Simulation view object is invalidated and cannot be used again to call getSharedMetatype

[mindThomas]

Description

I am frequently getting the following error when working with USDs with ActionGraphs:
Simulation view object is invalidated and cannot be used again to call getSharedMetatype
The error message is printed whenever the simulation is running. When this error is printed the ActionGraphs are no longer running.
I have not been able to correlate any changes made to the scene and the occurance of this error message. Sometimes I have just been starting and stopping the simulation a few times and then it suddenly enters this error state.

When the error appears, it is not enough just to restart the simulation. Sometimes it is enough to save the USD and start the simulation again. Other times I have to save and reopen the USD file for it to run without the error message. But why does this happen? I have a project with ActionGraphs only with regular Omnigraph nodes (hence without custom Python code).

Isaac Sim version

5.1.0

Operating System (OS)

Ubuntu 24.04

GPU Name

RTX 5090

GPU Driver and CUDA versions

Driver Version: 580.105.08, CUDA Version: 13.0

Logs

No response

Additional information

No response

[PeterL-NV]

This error occurs when the PhysX simulation view object becomes invalidated between simulation runs, causing ActionGraphs to lose their connection to the physics engine.

Root Cause

The error "Simulation view object is invalidated and cannot be used again to call getSharedMetatype" happens because:

• Physics View Invalidation: When you stop and restart the simulation, PhysX tensor views (simulation views) are destroyed and need to be recreated.
• ActionGraph State: OmniGraph nodes that use physics simulation views hold references that become stale after a timeline stop.
• Inconsistent Initialization: Sometimes the views aren't properly reinitialized when the simulation restarts.

Solutions

1. Ensure Proper Reinitialization in Custom Nodes
   If you have custom ActionGraph nodes accessing physics views, implement proper lifecycle management:

class MyActionGraphNode:
    def __init__(self):
        self._physics_view = None
    
    def on_play(self):
        # Recreate physics view on each play
        self._physics_view = omni.physics.tensors.create_simulation_view("cuda:0")
        self._physics_view.set_subspace_roots("/World")
    
    def on_stop(self):
        # Clear the view reference
        self._physics_view = None
    
    def compute(self, db):
        # Always check if view is valid before use
        if self._physics_view is None:
            return
        
        _Use the view_
        # ...

2. Check for Common Issues

• USD stage modifications: Avoid modifying prims with physics APIs while simulation is running
• Extension load order: Ensure physics extensions load before ActionGraph nodes that use them
• Multiple simulation contexts: Don't create multiple SimulationView instances
• Verify ActionGraph execution order: Physics nodes should run after physics step

The key is ensuring that every time simulation stops and restarts, your physics simulation views are properly recreated rather than reused.

[mindThomas]

Thanks for your prompt reply.

1. I am not using any custom nodes.
2. I have experienced this even for simple scenes with just one ActionGraph with a single "Isaac Read Effort Node" and where I just start and stop the simulation.

How do I ensure that "Physics nodes should run after physics step"? Does the On Playback Tick ensure this?