Explainable Machine Learning Models and Architectures

3Efficient Architectures and Trade-Offs for FPGA-Based Real-Time Systems

L.M.I. Leo Joseph*, J. Ajayan, Sandip Bhattacharya and Sreedhar Kollem

Department of Electronics and Communication Engineering, School of Engineering, SR University, Warangal, Telangana, India

Abstract

Real-time monitoring gives accurate timely information, crucial for any production enhancement and decision-making. The most recent technology for field-programmable gate arrays (FPGA) has blurred the gap between hardware and software. By combining measurements from many sensor systems, it is possible to circumvent issues which exist when employing stand-alone systems. This chapter emphasizes the significance of design strategies and algorithm development from several vantage points, as well as their implementation in diverse real-time systems, employing reconfigurable and hybrid approaches. Employing a configurable soft-core processor, on an FPGA will give the adaptability necessary for interfacing with a variety of sensors.

Keywords: FPGA, real-time systems, hybrid architectures, hardware acceleration, SVM, CNN, Zynq, Pynq

3.1 Overview of FPGA-Based Real-Time System

The system whose response seems accurate in a defined time or real time is a real-time system [1]. Flight control mechanisms and real-time monitors are two examples of real-time systems. Real-time systems find applications in different forms pertaining to different application. Real-time systems are classified into two types based on the ...