A scalable approach to enabling intelligent and adaptive robotics

As demand for feature-rich, AI-enabled robots explodes, one approach offers the most capability and development efficiency

Robotics, when compared to other industrial automation systems, is the equivalent of a pizza with all the toppings, because of the emerging need to concurrently pull from all subject matters across the high-tech landscape. Industrial controls, communications, machine vision, artificial intelligence, human-machine interfaces, cyber-security and functional safety are now common technology considerations when building cobots, industrial robots, or other commercial applications of robotics. Ignoring even crucial topics such as material science and other physical aspects of these machines, the embedded electronics alone make robotics development and deployment a complex and time-consuming mission.

The bottleneck in deploying robotics with all these functions is the integration of components and planning of behaviours. Often, an adjustment in one area has a ripple effect on the fine-tuning in other areas, sending the robotics developers themselves into loops.

The Xilinx Zynq portfolio of SoCs (System-on-Chips) significantly reduces the integration burden by enabling a modular approach to robotics, by providing a common embedded hardware and software platform that can be reused and scaled to dramatically reduce time-to-market, form factor and total cost of ownership, while maximising intelligence and adaptability across all intelligent nodes of these assets. Furthermore, robotic platforms, like other industrial IoT edge platforms, require common building blocks such as network interfaces, imaging filters, operating system drivers, cyber-security solutions, support for popular analytics and machine-learning frameworks and much more. These are typically non-differentiating items and Xilinx is focused on delivering these critical elements of the overall solution as represented in the industrial IoT solution stack (see image).

The Xilinx Industrial IoT Solution Stack, starting from the chip level up to the application level, makes a number of key robotics functions accessible, such as precise, deterministic control over multiple axes of motion via parallel processing of control loops and connectivity over diverse Industrial Ethernet standards (for example, CC Link, EtherCAT, EtherNet/IP, EthernetPOWERLINK, Mechatrolink-III, Profinet, Sercos-III, others), including support for deterministic ethernet in the form of Time-Sensitive Networking (TSN). Additionally, Xilinx provides support for numerous and diverse sensor inputs, from simple MEMS sensors such as accelerometers to high-resolution 4K cameras. These diverse sensor inputs can then be used for onboard artificial intelligence that supports predictive control, object detection, self-diagnosis with self mitigation and other advanced inferential tasks via Python libraries such as NumPy, ROS, or even embedded deep neural networks. All these intelligent behaviours would still be governed by world-class compliance for IEC 61508 SIL 3 functional safety and IEC 62443 cyber-security, also part of the Xilinx and ecosystem solution set. A key benefit of this physical and functional integration is a smaller physical and power footprint.

The Xilinx Industrial IoT Solution Stack is just the beginning of Xilinx’s push into modular robotics, with more solutions on the horizon powered by the Zynq portfolio of SoCs, which feature one to four Arm application processors and lots of other goodies. What makes Zynq unique is its ability to take virtually any task and assign it to hardware or software, or a custom blend between the two domains, using the Xilinx SDx™ or the PYNQ frameworks. This famously gives the user three degrees of freedom: programmable hardware, programmable software and optimised accelerator blocks for signal processing, encryption, and much more. Don’t feel self-conscious about the embarrassment of resources available in Zynq when compared with the traditional approach of just programmable software on a lightweight embedded processor tethered to a clunky PC. Without this flexibility and capability, you might not be able to fit all the toppings you want on your pizza.

To learn more about the Xilinx Zynq portfolio of SoCs and their value in robotic platforms and artificial intelligence, please be sure to check out these informative whitepapers:

Key Attributes of an Intelligent IIoT Edge Platform

The Value of Python Productivity:  Extreme Edge Analytics on Xilinx Zynq Portfolio

Deep Learning on Xilinx Devices

Originally published in Business Reporter Online: June 2018