Improve GCS graphviz #21530

bernhardpg · 2024-06-06T12:40:20Z

Adds a few features to the GCS Graphviz code when a result is passed, and additionally make it possible to turn on/off the visualization of different quantities.j

Changes:

show_vars: Decide to show vertex var results
show_costs: Decide to show costs AND also plot vertex costs (these were not currently shown)
show_flows: Plots the flow values as \phi = .. and also dims the edge based on the flow value
active_path: If provided, the graph will show the path in red on the graph. This is useful when debugging workflows with multiple different paths through the graph, to quickly be able to show a path in the graph.

This change is

RussTedrake

Reviewed 3 of 3 files at r1, all commit messages.
Reviewable status: 2 unresolved discussions, needs platform reviewer assigned, needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

geometry/optimization/graph_of_convex_sets.cc line 483 at r1 (raw file):

      graphviz << ", style=\"dashed\"";
      graphviz << "];\n";
    }

we need to confirm this functionality in the GraphViz test in graph_of_convex_sets_test

bindings/pydrake/geometry/geometry_py_optimization.cc line 882 at r1 (raw file):

            py::arg("precision") = 3, py::arg("scientific") = false,
            py::arg("show_vars") = true, py::arg("show_costs") = true,
            py::arg("show_flows") = true, py::arg("active_path") = std::nullopt,

we need the py test to call these values (to make sure they are spelled correctly).

ggould-tri

Assigning +@RussTedrake on this since he's already commenting here. However, Russ is out for the next week; if you would like a quicker turnaround we can assign a different reviewer.

Reviewable status: 2 unresolved discussions, LGTM missing from assignee RussTedrake(platform), needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

ggould-tri

One note on argument design below.

Reviewed 3 of 3 files at r1, 2 of 2 files at r2, all commit messages.
Reviewable status: 3 unresolved discussions, LGTM missing from assignee RussTedrake(platform), needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

geometry/optimization/graph_of_convex_sets.h line 594 at r2 (raw file):

      bool show_vars = true, bool show_costs = true, bool show_flows = true,
      const std::optional<std::vector<const Edge*>>& active_path =
          std::nullopt) const;

minor: With the large number of optional boolean arguments here, combined with C++'s lack of keyword arguments, the likelihood of user error -- setting the wrong parameter -- is very high (the /*foo=*/ comments in the test are indicative of this).

Per https://drake.mit.edu/styleguide/cppguide.html#Function_Argument_Comments , "Consider changing the function signature to replace a bool argument with an enum argument" or "For functions that have several configuration options, consider defining a single class or struct to hold all the options"

RussTedrake

pending CI. I'll wait for you to respond to Grant's suggestion before passing to platform review.

Reviewed 2 of 2 files at r2, all commit messages.
Reviewable status: 1 unresolved discussion, needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

…e GCS

bernhardpg

All comments addressed.

@RussTedrake Two tests are currently failing. One test is failing because of linting issues. How do you typically go about debugging this?

The other test is failing because of what I believe is a bug somewhere in GcsTrajectoryOptimization. We are getting the following error:

Traceback (most recent call last):
  File "/tmp/bazel-working-directory/drake/bazel-out/k8-opt/bin/bindings/pydrake/planning/py/trajectory\_optimization\_test.runfiles/drake/bindings/pydrake/planning/test/trajectory\_optimization\_test.py", line 589, in test\_gcs\_trajectory\_optimization\_2d
    gcs.GetGraphvizString(result=result,
ValueError: GetVariableValue: v\_ell0 is not captured by the variable\_index map.

which is caused by this line, where are we are now also showing the vertex costs (which we were not before).

If this is indeed a bug with GcsTrajectoryOptimization I suggest changing the failing test with graphviz_options.show_costs = False so we can merge this PR and put the bug fix (which is not really related to this PR) in its own PR. Alternatively, we can wait with merging this PR until the bug is addressed in a separate PR.

Reviewable status: 1 unresolved discussion, needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

geometry/optimization/graph_of_convex_sets.h line 594 at r2 (raw file):

Previously, ggould-tri wrote…

minor: With the large number of optional boolean arguments here, combined with C++'s lack of keyword arguments, the likelihood of user error -- setting the wrong parameter -- is very high (the /*foo=*/ comments in the test are indicative of this).

Per https://drake.mit.edu/styleguide/cppguide.html#Function_Argument_Comments , "Consider changing the function signature to replace a bool argument with an enum argument" or "For functions that have several configuration options, consider defining a single class or struct to hold all the options"

Agreed. Done.

geometry/optimization/graph_of_convex_sets.cc line 483 at r1 (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

we need to confirm this functionality in the GraphViz test in graph_of_convex_sets_test

Done.

bindings/pydrake/geometry/geometry_py_optimization.cc line 882 at r1 (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

we need the py test to call these values (to make sure they are spelled correctly).

Done.

RussTedrake · 2024-07-03T12:56:29Z

One test is failing because of linting issues. How do you typically go about debugging this?

You should be running the tests locally. bazel test //.... If you want to run the linters specifically, then use bazel test --config=lint //... . The error messages often tell you how to fix it.

Taking a look at the getvariable error

RussTedrake

Reviewable status: 2 unresolved discussions, needs at least two assigned reviewers, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits), missing label for release notes (waiting on @bernhardpg)

geometry/optimization/graph_of_convex_sets.h line 594 at r2 (raw file):

Previously, bernhardpg wrote…

Agreed. Done.

Ah. This is now a breaking change. Technically GCS is still marked as experimental, so we can make breaking changes to the API. But it would be kinder/better to also keep the original function signature, but implement that as a one-liner that uses this entry point. Then you can deprecate the original entry point for eventual removal.

geometry/optimization/graph_of_convex_sets.cc line 976 at r3 (raw file):

  }

  auto IncludesCurrentTranscription =

you've got some strange commit history here. this was introduced in a commit which landed last night.

RussTedrake · 2024-07-03T13:03:21Z

geometry/optimization/graph_of_convex_sets.h line 594 at r2 (raw file):

Previously, RussTedrake (Russ Tedrake) wrote…

Ah. This is now a breaking change. Technically GCS is still marked as experimental, so we can make breaking changes to the API. But it would be kinder/better to also keep the original function signature, but implement that as a one-liner that uses this entry point. Then you can deprecate the original entry point for eventual removal.

https://drake.mit.edu/doxygen_cxx/drake__deprecated_8h.html#a847d827d084f28e4ae1788a0ae75be96
(btw -- the strategy then would be to leave the original tests in tact behind a gcc diagnostic push/pop which you can find in other unit tests with deprecation, and simply duplicate the tests with the change to the new API. when the deprecation is removed, then we remove the tests as well)

RussTedrake · 2024-07-03T13:05:29Z

bindings/pydrake/geometry/test/optimization_test.py line 819 at r3 (raw file):

            "source",
            spp.GetGraphvizString(
                result=result, options=graphviz_options, active_path=[edge0]

actually there is a way to accept the struct fields as arguments here. if we did that, then python users wouldn't have to change their code. in that case, I think we could avoid the deprecation completely (I suspect almost all users of the GetGraphVizString are doing that from python).

RussTedrake · 2024-07-03T13:14:27Z

bindings/pydrake/geometry/geometry_py_optimization.cc line 722 at r3 (raw file):

    const auto &cls_doc = doc.GcsGraphvizOptions;
    py::class_<GcsGraphvizOptions> gcs_options(m, "GcsGraphvizOptions");
    gcs_options.def(py::init<>())

consider the use of ParamInit (e.g. as used in the ContactVisualizerParams). I know the other structs here don't use them yet, but perhaps they should...?

But i think it does not actually make the original identical syntax go through.

So we probably should deprecate the python bindings, too. I would be ok with deprecating only the python bindings, and letting the c++ be a breaking change.

bernhardpg

Addressed all your comments @SeanCurtis-TRI , except for the discussion commit message which is still unresolved.

Reviewable status: 3 unresolved discussions, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits)

-- commits line 41 at r4:

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

What you propose is one solution. I'd be equally content if the author would write their own take on the message and post it in this discussion. I'd be equally content if the feature reviewer were to do so. I'd be happy to use it in squashing, but I'd rather not have to come up with the message myself.

That said, it seems reasonable that the current PR title and summary seems consistent with what the commit message should be. (Possibly with the added note about C++ API breakage).

So what should I do here?

I think the commit message should be something like:
Extend the GraphOfConvexSets Graphviz functionality.

Should the commit also include a detailed commit body?

geometry/optimization/graph_of_convex_sets.h line 130 at r4 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

BTW On all of these comments, the line wrapping is incredibly odd. Typically, lines fill out as much as they can before wrapping. This isn't strictly a defect, but it will almost definitely be touched by someone else in the future. Perhaps do it now?

Done.

geometry/optimization/graph_of_convex_sets.h line 130 at r4 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

nit: You might consider making this serializable (it'll make it easier to bind and allow us to include this set of options in yaml files). It would look something like:
#include "drake/common/name_value.h"

struct GcsGraphvizOptions {
  template <typename Archive>
  void Serialize(Archive* a) {
    a->Visit(DRAKE_NVP(show_slacks));
    a->Visit(DRAKE_NVP(show_vars));
    ... etc
  }
  ...
};
You can find lots of examples in the structs defined in any *_config.h file.

Done.

geometry/optimization/graph_of_convex_sets.h line 141 at r4 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

BTW Are you worried about an expected edge not appearing because it has zero flow? Might it not confuse a user? An edge not being drawn can mean two things introducing ambiguity.

Good point. Solved by showing edges with zero flow as almost fully transparent instead of invisible.

geometry/optimization/test/graph_of_convex_sets_test.cc line 2970 at r4 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

BTW This test is fine, but it would be a bit stronger if you showed that there were no "color" if you don't pass a path and that, for the options values, it is only adding the path that introduces "color=".

Done.

bindings/pydrake/geometry/geometry_py_optimization.cc line 721 at r4 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

BTW If you added the ability to serialize GcsGraphvixOptions then you can use this paradigm:

  {
    using Class = GcsGraphvizOptions;
    constexpr auto& cls_doc = doc.GcsGraphvizOptions;
    py::class_<Class> cls(m, "GcsGraphvizOptions", cls_doc.doc);
    cls.def(ParamInit<Class>());
    DefAttributesUsingSerialize(&cls, cls_doc);
    DefReprUsingSerialize(&cls);
    DefCopyAndDeepCopy(&cls);
  }

This also future proofs it against new properties.

Done.

RussTedrake

@SeanCurtis-TRI -- this one is waiting for you, too.

Reviewed 3 of 5 files at r5, 2 of 2 files at r6, all commit messages.
Reviewable status: 1 unresolved discussion, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

-- commits line 41 at r4:

Previously, bernhardpg wrote…

So what should I do here?

I think the commit message should be something like:
Extend the GraphOfConvexSets Graphviz functionality.

Should the commit also include a detailed commit body?

Normally a few sentences of description would be appropriate for the commit body, if they help to explain.

But perhaps a good guiding principle is thinking about the fact that the monthly release notes author needs to be able to crawl through and understand/catalog all of the commits to produce e.g.: https://drake.mit.edu/release_notes/v1.30.0.html .
That doesn't require much text, but does require clarity.

Cleaning documentation and struct ordering.

SeanCurtis-TRI

+(release notes: feature)

(Note: in addition to the breaking change, this also includes a new feature.)

Reviewed 3 of 5 files at r5, 2 of 2 files at r6.
Reviewable status: 3 unresolved discussions, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

-- commits line 41 at r4:

Previously, RussTedrake (Russ Tedrake) wrote…

Normally a few sentences of description would be appropriate for the commit body, if they help to explain.

But perhaps a good guiding principle is thinking about the fact that the monthly release notes author needs to be able to crawl through and understand/catalog all of the commits to produce e.g.: https://drake.mit.edu/release_notes/v1.30.0.html .
That doesn't require much text, but does require clarity.

@bernhardpg I'd vote that we simply use the PR title and summary as the commit message. If you agree with that, I can copy and paste it when we squash merge thing.

If you want to take action, you can squash it all locally and edit the commit message yourself before force pushing it again.

geometry/optimization/graph_of_convex_sets.h line 130 at r4 (raw file):

Previously, bernhardpg wrote…

Done.

The pattern used elsewhere (and required by the style guide) would have the Serialize() function come before the declaration of the members.

I recognize that you've matched the previous struct in this file, but, in this case, that's a bad example.

geometry/optimization/graph_of_convex_sets.h line 136 at r6 (raw file):

  /** Determines whether the solution values for decision variables in each set
   * are shown. */

nit: The styleguide requests consistent formatting of comments in a file. I'm betting your IDE injected leading * on these comments.

geometry/optimization/graph_of_convex_sets.h line 670 at r6 (raw file):

  void ClearAllPhiConstraints();

  /**

nit: New line injected here.

geometry/optimization/test/graph_of_convex_sets_test.cc line 2905 at r6 (raw file):

GTEST_TEST(GcsGraphvizOptionsTest, Serialize) {
  GcsGraphvizOptions options;

FYI You may not be aware of designated initializers, but they are fun:

const GcsGraphvizOptions options {
    .show_slacks = false,
    .show_vars = false,
    .show_flows = false,
    .show_costs = false,
    .scientific = true,
    .precision = 5};

Gives you the benefit of defining values with clear semantics at the same time you construct without needing to leave a mutable instance flapping around.

(This is just informational, what you have here isn't defective.)

SeanCurtis-TRI

Reviewed 1 of 1 files at r7, 1 of 1 files at r8.
Reviewable status: 1 unresolved discussion, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

geometry/optimization/graph_of_convex_sets.h line 136 at r6 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

nit: The styleguide requests consistent formatting of comments in a file. I'm betting your IDE injected leading * on these comments.

I submitted a commit to fix this.

geometry/optimization/graph_of_convex_sets.h line 670 at r6 (raw file):

Previously, SeanCurtis-TRI (Sean Curtis) wrote…

nit: New line injected here.

I submitted a commit to fix this.

SeanCurtis-TRI

Reviewed all commit messages.
Reviewable status: 1 unresolved discussion, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

jwnimmer-tri · 2024-07-15T14:32:29Z

@drake-jenkins-bot mac-arm-ventura-clang-bazel-experimental-release please

SeanCurtis-TRI

Reviewed 1 of 1 files at r9, all commit messages.
Reviewable status: 1 unresolved discussion, LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

RussTedrake

Reviewed 1 of 1 files at r7, 1 of 1 files at r9, all commit messages.
Reviewable status: LGTM missing from assignee SeanCurtis-TRI(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

SeanCurtis-TRI

@bernhardpg This has been ready to merge for a bit. I'm going to squash it with your PR summary as the commit message. This is your last change to weigh in.

Reviewable status: commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

bernhardpg

Sorry for the slow response. That sounds good to me!

Reviewable status: commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

SeanCurtis-TRI

No worries. You weren't alone. :)

Reviewable status: commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on @bernhardpg)

SeanCurtis-TRI · 2024-07-23T13:22:55Z

+(status: squashing now)

… restriction **This is a breaking change** (Note that GCS is still marked experimental) During the review of RobotLocomotion#21530, we realized that cost slack variables were not being written to the MathematicalProgramResult in a consistent way between SolveShortestPath and SolveConvexRestriction. This PR addresses this by: 1) The "ell" slack variables associated with the AddCost methods are not returned to the user. Users should call vertex->GetSolutionCost() or edge->GetSolutionCost() to retrieve them. SolveConvexRestriction doesn't use ell, and it's an implementation detail that shouldn't have ever been exposed to the user. 2) GetSolutionCost and GetSolution are now more robust... if the placeholder variables are not in the MathematicalProgramResult, they return the implied values. 3) SolveConvexRestriction no longer tries to write NaN values for the unused vertices on every solve. The more robust GetSolution methods make this reasonable, and now that we've been generating large graphs this has become more important. Users may need to change calls from e.g. `result->GetSolution(v->x())` to `v->GetSolution(result)`. 4) The artificial `graphviz_options.show_costs = False` that was needed in `trajectory_optimization_test.py` to cover up this issue is now safely removed. Resolves RobotLocomotion#20443.

… restriction (#21833) **This is a breaking change** (Note that GCS is still marked experimental) During the review of #21530, we realized that cost slack variables were not being written to the MathematicalProgramResult in a consistent way between SolveShortestPath and SolveConvexRestriction. This PR addresses this by: 1) The "ell" slack variables associated with the AddCost methods are not returned to the user. Users should call vertex->GetSolutionCost() or edge->GetSolutionCost() to retrieve them. SolveConvexRestriction doesn't use ell, and it's an implementation detail that shouldn't have ever been exposed to the user. 2) GetSolutionCost and GetSolution are now more robust... if the placeholder variables are not in the MathematicalProgramResult, they return the implied values. 3) SolveConvexRestriction no longer tries to write NaN values for the unused vertices on every solve. The more robust GetSolution methods make this reasonable, and now that we've been generating large graphs this has become more important. Users may need to change calls from e.g. `result->GetSolution(v->x())` to `v->GetSolution(result)`. 4) The artificial `graphviz_options.show_costs = False` that was needed in `trajectory_optimization_test.py` to cover up this issue is now safely removed. Resolves #20443.

bernhardpg requested a review from RussTedrake June 6, 2024 12:40

RussTedrake reviewed Jun 7, 2024

View reviewed changes

ggould-tri added the component: planning and control Optimization-based planning and control, and search- and sampling-based planning label Jun 11, 2024

ggould-tri assigned RussTedrake Jun 11, 2024

ggould-tri reviewed Jun 11, 2024

View reviewed changes

ggould-tri reviewed Jun 12, 2024

View reviewed changes

RussTedrake reviewed Jun 16, 2024

View reviewed changes

bernhardpg added 13 commits July 3, 2024 11:28

Print vertex costs

ec4c49c

Write simple code for plotting flows (with color weighted by flow)

4140f17

Make the GCS graphviz more modular

29d5639

Fix bug in graphviz code

abb0f7e

Pass in optional argument 'active_paths' that highlights a path in th…

d6c7fde

…e GCS

Move arguments to function call

60e32ff

Fix minor formatting issues

fa2ecbe

Add c++ tests

a783154

Add python tests

b2a9067

Update test

6d17075

Replace bool arguments with a struct

949dda8

Update tests

7a1ea80

Test setting the struct values through the python bindings

ad0ee7b

bernhardpg force-pushed the improve-gcs-graphviz branch from a042e2e to ad0ee7b Compare July 3, 2024 09:41

Fix linting problems

38c8f1d

bernhardpg commented Jul 3, 2024

View reviewed changes

RussTedrake reviewed Jul 3, 2024

View reviewed changes

bernhardpg added 2 commits July 4, 2024 17:10

Fix linting errors

fdec9d8

Make it possible to pass GraphvizOptions as kwargs in python

2f6f4e2

bernhardpg added 4 commits July 10, 2024 14:28

Test serialization

79446f6

Show edges with zero flow as almost invisible instead of fully invisible

20af62f

Make test stronger by testing for absence of path color etc

762d779

Simplify code according to PR feedback

e646a23

bernhardpg commented Jul 10, 2024

View reviewed changes

bernhardpg added 2 commits July 10, 2024 15:34

Fix non-grammatical comment

e3d3cdb

Fix linting errors

2e36b3d

RussTedrake reviewed Jul 13, 2024

View reviewed changes

SeanCurtis-TRI added 2 commits July 15, 2024 05:42

Formatting clean up

46c6ce2

Cleaning documentation and struct ordering.

Use designated initializer in bindings

751ae8b

SeanCurtis-TRI added the release notes: feature This pull request contains a new feature label Jul 15, 2024

SeanCurtis-TRI reviewed Jul 15, 2024

View reviewed changes

Undo added lint

ced2a06

SeanCurtis-TRI reviewed Jul 15, 2024

View reviewed changes

RussTedrake reviewed Jul 16, 2024

View reviewed changes

SeanCurtis-TRI reviewed Jul 22, 2024

View reviewed changes

bernhardpg commented Jul 23, 2024

View reviewed changes

SeanCurtis-TRI reviewed Jul 23, 2024

View reviewed changes

SeanCurtis-TRI added the status: squashing now https://drake.mit.edu/reviewable.html#curated-commits label Jul 23, 2024

SeanCurtis-TRI merged commit 148bc2d into master Jul 23, 2024
9 checks passed

SeanCurtis-TRI deleted the improve-gcs-graphviz branch July 23, 2024 13:24

RussTedrake mentioned this pull request Aug 22, 2024

[GCS] Resolve inconsistencies in results from shortest path vs convex restriction #21833

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve GCS graphviz #21530

Improve GCS graphviz #21530

bernhardpg commented Jun 6, 2024 •

edited by sherm1

Loading

RussTedrake left a comment

ggould-tri left a comment

ggould-tri left a comment

RussTedrake left a comment

bernhardpg left a comment

RussTedrake commented Jul 3, 2024

RussTedrake left a comment

RussTedrake commented Jul 3, 2024

RussTedrake commented Jul 3, 2024

RussTedrake commented Jul 3, 2024

bernhardpg left a comment

RussTedrake left a comment

SeanCurtis-TRI left a comment

SeanCurtis-TRI left a comment

SeanCurtis-TRI left a comment

jwnimmer-tri commented Jul 15, 2024

SeanCurtis-TRI left a comment

RussTedrake left a comment

SeanCurtis-TRI left a comment

bernhardpg left a comment

SeanCurtis-TRI left a comment

SeanCurtis-TRI commented Jul 23, 2024

Improve GCS graphviz #21530

Improve GCS graphviz #21530

Conversation

bernhardpg commented Jun 6, 2024 • edited by sherm1 Loading

RussTedrake left a comment

Choose a reason for hiding this comment

ggould-tri left a comment

Choose a reason for hiding this comment

ggould-tri left a comment

Choose a reason for hiding this comment

RussTedrake left a comment

Choose a reason for hiding this comment

bernhardpg left a comment

Choose a reason for hiding this comment

RussTedrake commented Jul 3, 2024

RussTedrake left a comment

Choose a reason for hiding this comment

RussTedrake commented Jul 3, 2024

RussTedrake commented Jul 3, 2024

RussTedrake commented Jul 3, 2024

bernhardpg left a comment

Choose a reason for hiding this comment

RussTedrake left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

jwnimmer-tri commented Jul 15, 2024

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

RussTedrake left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

bernhardpg left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI left a comment

Choose a reason for hiding this comment

SeanCurtis-TRI commented Jul 23, 2024

bernhardpg commented Jun 6, 2024 •

edited by sherm1

Loading