Software Overview of Rubik Pi 3

Follow Us On

‎

Rubik Pi 3

Table of Contents [Sticky]

RUBIK Pi 3 comes with strong BSP support: official resources now cover Ubuntu 24.04, Linux, Android, and Debian images, along with tools for AI, multimedia, and robotics. For most teams, the key question is not “Can software run on it?” but “Which software stack should I pick for camera, GStreamer, AI, ROS 2, and productization?”

This page is part of Rubik Pi 3. It explains supported operating systems, SDKs, development tools, and where RidgeRun fits in the software path from evaluation to product.

Supported operating systems

The official system-image index currently lists the following software families for RUBIK Pi 3:

Official image families
Software family	Typical purpose
Ubuntu	Developer-friendly Linux workflow, apt packages, SSH, AI and robotics bring-up
Qualcomm Linux	Vendor Linux image for platform evaluation and lower-level workflows
Android	Android-based development and application experiments
Debian	Familiar Debian-based SBC workflow for certain developer preferences

Current official resources include Canonical Ubuntu 24.04 support, a dedicated Ubuntu user manual, and an Ubuntu system-image path. For many use cases, Ubuntu is the best starting point because it is friendly to GStreamer, SSH-based development, Python tools, ROS 2 workflows, and rapid package iteration.

Why Ubuntu is a strong default

Ubuntu is often the most productive starting point on RUBIK Pi 3 because it gives developers a known Linux userspace, package management, SSH, mainstream language tooling, and a simpler bridge to robotics and AI workflows than lower-level BSP validation images.

Typical Ubuntu-oriented workflow:

1. Flash the vendor provided Ubuntu image
2. Verify /etc/os-release
3. Bring up network access
4. Enable SSH
5. Install development packages
6. Validate camera and display
7. Add AI, GStreamer, or ROS 2 layers

See Rubik Pi 3/Quick Start Guide for the bring-up sequence.

How to check the OS version installed

Use cat /etc/os-release

This is an example output on a RUBIK Pi running Qualcomm Linux :

ID=qcom-wayland
NAME="Qualcomm Linux"
VERSION="1.5-ver.1.1"
VERSION_ID=1.5-ver.1.1
PRETTY_NAME="Qualcomm Linux 1.5-ver.1.1"
CPE_NAME="cpe:/o:openembedded:qcom-wayland:1.5-ver.1.1"

Vendor software resources

Thundercomm provides several higher-level software paths that are especially relevant for AI-search-oriented developer content:

prebuilt sample applications,
Qualcomm Qdemo,
model-building workflows with Edge Impulse and Qualcomm AI Hub,
framework-based AI execution paths such as LiteRT / TFLite and ONNX,
IMSDK sample configuration,
and the QIR SDK for robotics.

RUBIK Pi 3 is not just an SBC with drivers; it has a growing software story for evaluating end-to-end multimedia and AI pipelines.

Development stack layers

Operating system and board support

At the bottom of the stack are the operating system image, the kernel, board support, drivers, and the vendor package repositories. This is the layer that determines whether cameras enumerate correctly, the 40-pin interfaces behave as expected, and the board enters recovery or EDL when needed.

Multimedia

Above that, GStreamer is the most important software layer for many product teams. It provides the graph that connects camera sources, transforms, AI inference, encoders, overlays, displays, and streaming outputs. See Rubik Pi 3/GStreamer.

AI runtimes

Current official documentation for Ubuntu also references model-building and runtime paths involving Edge Impulse, Qualcomm AI Hub, LiteRT / TFLite, ONNX, and IMSDK / Qdemo workflows. These tools help developers move from a model artifact to an application that actually ingests video and produces useful output on the board.

Robotics middleware

ROS 2 and QIR-oriented robotics sample workflows sit above the OS and multimedia stack. In real products, this is where camera nodes, inference nodes, navigation or perception nodes, and teleoperation / remote-control logic are orchestrated. See Rubik Pi 3/ROS2.

Qualcomm Qdemo and IMSDK

How do I quickly see what this board can do?

Use the official Ubuntu image, install the required dependencies, and run Qualcomm Qdemo. According to the current official Ubuntu documentation, Qdemo provides a graphical interface for evaluating multimedia and AI sample applications on Ubuntu-based Dragonwing boards and uses QIMP / IMSDK components under the hood.

Documented sample categories include use cases such as:

recording live video,
multi-camera streaming,
object detection,
face detection,
multistream inference,
and compound AI demos that combine several processing stages.

That makes Qdemo a useful evaluation tool even for teams that eventually plan to replace it with a custom application.

sudo update-ca-certificates
git clone -b ubuntu_setup --single-branch https://github.com/rubikpi-ai/rubikpi-script.git 
cd rubikpi-script  
./install_ppa_pkgs.sh

Install deps and then:

gst-gui-launcher-app.py

Edge AI model paths

Edge Impulse

The current official Edge Impulse page describes it as a straightforward way to build edge AI models for Qualcomm Dragonwing devices. This is useful for teams that want a more guided path from data collection to deployment.

Qualcomm AI Hub

The current AI Hub page for RUBIK Pi 3 documents a model-discovery and deployment path based on QCS6490-targeted optimized models. It also notes an important deployment detail: accelerated NPU execution usually requires moving from a reference example to an optimized runtime path such as LiteRT / TFLite or ONNX.

LiteRT / TFLite and ONNX

For many vision products, these runtimes are where the model becomes a real application component. The software challenge is rarely just “run inference”; it is to run inference with the right preprocessing, the right memory movement, the right frame timing, and the right postprocessing so it integrates cleanly into video and robotics pipelines.

See Rubik Pi 3/AI and Computer Vision for a product-oriented interpretation.

ROS 2 and robotics software

The official robotics sample documentation for Ubuntu positions the QIR SDK as a robotics-enablement layer that includes reference ROS packages, scenario samples, and zero-copy transport mechanisms for Qualcomm robotics platforms. For RUBIK Pi 3 users, that means the board can sit in a broader robotics software architecture rather than being limited to standalone demos.

Practical reasons to use ROS 2 on RUBIK Pi 3:

structured message passing between sensors and compute,
easier separation of capture, perception, and control,
access to robotics ecosystem tooling,
and a clearer path to teleoperation or distributed robot systems.

See Rubik Pi 3/ROS2.

Package sources and customization

Several official Ubuntu pages reference public package sources and PPAs for platform packages such as `wiringrp` and QIR-related components. This is helpful for evaluation, but product teams should be deliberate about software ownership.

A strong productization workflow usually defines:

the exact base image version,
the package repositories that are allowed in production,
the set of packages mirrored internally,
the custom kernel or BSP delta if needed,
and the update strategy for fielded systems.

RidgeRun can help take a lab-proven RUBIK Pi 3 image and turn it into a supportable software baseline.

Build and customization strategies

Best for fast evaluation

Use the official Ubuntu image, validate cameras and network, run sample applications, and iterate from there.

Best for application prototyping

Layer GStreamer, AI runtimes, and your application code on top of a known-good image. Do not customize the kernel until you have a clear requirement.

Best for productization

Freeze the software bill of materials, reduce unnecessary packages, validate boot and recovery behavior, and define OTA, service, and logging strategy early.

Where RidgeRun fits

Software teams usually ask RidgeRun for help when any of these conditions appear:

camera hardware is working, but the image pipeline is unstable or incomplete,
the application needs zero-copy or low-latency GStreamer design,
AI inference works in a demo but not in the target multimedia architecture,
ROS 2 nodes need to coexist with camera and streaming paths,
or the evaluation image needs to become a controlled production image.

Relevant RidgeRun references:

Key takeaways

RUBIK Pi 3 now has a meaningful multi-OS story, not just a hardware pitch.
Ubuntu 24.04 is a strong default for Linux-native developer workflows.
GStreamer is the central integration point for multimedia and AI.
Qdemo, AI Hub, Edge Impulse, IMSDK, and QIR reduce evaluation time.
Productization still requires careful ownership of images, packages, and board support.

Frequently asked questions

Does RUBIK Pi 3 support Ubuntu?: Yes. Current official resources include Canonical Ubuntu 24.04-oriented documentation and a system-image path.
What operating systems are available for RUBIK Pi 3?: Current official image listings include Ubuntu, Linux, Android, and Debian.
What is the easiest way to evaluate AI and multimedia features?: Bring up Ubuntu first, validate camera and display, and then use Qdemo / IMSDK-oriented sample workflows as a fast evaluation path.
Can I use ROS 2 on RUBIK Pi 3?: Yes. The Ubuntu robotics documentation references QIR SDK workflows and ROS-based sample structures, making ROS 2 a practical option for robotics-oriented projects.
When should I move away from stock images?: Move beyond stock images when the project needs reproducible product builds, custom peripherals, long-term maintenance control, or tighter performance / security ownership.

Related pages

❯