A Cyber-Physical Systems Approach to Personalized Medicine: Challenges and Opportunities for NoC-based Multicore Platforms
Department of Electrical Engineering, University of Southern California, 3740 McClintock Avenue, Los Angeles, USA.
This paper describes a few fundamental challenges concerning the
design of Network-on-Chip (NoC) based multicores as the
backbone of cyber-physical systems (CPS) for personalized
medicine. One fundamental challenge in designing such CPS
architectures is the need for a unifying mathematical description
of the dynamical interactions between bio-physiological processes
and cyber states. Another fundamental challenge is to build a
rigorous mathematical optimization framework that allows the
CPS to adapt to varying workloads and demands. To enable largescale
parallelism, we need a rigorous understanding of the CPS
workloads that can guide the design and optimization of wired and
wireless NoCs. We advocate for the development of goal-oriented
self-organization algorithms that seek to both optimize specific
design cost functions and maximize information about future
system state. It is necessary to identify basic local rules of
interaction not only for solving large scale optimization problems
in a distributed fashion, but also for inducing an overall degree of
autonomy and intelligence in the CPS architecture.
Keywords: Cyber-physical systems, Networks-on-chip, Multicore platforms, Highly-variable workloads, Scalability, Adaptive autonomous
systems, Goal-oriented self-organization, Personalized medicine,
Non-stationary fractal behavior, Real-time performance guarantees.
Full Text (PDF)