AI at the Edge | element14 Korea

If only it had a Brain

Artificial Intelligence at the edge

Artificial Intelligence and Machine Learning on the network’s edge require workload-efficient computing at the lowest power and thermal headroom for operation in constrained environments. For this application, mid-range density FPGAs are an excellent option for implementing designs. The challenge is how to deploy scalable and extensible neural networks and how to enable software engineers with FPGA-based AI solutions performing efficient inferencing.

Microchip FPGAs enabling Artificial Intelligence at the Edge

Microchip’s newest video and image processing solutions portfolio is enabled by the PolarFire^® FPGA Video and Imaging Kit. The new kit enables developers to implement Smart Embedded Vision systems with the lowest power and smallest form factor for edge applications leveraging artificial intelligence and high-resolution imaging. It also helps developers get started with the RISC-V based PolarFire SoC and MATLAB^®/Simulink^® on FPGA-in-the-Loop simulation. The PolarFire family continues to hit key milestones.

Smart Embedded Vision opens up new possibilities for implementing systems that rely on visual data to make decisions across a broad spectrum of applications. Machine vision, thermal imaging, video surveillance, robotics, machine learning inference on the edge and HMI all rely on cameras and displays that demand low power while supporting high-speed interfaces as well as security in data and in design so that IP is protected. The inherent parallel processing and high-speed I/O capabilities of FPGAs make them ideal processing platforms for delivering the high data throughputs needed for both high-resolution imaging and machine learning algorithms.

Industrial

Microchip provides FPGA imaging and video solutions to enable the evaluation of multiple protocols and the development of a host of image and video processing applications. As a best-in-class imaging and video platform, Microchip’s solutions come with a complete ecosystem, including comprehensive application-specific hardware, optimized intellectual property suite for image processing, sample reference designs, demonstration designs and collateral. Compared to MCU, CPU, GPUs and AI ASICs, Microchip FPGAs:

Offer large DSP compute capacity (up to 1480 x18 x 18 Math Blocks) vs. MCU/ MPU/ CPU
Have lower power dissipation (~3-4W Core Power) vs. CPU/ GPU (>20W)
Have up to 50% lower power dissipation over competitive mid-density FPGAs
Offer scalability based on required performance
Integrate video, connectivity, security etc. vs. ASICs

Transportation

Advanced driver assistance systems (ADAS) will be the driving force behind achieving Vision Zero, the multi-national project to reduce the number of fatalities and serious injuries caused by road traffic to zero. Microchip is enabling ADAS through its expertise in FPGA’s, Smart Embedded Vision, sensor connectivity and signal conditioning, high-speed data transfer, timing solutions and commitment to functional safety. As the level of autonomy progresses, the number of ADAS sensors in a car will increase from two or three to over 30, including forward/rear/side facing cameras, forward-looking radar and LiDAR. Microchip’s expertise in safety-critical connectivity will form the automotive superhighway for ADAS. We enable you to create secured, connected systems to detect any bumps in the road, with the speed and safety needed to support emerging ADAS applications.

Modern vehicles are the sum of tens of thousands of components, each one of which has to be designed with safety, security and reliability in mind.

The transport vehicle is becoming increasingly connected, both in-vehicle and to today’s rapidly expanding IoT world, enabling a smarter, greener, safer future.

Microsemi AI solution for ADAS is based on FPGAs, which provide core data acquisition, processing and display functions while providing industry’s best reliability and security in small footprint packages crucial for building differentiated, secure and reliable advanced driver assist systems. There is support for various communications interfaces for capturing data like Camera, Radar etc. These are available as hard or soft IPs. Customization of sensor interface as per design is supported. Microsemi FPGAs offer support for complex algorithms required for multi sensor input, image processing at high speeds. FPGAs offer crucial advantage over DSPs in providing parallel processing, enabling faster responses to potential hazards.