Ground Autonomy Evolution - Autonomous and Autonomic Systems - page 69

3. Migration
4. Persistence
5. Communication
6. Cloning/spawning
Autonomy/semiautonomy is the ability of an agent to respond to a dynamic
environment without human intervention, thereby improving the productivity
of the user. When presented with a request, it can use its own strategy to
decide how to satisfy the request. Each agent is capable of a type of learning
that enables it to more responsively interact with its user community over a
period of time. Learning also enables the agent to keep abreast of changes in
its operational environment. The dynamic behavior of the agents is triggered
by either command or event-driven stimuli. Migration is the ability of an agent
to relocate to other nodes to accomplish its tasks. This ability can support
load balancing, improve eciencies of communication, and provide unique
services that may not be available at a local node. Persistence is the ability
to recover from environmental crashes and support time-extended activities,
thereby reducing the need for constant polling of the agent's welfare by the
user and providing better use of the system's communication bandwidth. A
communication ability provides an agent with the mechanisms for supporting
agent-agent and user-agent interactions either through an ACL or a domain-
specific natural language. Spawning is an agent's ability to create other agents
to support the parent agent, thereby promoting dynamic parallelism and fault-
tolerance. It is our opinion that these capabilities are necessary for building
autonomous satellite control centers.
The agent architecture described above is generic enough for use in au-
tomating the operations of the control center of any spacecraft. The Explorer
Platform's (EP's) [ 137 ] satellite control center was selected as a domain to
test the feasibility of AFLOAT. Figure 4.2 depicts the interactions between
different components of the EP satellite control center. The elements shown
in the diagram are similar to those found in a typical satellite operations con-
trol facility. A taxonomy of the EP subsystems and data extraction system
is also shown in Fig. 4.2 . The diagram describes the interactions between the
physical model (i.e., elements above the mnemonics database) and the logical
model (i.e., elements below the mnemonics database) of the components of
the EP system.
The main goal of the AFLOAT project was to implement a multiagent
architecture that could interface directly with sources of data from a satellite
and process and reason with the data to support autonomous operation of the
control center. This was achievable with the aid of a data server agent that
could interface directly with satellite telemetry and provide the information
as mnemonics to other specialist agents. Due to operational restrictions, mag-
netic tapes were used to transfer satellite data to a workstation for processing
by a data server agent. Even with that restriction, it was possible to demon-
strate the feasibility of an automated agent-based satellite operations control
center.