Your browser is not supported. Please update it.

22 March 2019

Qualcomm applies its integration skills to a broad robotics platform

Qualcomm has targeted robotics as a use case for some of its chips for years now, but it has gone one step further and is promising an integrated solution, with a hefty AI engine, which supports all three of a robot’s activities – sensing, thinking and acting.

This was how the company’s product line manager for robots, drones and intelligent machines, Dev Singh, described the new Robotics RB3 platform, which launched at the Embedded World show in Nuremberg two weeks ago. The offering has been in development for two years and is being tested by early customers, including Samsung.

The aim, said Singh, is to make it easier and cheaper for companies to develop robots so that they can proliferate in the home, in industry and in entertainment. The platform can enable low end, mass market robots, but can also scale to very sophisticated models, which may be 5G-enabled to support free mobility across large campuses or factory complexes.

He told EETimes: “We already had the thinking part, so we added sensors, which is an important part of the equation. For robotics, you also need a heterogenous computing platform.”

Qualcomm has taken the robotics chips it already has and added options for AI processing, edge computing and various connectivity standards. While the firm’s products are used in some commercial robots today, mainly in consumer-focused segments like multimedia or household help, the objective with RB3 is to “help spur the fast development and commercialization of a new generation of useful and intelligent robots in agriculture, consumer, delivery, inspection, service, smart manufacturing, warehousing and logistics, and other applications.”

The platform is based on the Qualcomm SDA/SDM845 system-on-chip (SoC), and also includes:

  • LTE connectivity including support for the USA’s CBRS shared spectrum, for private networks, plus WiFi
  • An AI Engine for on-device machine learning and computer vision
  • Hi-fidelity sensor processing for perception
  • Odometry for localization, mapping and navigation
  • Support for Linux and Robot Operating System
  • Support for a range of software development kits (SDKs) including Qualcomm’s Neural Processing SDK, Computer Vision Suite, and Hexagon DSP SDK, plus Amazon’s AWS RoboMaker

Later this year, 5G will be added to push the platform into low latency industrial robotics applications, and Ubuntu Linux will also be included as an OS choice.

This integrates an eight-core CPU based on the Kryo micro-architecture clocked at up to 2.8 GHz; with an Adreno 630 Visual Processing Subsystem and a Hexagon 685 digital signal processor with Hexagon Vector Extensions for on-device AI processing and computer vision.

A development kit based on the RB3 Platform is available now starting at $499 and commercial products are expected towards the end of 2019.