|
The
design of embedded systems has been for years more a craft
than an art or a science. The design methods used suffered
from informal specifications, lack of adequate support for
verification, fairly long design times. This situation has
become untenable as their complexity and safety, cost and
power consumption requirements posed upon them has scaled
up. The situation has been made even more difficult by the
increasing degree of integration in the semiconductor
industry that has made possible to build Systems-on-Chips
(SOC) with unparalleled compute power. In too many cases,
errors in conception and implementation of embedded
controllers have caused dramatic problems especially in the
area of space exploration and applications (in particular,
the Ariadne and the Mars Lander cases). On the other hand,
the opportunities offered by technology for embedded
controllers are immense: for example, the stability problem
of the Mercedes Class A vehicle was solved relatively
quickly by devising new control algorithms cleverly
implemented in software on the powerful platform that was
dedicated to suspension control. This solution would not
have been possible just a few years ago.
Our
research agenda is to introduce well-founded methods
for the design of embedded controllers for safety
critical application from specification to
implementation. Much research and development work
has been done in each of the disciplines that have
to be tapped for the solution of the problem.
However, a holistic view of the problem is still
missing and this is our first important innovation.
We believe that realizing this vision for
embedded systems requires combining the strengths of
several communities of interest: researchers in
system level design and Electronic Design Automation
(EDA); researchers in embedded SW and systems; and
tool companies in both the EDA and ESW domains.
These worlds have traditionally been isolated; it is
time for the worlds to collide.
We
wish to combine expertise not only in different
disciplines, but also from different scientific and
industrial cultures: Europe and US. Both industrial
and scientific communities realize the importance of
embedded systems in their roadmap. In Europe, two
networks of excellence have been recently started:
one on real-time systems and one in advanced
control.
In
U.S., we mention the efforts by DARPA (particularly
relevant examples are the two programs: Model Based
Integration of Embedded Software (MoBIES) and
Software Enabled Control (SEC)) and the ones by the
Marco Program initiated by the Semiconductor
Industry Association (particularly relevant to the
theme of COLUMBUS are the Gigascale Silicon Research
Center and the Circuit, Software and Components
Center). The second innovation comes, then, from
the attempt at merging Europe and US expertise in
the field. This forms the third leg in an
overall plan for science in embedded controller
design in Europe with the two networks of excellent
being the other two. We believe we can achieve
success in this area because of the leadership roles
that the proponents have in the world and of their
long lasting research interactions among the
participants across the Ocean. UCAM, AQUI, INRIA and
their subcontractors, are present in the steering
committee of both networks of excellence. UCB and VU
are in a leadership positions in the DARPA and MARCO
programs. The bridge between the UCAM, AQUI and UCB
has been ensured by frequent visits and joint
research carried out in the past. VU has been
interacting on a more recent basis but the research
vectors are perfectly aligned.
In
the remainder of this section, we analyze our
innovation claims in three main areas: control,
embedded software and platforms. |
Innovation