top of page
Writer's pictureLuis G. Leon-Vega

Build your own drone with Qualcomm RB5/RB6 and RidgeRun

Updated: Jun 10




The Qualcomm RB5 is an exciting new low-power embedded system that shows great promise for drone applications. With its optimized combination of performance and power efficiency, the RB5 brings capabilities previously only found in higher-end systems down to the mid-range market. 


Equipped with features like an octa-core CPU, GPU, DSP, AI acceleration, and 16GB of RAM, the RB5 has the power needed to take on intensive workloads like video processing, computer vision, and neural network inferencing. This opens up new possibilities for on-device processing in autonomous drones and other robotics applications where power consumption needs to be minimized.


With its compelling blend of mid-range pricing and high-end features, the Qualcomm RB5 represents an exciting step forward for embedded drone development. Its potential is only starting to be explored, but the RB5 is already demonstrating itself as a highly promising chipset for the next generation of smart, capable drones.


Drone Demo Overview


We have developed a drone demo that showcases the capabilities of the Qualcomm RB5 when integrated with RidgeRun's optimized products. The demo involves transmitting video and metadata from the drone over a network connection using GStreamer pipelines. 


Specifically, we are using the RTSPSink and GstSEI GStreamer plugins to transmit the video stream and embed metadata in the stream. This allows us to send real-time video and telemetry data from the drone to a ground station computer.


We are also utilizing RidgeRun's GstPTZR plugin to perform some basic image processing on the video feed from the drone's camera. GstPTZR enables tasks like pan/tilt/zoom/rotate within a GStreamer pipeline without needing extra encoding/decoding steps. In our demo, we use it to crop the video stream from the drone before transmitting it over the network.


By leveraging the RB5's powerful processing capabilities and RidgeRun's GStreamer integration, we can transmit high quality video and metadata from the drone in real-time while also performing image enhancement on the video feed. The demo showcases how the Qualcomm RB5 is well suited for drone applications when paired with RidgeRun's multimedia optimizations.


Demo Implementation




The RB5's power enabled us to run multiple components for video streaming, metadata transmission, and image processing in our drone demo. A key element was utilizing GStreamer pipelines for efficient streaming and computer vision.


We leveraged RTSPSink in GStreamer to transmit the video feed from the drone camera over the network via RTSP. This provided low latency and reliability for the video stream. 


For transmitting drone telemetry and metadata, we used GstSEI in the GStreamer pipeline. This allowed embedding metadata like GPS coordinates, altitude, etc directly in the video stream using SEI NAL units. The metadata could then be parsed on the receiving end.


To perform image processing on the video feed, we used GstPTZR. This enabled running effects like panning, tilting, zooming and rotation in real-time on the drone's video feed before transmission.


Overall, GStreamer provided an optimized streaming media framework that could take advantage of the RB5's multicore processing, integrated GPU and DSP. This was instrumental in transmitting high quality video and metadata from the drone.


AI Potential


The Qualcomm RB5 offers immense potential for AI inference capabilities that have not yet been fully utilized. While the demo showcased real-time video streaming and processing, the RB5 is capable of much more advanced on-device AI.


You can get more information at www.ridgerun.ai about our Embedded AI solutions and connect with the experts.


Demo Results


The capabilities of the Qualcomm RB5 platform integrated with RidgeRun's optimized software are demonstrated in the following video:



We are able to achieve real-time video streaming and metadata transmission over the network using GStreamer plugins like RTSPSink and GstSEI. The onboard image processing with GstPTZR also showcases the potential for more advanced features like AI inference on the RB5.

The optimized combination of RidgeRun software and Qualcomm's high-performance RB5 delivers an ideal embedded solution for drone and robotics applications needing substantial compute resources while maintaining low power consumption.


The full source code for the drone demo implementation is also available on demand. Please, write to www.ridgerun.com/contact for more information.


We also provide a wiki with documentation, tutorials, and API references for getting started with the RB5. Whether you want to replicate the demo or build your own custom UAV, robotics, or IoT applications, this knowledge base provides a valuable starting point.


Conclusion


The Qualcomm RB5 shows great promise for drone applications due to its impressive performance capabilities in a low-power embedded system. This demonstration integrating the RB5 with RidgeRun optimized products illustrates the potential for real-time video streaming and analytics in an aerial system. 


Key highlights from the demo include transmitting HD video and metadata over a network connection in real time while performing image processing algorithms on the video stream. This was accomplished using GStreamer pipelines leveraging RTSPSink, GstSEI, and GstPTZR.


With the powerful heterogeneous compute available on the RB5 system-on-chip, including an AI accelerator, there is significant headroom to integrate more advanced features into similar drone pipelines. AI inference at the edge on streaming video data is an exciting next step to enable intelligent drones that can analyze their surroundings in real time. Additional capabilities like computer vision, object detection and tracking, and autonomous navigation are now within reach for small, low-power drones thanks to system-on-chips like the Qualcomm RB5.


This demonstration only begins to tap into the possibilities enabled by the impressive performance per watt of the RB5. With RidgeRun's optimization expertise in embedded Linux software stacks and the Qualcomm RB5 providing a capable hardware foundation, innovative new drone applications are on the horizon.

63 views
bottom of page