Daily Digest

Loading brief…

• [Bug] IMU acceleration still omits the centre-of-mass to sensor-mount lever arm (follow-up to #735)

  github:isaac-sim/IsaacSim9/12/2026crashes-stabilityv6.1.0crashrlrenderinghardwaresensorsisaac-sim

  ### Description Follow-up to #735, which was closed as fixed in 6.1.0. Verified against tag `v6.1.0` (`7c206f7`): **the transport term is genuinely fixed, but a second term in the same computation is still missing**, so `IsaacImuSensor` / `IsaacReadIMU` still

• Enable CPU code vectorization

  github:NVIDIA/warp9/12/2026other

  The LLVM-based JIT-compiler for the CPU does not currently vectorize any code; no SIMD instructions get generated. The loop and SLP (superword-level parallelism) vectorization passes simply don't run, but also, the C++ code we generate isn't very vectorization

• [Bug Report] OVRTX duo-camera defaults return stale views; pose refresh leaves temporal artifacts

  github:isaac-sim/IsaacLab9/11/2026training-infrav6.1.0rlrenderinghardwaresensorsenv-apiintegration

  ### Describe the bug Two `Camera` sensors with equal OVRTX renderer configs share the same native camera binding and render product. With the Kuka duo-camera task's default `update_latest_camera_pose=False`, the base and wrist sensors return almost the same s

• Avoid null writes in deterministic component adjoints

  github:NVIDIA/warp9/11/2026hardware-integrationhardwarewarp

  ### Bug Description A deterministic custom adjoint that accumulates into a component of an array adjoint can write through a null array descriptor when that input does not require gradients. The update should be inactive when no backward destination exists.

• [Proposal] Reuse authored Newton MPM Points as dynamic render targets

  github:isaac-sim/IsaacLab9/11/2026asset-pipelineusdrenderingdxisaac-simisaac-labnewton

  ### Proposal After https://github.com/isaac-sim/IsaacLab/pull/7637, Newton MPM spawners author the schema-valid simulation `UsdGeom.Points` at `<asset>/geometry/points`. The simulation Points remain invisible, and `MPMObject` creates a second mutable `<asset>

• Can a light be excluded from specific cameras / render products? (IR projector bleeds into the RGB sensor)

  github:isaac-sim/IsaacSim9/10/2026asset-pipelineusdrenderinghardwaresensorsdxisaac-sim

  ### Description ## Topic Description Is there a supported way to make a `UsdLux` light illuminate some cameras but not others — i.e. per-camera or per-render-product light exclusion? I am simulating an active-stereo depth camera whose IR dot projector is, on

• [BUG] CUDA 13.1+ can make later Warp kernels fail on sm_120

  github:NVIDIA/warp9/10/2026renderingrenderinghardwarewarp

  ### Bug Description On `sm_120` systems with Warp built against CUDA Toolkit 13.1 through 13.4, compiling one CUDA module can cause an unrelated module compiled later in the same process to fail with CUDA error 700. Both modules compile successfully. The ear

• [Regression] Isaac Lab scene creation slows after Newton 1.5.1 to 1.6 development build

  github:newton-physics/newton9/10/2026renderingrenderingsensorsdocsisaac-labnewtonwarp

  ## Summary The Isaac Lab performance dashboard shows a reproducible initialization-time regression beginning with the September 4 daily build. The increase is isolated to configurations that instantiate Newton, occurs almost entirely during scene creation, an

• Fix missing block synchronization in tile Cholesky and matmul adjoints

  github:NVIDIA/warp9/10/2026asset-pipelineusdhardwarewarp

  ### Bug Description Two cooperative tile-math paths are missing block-wide synchronization between a shared-memory read phase and a subsequent in-place write phase. This can produce incorrect results when a CUDA tile block contains more than one warp. ### In

• Track behavior-preserving modularization of large Newton modules

  github:newton-physics/newton9/9/2026crashes-stabilitycrashrlusdhardwaredocsintegration

  Newton has several large files that combine distinct responsibilities or entire multi-stage algorithms. Refactor them along clear ownership boundaries while preserving the public API and all functionality. This tracker has **six work issues**. Implementation

• [Bug Report] Camera pose writes do not reach the renderer on the Newton backend

  github:isaac-sim/IsaacLab9/9/2026crashes-stabilitycrashsim2realusdrenderinghardwaresensors

  ### Describe the bug `Camera.set_world_poses` moves the camera on PhysX but not on Newton. Under Newton, `camera.data.pos_w` reports the new pose while the rendered image stays at the old one. ### Steps to reproduce Run the [regression test here](https://gi

• [Kamino][P-ADMM] Bounded implicit-PD effort constraints make DR Legs diverge to NaN

  github:newton-physics/newton9/8/2026training-infrarlrenderinghardwarelocomotionisaac-labnewton

  ## Summary Kamino P-ADMM intermittently produces explosive velocities and eventually a fully non-finite articulation state for Isaac Lab's DR Legs robot. The regression was bisected to #3990: - Good: [`bc43a55`](https://github.com/newton-physics/newton/comm

• GUI stays frozen while PhysX simulation continues to advance

  github:isaac-sim/IsaacLab9/8/2026crashes-stabilitycrashusdrenderinghardwareisaac-simisaac-lab

  ## Describe the bug When I run an interactive state-based rollout in Isaac Sim, the GUI viewport stays completely static for the whole episode. The robot, clutter, and floor all look frozen, and an object that should fall under gravity never moves. Howeve

• [BUG] Explicit SolverMuJoCo nconmax/njmax are overridden by conservative contact estimates

  github:newton-physics/newton9/7/2026crashes-stabilitycrashhardwareintegrationisaac-labmujoconewton

  ## Bug Description Newton 1.6.0rc1 changes `SolverMuJoCo(use_mujoco_contacts=False)` so explicitly configured `nconmax` and `njmax` values are overridden by conservative topology-based estimates. The regression was introduced by a2ad5d170ad540e1757fc55ba3630

• Severe FPS drop when playing physics with a SimReady kitchen asset: 62 FPS paused vs 19-24 FPS playing on GB10 — physics step costs ~26ms even with GPU dynamics and most props frozen

  github:isaac-sim/IsaacSim9/7/2026crashes-stabilitycrashusdrenderinghardwaredeploymentisaac-sim

  ### Description ### Environment - Isaac Sim version: 6.0.1 (container, WebRTC livestream, 1280x720) - Hardware: NVIDIA GB10 (DGX Spark, aarch64), Driver 580.159.03, CUDA 13.0 - Scene: Lightwheel open-source SimReady kitchen asset [https://drive.google.com/fi

• Extend implicitfast gyroscopic stabilization to free bodies with fixed descendants

  github:google-deepmind/mujoco9/6/2026docs-onboardingdocsmujoco

  ### Problem `implicitfast`'s local gyroscopic stabilization excludes a free body when it has massive fixed descendants. Consequently, two representations of the same rigid object can have substantially different numerical stability: moving an inertial element

• [Bug][PhysX backend] Joint effort targets silently not applied on isaacsim 6.0.1.0 (Newton backend works)

  github:isaac-sim/IsaacLab9/5/2026crashes-stabilityv6.0.1crashrlrenderingenv-apidocsintegration

  ### 🐛 Bug Report Joint effort targets are **silently not applied** by the PhysX backend on the isaacsim 6.0.1.0 stack: a constant effort command produces **zero** cart motion. The Isaac Lab-side buffers and actuator bookkeeping are all correct — the effort r

• Integer floor division (`//`) truncates toward zero instead of flooring

  github:NVIDIA/warp9/4/2026hardware-integrationhardwarewarp

  ### Bug Description Integer `//` in a Warp kernel rounds toward zero rather than toward negative infinity, so it returns the wrong value whenever exactly one operand is negative. There is no error or warning. `ast.FloorDiv` maps to the `floordiv` built-i

• wp.map() raises UnboundLocalError when out is a list for a single-output function

  github:NVIDIA/warp9/4/2026hardware-integrationhardwarewarp

  ### Bug Description `wp.map()` raises `UnboundLocalError` when the mapped function returns a single value and `out` is anything other than a bare `wp.array`. The signature and docstring accept `out: Array | list[Array] | None` ("Optional output array(s)"), bu

• fem.integrate() rejects a flat scalar output array and accepts a mis-shaped 2-D one

  github:NVIDIA/warp9/4/2026hardware-integrationhardwareintegrationwarp

  ### Bug Description `fem.integrate()` validates a caller-provided scalar `output` array for a vector-valued test field with `and` where it needs `or`: ```python # warp/_src/fem/integrate.py, integrate() -> linear form branch if type_size(output_dtype) != tes

• Scaled BSR expression on the left of +/- applies the scale to the other operand

  github:NVIDIA/warp9/4/2026hardware-integrationhardwarewarp

  ### Bug Description `_BsrScalingExpression.__add__` and `__sub__` apply the scale to the wrong operand, so `(s * A) + B` evaluates to `s * B + A` and `(s * A) - B` evaluates to `A - s * B`. The reflected forms (`B + (s * A)`, `B - (s * A)`) are correct, so th

• Harden empty CPU array conversion with NumPy before 2.4

  github:NVIDIA/warp9/2/2026feature-requestsfeature-requestwarp

  ## Bug description NumPy versions before 2.4 fail when an object exposes a zero-sized array through `__array_interface__` with a NULL data pointer. NumPy incorrectly sends the object through its scalar conversion path, which causes empty CPU Warp arrays to fa

• [BUG] `array_scan` silently corrupts its output on device OOM: unchecked `wp_alloc_device` return launches CUB on a NULL buffer

  github:NVIDIA/warp9/2/2026crashes-stabilitycrashhardwaredeploymentwarp

  ### Bug Description ## Summary `wp.utils.array_scan` can return normally having written garbage into its output array, with no exception raised and no return value to check. On a device OOM, `scan_device` passes an unchecked NULL temporary buffer to `cub::D

• [Proposal] Single-environment OVRTX LiDAR adapter through the shared renderer lifecycle

  github:isaac-sim/IsaacLab9/1/2026crashes-stabilitycrashrenderinghardwaresensorsdocsintegration

  ### Discussed in https://github.com/isaac-sim/IsaacLab/discussions/7094 <div type='discussions-op-text'> ^{Originally posted by **StriverAlex** August 14, 2026} ### Proposal I have a working proof of concept for exposing OVRTX 0.4 PointCloud LiDAR

• [BUG] SolverMuJoCo deterministic mode overflows scatter buffer in _update_gradient_JTCJ_dense

  github:newton-physics/newton9/1/2026renderingrenderinghardwaremujoconewtonwarp

  ### Bug Description `SolverMuJoCo` fails on its first simulation step in Warp run-to-run deterministic mode. The attached self-contained reproduction builds a synthetic articulated chain, replicates it to exercise the batched dense-contact path, and raises:

• Validate NanoVDB metadata when loading volumes

  github:NVIDIA/warp8/31/2026crashes-stabilitycrashhardwarewarp

  Warp checks the NanoVDB grid signature when creating a volume, but it trusts several metadata fields in the supplied buffer. We should reject malformed data before creating the volume descriptor or returning any of its metadata. The fixed-size grid and featur

• SolverMuJoCo silently publishes stale body_qdd / body_parent_f when disable_sensors=True

  github:newton-physics/newton8/31/2026hardware-integrationhardwaresensorsintegrationisaac-labmujoconewton

  ## Summary Constructing `SolverMuJoCo(..., disable_sensors=True)` on a model whose `State` requests `body_qdd` or `body_parent_f` produces values that are never refreshed. No error, no warning — the arrays are allocated and written every step from stale MuJoC

• Capture-safe notify scope for passive DOF properties (dof_frictionloss/damping/solref/solimp) in SolverMuJoCo

  github:newton-physics/newton8/31/2026sim2realsim2realhardwareintegrationmujoconewtonwarp

  ## Description Actuator models with load-dependent dry friction (BAM-class hobby-servo models, the LuGre prototype in #4071, mjlab's BAM implementation) don't clip torque in the drive — they write the friction budget into `dof_frictionloss` / `dof_damping` an

• [BUG] Unauthored joint-limit gains convert to an inertia-dependent jnt_solref, and the constants needed to override it are private

  github:newton-physics/newton8/31/2026sim2realsim2realusdrenderinghardwaredxintegration

  ## Bug Description An asset that authors no MuJoCo limit attributes inherits `ModelBuilder.JointDofConfig()`'s generic force-space limit gains, `limit_ke = 1e4` and `limit_kd = 1e1`. `SolverMuJoCo` converts those into MuJoCo's `jnt_solref` at runtime in `upda

• [BUG] Stateful actuator state does not advance across CUDA graph replays when the captured region has an odd step count

  github:newton-physics/newton8/31/2026renderingrenderinghardwaredxmujoconewtonwarp

  ## Bug Description The documented pattern for stateful actuators — allocate two `Actuator.state()` objects and swap them host-side after each `Actuator.step` — is not correct under CUDA graph capture. The swap happens while kernels are being *recorded*, so it

• [REQ] Expose application extension handles from ViewerViser

  github:newton-physics/newton8/31/2026sensors-perceptionsensorsintegrationnewton

  ### Description Expose a small, typed, non-owning context from `newton.viewer.ViewerViser` for applications that need to add Viser scene nodes and GUI controls to the existing viewer session. The context would contain only the existing Viser scene API, GUI A

• [BUG] L-BFGS IK uses a static initial inverse-Hessian scale, so floating-base convergence degrades with distance

  github:newton-physics/newton8/30/2026hardware-integrationhardwarenewtonwarp

  ## Bug Description `IKOptimizerLBFGS` uses a **fixed scalar** for the initial inverse-Hessian approximation `H0`. In the two-loop recursion (`_compute_search_direction_template` in `newton/_src/sim/ik/ik_lbfgs_optimizer.py`), the entire vector is scaled by on

• R39.2 common apt repo: Packages index is newer than the signed Release — apt update fails with "File has unexpected size"

  nvidia-forum:robotics-edge-computing9/8/2026▲ 2other
• JP7.2.1 massflash 5 failed

  nvidia-forum:robotics-edge-computing9/10/2026▲ 1other
• ( R39.2.0 ) apt-get error: File has unexpected size… Mirror sync in progress?

  nvidia-forum:robotics-edge-computing9/8/2026▲ 1other
• Boot failed

  nvidia-forum:robotics-edge-computing9/4/2026▲ 1other
• [Feature] IsaacArticulationController: add scale / offset / clip stage for joint commands

  github:isaac-sim/IsaacSim9/12/2026crashes-stabilitycrashrlrenderinghardwarefeature-requestisaac-sim

  ### Isaac Sim Version 6.1.0 ### Operating System Ubuntu 24.04 ### GPU Information - Model: RTX 5090 - Driver Version: 595.84 ### Topic Description `IsaacArticulationController` applies joint commands verbatim. There is no scale, offset, or clip stage, s

• [Bug Report] AppLauncher's Python SIGSEGV handler turns any segfault into an unkillable infinite fault loop with no core dump

  github:isaac-sim/IsaacLab9/12/2026crashes-stabilitycrashrenderinghardwareperceptionisaac-simisaac-lab

  ### Describe the bug `AppLauncher` installs a **Python** signal handler for `SIGSEGV` (and `SIGABRT`): https://github.com/isaac-sim/IsaacLab/blob/main/source/isaaclab/isaaclab/app/app_launcher.py#L355-L359 ```python # Set up signal handlers for graceful shu

• [REQ] Preserve addressable MPM particle groups through ModelBuilder composition

  github:newton-physics/newton9/11/2026integrationintegrationisaac-labnewton

  ## Description `ModelBuilder.add_usd()` returns `path_particle_map`, which maps each imported MPM `UsdGeom.Points` prim to a source-builder half-open particle range. Those ranges are intentionally build-time snapshots. When that source builder is later passed

• Jetson anvil L4T35.4.1 cause analysis to bootchain error

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• [Bug] RectLight isProjector pattern only resolves when the camera is stationary (RTPT radiance cache)

  github:isaac-sim/IsaacSim9/10/2026asset-pipelineusdrenderinghardwaresensorsisaac-sim

  ### Description ## Topic Description RectLight used as a pattern projector (`isProjector` + `projector:directionTexture:file`) only resolves its dot pattern while the camera is nearly stationary under `RealTimePathTracing`. Once the camera moves, the dots wa

• Isaac Sim 6.1: null-pointer SIGSEGV in debug-draw clear on stage replacement

  github:isaac-sim/IsaacSim9/10/2026crashes-stabilityv6.1crashusdrenderinghardwareperceptionintegration

  ### Description ## AI note Initially encountered and minimal reproducible example found by Astra. Crash confirmed and issue filed manually. ## Summary Replacing a stage after initializing and manually stepping a GPU rigid cube crashes in `PrimitiveDrawingH

• Newton backend: `Articulation.set_joint_position_targets()` triggers Warp CUDA error 710 (device-side assert) and kills the entire CUDA context — physics, RTX rendering and WebRTC — Isaac Sim 6.1.0

  github:newton-physics/newton9/10/2026crashes-stabilityv6.1.0crashusdrenderinghardwaredeploymentdx

  **Post title:** Newton backend: `Articulation.set_joint_position_targets()` triggers Warp CUDA error 710 (device-side assert) and kills the entire CUDA context — physics, RTX rendering and WebRTC — Isaac Sim 6.1.0 ## Summary On Isaac Sim 6.1.0 with the Newto

• Newton's importer has no support for UsdGeomPointInstancer

  github:isaac-sim/IsaacSim9/10/2026crashes-stabilitycrashusdrenderinghardwareisaac-simnewton

  ### Description I was using the point instancer to generate a pile of rocks for PhysX, and it works with PhysX. However, the generated items using instancing are completely ignored by Newton even though they are visible. From Claude: confirmed in newton/_src

• [BUG] Replicated same-body collision filters expand to 92 GiB during finalize

  github:newton-physics/newton9/10/2026crashes-stabilitycrashusdnewtonwarp

  ### Bug Description On current `main`, `ModelBuilder.replicate()` keeps collision-filter blocks compact in the builder, but `ModelBuilder.finalize()` expands every replicated same-body pair into one dense NumPy array. The allocation scales as `world_count * C

• SolverMuJoCo: CUDA illegal memory access after ~3 steps with a non-replicated (flat) multi-environment model

  github:newton-physics/newton9/10/2026crashes-stabilityv3.0crashrenderinghardwarelocomotionintegrationisaac-lab

  ## Description Stepping a `SolverMuJoCo`-backed simulation with a **non-replicated, multi-environment ("flat") model** (Isaac Lab's `replicate_physics=False`, where all environments share a single Newton "world" and the environment index is carried by the art

• [Bug Report] Errors when running Isaac Sim from pre-built Docker image

  github:isaac-sim/IsaacLab9/10/2026crashes-stabilityv6.0.1crashusdrenderinghardwaredeploymentintegration

  I get plenty of errors when running Isaac Sim 6.0.1 inside ``nvcr.io/nvidia/isaac-lab:3.0.0-beta2`` pre-built Docker image. ### Steps to reproduce I have followed the steps depicted in https://isaac-sim.github.io/IsaacLab/v3.0.0-beta2/source/deployment/docke

• Investigate intermittent MuJoCo line-search exhaustion in conveyor test

  github:newton-physics/newton9/10/2026othermujoconewton

  `test_long_running_stability_cuda_0` intermittently emits: ``` linesearch iterations limit reached - please increase ls_iterations to 100 ``` This caused [GPU CI job 102777142860](https://github.com/newton-physics/newton/actions/runs/34447817496/job/10277714

• The system fails to boot at the UEFI stage，after Update the db/dbx Keys with a Capsule Update

  nvidia-forum:robotics-edge-computing9/10/2026other
• Tutorial 4: Add Camera and Sensors to a Robot -> Error in camera reference

  nvidia-forum:simulation9/10/2026sensors-perceptionsensorsdocs
• GPU semaphore acquire timeout error

  nvidia-forum:robotics-edge-computing9/9/2026other
• Kernel array slicing fails to compile under wp.config.llvm_cuda: byte_offset_helper is host-only

  github:NVIDIA/warp9/9/2026hardware-integrationhardwarewarp

  # Bug Description Slicing an array inside a kernel fails to compile when `wp.config.llvm_cuda` is enabled. The kernel below compiles and runs under the default NVRTC path, but is rejected by the Clang-based CUDA compiler. `view()` in `warp/native/array.h` is

• CUDA symbol lookup fails after switching block sizes for a unique kernel with `wp.static()`

  github:NVIDIA/warp9/9/2026renderingrenderinghardwarewarp

  ## Bug description A `module="unique"` kernel that calls a helper containing `wp.static(values[i])` fails when querying occupancy after a launch with a different block size. The first occupancy query succeeds, and the launch with `block_dim=64` produces the

• Extract ModelBuilder composition and fixed-joint collapse algorithms

  github:newton-physics/newton9/9/2026crashes-stabilitycrashrlnewton

  Consolidated into #4192. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• AGX Orin (L4T 36.3) — initrd flash kernel fails to boot, USB device mode dead

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• OTA upgrades on external device nvme1n1 failed

  nvidia-forum:robotics-edge-computing9/9/2026other
• [Bug] Re-enable IsaacContrib-DrLegs-Walk in test_contrib_environments after fixing Kamino NaN divergence

  github:isaac-sim/IsaacLab9/7/2026crashes-stabilitycrashrllocomotiondxisaac-labnewton

  ## Summary `IsaacContrib-DrLegs-Walk` (and its deprecated alias `Isaac-DrLegs-Walk-v0`) is skipped in `source/isaaclab_tasks/test/contrib/test_contrib_environments.py` as of #7623 because the Kamino P-ADMM solver intermittently diverges to NaN under random ac

• URDF inertial origin rotation is lost during compilation, changing rotational dynamics

  github:google-deepmind/mujoco9/7/2026asset-pipelineusddocsmujoco

  ### Intro Hi! I use MuJoCo for robot simulation and found this while checking the inertial properties of imported URDF models. ### My setup - MuJoCo **3.12.0**, official Python package; also reproduced on **3.3.7**. - Python 3.11.15, NumPy 2.2.6. - Ubuntu 2

• USB2_0 Hub error

  nvidia-forum:robotics-edge-computing9/7/2026other
• Sporadic "NvRmGpuLibOpen failed, error=4" after boot on Orin Nano with Jetpack 7.2

  nvidia-forum:robotics-edge-computing9/7/2026hardware-integrationhardware
• [Bug Report] WebRTC Streaming Client connects but shows black screen with isaac-lab:3.0.0-beta2-post1 container

  github:isaac-sim/IsaacLab9/7/2026asset-pipelinev5.1usdrenderinghardwaredeploymentdocsisaac-sim

  ## Describe the bug When running Isaac Lab from the `nvcr.io/nvidia/isaac-lab:3.0.0-beta2-post1` docker image with WebRTC livestreaming enabled (`--livestream 2`), the Isaac Sim WebRTC Streaming Client successfully establishes a connection (signaling complete

• A DCE error on Jetson Thor caused the system to freeze abnormally

  nvidia-forum:robotics-edge-computing9/7/2026crashes-stabilitycrashhardware
• Build fails when the repository path contains a space

  github:NVIDIA/warp9/6/2026otherwarp

  ### Bug Description Building Warp from source fails to link when the repository path contains a space. Several paths are interpolated into shell command strings unquoted, so the shell splits them on whitespace. Cloning into a directory such as `/home/tushar/

• Warp FEM topology optimization example

  github:NVIDIA/warp9/5/2026renderingrenderingdocsfeature-requestwarp

  ## Category <!-- Please check all applicable options from the list below (use [x] in Markdown) --> - [ ] Report an error in the documentation. - [x] Request for something to be documented. - [x] Suggestion to improve the documentation. - [ ] Other (please exp

• CUDA: mixed-width texture sampling returns incorrect values on sm_89

  github:NVIDIA/warp9/4/2026renderingrenderinghardwarenewtonwarp

  On an NVIDIA L40 (`sm_89`), a Warp kernel can return incorrect values when the same function contains base-level texture samples with different return widths. We found this while reducing [Newton issue #4147](https://github.com/newton-physics/newton/issues/414

• [Custom Board] DisplayPort output fails after OS warm reboot with DCE RPC timeout errors (NV_ERR_TIMEOUT)

  nvidia-forum:robotics-edge-computing9/4/2026other
• APIC capture_save may snapshot past allocation bounds for offset torch views

  github:NVIDIA/warp9/4/2026renderingrenderinghardwarewarp

  ### Bug Description ## Description `wp.from_torch()` reports an incorrect capacity for some offset or strided Torch views. APIC treats the view pointer as the region base and snapshots `capacity` bytes from that pointer, which can extend beyond the underlyin

• Failed to open I2C device when using SIPL

  nvidia-forum:robotics-edge-computing9/3/2026other
• [Bug Report] IsaacLab installation encountered an error

  github:isaac-sim/IsaacLab9/3/2026renderingrenderinghardwaredeploymentintegrationisaac-simisaac-lab

  ### Describe the bug The installation and startup of isaac-sim are normal, but an error occurred during the installation of IsaacLab. Please refer to the attachment [IsaacLab_install.log](https://github.com/user-attachments/files/31815703/IsaacLab_install.lo

• [BUG] Mesh texture-SDF returns wrong values outside the narrow band on sm_89 GPUs since #3904

  github:newton-physics/newton9/3/2026crashes-stabilitycrashhardwarenewtonwarp

  ## Bug Description Since #3904 ("Accelerate SDF and hydroelastic collision", eaed8a07), mesh texture-SDF queries outside the narrow band can return wrong values on sm_89 GPUs (NVIDIA L40 in our case). A watertight unit box queried 0.5 m outside its +X face re

• Introduce an explicit tile storage model for Warp tiles.

  github:NVIDIA/warp9/3/2026hardware-integrationhardwaredxmujoconewtonwarp

  ### Description Today a tile's storage class (register vs. shared) is mutable and gets rewritten in place by the operations that consume it — e.g. a shared-requiring op sets `arg.type.storage = "shared"` on a *caller's* tile, and `tile_transpose()` / view res

• Nvsipl_camera error on Jetson AGX Orin with JetPack 7.2

  nvidia-forum:robotics-edge-computing9/3/2026hardware-integrationhardware
• OpenGLRenderer.render_mesh() can crash on an invalid face index

  github:NVIDIA/warp9/3/2026crashes-stabilitycrashrenderingperceptionwarp

  ## Bug description `OpenGLRenderer.render_mesh()` accepts face indices that fall outside the `points` array. It passes those indices directly to the rendering kernels. With smooth shading enabled, the normal-generation kernel uses each face index to read fro

• Intermittent PCIe power-up error

  nvidia-forum:robotics-edge-computing9/2/2026other
• SolverVBD: ALM-based contact forces for Divide and Truncate (DAT)

  github:newton-physics/newton9/2/2026crashes-stabilitycrashrl

  ## Motivation Divide and Truncate (DAT) guarantees penetration-free states by truncating per-DOF motion against division planes. Combined with penalty contacts this is structurally self-contradictory: a penalty contact can only produce force from penetration,

• [Bug] Image publish sizes buffers as width*channels*byteDepth: nv12 unpublishable, nv21/nv24 mis-sized, encoding token list blocks other formats

  github:isaac-sim/IsaacSim9/2/2026renderingrenderinghardwaredeploymentsensorsperceptionisaac-sim

  ### Description `Ros2ImageMessageImpl::generateBuffer` sizes every ROS 2 image message on the assumption that an encoding has an integer number of bytes per pixel: ```cpp // source/extensions/isaacsim.ros2.core/library/backend/Ros2Message.cpp:472-491 int cha

• `array_sum()` and `array_inner()` mishandle negative axes

  github:NVIDIA/warp9/2/2026otherwarp

  ## Bug Description `wp.utils.array_sum()` and `wp.utils.array_inner()` produce incorrect results when passed an in-range negative `axis`. This reproduction uses an explicit count of one so it does not risk reading outside the input allocation: ```python imp

• [Proposal] Promote the shared environment-test helpers into the installed surface

  github:isaac-sim/IsaacLab9/2/2026training-infrarlrenderingmanipulationlocomotionisaac-simisaac-lab

  ### Proposal Promote the reusable environment- and rendering-test helpers that currently live at the root of `source/isaaclab_tasks/test/` into the installed surface — for example an `isaaclab_tasks.testing` module — so in-tree suites and external projects ge

• Retire njmax_nnz joint-limit workaround once mjwarp #1630 is released

  github:newton-physics/newton9/2/2026other

  `SolverMuJoCo` recomputes joint-limit nonzeros itself and adds them to mujoco_warp's `_default_njmax_nnz` estimate (`solver_mujoco.py`, near # work around buffer under-sizing until the fix is released (mjwarp `#1630`)). This workaround should be retired once

• Error in flashing

  nvidia-forum:robotics-edge-computing9/2/2026other
• Build from source fails to find MSVC on Build Tools-only installations (vswhere omits Build Tools without -products *)

  github:NVIDIA/warp9/2/2026hardware-integrationhardwarewarp

  ### Bug Description Building Warp from source on Windows fails with `Warp build error: Could not find MSVC compiler` on machines where the only Visual Studio product installed is Visual Studio Build Tools, even though a working MSVC toolchain is present. `fi

• Jetson Orin NX 16GB - Jetson ISO Customization for Autoinstall. But fails

  nvidia-forum:robotics-edge-computing9/2/2026hardware-integrationhardware
• ISP device node error after upgrading AGX Orin from JetPack 6.2 to JetPack 7.2

  nvidia-forum:robotics-edge-computing9/1/2026hardware-integrationhardware
• IsaacSim with Newton Physics Engine doesn't show simulation working visually

  github:isaac-sim/IsaacSim9/1/2026renderingrenderinghardwaredeploymentisaac-simnewtonfigure

  ### Description While trying to execute code in the script editor for an example to work in IsaacSim with Newton Physics, the example executes in the terminal but the scene never changes. The scene stays still and even after complete execution of the example

• Provide an import .lib for Warp's .dll(s) on Windows

  github:NVIDIA/warp9/1/2026crashes-stabilitycrashwarp

  Currently a native C++ application that wants to use Warp's `.dll` files directly (e.g. to use APIC `.wrp` files without Python code) has to dynamically load them and resolve each symbol individually using `GetProcAddress()`. It's a significant amount of boile

• Error creating mfi archive for massflash Orin NX JetPack 7.2.1

  nvidia-forum:robotics-edge-computing9/1/2026hardware-integrationhardware
• [REQ] Add an asOverDog Kamino locomotion example

  github:newton-physics/newton9/1/2026training-infrarlusdhardwarelocomotionnewtonwarp

  ## Description [asOverDog](https://github.com/asOverDog/asoverdog_newton) is a quadruped platform with interchangeable Bennett, planar, and spherical overconstrained closed-chain leg mechanisms. The system was introduced in an [ISRR 2024 paper](https://arxiv.

• [REQ] Convenience API for self-collision filter groups within an articulation

  github:newton-physics/newton8/31/2026crashes-stabilitycrashusdisaac-labnewton

  ### Description Provide a convenience API on `ModelBuilder` for filtering self-collisions between *sets* of links within an articulation, beyond the existing parent-child filtering. For example: ```python # Filter all shapes of the listed bodies against each

• [Bug Report] omni.flowusd.ui is missing in 6.0.1 under Windows Build

  github:isaac-sim/IsaacLab8/31/2026renderingv6.0.1renderingintegrationisaac-sim

  I'm following the steps to use flowusd in Isaac Sim based on different youtube tutorials. All of them are based on older versions of Isaac Sim. And currently there is only omni.flowusd available for Isaac Sim 6.0.1. omni.flowusd.ui is missing. There is no othe

• [Proposal] Standardized robot fault injection for robustness evaluation

  github:isaac-sim/IsaacLab8/31/2026crashes-stabilitycrashrlmanipulationlocomotionsensorsdx

  ### Proposal Add a reusable, vectorized fault-injection layer for Isaac Lab so policies can be trained and evaluated under persistent and intermittent robot failures, not only stochastic noise and parameter randomization. A first version could support sensor

• VIDIOC_DQBUF fails with I/O error / VI capture timeout on TC358743 (Geekworm C790) — persists in both 2-lane and 4-lane CSI config

  nvidia-forum:robotics-edge-computing8/31/2026other
• SIGSEGV in libomni.physx.tensors on RigidBodyView.get_transforms() of a kinematic anchor welded to an articulation via a maximal-coordinate loop FixedJoint (6.0.1)

  github:isaac-sim/IsaacSim8/31/2026crashes-stabilityv3.0.0crashusdrenderinghardwareintegrationisaac-sim

  ### Description Welding a standalone floating articulation to a standalone **kinematic** rigid-body anchor with a **maximal-coordinate (loop) `UsdPhysics.FixedJoint`**, then calling `RigidBodyView.get_transforms()` on the kinematic anchor **after `world.step(

• Profession – Issac Asimov (1957)

  hn:isaac sim9/11/2026▲ 12other
• Games look real now. We're in the future

  hn:gr00t9/1/2026▲ 6other
• Isaac Asimov's "Profession"

  hn:isaac sim9/12/2026▲ 5other
• An UART3 (H62, K60) Be Used as a General-Purpose UART, or Is It Reserved for Debugging Only?

  nvidia-forum:robotics-edge-computing9/9/2026▲ 2other
• Investigate remote cache support for kernel and LTO artifacts

  github:NVIDIA/warp9/10/2026▲ 1docs-onboardingdocswarp

  ### Description Warp's kernel cache currently expects a local filesystem path. Investigate support for storing and retrieving compiled kernel and MathDx LTO artifacts from a remote location. This is primarily for cloud jobs with ephemeral disks. Without a pe

• Isaac Sim 6.1.0 - Release Notes missing

  github:isaac-sim/IsaacSim9/10/2026▲ 1renderingv6.1.0renderinghardwareisaac-sim

  ### Description Today, I have received a notification that Isaac Sim 6.1.0 has been released: https://github.com/isaac-sim/IsaacSim/releases/tag/v6.1.0 Unfortunately, the link to the release notes for v6.1.0 doesn't work, and therefore it's not possible to d

• Jetson Orin Nano Developer Kit – Unknown Password for Existing jetson User

  nvidia-forum:robotics-edge-computing9/10/2026▲ 1hardware-integrationhardware
• IsaacSim Newton Physics: Possible Memory Leak on Repeated Stop and Play Simulaiton

  github:isaac-sim/IsaacSim9/8/2026▲ 1crashes-stabilitycrashrenderinghardwareisaac-simnewton

  ### Description Refer to attached file on replicating the issue: [Issue.zip](https://github.com/user-attachments/files/31939297/Issue.zip) **What is the Issue:** Memory usage is keep on increasing whenever I repeatedly start and stop the simulation as long t

• Difference Between Cortex-A78A (Hercules) in DRIVE AGX Orin and Cortex-A78AE in Jetson AGX Orin

  nvidia-forum:robotics-edge-computing9/7/2026▲ 1hardware-integrationhardware
• Request for Documentation on Jetson Thor Safety MCU (STM32) Architecture and Communication

  nvidia-forum:robotics-edge-computing9/7/2026▲ 1hardware-integrationhardwaredocsfeature-request
• Jetson Orin Nano does not boot into Linux without display attached to displayport port

  nvidia-forum:robotics-edge-computing9/5/2026▲ 1hardware-integrationhardware
• Jetson Orin NX + Sony IMX900 Direct CSI-2: System Hangs on V4L2 STREAMON / Frame Capture

  nvidia-forum:robotics-edge-computing9/3/2026▲ 1crashes-stabilitycrashhardware
• ROS 2 Autonomous Mobile Robot with TensorRT Vision, LiDAR and Voice AI on Jetson Orin Nano

  nvidia-forum:robotics-edge-computing9/3/2026▲ 1hardware-integrationhardwaredeploymentsensors
• P3767-0005 rev T.1: "Applet/MB2 is not running" after RCM blob — same on JetPack 6.2 and 7.2.1

  nvidia-forum:robotics-edge-computing9/2/2026▲ 1other
• Link to the current sources for Linux on Tegra?

  nvidia-forum:robotics-edge-computing9/1/2026▲ 1other
• Nsys report not found (generated) within DL Designer on Jetson Thor with Jetpack 7.2.1

  nvidia-forum:robotics-edge-computing9/1/2026▲ 1hardware-integrationhardware
• Imx296 driver for nvidia jetson orin nano/nx

  nvidia-forum:robotics-edge-computing8/31/2026▲ 1hardware-integrationhardware
• My experience with WiFi adapters on Orin Nano - RTL8822CE vs QCNFA765

  nvidia-forum:robotics-edge-computing8/30/2026▲ 1hardware-integrationhardware
• Jetson Orin Nano custom carrier HDMI shows distorted output at low resolutions and no display at 1080p or higher

  nvidia-forum:robotics-edge-computing9/12/2026hardware-integrationhardware
• Waveshare OV9281 1MP USB Camera (A) via Type-C to Type-C on Jetson Orin Nano — Will it work?

  nvidia-forum:robotics-edge-computing9/12/2026hardware-integrationhardwaresensors
• How to communicate stm32 with jetson agx orin over the I2S Peripheril

  nvidia-forum:robotics-edge-computing9/12/2026hardware-integrationhardware
• Nvidia jetson orin nx GPIO access on coustom carrier board

  nvidia-forum:robotics-edge-computing9/12/2026hardware-integrationhardware
• [MuJoCo Warp] Selective reset_data corrupts packed contact membership across worlds

  github:google-deepmind/mujoco9/12/2026crashes-stabilitycrashusdrenderinghardwaredocsmujoco

  ### Intro Hi! I am testing MuJoCo Warp's Python APIs for batched simulation and selective world resets. I found a contact-bookkeeping discrepancy affecting worlds that were not selected for reset. The affected component is `google-deepmind/mujoco_warp`. ###

• [BUG] ArticulationView keeps stale views after source array replacement

  github:newton-physics/newton9/11/2026renderingrenderinghardwarenewtonwarp

  ### Bug Description `ArticulationView` keeps a cached raw-pointer view after an array on the same `State`, `Model`, or `Control` object is replaced. A later getter reads the old allocation and a later setter writes it, while the current source array remains u

• Jetson Orin Nano USB Target Flashing method crashes

  nvidia-forum:robotics-edge-computing9/11/2026crashes-stabilitycrashhardware
• Blue noise sampling on surfaces

  github:NVIDIA/warp9/11/2026otherwarp

  Blue noise sampling or poisson disk sampling is a way to generate random samples that don't clump together. This is useful for stratified integrations, procedural geometry generation and machine learning feature sampling on surfaces. It'd also be useful to el

• Oriented bounding box

  github:NVIDIA/warp9/11/2026othernewtonwarp

  Oriented bounding boxes are a useful bounding primitive, used for example in Newton. Newton currently computes them via convex hull (scipy), pca / inertia tensor, then dense sampling around major axis (warp). Besides being off device (and convex hull is exp

• Long-run Argus stability issues on our Jetson AGX Orin , JetPack 6.2.1 / L4T r36.4.4, kernel 5.15.148-tegra setup with 9 IMX390 cameras through UB960

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Sub-second voice latency on a Jetson: Little Gemma stack demo

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Evaluate augmented-assignment targets before the RHS

  github:NVIDIA/warp9/11/2026hardware-integrationhardwaredocswarp

  ### Bug Description Array component augmented assignments do not fully preserve Python evaluation order when the right-hand side mutates the selected component. Python augmented assignment evaluates and reads the target before evaluating the right-hand side.

• Preserve ordering between deterministic atomics and stores

  github:NVIDIA/warp9/11/2026hardware-integrationhardwarewarp

  ### Bug Description Deterministic floating-point atomics are recorded during the kernel and reduced afterward. A normal or component store to overlapping storage still executes during the kernel, which can reverse source program order even for a one-thread la

• Pixhawk Jetson Baseboard Bundle won't turn on after flashing older SDK version

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Rst_source: 0x17, rst_level: 0x1

  nvidia-forum:robotics-edge-computing9/11/2026other
• Prevent Thor users from flashing OS

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• [Isaac Sim 6.0.1] omni.physx.fabric does not output PBD particle positions to Fabric — particles forced through the USD write-back, capping FPS

  nvidia-forum:simulation9/11/2026crashes-stabilityv6.0.1crashusdisaac-sim
• Filter shape pairs attached to the same body in broad phases

  github:newton-physics/newton9/11/2026crashes-stabilitycrash

  The current `CollisionPipeline` broad phases do not filter out shapes attached to the same body. The model builder must therefore emit explicit shape collision filter pairs for any pair of colliders attached to the same body, growing the size of the filter pai

• Secure deployment inquiry for Isaac Sim WebRTC remote access

  nvidia-forum:simulation9/11/2026deploymentdeploymentisaac-sim
• FPGA与JETSON 无法通过PCIE 连接

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• [Jetson Orin NX 16GB] Rename network interface

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Does NVCSI on Jetson AGX Thor support CSI-2 lane rates above 1.5 Gbps/lane, and how should it be configured?

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• How to self rebuild kernel？

  nvidia-forum:robotics-edge-computing9/11/2026other
• Can partition 1 (APP1) be copied from a Jetson Nano SSD to a Jetson Orin NX SSD?

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• What is the difference between DRAM Memory ECC and Alt-link ECC?

  nvidia-forum:robotics-edge-computing9/11/2026other
• What is the purpose of the UFS device in the IGX Thor?

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• R39.2.1 用T5000 模拟t3000性能 烧录问题

  nvidia-forum:robotics-edge-computing9/11/2026other
• Jetson AGX Orin - AQR113C SPI Flash Firmware

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Can Alt-Link ECC be enabled on Jetson AGX Thor T4000?

  nvidia-forum:robotics-edge-computing9/11/2026hardware-integrationhardware
• Newton Learning Path: Set Up Your Environment and Build Your First Simulation

  nvidia-forum:robotics-edge-computing9/11/2026othernewton
• What’s New in Isaac Lab-Arena

  nvidia-forum:robotics-edge-computing9/11/2026otherisaac-lab
• Jetson Orin Nano HDMI display intermittently works with MAX96751+MAX96752

  nvidia-forum:robotics-edge-computing9/10/2026hardware-integrationhardware
• Can a light be excluded from specific cameras / render products? (IR projector bleeds into the RGB sensor)

  nvidia-forum:simulation9/10/2026renderingrendering
• About partitions flash and erase

  nvidia-forum:robotics-edge-computing9/10/2026other
• RectLight isProjector pattern disappears while the camera moves in RealTimePathTracing (radiance cache convergence?)

  nvidia-forum:simulation9/10/2026sensors-perceptionsensors
• [REQ] Brighten Viser Viewer Default Light

  github:newton-physics/newton9/10/2026asset-pipelineusdrenderingdocsisaac-labnewton

  ### Description Request to brighten the default lighting (I believe its a direct overhead light) in the Viser Viewer. Or: to expose a light intensity parameter. ### Motivation / Use Case The Viser viewer lighting is much darker in Isaac Lab scenes, compared

• [Question] IsaacLab: BeyondMimic/ProtoMotions G1 tracker transfers to MuJoCo, but a PhysX-trained AMP "run" actor layered on top does not – has anyone solved hierarchical Isaac→MuJoCo sim-to-sim?

  github:isaac-sim/IsaacLab9/10/2026crashes-stabilitycrashrlusddeploymentlocomotionisaac-lab

  TL;DR – I have a Unitree G1 hierarchical controller trained in Isaac Lab (PhysX): a high-level switcher → 64-D "target-frame" AMP actors (run/jump) → a FROZEN BeyondMimic/ProtoMotions tracker (157-D obs → 29 joint targets, with BeyondMimic PD scaling). Re-impl

• BeyondMimic/ProtoMotions G1 tracker transfers to MuJoCo, but a PhysX-trained AMP "run" actor layered on top does not – has anyone solved hierarchical

  nvidia-forum:simulation9/10/2026crashes-stabilitycrashmujoco
• performance issue with many meshes

  github:isaac-sim/IsaacSim9/10/2026crashes-stabilitycrashusdrenderinghardwareintegrationisaac-sim

  ### Description IsaacSim has terrible performance when you add a lot of meshes. I recognize that I'm supposed to use instancing if possible, and I'll switch to that soon. However, it seems like there might be some low-hanging fruit for improving this performa

• JetPack 6.x (L4T R36.4.3): How to completely disable PXE / Network Boot in UEFI?

  nvidia-forum:robotics-edge-computing9/10/2026other
• Jetson Orin NX 8GB hard reboots during TensorRT engine build on fresh JetPack 7.2.1 / DeepStream 9.1

  nvidia-forum:robotics-edge-computing9/10/2026hardware-integrationhardware
• How to enable GPIO function for SPI5_SCK and GPIO47 via Busybox

  nvidia-forum:robotics-edge-computing9/10/2026other
• SCF_AutocontrolACSync timeout after ~3.5 hours with synchronized dual GMSL cameras on JetPack 5 / R35.6.0

  nvidia-forum:robotics-edge-computing9/10/2026other
• Jetson Thor针对 SIPL 和 Argus 架构tuning生成的文件是否可以通用？还是说需要分别进行两次独立的 Tuning 流程？

  nvidia-forum:robotics-edge-computing9/10/2026hardware-integrationhardware
• [Question] Update RSL-RL dependency to v5.5.1 for Isaac Lab 3.0.0?

  github:isaac-sim/IsaacLab9/10/2026training-infrav3.0.0rlisaac-simisaac-lab

  The upcoming Isaac Lab 3.0.0 release currently pins [`rsl-rl-lib==5.0.1`](https://github.com/isaac-sim/IsaacLab/blob/99f1423e5d4a26216c0eedeb2aa78099a8c3a7d1/source/isaaclab_rl/setup.py#L56). Are there plans to update the required version to [`RSL-RL v5.5.1`]

• IGX Thor devkit Wi-Fi issue

  nvidia-forum:robotics-edge-computing9/10/2026hardware-integrationhardware
• Orin Rootfs A/B issue caused by unexpected power loss,

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Jetson Orin Nano开发套件上MCU (EFM8SB10F2G)程序下载链接

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Clarification on Jetson Thor DP++ Support and Pinmux Settings with PI3AUX221

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Afeter try to update, jetpack 6.2.2, No boot no display

  nvidia-forum:robotics-edge-computing9/9/2026other
• tile_matmul(): set the cuBLASDx Alignment and StaticBlockDim operators for faster shared-memory GEMMs

  github:NVIDIA/warp9/9/2026renderingrenderinghardwaremujoconewtonwarphumanoid

  ### Description `wp.tile_matmul()` builds its cuBLASDx GEMM (`wp_cuda_compile_dot` in `warp/native/warp.cu`) without setting the cuBLASDx `Alignment` operator, so cuBLASDx assumes the natural alignment of the element type (4 bytes for `float32`) for every sha

• `nvv4l2decoder SIGSEGV inside libnvmmlite_video/libnvvideo decoding 1080i H.264 (L4T 35.6.5, AGX Xavier)`

  nvidia-forum:robotics-edge-computing9/9/2026other
• Isaac Sim picture guide may be older picture

  nvidia-forum:simulation9/9/2026otherisaac-sim
• Recurring SIGSEGV inside libnvmmlite_video / libnvvideo while decoding 1080i H.264 broadcast streams with nvv4l2decoder (L4T 35.6.5, AGX Xavier)

  nvidia-forum:robotics-edge-computing9/9/2026other
• Separate coupled ADMM setup and constraint-family implementation

  github:newton-physics/newton9/9/2026hardware-integrationhardwarenewton

  Consolidated into #4200. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Extract ViewerBase model and geometry bookkeeping into private helpers

  github:newton-physics/newton9/9/2026renderingrenderingnewton

  Consolidated into #4200. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Separate collision setup and narrow-phase kernel families

  github:newton-physics/newton9/9/2026crashes-stabilitycrashrlhardwaredeploymentnewtonwarp

  Consolidated into #4200. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Organize VBD rigid kernels into cohesive kernel families

  github:newton-physics/newton9/9/2026renderingrenderinghardwareintegrationnewtonwarp

  Consolidated into #4200. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Refactor other large implementation and test files

  github:newton-physics/newton9/9/2026crashes-stabilitycrashusdrenderinghardwaredocsintegration

  Part of #4191. No public API or functionality changes; the shared compatibility requirements in the tracker apply. Track the remaining large-file cleanup in one follow-up issue. The main builder, USD/MJCF parser, MuJoCo solver, and MuJoCo test files have dedi

• Split MJCF ingestion into explicit parsing and builder phases

  github:newton-physics/newton9/9/2026crashes-stabilitycrashrlusdrenderingdocsintegration

  Part of #4191. No public API or functionality changes; the shared compatibility requirements in the tracker apply. `parse_mjcf()` occupies most of `newton/_src/utils/import_mjcf.py`, combining XML/include processing, defaults, assets, transforms, topology, an

• Rename and split test_mujoco_solver.py by behavior

  github:newton-physics/newton9/9/2026hardware-integrationhardwaremujoconewton

  Part of #4191. No public API or functionality changes; the shared compatibility requirements in the tracker apply. `newton/tests/test_mujoco_solver.py` contains dozens of test classes covering independent solver behaviors. Rename it to match the `SolverMuJoCo

• Separate MuJoCo state, contact, and model-update implementation

  github:newton-physics/newton9/9/2026othermujoconewtonwarp

  Consolidated into #4196. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Refactor solver_mujoco.py into registration, compilation, and runtime modules

  github:newton-physics/newton9/9/2026training-infrarlusdmujoconewtonwarp

  Part of #4191. No public API or functionality changes; the shared compatibility requirements in the tracker apply. `newton/_src/solvers/mujoco/solver_mujoco.py` combines the custom-attribute catalog, model compiler, state/contact bridging, and runtime updates

• Extract MuJoCo custom-attribute registration by domain

  github:newton-physics/newton9/9/2026asset-pipelineusdmujoconewton

  Consolidated into #4196. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Separate ModelBuilder construction domains behind the existing API

  github:newton-physics/newton9/9/2026crashes-stabilitycrashusddocsmujoco

  Consolidated into #4192. Its scope and validation requirements are now tracked there as part of the larger file refactor. This issue is closed to reduce tracking granularity; the implementation is not being marked complete. --- Part of #4191. No public API o

• Refactor builder.py into cohesive private modules

  github:newton-physics/newton9/9/2026crashes-stabilitycrashhardwaredeploymentnewton

  Part of #4191. No public API or functionality changes; the shared compatibility requirements in the tracker apply. `newton/_src/sim/builder.py` combines construction, custom attributes, composition, graph transformations, validation, and finalization in one l

• How should we tune Jetson AGX Thor for a CPU-bound web server workload (700 concurrent users), and can the GPU contribute anything?

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Bringing up PCIe-C5 on UPHY1 without MGBE — is there a supported path?

  nvidia-forum:robotics-edge-computing9/9/2026other
• Is UART4 Function Available on Jetson Thor UART4_TX/UART4_RX Pins?

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Manual Account Verification Request – Region Missing: Pakistan (+92)

  nvidia-forum:robotics-edge-computing9/9/2026other
• Realtime packages missing from rt-kernel r39.2 apt repo on JetPack 7.2.1 (Orin Nano)

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Regarding 25GbE configuration issues during Jetson Thor QSFP startup

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Isaac sim 6.0.1 version. object is slipping from the gripper while trying to pick in the reBot B601 DM robotic arm 6+2 DOF

  nvidia-forum:simulation9/9/2026manipulationv6.0.1manipulationisaac-sim
• [REQ] Implement CollisionPipeline.create_from_usd(scene_prim, model)

  github:newton-physics/newton9/9/2026asset-pipelineusdisaac-simisaac-labnewton

  ### Description Parse NewtonCollisionAPI attributes applied to the scene prim and use them to implement the function CollisionPipeline.create_from_usd(scene_prim, model) ### Motivation / Use Case Allow everything required for Newton simulation to be configu

• [REQ] Add NewtonCollisionAPI to newton_usd_schemas

  github:newton-physics/newton9/9/2026asset-pipelineusdisaac-simisaac-labnewton

  ### Description The goal is to support the implementation of the following: collision_pipeline = CollisionPipeline.create_from_usd(scene_prim, model) ### Motivation / Use Case Allow everything required for Newton simulation to be configured in USD. The go

• IMX412 Tegracam Driver on Orin Nano (P3768): I2C Succeeds but VI Capture Times Out (uncorr_err)

  nvidia-forum:robotics-edge-computing9/9/2026hardware-integrationhardware
• Flashing problem in ORIN_AGX with custom carrier board from connecttech

  nvidia-forum:robotics-edge-computing9/9/2026other
• IMX477 (CAM1) streaming corruption when the stock IMX219+IMX477 overlay is used on Jetson Orin NX

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• JetPack7.2.1 r39.2.1 T5000 限制功耗模式120w跑压力测试会达到130多w

  nvidia-forum:robotics-edge-computing9/8/2026other
• Jetson Thor PCIe SI test

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• Jetpack7.2 jetson thor T5000 spi如何软件控制cs

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• Replace bare assertions in runtime code

  github:NVIDIA/warp9/8/2026renderingrenderingwarp

  Several runtime paths use bare `assert` statements to validate user input or internal state. Python removes these checks when run with `-O`, which can allow invalid conditions to continue and produce less useful errors later. Warp should validate inputs and i

• Dependency errors pulling down torch

  nvidia-forum:robotics-edge-computing9/8/2026other
• AGX THOR include an SMCU sockect?

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• Using CAM_PWDN on Jetson Orin Nano P3768 as IMX586 XCLR reset

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• [REQ] Add viewer picking for particle-based deformables

  github:newton-physics/newton9/8/2026renderingrenderingmanipulationdocsnewton

  ### Description Newton's viewer picking currently supports rigid bodies only. `Picking.pick()` raycasts `Model.shape_*`, records a `pick_body`, and accepts a hit only when it resolves to a body index. The drag force is then applied through `State.body_f`. As

• Need Design Checklist and Datasheet of "NVIDIA Jetson AGX Orin SoM"

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• Loading MJCF in python that contains an actuator subclass default overwrites other subclass params

  github:google-deepmind/mujoco9/8/2026training-infrarlusddxdocsmujoco

  ### Intro I'm working on an RL environment, and wanted to compare various actuator settings' effects on the learning. I noticed that having defaults of one subclass of actuators in an MJCF sets the params of other subclasses defined. I know that the actuato

• Thor 双nvme盘启动问题？

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware
• Jetson Thor devkit USB PD enumeration

  nvidia-forum:robotics-edge-computing9/8/2026hardware-integrationhardware