An Overview of Cyber-Physical Systems' Hardware Architecture Concerning Machine Learning

Author(s)

Loubach, Denis S. ; IEEE

• Document type: Journal article
• Source: 2021 IEEE/AIAA 40TH DIGITAL AVIONICS SYSTEMS CONFERENCE (DASC); v. N/A, p. 6-pg., 2021-01-01.

Abstract

In this paper, we present the main cyber-physical systems' hardware architectures that also take into account the use or possible use of machine learning algorithms to improve the overall system performance. Our brief overview considers aspects including domain-specific architecture, runtime reconfiguration, separation of virtual and physical platforms, formal models of computation, and the use of machine learning algorithms. Here, we also regard the safety aspect since these key concepts and the application of autonomy seems to have some level of acceptance in digital avionics systems. Our study highlights the main benefits and drawbacks of analyzed architectures and their impact on safety. (AU)

• FAPESP's process: 19/27327-6 - Runtime reconfigurable hardware platform model
• Grantee: Denis Silva Loubach
• Support Opportunities: Regular Research Grants