Environmental Engineering Reference
In-Depth Information
the plan's steps in order for execution. The agenda kept track of which steps
were being executed, had finished executing, were idle, or were waiting for
execution. It updated the status of each step appropriately as the step moved
through the execution cycle. The agenda reported the plan's final completion
status to the Planner and Agent Reasoner when the plan was complete.
The Executive would execute the steps it received from the Agenda. A
step contained preconditions, an action, and possible postconditions. If the
preconditions were met, the action was executed. When executions finished,
the postconditions were evaluated, and a completion status was generated for
that step. The completion status was returned to the agenda, which allowed
for overall plan evaluation.
The execution component interacted with the agenda in the following way.
The agenda sent the first step to the execution component. This woke up the
Executive. The component then began executing that step. The Executive
then checked to see if another step was ready for execution. If not, the com-
ponent would go back to sleep until it received another step from the agenda.
The Modeling component would record state changes caused by a step
execution. When a plan was finished executing, the Agenda component sent
a completion status to the Reasoning component to indicate that the goal
established by the Reasoner had been accomplished. If the Agent Reasoning
component was dealing with data from the environment, it could decide either
to set a goal (for more deliberative planning) or to react quickly in an emer-
gency situation. The Reasoner could also carry on a dialog with another agent
in the community through the Agent Communication Perceptor/Effector.
A watch was also attached to the Executive. It monitored given condi-
tions during execution of a set of steps and the consequence if the condition
occurred. Watches allowed the planner to flag things that had to be particu-
larly looked out for during real-time execution. They could be used to provide
“interrupt” capabilities within the plan. An example of a watch might be to
monitor drift from a guide star during an observation. If the drift exceeds a
threshold, the observation is halted. In such a case, the watch would notify the
Executive, which in turn would notify the Agenda. The Agenda would then
inform the Reasoner that the plan had failed and the goal was not achieved.
The Reasoner would then formulate another goal (e.g., recalibrate the star
tracker).
Agent Communications
The agent communication component was responsible for sending and receiv-
ing messages to/from other agents. The component took an agent data object
that needed to be transmitted to another agent and converted it to a message
format understood by the receiving agent. The message format that was used
was based on Foundations of Intelligent Physical Agents (FIPA) [
110
]. The
message was then transmitted to the appropriate agent through the use of a
NASA-developed agent messaging protocol/software called Workplace [
7
].
