ATRON is a "homegeneous unit modular self-reconfigurable robot system",
more commonly known as a shape-shifting robot. Developed by Henrik Hautop
Lund and other researchers at the Maersk Institute
ATRON is a cubic lattice of self-contained modules
that can attach to each other, share power, and communicate to form a
larger robot of the desired shape. Each module is composed of two
hemispheres, connected by a slip-ring so they can rotate
independently while still transferring power and communicating with each
other. The "northern" hemisphere of each module contains an
128 microcontroller. Each module also contains internal and external sensors including a 2-axis accelerometer. For more
details there is a short technical
paper (PDF format) available describing the ATRON system. It
includes photos and diagrams of the modules. Other papers on
ATRON and self-reconfiguring robotics in general are available as is
list of other
self-reconfiguring robot projects. There is
also an ATRON
simulator. For the curious, the name is
a combination of Atom and elecTRON and a play on the name of a previous
shape-shifting robot called MTRAN (Modular TRANsformer). And, don't
forget to see the recent NewScientist
article submitted by The Swirling
Brain that brought ATRON to my attention.
Glad to see this get some press. I had a chance to meet Kasper Stoy,
Hendrik Lund, and others who were working on the ATRON project last
month in Tsukuba, Japan. We were having a satellite workshop on modular
robotics in the shadow of the IROS2004 robotics conference at Sendai, Japan.
I also work in the lab of the CONRO project, which you'll see on that
link to the list of other modular robot projects.
I got to handle and look at the ATRON module up close. At the time,
they only had a couple working modules and the rest were being hastily
assembled by their army of Chinese students back in Denmark. The thing
that I was truly impressed with were their mechanical connectors. The
connectors are a pair of interlocking levers that have some capacity to
correct for misalignment. So, even if your modules are poorly
misaligned with respect to each other, the connector can correct for
this and bring them back into alignment.
The other useful innovation was their power distribution system. Power
in modular robots is a true sore because you either have to have onboard
batteries that add lots of weight and need to be replaced frequently, or
you have to have all your modules tethered, which makes it really hard
to do shape-shifting and locomotion. Their power distribution system
means that modules can share their battery power with other modules its
connected to. This makes things dramatically easier, and I'm glad they
found a way to do it.
Communication between the modules is achieved by IR emitter/detector
pairs. This seems to be a common method among modular robot builders
because it reduces the need for making wire contacts with the connectors.
Kasper Stoy, one of the project members, used to be a lab member of the
ISI Polymorphic Robotics groups, where I now work, and he graduated here
and got his PhD. We're seeding the Earth and pretty soon, you'll see
shape-shifting robots everywhere you go! They're only starting to get
more and more popular with time.