Is this page helpful?

[isaacsim.core.experimental.primdata] Isaac Sim Core Experimental Prim Data#

Version: 0.3.2

Overview#

The isaacsim.core.experimental.primdata extension provides the default C++ provider implementation for the IPrimDataReader and IPrimDataReaderManager Carbonite interfaces defined in isaacsim.core.experimental.prims. It delivers high-performance, read-only access to prim transform, rigid body, and articulation data from both PhysX and Newton physics backends, with support for CPU and GPU (CUDA) buffer storage.

The interface contracts remain in isaacsim.core.experimental.prims so consumer extensions can depend on stable shared Carbonite interfaces while selecting the concrete provider through extension enablement.

Key Components#

PrimDataReader#

The PrimDataReader implements the IPrimDataReader interface, managing the lifecycle of data views for different prim types. It routes data access by physics engine:

  • PhysX: Direct C++ TensorApi calls for zero-copy GPU buffer access without Python involvement

  • Newton: Python callbacks fill C++-owned buffers via registered field callbacks

  • Transforms: Engine-agnostic Fabric/USDRT hierarchy queries for world and local poses

The reader supports three view types:

  • XformDataView: Read-only access to world positions, world orientations, local translations, local orientations, and local scales for transform prims

  • RigidBodyDataView: Extends transform data with linear and angular velocities for rigid body prims

  • ArticulationDataView: Extends transform data with DOF positions, velocities, efforts, root transforms, root velocities, DOF types, DOF names, link hierarchy, Jacobians, and mass matrices for articulated robots

PrimDataReaderManager#

The PrimDataReaderManager implements the IPrimDataReaderManager interface, centralizing lifecycle management for the shared IPrimDataReader instance. It subscribes to physics simulation events (stop/resume) to trigger automatic reinitialization, ensuring that sensor plugins and nodes do not need to call initialize() independently.

Buffer Management#

The BufferRegistry module provides per-field buffer state management through FieldEntry structures. Each named field in a view owns a device buffer (CPU or GPU), an optional CPU staging buffer for host-side access, and a fill callback with dirty tracking based on physics step counts. This lazy evaluation ensures buffers are only updated when accessed and stale.

Functionality#

Lazy Evaluation and Dirty Tracking#

Field data is fetched on demand. Each field tracks the last physics step at which it was filled. When a consumer requests data, the reader compares the current step count against the field’s last update and runs the fill callback only if the data is stale, minimizing redundant physics queries.

Host and Device Access#

All view types provide both device-pointer and host-pointer access methods. For GPU-resident data, host variants automatically copy from the device staging buffer to a CPU staging buffer, caching the result until the next physics step invalidates it.

Physics Engine Integration#

For PhysX views, the reader creates native IArticulationView and IRigidBodyView handles through the TensorApi, enabling direct C++ buffer fills without crossing the Python boundary. For Newton views, the buffer infrastructure supports Python-registered callbacks that write into the same C++-managed buffers.

Enable Extension#

The extension can be enabled (if not already) in one of the following ways:

Define the next entry as an application argument from a terminal.

APP_SCRIPT.(sh|bat) --enable isaacsim.core.experimental.primdata

Define the next entry under [dependencies] in an experience (.kit) file or an extension configuration (extension.toml) file.

[dependencies]
"isaacsim.core.experimental.primdata" = {}

Open the Window > Extensions menu in a running application instance and search for isaacsim.core.experimental.primdata. Then, toggle the enable control button if it is not already active.