As mentioned in Section 3.2, the communication between planner and scheduler can
be seen as policy suggestions by the planner about scheduling variables, their domains
and constraints. The scheduler responds by flagging success or failure with the suggested
parameters. If scheduling method fails to allocate resources in the context of given
resources, time limit and nature of allocation policy, the responsibility transfers to the
planner to change any of the permissible parameters and try again. The planner also has
the option to take up nonabstracted planning at any stage. If resource allocation succeeds,
the schedule and the allocation policy are used to derive an executable plan.
The different classes described in the previous section can be interpreted in the form
of resource allocation policies. The different policies supported include maintaining the
concurrencyof the plan, serializing the plan and inserting actions to free and reallocate the
resources. The DCSP formulation allows the scheduler to interpret the resource allocation
policy prescribed by the planner (see Fig. 9) in terms of constraints on the values of
variables.
Table 6 summarizes the different policies and what they imply in terms of legal values
of variables. Maintaining concurrency of the plan corresponds to all actions A i in the