Skip to content

v0.3.0 - Integrated CudaNav Release

Latest
Latest

Choose a tag to compare

View all tags
@rsasaki0109 rsasaki0109 released this 30 Jul 14:13
v0.3.0
67f1345
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v0.3.0 - Integrated CudaNav Release

CudaRobotics v0.3.0 turns the standalone CUDA building blocks into an
integrated navigation stack and publishes the same source version across the
Python and ROS 2 surfaces.

Highlights

  • Integrated CudaNav pipeline: KISS-ICP-style LiDAR odometry, voxel mapping,
    ESDF generation, Nav2 costmap integration, and the CUDA MPPI controller.
  • ROS 2 Jazzy packages for messages, shared CUDA infrastructure, odometry,
    mapping, ESDF, costmap, control, and launch/bringup.
  • Commit-bound Build, Python-package, ROS 2, CudaNav, and systems-paper
    evidence workflows.
  • Reproducible CudaNav scenario harnesses, real-dataset evaluation contracts,
    profiling/ablation reports, and independently validated evidence archives.
  • Python cudarobotics 0.3.0 source distribution and portable CPython 3.10
    and 3.12 manylinux x86_64 wheels.
  • Colab quickstart pinned to the immutable v0.3.0 tag.

Installation

Linux x86_64 users can download the wheel matching CPython 3.10 or 3.12 from
the GitHub release:

python -m pip install ./cudarobotics-0.3.0-*.whl

Building from the source archive requires CUDA Toolkit 12 or newer and CMake
3.18 or newer:

python -m pip install ./cudarobotics-0.3.0.tar.gz

ROS 2 users should build the packages under ros2_ws/src on ROS 2 Jazzy /
Ubuntu 24.04 with CUDA 12.6.

Evidence and limitations

  • Release assets are validated against the exact tagged commit before
    publication and are re-downloaded after publication for a final integrity
    check.
  • The systems-paper archive records the accepted, partial, or blocked status
    of every evaluation claim. Optional multi-GPU scaling remains partial and is
    not a v0.3.0 release blocker.
  • The public ERL Prueba2 rosbag is retained as recorded-motion evidence and an
    explicit negative quality result; it is not claimed as closed-loop success.
  • GPU wheels require a compatible NVIDIA driver at runtime. CUDA Toolkit and
    GPU availability are still required for source builds and GPU execution.

The future v1.0 support contract remains tracked separately in
docs/v1_support_matrix.json; v0.3.0 does not claim v1 readiness.

Assets 7
Loading