Add Vizard visualization support for MuJoCo multi-body scenes by ReeceHumphreys · Pull Request #1287 · AVSLab/basilisk

ReeceHumphreys · 2026-02-24T17:38:02Z

Tickets addressed: Closes Support MuJoCo message passthrough to Vizard #1122
Review: By commit
Merge strategy: Merge (no squash)

Description

TODO

Verification

TODO

Documentation

TODO

Future work

TODO

ReeceHumphreys · 2026-02-24T17:46:50Z

@juan-g-bonilla definitely want your feedback on this as the resident MuJoCo guy. Want to make sure not doing anything dumb here haha.

enableUnityVisualization now accepts MJScene objects in scList.

juan-g-bonilla · 2026-02-24T17:52:33Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJScene.swg

+struct MJGeomInfo {
+    std::string bodyName;
+    int type;
+    std::vector<double> size;
+    std::vector<double> pos;
+    std::vector<double> quat;
+    std::vector<double> rgba;
+};


This is interesting, you shouldn't have to define this here since it's already in MJScene.h. Duplicating the definition might lead to divergence.

juan-g-bonilla · 2026-02-24T17:53:47Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJSpec.h

+    std::vector<double> size;  ///< Size parameters (3 elements).
+    std::vector<double> pos;   ///< Position in body frame (3 elements).
+    std::vector<double> quat;  ///< Quaternion in body frame (4 elements: w, x, y, z).
+    std::vector<double> rgba;  ///< Color (4 elements, 0-1 range).


Any chance to make these std::array? They are fixed-size after all.

juan-g-bonilla · 2026-02-24T17:56:12Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJSpec.h

+    std::vector<double> size;  ///< Size parameters (3 elements).
+    std::vector<double> pos;   ///< Position in body frame (3 elements).
+    std::vector<double> quat;  ///< Quaternion in body frame (4 elements: w, x, y, z).
+    std::vector<double> rgba;  ///< Color (4 elements, 0-1 range).


If not changed to std::array, consider initializing to their correct size:

std::vector<double> size(3);

juan-g-bonilla · 2026-02-24T18:08:30Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJSpec.cpp

+        info.size = {m->geom_size[i * 3], m->geom_size[i * 3 + 1], m->geom_size[i * 3 + 2]};
+        info.pos = {m->geom_pos[i * 3], m->geom_pos[i * 3 + 1], m->geom_pos[i * 3 + 2]};
+        info.quat = {m->geom_quat[i * 4], m->geom_quat[i * 4 + 1],
+                     m->geom_quat[i * 4 + 2], m->geom_quat[i * 4 + 3]};
+        info.rgba = {static_cast<double>(m->geom_rgba[i * 4]),
+                     static_cast<double>(m->geom_rgba[i * 4 + 1]),
+                     static_cast<double>(m->geom_rgba[i * 4 + 2]),
+                     static_cast<double>(m->geom_rgba[i * 4 + 3])};


This is just preference, but I prefer algorithm-oriented style to these operations:

std::copy_n(m->geom_size + i * 3, 3, std::begin(info.size));

if you use std::array,

or:

std::copy_n(m->geom_size + i * 3, 3, std::back_inserter(info.size));

if you use std::vector.

Most of mujoco code is written with algorithms.

And:

std::transform(m->geom_rgba + i * 4, m->geom_size + (i+1) * 4, std::begin(info.rgba), [](float v) { return static_cast<double>(v); });

juan-g-bonilla · 2026-02-24T18:08:45Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJSpec.cpp

+                     static_cast<double>(m->geom_rgba[i * 4 + 2]),
+                     static_cast<double>(m->geom_rgba[i * 4 + 3])};
+
+        geoms.push_back(std::move(info));


nice std::move

If you want, you could do in the beginning of the loop:

for (int i = 0; i < m->ngeom; i++) { int bodyId = m->geom_bodyid[i]; if (bodyId == 0) continue; auto& info = geoms.emplace_back();

and eliminate the push back at the end (plus a spurious allocation of MJGeomInfo at the start of the loop for the case bodyId == 0.

juan-g-bonilla · 2026-02-24T18:14:40Z

src/simulation/mujocoDynamics/_GeneralModuleFiles/MJSpec.h

+ */
+struct MJGeomInfo {
+    std::string bodyName; ///< Name of the body this geom belongs to.
+    int type;             ///< MuJoCo geom type (e.g. mjGEOM_BOX=6, mjGEOM_SPHERE=2, mjGEOM_CYLINDER=5, etc.).


Probably good to initialize to int type = -1. Don't like garbage by default.

Actually mjGEOM_NONE would be the more consistent initial value here to follow the mujoco convention.

juan-g-bonilla · 2026-02-24T18:15:56Z

src/tests/data/hypso1.tle

+HYPSO 1
+1 51053U 22002BX  25279.47924866  .00028852  00000-0  56053-3 0  9998
+2 51053  97.3197 349.0390 0004257 136.0807 224.0782 15.47578936207406


sorry to be like this, but these files should probably be added on their own commit since they're not related to "Add body and geom introspection API to MJScene and MJSpec"

Actually, this files might not be related to the PR at all haha

juan-g-bonilla · 2026-02-24T18:24:24Z

src/utilities/vizSupport.py

+            # Position offset in body frame
+            offset = list(geom.pos)
+
+            # Convert quaternion (w, x, y, z) to 3-2-1 Euler angles (z, y, x)


Consider using EP2Euler321 from RigidBodyKinematics.py (but double check that function is doing what you expect!).

juan-g-bonilla · 2026-02-24T18:26:41Z

src/utilities/vizSupport.py

+            rgba = list(geom.rgba)
+            color = [int(c * 255) for c in rgba]


can you do?:

color = [int(c * 255) for c in geom.rgba]

juan-g-bonilla · 2026-02-24T18:30:59Z

src/utilities/vizSupport.py

+                # Walk up the MuJoCo body tree to find the nearest ancestor that 
+                # is in bodyNames


I don't believe that's what's going on?

Ok upon further inspection that is what's going on but simply because bodyNames retrieves all bodies, so find the nearest ancestor that is in bodyNames just means get parent body. I think the comment is more confusing than helpful.

juan-g-bonilla · 2026-02-24T18:34:26Z

src/utilities/vizSupport.py

+                if sc.getBody(name).isFree():
+                    hubName = name


MuJoCo scenes support multiple top-level spacecraft. The intended pattern is to support multiple spacecraft by having multiple free bodies in the same scene, not multiple scenes. It'd be great to support that.

juan-g-bonilla · 2026-02-24T18:35:05Z

src/utilities/vizSupport.py

+                # Walk up the MuJoCo body tree to find the nearest ancestor that 
+                # is in bodyNames
+                parentName = sc.getBodyParentName(name)
+                scData.parentSpacecraftName = parentName if parentName else hubName


You can do:

scData.parentSpacecraftName = sc.getBodyParentName(name) or hubName

matter of taste I guess.

juan-g-bonilla · 2026-02-24T18:40:57Z

src/utilities/vizSupport.py

+            # Add the hub body first as the top-level spacecraft
+            scData = vizInterface.VizSpacecraftData()
+            scData.spacecraftName = hubName
+            spacecraftParentName = hubName


I think you actually never get to use spacecraftParentName since you continue out of this loop.

juan-g-bonilla · 2026-02-24T18:45:27Z

src/utilities/vizSupport.py

+            _createCustomModelsFromMJScene(scSim.vizMessenger, sc)
+
+            c += 1
+            continue


If you're feeling up for a bigger refactor (you can just ask Claude or Codex to help you), I would try to split enableUnityVisualization from this kilometric function into several functions:

for sc in scList: # Check if this is an MJScene object if mujocoFound and isinstance(sc, mujoco.MJScene): _handleMJScene(...) else: if isinstance(sc, spacecraft.Spacecraft): _handleSpacecraft(...) else: _handleEffector(...) _handleSubEffectors()

(not necessarily like that, but you get the idea)

juan-g-bonilla

I have been summoned! Great PR. I left a lot of comments because reviewing PRs is my happy place, but it's mostly about how I would write some code rather than any "mistakes". Feel free to accept, ignore, or argue about any of my comments (I love an argument).

ReeceHumphreys self-assigned this Feb 24, 2026

ReeceHumphreys added the enhancement New feature or request label Feb 24, 2026

ReeceHumphreys added this to Basilisk Feb 24, 2026

ReeceHumphreys force-pushed the feature/mujoco-vizard branch from d2bc5fc to e553971 Compare February 24, 2026 17:46

ReeceHumphreys requested a review from juan-g-bonilla February 24, 2026 17:46

ReeceHumphreys added 4 commits February 24, 2026 10:47

Add body and geom introspection API to MJScene and MJSpec

bc20912

Add MJScene support to vizSupport.enableUnityVisualization

48187b6

enableUnityVisualization now accepts MJScene objects in scList.

Add Vizard live stream support to scenarioDeployPanels

72a0f91

Add release snippet

ffa76e5

ReeceHumphreys force-pushed the feature/mujoco-vizard branch from e553971 to ffa76e5 Compare February 24, 2026 17:47

juan-g-bonilla reviewed Feb 24, 2026

View reviewed changes

		# Walk up the MuJoCo body tree to find the nearest ancestor that
		# is in bodyNames

Conversation

ReeceHumphreys commented Feb 24, 2026

Description

Verification

Documentation

Future work

Uh oh!

ReeceHumphreys commented Feb 24, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

juan-g-bonilla Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

juan-g-bonilla Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

juan-g-bonilla left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

juan-g-bonilla Feb 24, 2026 •

edited

Loading

juan-g-bonilla Feb 24, 2026 •

edited

Loading