Thursday's keynote was given by William O'Mullane from the European
Space Agency's Space Astronomy Center in Madrid. His talk, entitled
"The Gaia Satellite and Data Processing", covered an interesting mix
of science, software development, and experiences gained in their quest
to deploy part of the data processing on 1000 Amazon EC2 compute nodes.
The goal of the Gaia mission is use iterative satellite observations,
made over several years, to produce essentially a 3D map of our Milky
Way galaxy at unprecendented resolution.
The level of precision they are hoping for can be compared to being
able to observe a Euro coin in the hand of an astronaut standing
on the moon as seen from the Earth. Such precise measurements are
not possible from the Earth because of problems such as atmospheric
effects on light and thermal effects on telescopes - causing expansion
Among the many potential benefits of this project, it is expected
that 10,000-20,000 exo-planets will be discovered
by 2020 - on
average 10 planets per day, and of these maybe 5000 of their orbits.
Although these would all be uninhabitable gas giants, finding them
would also help with finding nearby earth-sized planets. Gaia will
be launched into an L2 orbit, pulled by the earth as it circles the
sun. Its 2 telescopes rotating, every 6 hours, will see everything
there is to see about 80 times in the 5 year mission. The information gathered
on the measured objects include astrometry (sizes), photometry
(colors) and radial velocities (movement) with iterative improvements
of accuracy in each of these characteristics providing feedback
that can improve accuracy of the others.
About 30GB per day of compressed data are expected to be downlinked
eventually including information on up to 1 billion stars and other
objects. This is expected to eventually lead to about 1/2 Petabyte of
data. Among the many science challenges include calibration of an
instrument that will be shaken during launch, extrmemly precise
time measurement, and accounting for the effects of general
relativity on gravity bending light around large planets passing
between the telescope and observed objects.
Coding for the project began 10 years ago and must last for another
10 years and feel fortunate that in the struggle over programming
languages Java was the ultimate winner. There are currently 400
people from over 20 institutions involved in development and
responsibility for data processing is to be distributed among
9 different institutions all over Europe. Collaboration and integration
is very important because "a difference in the 15th decimal place
in the constant for the speed of light can screw things up". So
far everyone has been pleased with Java. In one example, a module
that had been left in C for performance was re-written in Java
by the original author and showed a 10x improvement. There are
minor drawbacks including the fact that long-known IEEE non-compliances
were never fixed and that NaNs are not trapped but have to be
tested for explicitly. One very positive experience has been
porting the system to Amazon's EC2, which took only about 20 person
days of effort and required changing only 4 lines of code. They
have recently attempted to scale up from 100 to 1000 nodes but
have experienced data feed problems and are now investigating
whether they lie with Amazon's network or Gaia code.
The final keynote presentation of Jazoon 2010 entitled "Software
in the service of handicapped people: Research and Development at
Otto Bock" was given by Hans-Willem van Vliet the director of
Research and Development. Otto Bock is the worldwide leading provider
of prostheses for people handicapped from missing limbs. The company
was started in Berlin in 1919 and provided hand-made artificial
limbs for World War I amputees. Today the company has 40 different
locations, 400 million Euro in turnover, and 4000 employees.
Worldwide there are 10 million amputees. In the USA there are about
200,000 amputees per year and in Europe about 125,000. The main
cause for amputees in the rich world is the consequence of diabetes
resulting from obesity.
There is a larger market for limbs for the
lower body than the upper body because mobility is so important.
Otto Bock has produced a revolutionary microprocessor controlled
knee joint that works remarkably better than the mechanical version
that it replaces because it involves less concentrated control
during use. This leads to less falls. Falls lead to insecurity,
which leads to less mobility, which leads to poorer health. They
use the V-model of development because with such devices, long
development is expected for a large amount of validation and testing
needed for medical devices. There are devices to help for many
limbs and many different types of devices ranging from more mechanical
and muscle-driven to more electrical and neuro-driven. Just last
month they introduced a new product - Germium - offering even more
realistic movement even in difficult situations such as climbing
up or down stairs. Although this hasn't technically been problematic
for amputees, it was the most desired feature because only at stairs
and similar situations did other people notice that they are
handicapped and this was the last hurdle toward movement that allows
amputees to better fit in to everyday environments.
Future plans involve innvation in various areas leading to limbs
with a greater number degrees of freedom and more natural control
through more sophisticated brain-computer interfaces. In this
area, the future looks bright.