As more and more devices connect to the internet, edge computing has grown in popularity in recent years. In contrast to transmitting data to a centralized data center, edge computing involves processing data closer to the source. Many advantages of this strategy include decreased latency, enhanced security, and reduced bandwidth needs. Yet, edge computing also comes with some particular difficulties, like low processing speed and memory. Xilinx has created a brand-new platform called Kria to manage AI for edge applications to overcome these difficulties.
What is Xilinx Kria?
A series of adaptive system-on-modules (SOMs) called Xilinx Kria can speed up AI at the edge. These SOMs are compact, production-ready cards that are simple to incorporate into various edge technologies. Examples include robots, embedded vision systems, and smart cameras. The Zynq UltraScale+ MPSoC, which combines an FPGA and a potent Arm CPU, is the foundation upon which we build the Kria SOMs. Thanks to this hardware configuration, the Kria SOMs can handle both conventional software processing and hardware acceleration for AI workloads.
There are two versions of the Kria SOMs: commercial and industrial. The industrial version is essential in hostile settings and demands ruggedized hardware. On the other hand, the commercial version is ideal for applications requiring high performance and low power consumption. In addition, both Kria SOM variants offer a choice of Yocto or Ubuntu Linux infrastructure, giving software programmers a comfortable working environment.
Benefits of Xilinx Kria
The versatility of Xilinx Kria is one of its key advantages. Developers may add AI capabilities to their products since the Kria SOMs are simple to connect to various edge devices. Also, various AI frameworks, such as TensorFlow, PyTorch, and Caffe, are supported by the Kria SOMs, making it simple for developers to use their chosen work methods and tools.
The performance of Xilinx Kria is an additional advantage. The Kria SOMs can perform well using less power than conventional CPUs or GPUs. They can handle traditional software processing and hardware acceleration for AI applications. As a result, the Kria SOMs are perfect for usage in battery-operated devices or other situations where power consumption is an issue.
Last but not least, Xilinx Kria is made simple to use. Developers may easily start using AI at the edge thanks to various software tools and frameworks in the Kria SOMs. Moreover, Xilinx offers a variety of reference designs and application notes that can assist designers in building and deploying AI-enabled edge devices fast.
How developers can get started with Xilinx Kria
Xilinx Kria developers can start by utilizing the software resources and tools. One choice is using Vivado, the beginning point for any unique FPGA design on a Xilinx processor. Another collection of software tools called PetaLinux is available for creating an embedded Linux image for Xilinx FPGAs. It is simpler to use Xilinx platforms when developers can take advantage of the Python language and libraries through the PYNQ framework. To assist developers in getting started with Kria, Xilinx also offers a variety of reference designs, application notes, and tutorials. Moreover, the Xilinx App Store offers Kria Accelerated Applications, which allow novice and seasoned Xilinx designers to avoid undertaking any FPGA design.
Minimum hardware requirements for developing with Xilinx Kria
The Kria System-on-Module (SOM) and a carrier card are the two pieces of hardware required at the absolute minimum for development with Xilinx Kria. The XCK26 Zynq MPSoC chip, which houses both the Processor Subsystem (PS) and Programmable Logic, is used by the Kria SOM (PL). The carrier card may be a KV260 from Xilinx or a special carrier card. To create the application software that runs on the APU and RPU PS targets, developers require a machine with software tools like Vivado, PetaLinux, or other open-source tools (like Yocto).
Vitis, PetaLinux, or other open-source software development tools may create the application software. Developers can also use the AMD-Xilinx PYNQ framework, an open-source project that simplifies the use of AMD-Xilinx platforms using Python and resources. Canonical’s Ubuntu Linux with AMD-adaptive Xilinx’s SoCs now has enterprise-grade Linux, accelerating the creation of Vision AI Edge Systems, Industrial and Healthcare IoT embedded devices, and other products.
Use cases of Xilinx Kria
A family of adaptive system-on-modules (SOMs) called Xilinx Kria takes advantage of AMD programmable technology’s strength, speed, and adaptability. These SOMs are production-deployable, adaptive modules that offer a complete board-level solution with a pre-built software stack, enabling speedy deployment. As a result, the Kria SOMs can hasten industrial-grade robotic solution development, upkeep, and commercialization. In this post, we’ll examine Xilinx Kria’s use cases and how it helps hasten the creation of AI edge applications.
Kria Robotics Stack (KRS)
An integrated collection of robot libraries and utilities called the Kria Robotics Stack (KRS) uses hardware to hasten the creation, upkeep, and commercialization of industrial-grade robotic solutions. It suggests a development strategy that is ROS 2-centric and utilizes ROS 2 as the Software Development Kit (SDK). In addition, a collection of pre-made software components from the KRS may help to create robotic applications. These elements include simulation and visualization tools, perception and control algorithms, and sensors and actuator drivers.
The Kria SOMs, which offer the hardware acceleration required to execute sophisticated robotic algorithms, are intended to be utilized with the KRS. The Xilinx Zynq UltraScale+ MPSoC, which combines a quad-core ARM Cortex-A53 CPU with a programmable logic fabric, is the foundation upon which we build the Kria SOMs. With this programmable logic fabric, we may execute algorithms for image processing and machine learning more quickly.
A collection of pre-made software components from the KRS may help to create robotic applications. These elements include simulation and visualization tools, perception and control algorithms, and sensors and actuator drivers. In addition, a collection of tools for developing and deploying robotic applications are also available through the KRS. These tools include a build system, a package manager, and a deployment tool.
Machine Vision AI
Systems-on-modules (SOMs) for Xilinx Kria for AI edge applications offer a cost-effective, production-ready FPGA chip with ample IO to speed up robotics and vision operations. As a result, the Kria SOMs can hasten the creation of AI-edge applications like robotics and machine vision.
The Xilinx Zynq UltraScale+ MPSoC is the foundation for building the Kria SOMs. It combines a quad-core ARM Cortex-A53 CPU with a programmable logic fabric. Moreover, the Kria SOMs include a selection of IO ports like Ethernet, USB, and HDMI that can connect to sensors and actuators.
The Kria Vision AI Starting Kit works well with the Kria SOMs. It offers a full development environment for machine vision and robotics applications. A carrier board, a collection of pre-built software components, and a Kria SOM are all included in the Kria Vision AI Starting Kit. These pre-built software components include tools for simulation and visualization, image processing and machine learning algorithms, and drivers for sensors and actuators.
An additional tool for developing and deploying robotics and machine vision applications is the Kria Vision AI Starting Kit. These tools include a build system, a package manager, and a deployment tool. In addition, the Xilinx Vitis AI development environment, which offers a collection of tools for creating and deploying machine learning models on the Kria SOMs, is helpful with the Kria Vision AI Starting Kit.
Accelerated Applications
Many applications, including robotics, machine learning, and image processing, can be accelerated using Xilinx Kria. The Xilinx Zynq UltraScale+ MPSoC, which combines a quad-core ARM Cortex-A53 CPU with a programmable logic fabric, is the foundation upon which we build the Kria SOMs. With this programmable logic fabric, algorithms for image processing and machine learning may execute more quickly. Moreover, the Kria SOMs include a selection of IO ports like Ethernet, USB, and HDMI that can connect to sensors and actuators.
Applications like smart cameras, embedded vision, security, retail analytics, smart cities, and machine vision can all be accelerated using Kria SOMs. A development platform that is ready to use right out of the box for AI-embedded SW and HW developers is the Kria KV260 Vision AI Starting Kit. It is ideal for Visual AI applications. Additionally, it is the quickest approach to creating distinctive deployment solutions in production volumes. With the Kria Robotics Stack, the Kria SOMs can also quicken the creation, upkeep, and commercialization of industrial-grade robotic solutions (KRS).
Healthcare
Medical imaging, patient monitoring, and drug development are just a few applications. They can accelerate in the healthcare industry with Xilinx Kria. Also, the sensors and actuators commonly helpful in patient monitoring can interface with using the Kria SOMs.
The Kria KV260 Vision AI Starting Kit can help to create healthcare applications focusing on vision. With a non-production K26 SOM put into a carrier board, the Starting Kit offers general-purpose development interfaces, flexible I/O connectivity via Ethernet and USB, and other features. In addition, the Xilinx App Store’s expanding ecosystem of accelerated applications enables developers of all stripes to launch applications in under an hour without prior FPGA knowledge. From there, preferred design environments can create personalization and uniqueness at every level.
Retail
Retail can speed up various applications, including smart cities, AI/ML computing, security, and future retail, using Xilinx Kria. In addition, the Kria SOMs can speed up the execution of algorithms frequently used in retail analytics. Examples include image processing and machine learning. In the Xilinx Kria product line, the Kria KV260 Vision AI Starting Kit is an evaluation/development platform ready to use immediately.
The Starting Kit is the preferred platform for creating vision-specific target apps in retail. It consists of a non-production K26 SOM on a carrier board. The board offers general-purpose development interfaces, expandability via Pmod connectors, and other features. In addition, the Xilinx App Store’s expanding ecosystem of accelerated applications enables developers of all stripes to launch applications in under an hour without prior FPGA knowledge. From there, preferred design environments can create personalization and uniqueness at every level.
The difference between the Kria K26 and KV260
For AI edge applications, Xilinx developed the Kria K26 and KV260 system-on-modules (SOMs). The Kria K26 SOM includes a customized, specially created XCK26 SoC tailored for better visual AI application acceleration and based on the Zynq UltraScale+ MPSoC architecture. In addition, it has a customizable logic fabric and a quad-core ARM Cortex-A53 Processor, which can speed up algorithm execution.
The Kria KV260 Vision AI Starting Kit can create sophisticated vision applications without the need for in-depth hardware design expertise. It features a carrier board and pre-built software components based on the Kria K26 SOM. As a result, the KV260 simplifies the deployment of FPGAs and offers fundamental capabilities. These are the capabilities needed to create platforms for vision applications using cameras and the K26. The Kria KV260 is a development kit that comprises a carrier board and pre-built software components, whereas the Kria K26 is a standalone SOM. This is the primary distinction between the two models.
The features of Xilinx kria
A portfolio of System-On-Modules (SOMs) called Xilinx Kria is ideal for edge applications in various use cases and production environments. For example, the Kria K26 SOM is suitable for edge vision applications that need adaptability to change conditions. Furthermore, given that it is available in Commercial and Industrial grades, it is the best choice for application production deployment. Examples include Smart Camera, Embedded Vision, and others in the fields of security, retail analytics, smart city, and machine vision.
Adaptability
The adaptability of the Kria K26 SOM is one of its primary characteristics. Since requirements can quickly change in edge applications, we can easily reprogram them to respond to changing requirements. This adaptability is possible via Xilinx’s programmable logic, which enables users to modify the SOM to suit their unique requirements.
SOM Starting Kit
The SOM Starting Kit is another option Xilinx provides in addition to the Kria K26 SOM. It is a platform for evaluation and early development for the Kria SOMs. The SOM Starting Kit comes with a carrier board, a Kria SOM, and all the software and documentation required to begin developing.
The Kria Robotics Stack (KRS), a ROS 2 superset for business, is another product from Xilinx. KRS allows roboticists to safeguard and profit from their accelerated ROS 2 packages. It provides robot libraries and tools for building software-defined robotics systems. It streamlines the creation of robotics applications by bridging the gap between roboticists and Xilinx.
The Xilinx Kria K26 SOM is a strong and adaptable System-On-Module for applications requiring edge vision. It is perfect for various applications, from Smart Cameras and Embedded Vision to Security, Retail Analytics, and Machine Vision. We can attribute this to its high-performance processing capabilities, versatility, and range of connectivity options. In addition, the Kria K26 SOM is a complete solution for creating and deploying AI and robotics applications. It supports Xilinx’s VitisTM AI development environment and the Kria Robotics Stack.
