Biomedical Engineering Reference
In-Depth Information
complete, its staircase disassembles (Figure 11c).
After the floor and the columns are complete, the
door is marked by depositing the lower sides of
its frame, consisting of a small and a medium
block each as shown in Figure 11d. The walls,
which consist of medium blocks, are then begun
along with the wall staircase discussed earlier
(Figure 11e). As these walls reach an appropriate
height, large blocks begin to climb the staircase,
and one such block eventually places itself over
the door, to form the top part of the door frame
(Figure 11f). When the walls are finished, large
blocks climb the staircase once again to build the
two-layer roof. The second roof layer (Figure
11g) forms perpendicularly to the roof blocks
below; it begins near the edges farthest from
the staircase and extends towards the staircase.
When the roof is complete and there are no blocks
left moving above it (variable change rules are
in place to ensure that this is the case), the wall
staircase disassembles (Figure 11h), resulting in
the final structure.
One factor that affects completion time is the
availability of different sized blocks in regions
where they are needed. We hypothesized that, as
in competitive foraging and collective transport,
block availability can be enhanced via the use of
collective, flock-like movement behaviors, which
allow a block that found the general region of
construction to “pull”' other members of the flock
there as well. We compared the use of cohesion and
alignment to create this “flocking” versus simu-
lations in which such forces were absent. These
forces indeed reduce the mean completion time
from 14,604.1 to 13,461.0, which is statistically
significant. The improvement is considerably more
pronounced in the absence of any global informa-
tion, as demonstrated by running further trials
where blocks have no knowledge of the relative
direction of the center of the world. In this case,
the reduction is from 67,666.5 to 43,338. While
blocks must spend a much greater amount of time
searching for the region of the world where they
are needed, a flock-like dynamic can partially, yet
significantly, compensate for the lack of global
guidance.
Interestingly, there is sometimes a tradeoff
between the goals of increasing block availability
and reducing interference. We found that the
latter can be particularly severe between blocks
of different sizes, and it is sometimes necessary
to coordinate the process (via variable change
rules) such that only blocks of one particular size
are actively assembling at a given point in time;
however, this means that blocks of other sizes are
temporarily unavailable.
We have thus shown an integrated methodol-
ogy for controlling the self-assembly of complex
structures in a non-trivial environment. While
the methodology has been shown to be success-
ful, the design of stigmergic and variable change
rules for assembling some desired structure has
proven to be a time-consuming and error-prone
task. The present focus of our research is on the
development of approaches for generating such
rules automatically, given a specification of the
target structure.
FUTURE TRENDS
Recent years have witnessed an increased degree
of interest in the engineering of large complex
systems consisting of numerous components in-
teracting in a non-linear fashion. The intrinsically
complex nature of such systems renders conven-
tional methods for their analysis inappropriate
and raises interests in alternative approaches.
One promising approach is presented by guided
self-organization. Inspired by examples of self-
organizing systems in nature, such as swarm-
ing insects or social animals, these systems are
characterized by the property that the global,
system-wide behavior emerges from local inter-
actions among their components, with no single
component in control of the system or aware of
its global state. Guiding this process into desired,
goal-oriented behavior is a promising artificial in-
Search WWH ::
Custom Search