Environmental Engineering Reference
In-Depth Information
While symbolic planners vary in detail, they generally start the planning
activity by examining the goals in light of the current environment and then
break the goals into a series of tasks. These tasks are themselves examined and
broken into simpler subtasks. This iterative refinement process continues until
a point is reached where the tasks define a series of steps at an appropriate level
of abstraction for plan implementation. The plan is executed and feedback is
generated on the success of the tasks. The higher level system is signaled when
objectives have been met. If a failure occurs in one or more of the tasks, the
planner modifies its plan and the cycle is repeated. If the planner exhausts all
of its options and still the objectives have not been met, the planner signals
to the higher level that it has failed.
Symbolic planners use many different strategies for choosing among the
potential plans. They often spend large amounts of computer resources gen-
erating plans and selecting the best ones. They have diculties in situations
where the decision cycle time is short, or where actions and failures are not
deterministic.
5.1.3 Reactive Planners
Reactive planners are specifically designed to make rapid choices in time crit-
ical situations. They can be designed around either symbolic or numeric rep-
resentations.
Figure
5.4
shows the structure of a reactive planner. Reactive planners
begin by evaluating the available plans in light of the current context and
then choosing the most appropriate. These plans are simple in nature and are
designed to execute immediately. Once the plan is selected and being executed,
the planner monitors the current situation and only changes the plan if the
Possible
Plans
Select Best
Plan
Goals
Execute
Plan
And
Return
Results
World
Model
Update
Model
Fig. 5.4.
Reactive planner
