[Bug] MjPDController attached to G1 articulation but robot remains fully unresponsive

[mulegan]

Summary

MjPDController attached to G1 articulation but robot remains fully unresponsive

Environment

• Unreal Engine version:5.7release
• OS (Windows / Linux) and version:Windows10 22H2
• URLab plugin version / commit SHA:567cbd907a570b820beb87fbddd69c356a6d86da
• Visual Studio version (Windows only):Visual Studio 2022
• MJCF Path ：RoboJuDo/assets/robots/g1/g1_29dof_rev_1_0.xml

Reproduction steps

1. Added AAMjManagerActor to the level, enabled Bridge Server, and configured Control Subscribe Endpointto tcp://0.0.0.0:5556
2. Imported the G1 MJCF to generate the blueprint, manually added the MjPDControllercomponent in the Blueprint's Components panel, and ensured Enabledand Auto Activateare checked
3. Spawned the G1 instance in the level, confirmed its Articulation Prefixexactly matches the target in the policy: g1_29dof_rev_1_0_C_1
4. Launched PIE, opened the Simulate Widget, and clicked Play (or pressed P) to trigger control logic

Expected behaviour

1. After clicking Play in the Simulate Widget, the widget enters running state, and actuator sliders update dynamically in sync with policy output.
2. G1 responds to policy control and executes corresponding motions..

Actual behaviour

1. Simulate Widget is unresponsive: After clicking Play, the widget remains in Pausedstate. All 29 actuator sliders stay static with no value updates.
2. Persistent collision scale warnings: UE outputs repeated LogPhysics: Warning: Scale3D is (nearly) zeroerrors. Even after manually setting all Geom_Sphere*collision components' Transform Scale to (1,1,1)in the editor, saving the blueprint, and relaunching PIE, the warning persists.
3. Auto-activation has no effect: Confirmed Enabledand Auto Activateare checked for the PDController. Restarting PIE does not change the behavior.

Relevant logs

Before submitting

• I searched existing issues and didn't find a duplicate.
• I'm on a supported Unreal Engine version (5.7 or newer), or I've noted why not in the environment block.
• I reproduced this on the latest main (or noted the commit I'm on).

[jonathanembleyriches]

Hey,

1. The Scale3D is (nearly) zero warnings are a confirmed importer bug, but purely cosmetic. When a geom inherits its size from a class (as the G1’s collision spheres do), the importer bakes a zero scale into the generated Blueprint’s component template, that’s why editing the instance transform doesn’t stick. This only affects the UE-side visual/log noise: the compiled MuJoCo model resolves class-inherited sizes correctly, so physics is unaffected. We’ll track a fix separately.

2. Your policy pipeline is on the old bridge interface, which is outdated and no longer a supported path. Hand-publishing to the control SUB socket (guessed prefix, per-joint keys) predates the current client and is effectively broken for this workflow, the payload format has changed, and a 29-DoF G1 also needs per-joint gains pushed (the built-in Kp=100/Kv=5 defaults won’t hold it). Update to the new typed Python client + policy adapter (URLab_Bridge PR #2). It includes a RoboJuDo→G1 adapter with bundled G1 29-DoF policies that handles prefix detection, control format, gains, and stepping automatically. That should fix the unresponsive robot and the static sliders.

3. Auto Activate is a red herring, controllers aren’t UE-activated components; the controller binds automatically at sim setup, gated only by its Enabled property.

If it still misbehaves after switching, please attach the full LogURLab / LogURLabNet output from PIE start so we can confirm the model compiled and the controller bound.

TL;DR: The robot not responding is your publisher using the old, unsupported bridge interface: switch to the new client/policy adapter (URLab_Bridge PR #2) and it should just work.

EDIT: I have merged the above mentioned PR now into the bridge repo to hopefully prevent this from happening again.

Buzz

[mulegan]

Hi, I'm testing with the latest urlab_bridge from main (including PR #2), and I'm still seeing functional issues with BeyondMimicPolicy despite it appearing to step forward.
The env self-check fails hard with no nonzero dof_pos, and all 29 DoF joints/actuators are reported as not found (e.g. left_hip_pitch_joint not found in art.joints). Key log snippet:

[Registry][robojudo.pipeline] RlLocoMimicPipeline, total: 3
07-17 02:10:16.673 [INFO] [robojudo.pipeline.base_pipeline] Using device: cpu
[Environment] Dynamic import of environment: URLabRoboJuDoEnv
07-17 02:10:16.846 [DEBUG] [robojudo.tools.kinematics] Loaded model from /home/administrator/urLab_Bridge/URLab_Bridge/RoboJuDo/assets/robots/g1/g1_29dof_rev_1_0.xml
07-17 02:10:16.846 [DEBUG] [robojudo.tools.kinematics] [MujocoKinematics] set fk with 29 joints
07-17 02:10:16.847 [INFO] [robojudo.utils.rotation] base set to pos: [0. 0. 0.], quat: [0. 0. 0. 1.]
URLabRoboJuDoEnv: dof joint 'left_hip_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_hip_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_hip_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_hip_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_hip_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_hip_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_knee_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_knee_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_ankle_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_ankle_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_ankle_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_ankle_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_knee_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_knee_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_ankle_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_ankle_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_ankle_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_ankle_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_elbow_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_elbow_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_elbow_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_elbow_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_yaw_joint' not found; ctrl will be dropped for this dof
[Controller] Dynamic import of controller: URLabUiCtrl
[Registry][robojudo.policy] AMOPolicy, total: 1
[Policy] Dynamic import of policy: AMOPolicy
07-17 02:10:16.850 [DEBUG] [robojudo.policy.base_policy] Loading jit from /home/administrator/urLab_Bridge/URLab_Bridge/RoboJuDo/assets/models/g1/amo/amo_jit.pt...
[Registry][robojudo.policy] BeyondMimicPolicy, total: 2
[Policy] Dynamic import of policy: BeyondMimicPolicy
07-17 02:10:16.931 [INFO] [robojudo.policy.beyondmimic_policy] [BeyondMimicPolicy] Using modelmeta as config ...
07-17 02:10:16.932 [INFO] [robojudo.utils.rotation] base set to pos: [-1.26505829e-03 2.29079742e-05 0.00000000e+00], quat: [ 0. 0. -0.07370886 0.9972798 ]
07-17 02:10:16.941 [INFO] [robojudo.policy.beyondmimic_policy] [BeyondMimicPolicy] Using modelmeta as config ...
07-17 02:10:16.945 [INFO] [robojudo.utils.rotation] base set to pos: [ 1.01630437 -0.00272518 0. ], quat: [0. 0. 0.00698378 0.99997561]
07-17 02:10:16.961 [INFO] [robojudo.policy.beyondmimic_policy] [BeyondMimicPolicy] Using modelmeta as config ...
07-17 02:10:16.964 [INFO] [robojudo.utils.rotation] base set to pos: [ 0.83411127 -0.00389521 0. ], quat: [0. 0. 0.07044943 0.99751535]
07-17 02:10:16.965 [DEBUG] [robojudo.tools.dof] [DoF] override default_pos with [-0.1, 0.0, 0.0, 0.3, -0.2, 0.0, -0.1, 0.0, 0.0, 0.3, -0.2, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
07-17 02:10:16.966 [DEBUG] [robojudo.tools.dof] [DoF] override stiffness with [150, 150, 150, 300, 80, 20, 150, 150, 150, 300, 80, 20, 400, 400, 400, 14, 14, 14, 14, 14, 16, 16, 14, 14, 14, 14, 14, 16, 16]
07-17 02:10:16.966 [DEBUG] [robojudo.tools.dof] [DoF] override damping with [2, 2, 2, 4, 2, 1, 2, 2, 2, 4, 2, 1, 15, 15, 15, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1]
07-17 02:10:16.967 [DEBUG] [robojudo.tools.dof] [DoF] override torque_limits with [88, 139, 88, 139, 50, 50, 88, 139, 88, 139, 50, 50, 88, 50, 50, 40, 40, 18, 18, 10, 10, 10, 40, 40, 18, 18, 10, 10, 10]
Beyondmimic Dance_wose: 0step [00:00, ?step/s] URLabRoboJuDoEnv: dof joint 'left_hip_pitch_joint' not found in art.joints; state will be zero for this dof0<?, ?step/s]
URLabRoboJuDoEnv: dof actuator for 'left_hip_pitch_joint' not found; ctrl will be dropped for this dof00:00<?, ?step/s]
URLabRoboJuDoEnv: dof joint 'left_hip_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_hip_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_hip_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_hip_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_knee_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_knee_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_ankle_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_ankle_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_ankle_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_ankle_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_hip_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_hip_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_knee_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_knee_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_ankle_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_ankle_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_ankle_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_ankle_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'waist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'waist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_shoulder_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_shoulder_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_elbow_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_elbow_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'left_wrist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'left_wrist_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_shoulder_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_shoulder_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_elbow_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_elbow_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_roll_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_roll_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_pitch_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_pitch_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: dof joint 'right_wrist_yaw_joint' not found in art.joints; state will be zero for this dof
URLabRoboJuDoEnv: dof actuator for 'right_wrist_yaw_joint' not found; ctrl will be dropped for this dof
URLabRoboJuDoEnv: self-check saw no nonzero dof_pos
07-17 02:10:17.148 [INFO] [robojudo.pipeline.rl_pipeline] Pipeline reset
07-17 02:10:17.153 [INFO] [robojudo.utils.rotation] base set to pos: [0. 0. 0.], quat: [0. 0. 0. 1.]
Beyondmimic Dance_wose: 0step [00:00, ?step/s]
Beyondmimic Violin: 0%| | 0/500 [00:00<?, ?step/s]
Beyondmimic Waltz: 0%| | 0/850 [00:00<?, ?step/s]

[mulegan]

[jonathanembleyriches]

Two fixes incoming:

1. The dashboard connected with a temporary bypass that skipped building the local MuJoCo model, and every joint/actuator name lookup comes from that model. So the policy adapter found none of the 29 joints and dropped all control, which is exactly your log. Fixed in fix(client): version-skew warning + compiled-XML model fallback URLab_Bridge#3: the client now handles the MuJoCo version skew itself and builds the model from the compiled XML when the MJB can't load.
2. The Geom_Sphere* / Scale3D is (nearly) zero warnings: the importer created typeless geoms as the wrong component class and baked a zero scale into the Blueprint template when a geom's size came from a class. Fixed in fix(import): geom default resolution, GLB normals, degenerate-transform guards #77, along with a mesh import regression that produced broken normals on every imported mesh.

Both PRs are up, working on my side (gif below). If you can test them on your setup and let me know, I'll get them merged.

Buzz

[mulegan]

Ran uv run urlab-ui on latest urlab_bridge main (+ PR #2), and the MJB loading block looks like this:
MuJoCo version skew: server=3.10.0 client=3.8.1
MJB load failed (mj_loadModel: failed to load from mjb)
falling back to the compiled-XML model
My pyproject.toml pins mujoco==3.8.1 (matched with mujoco-warp==3.8.1 under [mjlab]), but the UE-side URLab plugin appears to be built against 3.10.0, so the precompiled MJB is rightly rejected.
Offline sim / BeyondMimic stepping works fine (XML fallback), but the warning says puppet-mode behaviour may still diverge across versions — I haven't tested UE sync yet, so flagging this early.
Quick check:
Is the intention for the WSL side to move up to 3.10.0 (with mujoco-warp bumped in lockstep), or will the UE plugin eventually relock to 3.8.1?
Any known API breaks going 3.8 → 3.10? (Your pyproject.toml comment already called out the mjENBL_MULTICCD trap with mjlab 1.3.0 + warp 3.7.0.1, wondering if 3.10 has similar landmines.)
Fresh logs, just wanted to align on the version skew before I go deeper into puppet mode.

===========================================================================================
administrator@DESKTOP-DO5NEQQ:~/urLab_Bridge/URLab_Bridge$ uv run urlab-ui
07-17 13:27:20.838 [DEBUG] [robojudo] ========== robojudo-1.5.0 init done ==========

MuJoCo version skew: server=3.10.0 client=3.8.1. The MJB binary format is version-locked, so the local model will be built from the compiled XML instead; puppet mode may still behave differently across versions.
MuJoCo version skew: server=3.10.0 client=3.8.1. The MJB binary format is version-locked, so the local model will be built from the compiled XML instead; puppet mode may still behave differently across versions.
WARNING: Model and executable use different MuJoCo version

MJB load failed (mj_loadModel: failed to load from mjb); falling back to the compiled-XML model (server mujoco 3.10.0, client 3.8.1)
MuJoCo version skew: server=3.10.0 client=3.8.1. The MJB binary format is version-locked, so the local model will be built from the compiled XML instead; puppet mode may still behave differently across versions.
WARNING: Model and executable use different MuJoCo version

MJB load failed (mj_loadModel: failed to load from mjb); falling back to the compiled-XML model (server mujoco 3.10.0, client 3.8.1)

[jonathanembleyriches]

Hey,

First, could you please confirm that the recent PRs resolved your main issue and that the policy is now running successfully?

Regarding the Python version, we definitely need to bump pyproject.toml in the bridge to 3.10 (mjlab and mjwarp are already updated; the comments in my toml are just old). I don't foresee any issues with this bump. I'm tracking the dependency update here: #78. Feel free to send a PR if you want to hotfix this yourself sooner!

A bit of background on our versioning and puppet mode:

• Puppet mode: This should work regardless of versions, as all it does is update transforms.
• Bridge vs. UE Plugin: The bridge is currently experimental while we work out some teething/UX issues, which is why it sometimes lags behind the main plugin. Moving forward, the intention is to keep the bridge Python version tied to the UE plugin version

Buzz

[mulegan]

Hey，
Environment

• Host: Win10, UE5.7 (UnrealRoboticsLab main)
• WSL: Ubuntu 24.04, urlab-bridge fix/issue76-model-fallback
• BridgeServer pill is green, no obvious errors on the UE side.

Symptoms

• uv run urlab-ui launches fine, DearPyGui window pops up.
• All five tabs (Scene / Runtime / Camera / Policy / Debug) appear frozen / unresponsive.
• RobotList is empty​ — can't select a robot, can't run any policy.

The policy list itself shows up (e.g. beyondmimic_dance, unitree_12dof appear in the UI LOG), but RobotList stays empty, so I suspect the UE → WSL robot registration / state push isn't coming through.
Questions

1. Is fix/issue76-model-fallback expected to pair with UnrealRoboticsLab main, or should it pin to a specific UE plugin commit? Anything extra needed on the UE5.7 side (e.g. URLabRobot / G1 articulation in level with Publish enabled)?
2. Empty RobotList usually means ZMQ StateEP not receiving, or UE never registered the robot at all — which is it? I don't see a post-handshake robot-list push in the WSL logs.

Can provide UE Output Log (filtered URLab) and full urlab-ui startup log if helpful.

[mulegan]

[jonathanembleyriches]

on top of main you need both URLab-Sim/URLab_Bridge#3 in the bridge repo
and #77 in this repo.

Then please create a new level, add mjmanager and reimoprt the g1.
Then launch the ui dashboard and try again. If it still fails please provide all logs.

[mulegan]

I've strictly followed your instructions step by step:

1. Created a brand-new empty Level to avoid leftover legacy configs.
2. Located and placed the mjmanager Actor into the scene from the Content Browser.
3. Re-imported the G1 asset from the canonical path: RoboJuDo/assets/robots/g1 (inside the RoboJuDo submodule) per the updated workflow.
4. New observations around the freeze:

The UI freezes completely across all tabs immediately after clicking "Play in PIE"​ in the urlab-ui dashboard. Notably, the UI rendered correctly in its entirety prior to triggering PIE. I can occasionally recover by clicking the "Stop"​ button within the urlab-ui interface to terminate PIE playback.
Promising sign:​ The RobotList populated successfully​ immediately before/after the freeze. This confirms the robot registration phase is working correctly—the freeze appears to occur once the simulation loop starts, likely due to a blocking call during the PIE handshake or simulation tick that locks the UI thread.
I’ve attached the full WSL urlab-ui runtime log and filtered UE Output Log (keywords: URLab, mjmanager) to this comment. Let me know if you need any additional context!