Hybrid and Heterogeneous Models
The problem here is the integration of
deep-physics models with
the digital world.
The goal is technologies for
co-designing and analyzing
hybrid networked systems with integrated cyber,
engineered, and human elements.
The challenges center around the divergent
abstractions that are used for physical engineered
systems, biological systems, human systems, and
computation and networking.
Heterogeneity
Cyber-physical systems are intrinsically heterogeneous.
There are two distinct approaches to modeling heterogeneous
systems: (1) a grand unified theory (GUT) and (2) an
abstract semantics.
The former is about developing a modeling language and
conceptual framework into which heterogeneous modeling languages
and frameworks can be translated. The latter is about
developing interfaces between heterogeneous modeling
languages that are sufficient for interoperation,
but not so rich that the interface language itself
becomes a modeling language.
A GUT has the advantage of enabling model exchange between
tools, but the disadvantage that the semantic richness that
is required to be able to encompass all interesting heterogeneous
modeling languages makes analysis of models difficult.
An abstract semantics has the advantage of enabling
composition of domain-specific modeling languages that
are themselves sufficiently constrained that analysis
is still possible, but the disadvantage that engineers
must learn a multiciplicity of modeling languages and must
understand how they interact within an abstract semantics.
Multiform Time
Central to cyber-physical systems is the interaction
between cyber and physical components. These interactions
occur in time. A naive model of time, which assumes all
players in a system have access to a common, smoothly advancing
measurement of time, is usually not adequate. In the context
of CPS, models of time must be able to deal with non-homogeneous
measurements of time, where different parts of a system may
disagree on the current time of day. Models of time must
also be capable of making a semantic distinction between
continuously evolving processes and discrete state changes.