Toolspublished

Qualcomm Explains How Its Robotics Stack Becomes an SDK Archive

The QIR documentation pairs a wide robotics catalog with the less visible rules that package components, gather dependencies and set a ROS 2 build configuration.

By 2 min read
Qualcomm Explains How Its Robotics Stack Becomes an SDK Archive
Qualcomm Explains How Its Robotics Stack Becomes an SDK Archive

Listen to this story

The audio brief

About 1:29
0:001:29
Read transcript
Qualcomm’s new robotics documentation reveals the machinery that turns its component catalog into installable SDK archives. The important development is not another robot demo; it is a defined build-and-distribution path for the Qualcomm Intelligent Robotics SDK, or QIR. Its robotics layer uses four BitBake classes to package ROS-style components, collect runtime dependencies, create symbolic links in the finished filesystem, and assemble QIRP archives. The psdk-image class pulls together the toolchain, setup scripts, samples, runtime scripts, and runtime packages. A separate rdepends-collector class writes direct package-group dependencies to list files, which psdk-image then consumes when building the final archive. That makes the dependency path visible instead of leaving it as hidden build logic. The configuration enables ROS 2 Jazzy, with the distribution named “QCOM Robotics Reference Distro with ROS,” and supports Yocto’s wrynose release. The catalog reaches from MoveIt motion planning and Navigation2 to Cartographer SLAM, Orbbec cameras, RPLidar, perception, neural-network inference, and autonomous mobile robot services. Examples include AprilTag detection, depth and pose estimation, object detection, a ResNet-101 model, remote assistance, and simulated pick-and-place. The concrete constraint to watch is whether this packaging system can keep those broad integrations and their dependencies synchronized as the catalog grows.

Story brief

3 key points

Qualcomm’s newly published robotics-layer documentation turns its broad catalog into a defined build-and-distribution pipeline. Four BitBake classes package ROS-style components, gather runtime dependencies, create filesystem links, and assemble QIRP SDK archives containing toolchains, scripts, samples, and runtime packages. The stack supports ROS 2 Jazzy on Yocto’s wrynose release and spans navigation, perception,...

  1. 01

    The psdk-image class consumes dependency lists generated by rdepends-collector when assembling final QIRP archives.

  2. 02

    Supported configuration enables ROS 2 Jazzy as “QCOM Robotics Reference Distro with ROS.”

  3. 03

    Named images include qcom-robotics-image and qcom-robotics-proprietary-image, plus open-source and proprietary package groups.

Qualcomm’s robotics catalog spans navigation, perception, sensors and simulation. But its newly published QIR material chiefly spells out the machinery beneath that breadth: build rules that turn components and their dependencies into distributable SDK archives.

The Qualcomm Intelligent Robotics SDK combines Qualcomm Linux layers with dedicated robotics layers. Its meta-qcom-robotics-sdk documentation covers BitBake build classes, distribution configuration, recipes, package groups, images and SDK assembly.

The build path is explicit

Four BitBake classes divide the assembly work. They package ROS-style components, collect runtime dependencies, create symbolic links in the finished root filesystem and assemble distributable robotics SDK archives. The psdk-image class gathers a toolchain, setup scripts, sample content, runtime scripts and runtime packages into a QIRP SDK archive.

The primary ROS 2 configuration enables ROS 2 Jazzy and sets the distribution name to “QCOM Robotics Reference Distro with ROS.” The layer identifies wrynose as its supported Yocto version.

A catalog from motion planning to sensor services

The layer supplies appends and integrations for MoveIt motion planning, Navigation2, ROS perception and communication packages, containers, and support libraries including Ceres Solver, HDF5 and LXC. Its recipe catalog also includes Cartographer SLAM, data transport, OCR, Orbbec cameras, AMR services, audio, cameras, IMU support, neural-network inference, robot-base control, system monitoring, video and RPLidar.

Included application examples

  • AprilTag detection, depth estimation, face and hand detection, and pose estimation.
  • Object detection and segmentation, a ResNet-101 example, and remote assistance.
  • Simulated pick-and-place and simulated AMR simple motion.

Images and package groups are named in the layer

The specification lists qcom-robotics-image and qcom-robotics-proprietary-image recipes. It also names package groups for ROS 2, open-source robotics, proprietary robotics, and open-source packages with proprietary dependencies, alongside a qirp-sdk recipe.

Sources

  1. dragonwingdocs.qualcomm.comRobotics layers specifications - Qualcomm Dragonwing Documentation