Biomedical Engineering Reference
In-Depth Information
agents in two different ways: (1) by affecting the
weight of a piece (proportional to its size), and
(2) by affecting the number of agents that can
evenly attach to it.
In simulations, when agents are tasked with
delivering at least 10 pieces of product to each
destination, the time required to complete the
task increases depending on the size of the prod-
uct (Figure 9a), as expected due to the increased
weight of the pieces. However, the number of
pieces dropped at unintended locations (failed
deliveries) is higher for short pieces than for me-
dium-length pieces (Figure 9b). This is probably
caused by the fact that larger pieces can be carried
by a larger number of agents whose distribution
along the product gives them essentially different
visual fields, allowing the kind of coordination
explained before, making medium-sized pieces
easier to transport than short pieces. Experimental
details are given at Rodriguez, 2005.
existing local arrangements of nest elements
can trigger the deposition of additional material,
which, in turn, creates new arrangements, result-
ing in a self-coordinating construction process.
While a modified form of stigmergy has been
successfully used for assembling a wide range of
structures in an idealized, discrete space (Adam,
2005, Jones, 2003), extending these methods to
more complex environments having continuous
motion and physical constraints has proved to
be much more difficult. In order to overcome
this challenge we followed the same underlying
approach as in previous sections, and integrated
low-level stigmergic behaviors with reactive
movement dynamics and with higher level, FSM-
like coordination mechanisms (Grushin, 2006).
The basic agents/components of the self-as-
sembly processes that we study are small (1×1×1),
medium (2×1×1) and large (4×1×1) blocks, which
operate in a continuous world. Initially scatted at
random on the ground, blocks arrive and come to
rest adjacent to a
seed
or other already stationary
blocks. Individual blocks have no prespecified
locations within a structure, making blocks of the
same size interchangeable. Furthermore, a block's
decisions must be based strictly on local informa-
tion, since blocks can only detect other blocks
if their centers fall within a local neighborhood.
The task of assembling some desired structure
is made more challenging by the presence of
physical constraints within the simulated envi-
ronment. Importantly, blocks are impenetrable.
Furthermore, they are subject to a gravity-like
constraint in the form of a restriction on vertical
movement, necessitating the assembly, and sub-
sequent disassembly of temporary
staircases
. A
block can increase or decrease its vertical position
by a single unit by climbing onto or descending
from a stationary block.
Subject to these constraints, and similar to
the agents in the preceding sections, a block
influences its movement by defining an internal
“force”, which directly affects its acceleration,
velocity and position. The net force is computed
SELF-ASSEMbLy
Our final example, the self-assembly problem,
entails designing a set of control mechanisms
that individual agents/components can follow in
order to form some desired 3D
target structure
.
This problem has received considerable attention
in recent years (Adam, 2005, Fujibayashi, 2002,
Jones, 2003, Klavins, 2002, White, 2005), as its
study can provide a better understanding of the
fundamental relationship between individual
actions and emergent, systemic properties (Cama-
zine, 2001).
An important class of existing approaches to
self-assembly is inspired by the nest construction
behaviors of paper wasps (Bonabeau, 1999, The-
raulaz, 1995, Theraulaz, 1999), where the agents
and components are physically distinct. While
an individual wasp is not believed to “know” the
global properties of the nest that it builds, this
nest emerges via
stigmergy
, an implicit form of
communication through the environment, where
Search WWH ::
Custom Search