SoC aims at robotics apps with 4G/LTE connectivity, on-device ML

Qualcomm Technologies, Inc. announced its Qualcomm Robotics RB3 Platform – the company’s first integrated, comprehensive offering designed specifically for robotics. This purpose-built platform features ahighly optimized set of hardware, software and tools designed to help manufacturers to create the next generation of advanced consumer, enterprise and industrial robotics products.

Based on the Qualcomm SDA/SDM845 system-on-chip (SoC), the platform integrates key capabilities such as high-performance heterogeneous computing, 4G/LTE connectivity including CBRS support for private LTE networks, a Qualcomm AI Engine for on-device machine learning and computer vision, hi-fidelity sensor processing for perception, odometry for localization, mapping, and navigation, vault-like security, and Wi-Fi connectivity. The Qualcomm Robotics RB3 Platform also plans to introduce 5G connectivity support later this year to further enable low-latency and high-throughput industrial robotics applications.

The Qualcomm Robotics RB3 Platform is designed to allow for flexible design options for development and commercialization, from development board offerings for prototyping, to off-the-shelf system-on-module solutions for faster commercialization, to the flexibility for chip-on-board designs for cost optimization at scale. The platform currently supports Linux and Robot Operating System (ROS), while also including support for the Qualcomm® Neural Processing software development kit (SDK) for advanced on-device AI, the Qualcomm® Computer Vision Suite, the Qualcomm Hexagon DSP SDK, and Amazon’s AWS RoboMaker, with plans for Ubuntu Linux support.

The platform’s hardware development kit contains the new purpose-built robotics-focused DragonBoard™ 845c development board, based on the Qualcomm SDA/SDM845 SoC and compliant with the 96Boards open hardware specification to support a broad range of mezzanine-board expansions. Optional elements for the kit include a connectivity board; an image camera for superb hi-res photo, 4K video capture and AI-assisted detection and recognition of people and objects; a tracking camera for path planning and obstacle avoidance using visual simultaneous localization and mapping (vSLAM); a stereo camera for navigation; and a time-of-flight camera for people, gesture and object detection even in low light conditions.

Key Qualcomm Robotics RB3 Platform technical features:

• Heterogeneous computing architecture: For superior performance and power-efficiency, the Qualcomm SDA845/SDM845 SoC powering the platform is built on 10-nanometer (nm) LPP FinFET process technology. The SoC integrates an Octa Core Qualcomm® Kryo™ CPU with performance up to 2.8GHz, a Qualcomm® Adreno™ 630 Visual Processing Subsystem (including GPU, VPU and DPU), and a Qualcomm® Hexagon™ 685 DSP with Hexagon Vector Extensions (HVX), delivering sophisticated, on-device AI processing and mobile-optimized computer vision (CV) capabilities for perception, navigation and manipulation.
• Qualcomm AI Engine: Deep learning performance across the Qualcomm AI engine (CPU, GPU, and DSP) delivers up to three tera-operations-per-second (TOPS) performance, with the DSP itself delivering 1.2 TOPS at one watt with hardware acceleration. Includes the Qualcomm Neural Processing SDK, which encompasses analysis, optimization, and debugging tools designed to allow developers and manufacturers to port trained deep learning networks on the various heterogeneous compute blocks offered by the platform.
• Camera and video: Dual 14-bit Qualcomm Spectra™ 280 ISP supporting up to 32 MP single camera; support for up to 4K HDR video capture at 60 frames per second (fps).
• Security: The Qualcomm® Secure Processing Unit (SPU) offers a high level of security and robustness, while also providing high performance while maintaining power efficiency.  The SPU includes the following key components: secure boot, cryptographic accelerators, Qualcomm® Trusted Execution Environment (QTEE), and camera security. To address advanced AI, ML, and biometrics, Qualcomm SDA/SDM845 supports porting of virtualization software.
• Robust Sensor and Microphone Support: The platform includes support for sensors like a 6-axis inertial measurement unit (IMU) consisting of a 3-axis gyroscope and a 3-axis accelerometer; capacitive barometric pressure sensor; multi-mode digital microphones; and additional ports for supplementary sensors from TDK-InvenSense.
• Connectivity: Supports integrated 4G/LTE and CBRS, with 5G support planned to be enabled late this year; Wi-Fi integrated 802.11ac 2×2 with MU-MIMO; Tri-band Wi-Fi: 2.4 GHz and 5 GHz with Dual Band Simultaneous (DBS); and Qualcomm TrueWireless™ Bluetooth 5.0.

Commercial products based on the Qualcomm Robotics RB3 Platform are expected to be available in 2019. Both NAVER and LG are evaluating the Qualcomm Robotics RB3 Platform, and plan to showcase select robotics products based on the platform early next year.