTedrake for feature review and +@jwnimmer-tri for platform review

Review status: 0 of 12 files reviewed at latest revision, 5 unresolved discussions, some commit checks failed.

Comments from Reviewable

…tion instead of returning a nullptr when finding a non-existent link in a rigid body tree.

liangfok · 2016-06-04T17:06:46Z

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/1946/ and is thus ready for review.

Previously, liangfok (Chien-Liang Fok) wrote…

Two unit tests are failing: rigid_body_tree_test and load_model_test. I am able to replicate these failures locally and will debug them soon.

Review status: 0 of 14 files reviewed at latest revision, 5 unresolved discussions.

Comments from Reviewable

jwnimmer-tri · 2016-06-06T18:24:22Z

I started reading this, but TBH I'm not sure this is a direction I want to support. Robot instancing would be trivial with a different RBT design; bolting it onto this design is just more difficult-to-get-right complexity. If there is no immediate need (in other words, if this PR was only for my benefit), I'd rather stack this PR after a general cleanup / redesign of the RB/RBT interfaces. Adding one more hand-crafted underdocumented public integer into the mix doesn't seem worth it, if nobody is chomping at the bit for the feature. If there is a quick need for this, let me know and I can try to find a good compromise.

Previously, liangfok (Chien-Liang Fok) wrote…

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/1946/ and is thus ready for review.

Reviewed 1 of 13 files at r3.
Review status: 1 of 14 files reviewed at latest revision, 5 unresolved discussions.

Comments from Reviewable

sherm1 · 2016-06-06T18:43:14Z

@liangfok this still has status "do not review" -- should that be removed?

liangfok · 2016-06-06T19:06:25Z

@jwnimmer-tri: +1 for overhauling the rigid body tree design. I'm thinking there should be a "robot" abstraction that sits between the rigid body tree and the rigid body abstractions. However, I'm not sure this is enough since how would one identify a particular robot unless there were some unique ID included with each "robot" abstraction?

@sherm1: Oops. I will remove the the "do not merge" label.

amcastro-tri · 2016-06-13T19:21:30Z

Previously, liangfok (Chien-Liang Fok) wrote…

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2041/ and is thus ready for review.

Review status: all files reviewed at latest revision, 33 unresolved discussions.

drake/examples/Cars/car_simulation.h, line 61 [r9] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

What is the intention of the DRAKECARS_EXPORT?

For Windows compatibility. This file is part of a shared library.

Also, what do you think of the design I use in PR #2542 where a more specific system (Atlas) inherits from RigidBodySystem. A similar pattern is followed in the SpringMassSystem. Maybe worth updating the cars example to follow a similar pattern as we migrate towards a System 2.0 design?

I like it but would like to have a more thorough discussion: #2550

I was not proposing to make such a change in this PR though. Just suggesting the idea for when System 2.0 arrives. Thanks for creating #2550!

drake/examples/Cars/car_simulation.h, line 76 [r10] (raw file):

 *
 * The length and width of the terrain lies in the X-Y plane of the world
 * coordinate frame abd has lengths of \p box_width. The depth of the terrain is

abd->and

Comments from Reviewable

…e/multi_prius_demo

liangfok · 2016-06-13T19:35:44Z

I addressed all reviewer comments. PTAL.

Previously, amcastro-tri (Alejandro Castro) wrote…

:lgtm:

Review status: all files reviewed at latest revision, 33 unresolved discussions.

drake/examples/Cars/car_simulation.h, line 61 [r9] (raw file):

Previously, amcastro-tri (Alejandro Castro) wrote…

I was not proposing to make such a change in this PR though. Just suggesting the idea for when System 2.0 arrives. Thanks for creating #2550!

No problem!

drake/examples/Cars/car_simulation.h, line 76 [r10] (raw file):

Previously, amcastro-tri (Alejandro Castro) wrote…

abd->and

Fixed. Thanks. I also updated the text for clarity.

Comments from Reviewable

liangfok · 2016-06-13T21:46:13Z

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2047/ and is thus ready for review.

sherm1 · 2016-06-14T00:04:35Z

Platform review pass done.

Previously, liangfok (Chien-Liang Fok) wrote…

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2047/ and is thus ready for review.

Reviewed 1 of 14 files at r3, 2 of 7 files at r8, 5 of 6 files at r9, 1 of 2 files at r10, 1 of 1 files at r11.
Review status: all files reviewed at latest revision, 30 unresolved discussions.

drake/examples/Cars/car_simulation.h, line 76 [r11] (raw file):

 *
 * The Z-axis of the box matches the Z-axis of the world, i.e., positive Z
 * points up. The length and width of the terrain is \p box_size. The depth of

This still isn't completely clear. You talked about Z but were silent about X and Y. Should be explicit: length and width are aligned with X and Y and have the same dimension box_size; depth is aligned with Z.

drake/examples/Cars/car_simulation.h, line 80 [r11] (raw file):

 *
 * @param[in] rigid_body_tree The rigid body tree to which to add the terrain.
 * @param[in] box_size The length and width of the terrain.

aligned with world's x and y

drake/examples/Cars/car_simulation.h, line 81 [r11] (raw file):

 * @param[in] rigid_body_tree The rigid body tree to which to add the terrain.
 * @param[in] box_size The length and width of the terrain.
 * @param[in] box_depth The depth of the terrain. Note that regardless of how

aligned with world's z

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

Perhaps one source of confusion is the current PR enables the same SDF file to be loaded multiple times. Thus, final_model_id is not equal to the number of models in the SDF file since other SDF files may have been loaded previously.

Please explain that in the code.

drake/systems/plants/RigidBodyTree.h, line 729 [r8] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

(1) If that comment is correct why isn't this method called findBodyIndex()?

Historical reasons; this was the name of the method prior to this PR. Should I rename it to be FindBodyIndex() in this PR?

(2) Why would anyone think a body index is the ID of a model?

The only member variable within a RigidBody that hints at an "ID" is model_id. Thus, I wanted to ensure no one thought this method returned the model_id. Also, see your comment below about how there are two IDs that need to be disambiguated. I suppose changing the method name will clear up all confusion.

Yes, I think you should rename it to FindBodyIndex() in this PR since you are introducing a real "id" in this PR.

Comments from Reviewable

liangfok · 2016-06-14T02:12:00Z

I addressed all reviewer comments. PTAL.
I will add an entry to the CHANGELOG once this PR is ready for merge into master.

Previously, sherm1 (Michael Sherman) wrote…

Platform review pass done.

Review status: 10 of 21 files reviewed at latest revision, 30 unresolved discussions.

drake/examples/Cars/car_simulation.h, line 76 [r11] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

This still isn't completely clear. You talked about Z but were silent about X and Y. Should be explicit: length and width are aligned with X and Y and have the same dimension box_size; depth is aligned with Z.

Done.

drake/examples/Cars/car_simulation.h, line 80 [r11] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

aligned with world's x and y

Done.

drake/examples/Cars/car_simulation.h, line 81 [r11] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

aligned with world's z

Done.

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

Please explain that in the code.

Done. Please see the new Doxygen comments for `RigidBodySystem::addRobotFromSDF()`.

drake/systems/plants/RigidBodyTree.h, line 729 [r8] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

Yes, I think you should rename it to FindBodyIndex() in this PR since you are introducing a real "id" in this PR.

Done.

Comments from Reviewable

sherm1 · 2016-06-14T03:18:26Z

Review pass done -- almost there!

Reviewed 11 of 11 files at r12.
Review status: all files reviewed at latest revision, 28 unresolved discussions.

drake/examples/Cars/car_simulation.h, line 76 [r11] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

Done.

Looks great now, Liang -- thanks!

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

Done. Please see the new Doxygen comments for RigidBodySystem::addRobotFromSDF().

That's a nice addition to the API description! But I was referring to the mysterious implementation code, not the external behavior. Please add some comments explaining what's going on with the various chasing of model lists, counting of models, and arithmetic used finally to assign model ids to models. Maybe this seems obvious to you but your code is depending on a lot of knowledge that resides only in your head at the moment.

drake/systems/plants/RigidBodySystem.h, line 190 [r12] (raw file):

   * @param[in] floating_base_type The type of floating base to use to connect
   * the models within the SDF file to the world.
   * @param[in] welt_to_frame The frame used for connecting the models in the

typo

Comments from Reviewable

liangfok · 2016-06-14T04:53:41Z

I've addressed all reviewer comments. PTAL. There are still at least two things I need to do:

Update the CHANGELOG since this PR changes some APIs.
Add a unit test for RigidBodySystem that verifies the correct implementation of model_id when the same SDF is loaded multiple times and the SDF contains multiple models with at least one joint and one sensor each.

Review status: 19 of 21 files reviewed at latest revision, 27 unresolved discussions.

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

That's a nice addition to the API description! But I was referring to the mysterious implementation code, not the external behavior. Please add some comments explaining what's going on with the various chasing of model lists, counting of models, and arithmetic used finally to assign model ids to models. Maybe this seems obvious to you but your code is depending on a lot of knowledge that resides only in your head at the moment.

Done. I added additional documentation comments to this method.

drake/systems/plants/RigidBodySystem.h, line 190 [r12] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

typo

Fixed.

Comments from Reviewable

liangfok · 2016-06-14T06:00:32Z

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2068/ and is thus ready for review.

Review status: 19 of 21 files reviewed at latest revision, 27 unresolved discussions.

Comments from Reviewable

…e can be loaded multiple times.

liangfok · 2016-06-15T21:00:11Z

I added the unit test for RigidBodySystem and added a CHANGELOG. Thus this PR is ready for another round of review.

Review status: 14 of 26 files reviewed at latest revision, 26 unresolved discussions.

Comments from Reviewable

…e/multi_prius_demo

liangfok · 2016-06-15T22:38:13Z

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2094/ and is thus ready for review.

Review status: 13 of 26 files reviewed at latest revision, 26 unresolved discussions.

Comments from Reviewable

sherm1 · 2016-06-16T00:57:17Z

Liang, did this PR pick up some unrelated commits?

Reviewed 1 of 2 files at r13, 7 of 11 files at r15.
Review status: 21 of 26 files reviewed at latest revision, 27 unresolved discussions.

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, liangfok (Chien-Liang Fok) wrote…

Done. I added additional documentation comments to this method.

Great explanation -- thanks!

drake/systems/plants/RigidBodySystem.cpp, line 878 [r16] (raw file):

        "ID (" +
        std::to_string(final_model_id) + ")");
  }

@liangfok, it's great that you documented this now. As often happens when one tries to explain code to someone else, I see you found several bugs in this code! I think that makes it clear that this needs some unit tests in place to make sure the model ids are being generated correctly before this PR can go in.

Comments from Reviewable

liangfok · 2016-06-16T01:36:33Z

I just scanned through the commits associated with this PR and don't see anything out of the ordinary.

Review status: 21 of 26 files reviewed at latest revision, 27 unresolved discussions.

drake/systems/plants/RigidBodySystem.cpp, line 819 [r5] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

Great explanation -- thanks!

No problem! Thank you for the reviews.

drake/systems/plants/RigidBodySystem.cpp, line 878 [r16] (raw file):

Previously, sherm1 (Michael Sherman) wrote…

@liangfok, it's great that you documented this now. As often happens when one tries to explain code to someone else, I see you found several bugs in this code! I think that makes it clear that this needs some unit tests in place to make sure the model ids are being generated correctly before this PR can go in.

Yes, in fact, I discovered the bugs after I added a unit test called `rigid_body_system_test` (see `drake/systems/plants/test/rigid_body_system/rigid_body_system_test.cc` in this PR). I believe this unit test is sufficient to verify that the model IDs are being correctly set.

Comments from Reviewable

liangfok · 2016-06-16T14:13:38Z

This PR passed https://drake-jenkins.csail.mit.edu/job/experimental/2096/ and is thus ready for review.

Review status: 21 of 26 files reviewed at latest revision, 27 unresolved discussions.

Comments from Reviewable

sherm1 · 2016-06-17T03:12:15Z

ready to merge!

Reviewed 3 of 11 files at r15, 2 of 2 files at r16.
Review status: all files reviewed at latest revision, 28 unresolved discussions.

drake/systems/plants/test/rigid_body_system/rigid_body_system_test.cc, line 120 [r16] (raw file):

  EXPECT_NE(tree->findJoint("joint_2", 2), nullptr);
  EXPECT_NE(tree->findJoint("joint_2", 3), nullptr);
}

BTW, test looks good!

Comments from Reviewable

Added support for loading the same SDF file multiple times.

8bb2b98

liangfok added type: feature request team: software core priority: medium labels May 23, 2016

liangfok self-assigned this May 23, 2016

liangfok added status: do not merge status: do not review labels May 23, 2016

Added constants and comments.

64535eb

Removed commented out code.

1d1d157

liangfok mentioned this pull request May 26, 2016

RigidBodyTree::findLinkId(), RigidBodyTree::findJoint(), and RigidBodyTree::findJointId() Need model_name parameter #2456

Closed

Chien-Liang Fok added 8 commits May 31, 2016 17:31

Merge branch 'master' into feature/multi_prius_demo

aa51876

Added multi_car_sim unit test.

2efdbad

Merge branch 'master' of github.com:RobotLocomotion/drake into featur…

43c4fe2

…e/multi_prius_demo # Conflicts: # drake/examples/Cars/car_simulation.cc # drake/systems/plants/RigidBodySystem.cpp # drake/systems/plants/RigidBodyTreeSDF.cpp

Merge branch 'master' of github.com:RobotLocomotion/drake into featur…

2e32b17

…e/multi_prius_demo # Conflicts: # drake/examples/Cars/car_simulation.cc

Merge branch 'feature/multi_prius_demo' of github.com:liangfok/drake …

694f290

…into feature/multi_prius_demo

Added method SetRigidBodySystemParameters(). Renamed "multi_car_sim.c…

21ab78d

…pp" to be "multi_car_sim_lcm.cpp".

Removed model name postfix. Enable multiple models with the same name…

7e9323e

… to co-exist in the rigid body tree. They are differentiated by the model ID.

Fixed bug that resulted in multi_car_sim_lcm unit test running forever.

c17a62b

liangfok assigned jwnimmer-tri and RussTedrake Jun 4, 2016

Chien-Liang Fok added 2 commits June 4, 2016 08:03

Removed collision element header inclusion due to recent PR revert.

db56e29

Rang clang-format.

36b34c6

Fixed two unit tests that broke due to me switching to throw an excep…

a1ebf50

…tion instead of returning a nullptr when finding a non-existent link in a rigid body tree.

Chien-Liang Fok added 2 commits June 13, 2016 15:30

Merge branch 'master' of github.com:RobotLocomotion/drake into featur…

504ef8a

…e/multi_prius_demo

Clarified comment.

8ea36b0

liangfok mentioned this pull request Jun 13, 2016

Should Cars Example Subclass RigidBodySystem? #2550

Closed

Updated doxygen comments. Renamed findLinkId to be FindBodyIndex.

3b37c98

Fixed some comment typos. Added more comments.

a1d0c9b

Polished comment some more.

ed5cc2d

jwnimmer-tri mentioned this pull request Jun 14, 2016

examples/Cars refactorings #2526

Merged

Chien-Liang Fok added 5 commits June 14, 2016 21:26

Merge branch 'master' into feature/multi_prius_demo

e66b588

Fixed bug counting the number of models in an SDF.

2f36eb2

Ran clang-format.

c7b6f71

Added unit test for rigid body system verifying that the same SDF fil…

7bfeba8

…e can be loaded multiple times.

Updated CHANGELOG.

45f39c5

Merge branch 'master' of github.com:RobotLocomotion/drake into featur…

b4fc078

…e/multi_prius_demo

sherm1 merged commit 835f277 into RobotLocomotion:master Jun 17, 2016

liangfok mentioned this pull request Jun 17, 2016

Add Support for Loading Same URDF Multiple Times Into Same Rigid Body System #2583

Closed

liangfok mentioned this pull request Jul 20, 2016

Adds Support for Loading URDF Multiple Times #2859

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added support for loading the same SDF file multiple times. #2426

Added support for loading the same SDF file multiple times. #2426

liangfok commented May 23, 2016 •

edited

Loading

RussTedrake commented May 23, 2016

liangfok commented Jun 4, 2016

Added support for loading the same SDF file multiple times.

liangfok commented Jun 4, 2016

liangfok commented Jun 4, 2016

jwnimmer-tri commented Jun 6, 2016

sherm1 commented Jun 6, 2016

liangfok commented Jun 6, 2016 •

edited

Loading

amcastro-tri commented Jun 13, 2016

liangfok commented Jun 13, 2016

liangfok commented Jun 13, 2016

sherm1 commented Jun 14, 2016

liangfok commented Jun 14, 2016

sherm1 commented Jun 14, 2016

liangfok commented Jun 14, 2016

liangfok commented Jun 14, 2016

liangfok commented Jun 15, 2016

liangfok commented Jun 15, 2016

sherm1 commented Jun 16, 2016

liangfok commented Jun 16, 2016

liangfok commented Jun 16, 2016

sherm1 commented Jun 17, 2016

Added support for loading the same SDF file multiple times. #2426

Added support for loading the same SDF file multiple times. #2426

Conversation

liangfok commented May 23, 2016 • edited Loading

RussTedrake commented May 23, 2016

liangfok commented Jun 4, 2016

Added support for loading the same SDF file multiple times.

liangfok commented Jun 4, 2016

liangfok commented Jun 4, 2016

jwnimmer-tri commented Jun 6, 2016

sherm1 commented Jun 6, 2016

liangfok commented Jun 6, 2016 • edited Loading

amcastro-tri commented Jun 13, 2016

liangfok commented Jun 13, 2016

liangfok commented Jun 13, 2016

sherm1 commented Jun 14, 2016

liangfok commented Jun 14, 2016

sherm1 commented Jun 14, 2016

liangfok commented Jun 14, 2016

liangfok commented Jun 14, 2016

liangfok commented Jun 15, 2016

liangfok commented Jun 15, 2016

sherm1 commented Jun 16, 2016

liangfok commented Jun 16, 2016

liangfok commented Jun 16, 2016

sherm1 commented Jun 17, 2016

liangfok commented May 23, 2016 •

edited

Loading

liangfok commented Jun 6, 2016 •

edited

Loading