Environmental Engineering Reference
In-Depth Information
As long as the constellation produces, i.e., delivers its end purpose on its own
without any external support, it is viable and autonomous. Some members
of the group may fail, but the remainder will continue to produce, thereby
maintaining the essential nature of the constellation.
The second and final step (
progressive autonomy
) in the proposed overall
plan to realize space-based autonomy is to migrate the spacecraft surrogate
community of agents to the actual spacecraft. As discussed in previous chap-
ters, this is a nontrivial step. A major step in the direction of actual onboard
spacecraft autonomy is to have the agent community demonstrate its cor-
rectness in actual ground-based spacecraft control centers [
178
,
184
]. This is
There are many issues that need to be addressed before this becomes a
reality. Some of the major issues are as follows:
•
Adaptation to resource constraints.
As an example, a spacecraft subsys-
tem agent must be able to exist and operate within the microprocessor
associated with the subsystem. This is where the concept, which we call
“economical intelligence,” comes into play. Reasoning code and knowledge
and information structures and management need to be “optimized” in
order to function properly in the resource-constrained environment of a
spacecraft subsystem microprocessor.
•
Integration with existing subsystem autonomy.
As discussed earlier, most
spacecraft subsystems already have a degree of autonomy built into their
operations. This is realized usually through the use of expert systems or
state-based technologies. A subsystem agent should be able to take advan-
tage of the existing capability and build upon it. The existing capability
would become an external resource to the subsystem agent that would be
used to realize a higher level of autonomy for the subsystem. The agent
would need to know about the external resource and how to use it, i.e.,
factor its information into its reasoning process.
•
Real-time activity.
Most situations experienced by a spacecraft require
real-time attention. If the situation is not readily handled by built-in sub-
system autonomy, the associated subsystem agent will need to respond in
real-time. This will require the agent to have a working reflexive behavior.
9.4.3 Autonomicity in Constellations
A step beyond an autonomous constellation is an autonomic constellation -
a collective of autonomous agents that are self-governing and learn individual
sequences of actions so that the resultant sequence of joint actions achieves a
predetermined global objective. As discussed earlier, this approach is partic-
ularly useful when centralized human control is either impossible or imprac-
tical, such as whenever timely and adequate communication with humans is
impractical. Constellations controlled by in situ “intelligent” spacecraft, in
