24-07-2020 | | By Robin Mitchell
Qualcomm announced the release of their latest robotic platform, the RB5, which provides multiple improvements over its predecessor, the RB3, while also integrating a whole range of new features. Learn all about the Qualcomm robotics RB5!
The robotics industry has come a long way since the first use of automated systems in industrial environments. Tasks completed by people can be done by robotic systems in a fraction of the time while not require a heated environment or even lighting. The ability for robotic systems to automate tasks makes them a desirable investment with their ability to reduce labour costs, operating costs, and improve consistency. However, many robotic systems lack intelligence and generally follow a basic pattern of operation. For example, robotic arm systems used to pack items into boxes may not recognise if an item has fallen out of its grip, or if the lid to that item has fallen off.
Artificial Intelligence (AI) can significantly improve the performance of a system thanks to its ability to learn from mistakes. For example, a robotic system equipped with a camera can use object recognition to determine if an object has been picked up correctly. The use of additional sensors can then be trained to identify if the arm has dropped an item, and an AI system can combine all data to make determinations as to the best way to pick up an object.
AI is a potent tool, but being powerful also means that it requires a decent amount of processing power. The amount of processing power needed depends on the complexity of the AI system, but generally speaking, practical AI applications struggle to run on microcontrollers and system-on-a-chips (SoCs) aimed at small portable devices. However, the advantages of AI in portable and edge devices has resulted in multiple silicon manufacturers creating AI cores that specifically run pre-trained algorithms. This means that the main processor can be left to run an operating system and other essential tasks, while the AI core handles artificial intelligence routines.
Qualcomm, like many others, recognise the importance of integrating AI systems into robotic devices and thus have developed their RB5 robotics platform. The RB5 platform is aimed at developing low-powered robotic systems that require 5G connectivity and AI capabilities in a single package. A range of software support packages such as Qualcomm's Neural Processing SDK and Qualcomm Computer Vision SDK help robotic systems work with AI and video-based applications while the Hexagon DSP SDK allows for optimisation of multimedia applications.
The hardware side of the Qualcomm RB5 robotics development kit platform is the QRB5165 processor which integrates a whole range of features for use in robotics. The core of the QRB5165 is the Qualcomm Kryo 585 CPU which integrates eight cores clocked up to 2.84GHz, while various DSPs are integrated for signal processing including a Qualcomm Hexagon tensor Accelerator, Qualcomm Hexagon 689 DSP, and a Qualcomm Hexagon Vector eXtensions (HVX). The AI capabilities of the 5th generation Hexagon Tensor AI engine allows for up to 15 trillion operations per second, allowing for both running pre-learned AI algorithms and the learning of new ones.
The robotic processor integrates Wi-Fi 6 and 5.1 Bluetooth communications while also supporting 5G connectivity. Visual stimuli is an incredibly important feature in modern robotics which is why Qualcomm developed the RB5 platform to handle up to 12 cameras using D-PHY and 18 cameras using C-PHY. The Qualcomm Spectra 480 image signal processor allows for 30fps from a single camera up to 64MP or dual cameras at 25MP. The image processor can also handle 8K video capture while supporting up to 10-bit colour depth.
Security on modern devices is becoming ever more critical, which is why Qualcomm has integrated a range of security feature into their RB5 platform. Such security features include Camera Security, Crypto Engine, Cryptographic Accelerator, and Secure Boot. Being targeted at robotic systems, the RB5 platform supports industrial temperatures ranging from -30ºC to +105ºC.
The RB5 platform integrates everything a designer could want in a smart robotic system with the capability to not only understand its environment but learn from it. The RB5 platform will allow engineers to create secure robotic systems that can operate in industrial environments while providing a wealth of valuable information about their performance.