Qualcomm Puts AI and 5G in the Hands of Robotics Designers
Surging 15 trillion operations per second, Qualcomm’s RB5 platform is bringing AI and deep learning workloads to the edge of robotics.
Qualcomm, a leader in wireless technology, has recently released the world's first 5G and AI-enabled robotics platform.
Qualcomm says the Robotics RB5 platform includes novel technology that will accelerate the development of power-efficient, high-computing robots and drones for applications ranging from industrial and enterprise applications to defense and military sectors.
RB5 development kit. Image used courtesy of Qualcomm
Specifically, some of the applications this platform can enhance include inventory robots, retail robots, pick-sort-place robots, cleaning robots, delivery drones, healthcare robots, and defense robots.
Qualcomm’s Robotics RB5 platform, punctuated by their QRB5165 processor, supports a number of connectivity protocols through Qualcomm's FastConnect subsystems: Wi-Fi 6, Bluetooth 5.1, 4G, and 5G.
AI, ML, and the Upgrade from RB3 to RB5
According to Qualcomm’s press release, the platform features artificial intelligence (AI) and machine learning (ML) capabilities while still minimizing power consumption. This platform has been designed specifically for applications in robotics, and it comes with hardware, software, and other development tools to help designers with such applications.
RB5 is said to be an improvement from the company’s RB3 platform, which is still used in industry today.
RB3 platform block diagram. Image used courtesy of Qualcomm
The RB5 is the first of its kind to implement 5G technology, and there are already over 20 early adopters. More than 30 companies are developing the necessary hardware and software for these technologies as well.
Tools for Comprehensive Robotics Designs
Qualcomm’s product page shows the myriad of capabilities for the RB5 platform. First, it supports software for Linux, Ubuntu, and ROS (Robot Operating Systems) 2.0. It also has neural processing and FastCV Computer Vision SDKs, which optimize deep learning performance and offer a mobile-optimized computer vision library, respectively.
The platform accounts for a variety of user interactions with a device, including gesture and facial recognition, text recognition, tracking, and augmented reality. In addition, it utilizes visual simultaneous localization and mapping (visual SLAM), which allows the robot to both map the area around it while locating itself in the map it has made using only cameras.
The platform utilizes multiple digital signal processing (DSP) units for computation, audio and sensor signals, ISPs, a dedicated AI engine called the Hexagon Tensor Accelerator and a dedicated computer vision block called the Enhanced Video Analytics Engine.
Highlights of the RB5 Platform
Some of the RB5 platform's key specifications include:
- CPU clock speed up to 2.84 GHz
- Dual 14-bit ISPs
- 16 GB, LPDDR5, LPDDR4x, 2133MHz, 2750MHz memory
- Dual-Band Simultaneous (DBS) capabilities
Qualcomm seemed to pay special attention to detail with the platform's cameras. RB5 includes a single camera with zero shutter lag (ZSL), 3fps, and up to 64 megapixels (MP). The dual camera in the platform also features ZSL, 30fps, and up to 25 MP. These cameras include security features as well.
The RB5 platform. Image used courtesy of Qualcomm
When it comes to video capabilities, Qualcomm says the platform is capable of up 8K video capture, 4K HD capture, simultaneous 64 MP photo capture, and 10-bit color depth video capture.
Security was paramount in designing this platform, and as such, Qualcomm prioritized built-in security in the cameras as well as a crypto engine, accelerator, and secure boot.
Simplifying the Complexity of Robotics Design
With this platform, Qualcomm says robotics designers can perform multiple high complexity processes while consuming a minimal amount of power. Beyond this, the connectivity power of 5G drives this device even further, allowing for capable, smart, and fast robot functionality.
Do you have experience in the robotics industry? How have you seen key design measures change with the rise of machine learning? Share your experiences in the comments below.