|
DAN SIEWIOREK
Buhl University Professor
Electrical and Computer Engineering, Computer Science
Director, Human Computer Interaction Institute
 www
My major interest is in the modular design and rapid prototyping of
reliable computing structures. Three research projects support this
interest: Mobile/Wearable Computers, Concurrent Design, and Reliable
Systems.
Mobile/Wearable Computers. The information processing industry is
undergoing a paradigm shift. In the 1960's information processing was
concentrated in mainframe computers operated by central staff and accessed
by custom-built programs executed in batch mode. By 1970 the invention
of the time-sharing operating system allowed users to interact with
information on-line. However, time-sharing systems were still centrally
based with most of the computing cycles devoted to information manipulation
rather than the human computer interface. With the advent of the personal
computer in the early 1980's a substantial portion of the computing
power could be dedicated to the single user. In the 1990's wearable
computers allow mobile users to remotely access information and collaborate
with exerts. We have built 20 generations of wearable computer systems
in such diverse areas as heavy vehicle maintenance, aircraft manufacturing,
plant operations, language translation, and medical monitoring. Example
systems can be found on www.cs.cmu.edu/~wearable. Systems involve hardware
architecture, software architecture, wireless communications, ergonomic
design, and human computer interaction. Research is in cooperation
with the Design Department, Electrical and Computer Engineering, and
the Human Computer Interaction Institute
Concurrent Design. The goal of the project is to support the generation
of designs from high level systems specifications into completely assembled
electronics, mechanical, and software systems. The goal is to reduce
design time by 1 to 2 orders of magnitude. The Concurrent Design methodology
has been used in all twenty generations of mobile computers described
above. Groups of up to 30 designers representing up to five disciplines
designed and fabricated multiple copies in less than four months, and
developed tools to support the concurrent design process. For all levels
of design there are common issues that must be addressed including
design data bases, design information representation, human-computer
interfaces, simulation/validation/ verification, automatic synthesis,
test generation, and design selection criteria.
Reliable Systems. For over two decades, computer system design and
evaluation has been based upon performance benchmarks. Comparable benchmarks
do not exist for evaluating the quality and robustness of computer
hardware/software systems. This project is developing a family of portable
benchmarks for a variety of operating systems and programming language
environments. The benchmarks are based upon over a decade of experimentation
with fault injection including the next generation air traffic control
system. Information can be found on www.ices.cmu.edu/ballista. Another
goal of the project is to develop technology to enable the construction
of reliable systems from commercial-off-the-shelf (COTS) hardware and
software. One approach is to add middleware between the applications
software and the operating system. Sentries can perform error detection,
data logging, and automatic retry, all transparent to the application
software. Studies indicate that the majority of system downtime is
due to human errors in either design or operation. This project also
explores the design of software systems and interfaces to reduce human
errors. Other research issues include automatic generation of assertion
checks, checkpointing/rollback, and redundancy through replication.
Research is in cooperation with Roy Maxion and Phil Koopman.
|
