Crocotta R&D - Alternate Reality - a Virtual World made of Particles
Click here to load reader
-
Upload
crocotta-rd-limited -
Category
Technology
-
view
221 -
download
0
description
Transcript of Crocotta R&D - Alternate Reality - a Virtual World made of Particles
Crocotta Research & Development Ltd Suite 5, 39 Irish Town, Gibraltar
“Be ambitious of climbing up to the difficult, in a manner inaccessible...”
We are a small team of international researchers with the aim of conducting technology leaps in exciting fields of exploration like visualization, virtual reality, virtual synthesis of matter, perception &
recognition, artificial intelligence, and robotics.
www.crocotta.co.uk [email protected]
+44 20 3239 7007
Crocotta Research & Development Ltd Suite 5, 39 Irish Town, Gibraltar
Registered number 107957 www.crocotta.co.uk
Crocotta Research & Development Ltd | Suite 5, 39 Irish Town, Gibraltar | registered no 107957 | www.crocotta.co.uk
Page 2
“ALTERNATE REALITY”
A VIRTUAL WORLD MADE OF PURE PARTICLES By Crocotta R&D
Robert Sugar1 Zsolt Peter Velitchko Filipov
ABSTRACT
3D visualization today has ever-expanding applications in science, education, engineering, medicine,
interactive multimedia like games, etc. Producers of graphics processing units (GPU) – are specialized
electronic circuits designed to rapidly manipulate and alter computer memory in such a way so as to
massively accelerate the visualization of 3D environments – bring ever faster products to the market
every six months which is rapidly increasing the possibilities of near future visualization/simulation
methods.
In this paper we discuss the short/medium term possibilities of a large scale, accurate, and quasi-
real-time virtualization method of real-world complex material structures and their interactions,
which would be a groundbreaking tool for making better predictions in various research/industrial
fields as well as for running more simulation tests in fully virtual environments without the use of
expensive testing facilities.
INTRODUCTION
In the last few years, particle-based object
representations have proven to be a flexible and
efficient alternative to mesh-based surface
representations. In fact, particles are more suited to
imitate real-world phenomena, as real-world matter is
composed of extremely small particles called atoms
and not of mesh-based surfaces.
However a quasi-real-time virtualization method of
real-world complex material structures in large scale
and with high accuracy has to face two major
technical difficulties, namely:
#1 Large amounts of computation
Though today’s microprocessors are getting amazingly
fast as GPUs’ parallel computational performance is
beyond teraflop level, the complex physical simulation
of large scale virtual environments filled of virtual
matter copied from real-life will definitely remain an
enormous job over the next few years, as even a
simple light scattering simulation inside such virtual
environments (via progressive photon mapping or
similar technic) requires hundreds of millions of
lightray calculations.
#2 Enormous amounts of data
Virtual representation of the uncompressed virtual
matter of one cubic meter of volume in high-
resolution (provided that a living cell is the smallest
element) requires approximately 109 cells (per cubic
centimetre) x 106 cubic centimetres x 100 bytes
(provided minimal cellular info fits to 100 bytes) =
about 89 petabytes of data. No known compression
method exists to reduce these enormous numbers to
processable levels, especially not without sacrificing
efficient read-/write-ability.
PROPOSAL
We present a novel method (a set of sub-methods) to
the quasi-real-time synthesis, compression,
Crocotta Research & Development Ltd Suite 5, 39 Irish Town, Gibraltar
Registered number 107957 www.crocotta.co.uk
Crocotta Research & Development Ltd | Suite 5, 39 Irish Town, Gibraltar | registered no 107957 | www.crocotta.co.uk
Page 3
simulation, and visualization of virtual matter, and of
virtualized environments in high-resolution with the
living cell being the smallest visible element.
Figure 1:
Simulation of embryonic stem cell structures (actual
screenshots from our recent test scenarios).
By the time our technologies go live in large scale
scenarios we expect GPUs to have crossed the level of
100,000 parallel cores, and data storage devices to
have crossed the petabyte level. Our expectations are
easy to prognosticate, even in short term, if we check
all the significant growth of computation power which
happened in the last 4-5 years.
STATUS
We have successfully achieved good results in the
fields of photon traced visualization, gradient field
reconstruction, particle interactions (including fluid
dynamics), and smart interpolation throughout
medium scale scenarios, and we have promising first
results with massive data compression related to
natural patterns in large scale scenarios.
Comparison with Nvidia demos
The fact that Nvidia is a well-known brand, and we
have been using their hardware as base-platform for
running our simulations, makes them an ideal
candidate we can compare with.
Nvidia is usually presenting static particle spaces (no
physics or any kind of interaction involved) which are
just primarily ray-casted at about 200 fps (frames per
second) on a Geforce GTX 460 with 384 parallel
working cores = 0.52 frames per second per core,
while we are capable of the same at over 8 fps per
core, which is approximately 15 times faster.
It is also worth to mention that our systems are
capable of fluid dynamics simulation, iso-gradient
extraction, primary ray-casting, and multi-light-source
photon tracing (a single light-source with 16,384
emitted photons per frame was used in the
comparison scenario) the same time at over 3 fps per
core which is a significant breakthrough. In all
comparison scenarios a cubic particle space of 32
particles side length was used.
Figure 2:
Comparison screenshots (left: Nvidia volume renderer
demo, right: ours).
Nvidia is usually dealing with cubic particle spaces
with side length up to 256 (made of approximately 16
million particles), we are dealing with particle spaces
with side length up to 100,000 (made of
approximately 1 quadrillion particles).
Crocotta Research & Development Ltd Suite 5, 39 Irish Town, Gibraltar
Registered number 107957 www.crocotta.co.uk
Crocotta Research & Development Ltd | Suite 5, 39 Irish Town, Gibraltar | registered no 107957 | www.crocotta.co.uk
Page 4
FIELDS OF APPLICATION
Entertainment
Combining with augmented reality the “holodeck”
dream of the science fiction movies can be brought to
reality on longer term.
Figure 3:
Fictitious showcase scenarios for environmental possibilities
in movies/games.
Concerning the movie production of large scale liquid
scenes (underwater scene), gaseous environments
(cloudy atmosphere, space nebula, and turbulent
weather phenomena) the editing time can be
dramatically cut, as well as the final rendering time.
Application for games is also in view, game concepts
to be published in our next paper soon.
Medicine
Combined with biological properties the simulating of
tumour growth, the observation of tissue changes,
foreign body obstruction, etc. within the virtual
human body could be possible on medium term.
Prediction & early warning
Current predictions of weather phenomena like super
cell formation, earthquakes, and volcanic activity can
be dramatically enhanced.
Industrial material development & testing
We will be able to replace expensive stress testing
procedures, like fatigue testing for materials, with
pure virtual methods on short/medium term.
Astrophysics
Our simulation environments may open up extra
perspectives concerning the studying of star birth,
black holes, as well as the behaviour of star and galaxy
clusters in large scale.
FUTURE
In the next period we aim to fully develop our
compression and pattern reconstruction methods,
nevertheless further improve our fluid dynamic and
other force field like technics to be able to fully realize
our goals.
MAJOR FEATURES
Flexibility
We have been paying special attention to keep the
flexibility of our algorithms in relation to speed and
accuracy throughout our systems which allows:
(a) the quick, quasi-real-time running of
simulations by deploying massive series of
reconstruction and data recovery technics in small
scales while maintaining higher accuracy in large
scales; such simulation mode is suited for applications
where speed and/or interactivity is more a relevant
factor, like in the entertainment industry for instance,
Crocotta Research & Development Ltd Suite 5, 39 Irish Town, Gibraltar
Registered number 107957 www.crocotta.co.uk
Crocotta Research & Development Ltd | Suite 5, 39 Irish Town, Gibraltar | registered no 107957 | www.crocotta.co.uk
Page 5
and,
(b) the accurate running of simulations where we
also maintain accuracy in small scales by sacrificing
part of the physics/rendering acceleration and data
compression benefits; such simulation mode is
required for scientific, engineering, and industrial
applications where accuracy has the absolute priority.
Structure synthesis
We developed new methods for the
synthesis/reconstruction of 3d natural patterns in
local neighbourhoods purely from 2d exemplars. The
exemplars can be electron microscope images
(however not restricted to) in order to closely mimic
real-world organic/inorganic material structures.
Extreme detail
We developed special algorithms for the
compression/decompression of the large amount of
data generated by our virtual cell-sized particles on a
multi-trillion level.
Simulation & visualization
Our virtual cell-sized particles can have tailor made
physical as well as biological properties which enables
the simulations of complex scenarios throughout
possible. We developed ultra-fast acceleration
algorithms for speeding up the physical simulating as
well as rendering processes.
REFERENCES
Robert Sugar1 is a scientist, researcher and IT
entrepreneur the same time. He has been starting
companies since 1996, ranging from software
companies, media companies, computer game
developer companies and internet companies.
He was born in 1978, and grew
up in Hungary. He graduated in
physics at the Lorand Eotvos
University (Budapest). First
software engineering was just
his hobby and later it has
become his full time
profession. His first development project was about
artificial intelligence and graphical visualization for
computer games back in 1996. He founded his own
game developing studio in 2001 - called Mithis
Entertainment - in the heart of Budapest for the
purposes of "AAA" game development. From a small
group of enthusiastic people, Mithis Entertainment
has become the biggest developer studio in Hungary
by 2005 and completed four big game titles which
were distributed world-wide by well-known multi-
national publishers.
Since his departure from the gaming industry in 2006
he has been focusing on the researching of cutting
edge technologies.
By Crocotta R&D, July 2012