IQR421
On the synthetic method
The human brain is the most complicated structure in the known universe.
It contains roughly 100 billion nerve cells, 3.2 million kilometers of
wires, a million billion connections, all packed in a volume of 1.5 liters
weighing a bit more than 1.5 kilogram, and consuming about 10 Watt of energy.
In order to understand its functions and properties we must bridge many
levels of description from molecule, cell and synapse to perception, cognition
and behavior. We pursue this challenge by constructing artificial
behaving systems and their "brains".
In the
early 18th century the Italian Philospher Giambattista Vico proposed his
principle: ``verum et factum convertuntur'' (the true and the made
are interchangeable). In other words ``we can only understand what we make''.
If we translate Vico's principle to the study of the brain it means that
in order to understand the brain we must construct one. We call our efforts
to follow this principle in the study of the brain and behavior ``Synthetic
Epistemology'', meaning that we want to understand how biological systems
acquire, retain, and express knowledge .
In our research we combine computer simulations and robots to study
the brain and how it relates to behavior. Our research addresses on one
hand ``complete systems'', models which describe the whole loop from sensing
to acting. In these models we try to understand overall principles of neuronal
organization. In this case the emphasis is on studying the behavior of
artificial creatures in the real world. In parallel we zoom in on particular
components of the nervous system. These models are very closely based on
our understanding of the anatomy and the physiology of the brain. In developing
these models we collaborate closely with neuroscientists. To facilitate
our research in synthetic epistemology we develop our own tools, such as
the simulation environment IQR421.
Introduction to IQR421
It is only recently that the type of technology needed to pursue synthetic
epistemology is available; digital computers. The tool we use to realize
our research is called IQR421. With IQR421
the user can define and analyze computer simulations of large scale neuronal
systems and combine them with sensing and behaving devices, such as cameras
and robots. IQR421 is the result of an active
research program in synthetic epistemology developed over the last 10 years.
IQR421
was
developed in our own research program since appropriate tools to pursue
our research goals did not exist.
IQR421 is a graphical programming environment
to specify, simulate, control, analyze, and document large scale heterogeneous
neuronal systems which can be interfaced to external devices such as robots
and cameras. It is developed under Linux using c and the X-motif graphics
environment. It has been tested under most common Unix environments (Solaris,
Irix, etc.)
Features
-
Graphical specification language. Draw the circuits you want to simulate
in terms of populations of neurons and their connections. Hook the simulated
neurons up to external devices. By using this approach the user can
remain focussed on working on the neuronal circuit without having to switch
to the level of a computer programming language.
-
Distributed computing using the TCP/IP protocol. Using the local area network
high computational performance can be achieved without having to worry
about the underlying communication mechanisms.
-
Interfaces to external devices (K-team
Khepera microrobot and the Koala mobile robot, video cameras, silicon retinae,
microphones, etc.)
-
On-line control and interaction, all properties of the system can be monitored,
analyzed, and changed while the simulation progresses.
-
Open compute engine through the use of dynamic link libraries (DLL).
Write you own customized routines in C or C++.
-
Integrated analysis tools using the ``Analysis Manager''. Record
data from any element of the system and perform analysis such as crosscorrelations
and pseudo phase plots.
-
Experimental protocol automation using the ``Protocol Manager''.
Record one experiment and subsequently rerun it automatically. Includes
systematic data acquisition and storage.
-
Document models and experiments. All graphics can be saved in PostScript
and other graphics formats. Automatically generate a full description of
any model in LaTeX format, including circuit diagrams, model equations,
and parameters used.
-
Track the behavior of external devices with TraX. TraX data can be integrated
in all other data sampled from the system under investigation. As many
TraX modules can be used as available hardware allows. TraX allows the
control of the environment (e.g. light sources) and other auxiliary devices.
-
IQR421 has been used in the study of learning
(DAC), models of the spinal cord, primary visual cortex, cerebellum, and
insect vision, and in the construction of the creative artifact RoBoser.
The IQR421 development team is:
-
Dr. Paul F. M. J. Verschure,
INI, Zürich, Switzerland, pfmjv@ini.phys.ethz.ch.
-
Ulysses Bernardet, INI, Zürich, Switzerland, ulysses@ini.phys.ethz.ch.
-
Dr. Mark Blanchard, INI, Zürich, Switzerland, jmb@ini.phys.ethz.ch.
At present we are developing IQR421 into a
commercial product which will facilitate other researchers and students
interested in synthetic epistemology, computational neuroscience, robotics,
or artificial intelligence to define their own studies. If you would
like to know more, please contact one of us.