Bench Talk

AMD Robotics Starter Kit Accelerates Design

(Source: metamorworks - stock.adobe.com)

Robotics is the combination of many engineering elements. Integrating new components and elements creates new changes to the framework. Another challenge is keeping high quality performance with framework changes. Adapting to these changes leads way for adaptive computing. Adaptive computing is the latest cutting-edge robotics component that improves robot’s response times, capability, and flexibility over industrial lifecycles. Robotics starter kits like the Kria™ KR260 Robotics Starter Kit from AMD Xilinx provide roboticists with building blocks to efficiently implement designs with ROS 2 . This can shorten design cycles by providing out-of-box ready development platforms that engineers can evaluate and prototype more rapidly, including machine vision, AI, robotics, industrial, communication and control.

While highly integrated processors, memory, and communications have simplified many of our responsibilities, the algorithms, and techniques used to make an autonomous robot extend far beyond a processor board. Adaptive computing goes beyond simple sequencers and state machines. AI and machine learning techniques help a machine figure out its next course of action when unexpected circumstances occur.

More recently, highly integrated System-on-Chip (SoC) and System-On-Module (SOM) solutions have allowed higher levels of sophistication using less space. A SOM can incorporate processors, I/O, wired and wireless interfaces, on-board memory, power management unit, and security in small form factor board. This saves all the time designing, prototyping, testing, and debugging basic features like Wi-Fi^®, USB, vision interfaces, and networking peripherals.

A critical area of concern is the use of sensors and the interpretations and decisions a robot will make from sensor data. The concern is not limited to mobile robots. With fewer workers and supply chain issues making manufacturing schedules dynamic, even factory robots and industrial machines need to meet the challenge.

Fortunately, engineers can tap into many powerful processors with single or multiple cores that perform individual and coordinated tasks. These can run open-source or developed software that allows design teams quickly evaluate various approaches and refine algorithms to suit their needs.

Robotics operating systems like ROS 2 provide software libraries and tools for navigation, motion control, machine vision, and 3D visualization tools like RVIS, which are very useful for implementing hazard identification and avoidance. The engineering team must still make choices best suited to the applications.

For example, do you implement high-resolution video systems and lighting to use and decode for navigation, or do you use Lidar, reflective optical, contact switches, or ultrasonic distance measurement for range finding and object avoidance? All of these are possible, and these approaches have modular technical kits that allow you to prototype, test, and evaluate the specific technology.

Often, combinations of technologies can leverage the benefits of each technology and make a better result than either technology by itself. For example, combining a high-end processor with a high-performance FPGA can allow the fixed program performance of the process to take advantage of the hardware acceleration of tasks possible with an FPGA. Not only can this provide better performance for the end product, but it also allows designers to quickly experiment and test varied approaches without extensive recoding or board layout iterations.

Adaptive computing conforms to the changing demands of robotics applications. Adaptive computing needs reliable base of rules and problem-solving techniques to adapt to new challenges safely, efficiently and intelligently.

The AMD Xilinx Kria KR260 Robotics Starter Kit targets adaptive computing by providing a high-level platform for robotics. Kria KR260 combines Arm® cores for application processing, programmable logic for real-time processing and control offered by the Kria K26 SOM and includes pre-built interfaces to quickly prototype robotics and industrial applications. Offered in Commercial and Industrial grades, the Kria K26 SOM (Figure 1) is ideal for production deployment in Robotics, Embedded Vision, and Machine Vision applications.

Figure 1: The AMD Xilinx Kria KR260 Robotics Starter Kits combines high-performance industrial interfaces and features for engineers to test and evaluate intelligent autonomous robot solutions. (Source: Mouser Electronics)

When getting started, no proprietary tools or AMD Xilinx development software are needed. AMD Xilinx states the entry level developers can be up and running in under an hour. The AMD Xilinx Apps Store provides accelerated applications like the ROS 2 Perception Node, and designers can use the ROS 2 framework and/or code from Python, C++, and FPGA RTL.

The Starter Kit is based on AMD Xilinx K26 SOM with 4GB of DDR4 memory, on-board power supply, boot option, TPM 2.0 for enhanced security and connectors that map IOs for carrier card (Figure 2). What's more, these clock up to 1.33GHz (TOPs). Logic at these speeds will consistently outperform code-driven decision-making. The K26 SOM would plug into a custom carrier card for Vision AI, Robotics, Industrial Communication, Control and many more applications.

Figure 2: The highly integrated KR260 Robotics Starter Kit contains many hardware resources for communications, computations, sensor interfacing, motion control, and adaptive learning. (Source: AMD Xilinx)

Standard JTAG programming and debugging provide access to all the embedded ports and peripherals like the 4xUSB 3.0 ports that can be used for camera interfaces like SLVS-EC and other interfaces that require up to 10GB/sec communications like 5G ports. A built-in 1920 x 1080 resolution display port can be invaluable for development and debugging. The two industrial Ethernet ports allow high-speed wired connectivity to Ethernet-enabled peripherals like motor control and position feedback systems.

The multiple Micro USB, UART/JTAG, Pmods and Raspberry Pi extension header are ideal for interfacing with numerous low bandwidth sensors and indicators for operational status. These development boards even have built-in Micro SD slots that can be used for development purposes and data logging.

As technology advances, so must the tools engineers have to develop. The Kria KR260 Robotic Starter Kit is the perfect way to get into autonomous robot development.

The hardware acceleration discussed here not only shortens design and development time but can also make the finished product perform much faster than code-only executed approaches. With all its robotics focused features and capabilities, KR260 provides an excellent development platform for the production ready K26 SOM to accelerate just about any robotic or intelligent industrial applications.