A New Approach for Advanced Life-Support Systems Control- II

The goal of BLSS is very ambitious, as
the fact of using living beings generates
a problem in the determinism of the
system. These systems are highly nonlinear,
with a high level of uncertainty in
their behaviour, making it impossible to
perform a complete analytical modelling
of the processes. It is therefore necessary
to develop, in parallel with the biochemical
and physiological studies, new
approaches to system control (Figure 1).
Model-based control systems will not be
successful since they cannot deal with
incomplete or inaccurate information.
For instance, the maturity level of the
crop must be assessed indirectly using
several variables (atmosphere gas composition,
plant colour, biomass, time
from seeding etc). In addition, these variables
will depend on a great number of
factors, making predictability very poor.
New control-system architecture to cope
with these problems can be effectively
implemented using a Multi-Agent
System (MAS). This approach allows
the problem to be broken down into
small parts, each dealing with specific
tasks but in a coordinated manner, performing
as an organization with a
common objective and sharing a set of
rules. In addition, designing this system
as a multi-agent network will allow specific
control solutions to be applied to
each part as needed. The different types
of controllers will become encapsulated
in the agent structure and only relevant
information will be shared to enable
monitoring and global control. Another
benefit will be the reconfiguration capability,
in cases of, for instance, failure of
part of the system or the need to adapt
the system to new objectives.