# Architecture
Source: [https://docs.qualcomm.com/doc/80-88500-4/topic/6_Architecture.html](https://docs.qualcomm.com/doc/80-88500-4/topic/6_Architecture.html)
The Robotics RB5 Platform supports both Linux Ubuntu software and Linux Embedded
software.
The figure shows the architecture of high-level operating system (HLOS – Linux Ubuntu) and
non-HLOS, and their interaction with various application software. For the hardware
architecture, see *QRB5165 architecture overview* in QRB5165 Technical Reference
Manual (80-PV086-5).
Figure : Software architecture (Linux Ubuntu)
Figure : Software architecture (Linux Embedded)
The Qualcomm Robotics RB5 platform includes support for multiple SDKs and tools, including
the Qualcomm Neural Processing SDK for AI, Qualcomm Robotics Vision SDK, Qualcomm Computer
Vision SDK, Hexagon DSP SDK, and robotics operating system (ROS).
These SDKs enable developers to implement the following features of robotics:
Note: Robotics features are not yet enabled on Linux Embedded
systems.
- Object tracking, detection, and classification
- Object depth estimation and obstacle avoidance
- Visual simultaneous localization and mapping (vSLAM)
- Stereo-vSLAM support
- Mono-vSLAM support
- Visual inertial odometry (VIO)
- Path planning and 3D-map formation
- Face detection and recognition
- Scene classification and understanding
- Voice activation and audio classification
- Deep learning
There are multiple subsystems within Qualcomm Robotics RB5. The following are some of the
salient features in this platform:
- Multimedia subsystems (MMSS) comprises display, graphics, audio,
video, and camera can be accessed from the user space
- Qualcomm GStreamer (GST) plugins to access the video and camera
subsystems using the Qualcomm Multimedia framework (QMMF)
- Qualcomm-customized Pulse GStreamer plugins can be used for the audio functionality
- Display can be Wayland; access is directly through the Wayland protocol for Weston
applications
- Support for ROS2.0; the Robotics Vision modules interact with the
ROS2.0 per the interface specification of ROS2.0
- Sensor data is available through the sensor ROS nodes
- Qualcomm-specific artificial intelligence (AI) and machine learning (ML) components are
included in the platform
- Qualcomm Neural Processing SDK is accessible via a GStreamer plugin
- Support for TensorFlow Lite (TFLite) and associated GST plugins are added to the platform;
standard TFLite applications can be executed in the shell
- Support for the Qualcomm Computer Vision SDK
- Support for the Qualcomm^®^ Hexagon™ DSP
- The standard Ubuntu-based application packages can be installed via
`apt-get` command on Ubuntu Platform. For Linux Embedded you can do local
installs using yocto IPK mechanism.
The QMMF provides access to low-level APIs using the GST elements. It is recommended to use
GStreamer for accessing MMSS.
For more information on Qualcomm GStreamer plugins, see QRB5165.LU.2.0 Qualcomm Intelligent
Multimedia SDK (IM SDK) Reference (80-88500-50).
Note: For more information on GStreamer, see GStreamer Tutorials at [https://gstreamer.freedesktop.org/documentation/tutorials/index.html?gi-language=python](https://gstreamer.freedesktop.org/documentation/tutorials/index.html?gi-language=python).
Last Published: Aug 18, 2023
[Previous Topic
Introduction](https://docs.qualcomm.com/bundle/publicresource/80-88500-4/topics/1_Introduction.md) [Next Topic
System](https://docs.qualcomm.com/bundle/publicresource/80-88500-4/topics/7_System.md)