The domain of C4-I process management affords many difficulties that are not present in traditional planning environments. The standard methodology of planning, schedule, execute, needs to be modified into a more continuous process. There must be a closed loop, with the results of execution monitoring being fed back into the planning process.
The assumption that all the relevant data is present and in an expected format breaks down. Information may be absent, and the system may need to plan to gather it, or the information may be implicit in data of another form, or the necessary information may not be available yet. Each of the these contingencies leads away from the traditional planning process that is given a fixed well defined problem to solve
Planning must be carried on simultaneously at many different levels. These different levels use very different ways of describing the world, but they must all be linked together, while progress at one level must not rely on complete progress at other levels. In particular higher levels of abstraction must not be forced to wait for lower levels to instantiate plans.
The paradigm of continuous planning also entails that the base-line of resources and objectives may change over time. The nature of the planning problem may develop in radically different ways depending on the information received as the execution is monitored. Thus when monitoring plans the system must be prepared to accept changes in what the underlying facts in the ground are, without needing to replan from scratch.
Continuous planning is done in real-time. There is not an arbitrarily large amount of time to work out problems in all details. For this reason, trade-offs between planning and acting must be made. If these trade-offs are to be made in a sensible way, meta-level planning that considers both the planning problem, and the meta-problem must be carried out.
In real world domains, information has an uncertainty associated with it. This uncertainty may be as a result of a lack of precision in the data, or the nature of the world, or because the information has come from an unreliable source. All these sources of uncertainty must be represented and taken into account if well-informed decision are to be made.
The planning domain necessarily deals with large amounts of unstructured data. It is not feasible to force a structure onto all the planners that the system must manage. In particular management must deal with highly concurrent, unsynchronized sub-systems.
In traditional planning, once the plan is made the p[lanner is finished. However, when execution monitoring is fed back, the planner can monitor for assumption violation, and replan to avoid that assumption. If this is to be effectively carried out, assumptions must be represented, detected, and monitored as more data becomes available.
The data that planner must use in his domain necessarily comes from a wide variety of sources, including importantly knowledge acquisition, and human experts. Thus the planner must be able to take information in a human understandable way, if it is to be informed by humans.
Successful planning depends on the use of many levels of abstraction. Planning must be capable of being carried out a higher levels without considering lower levels in detail, if the benefits of hierarchical planning are to ensue. The relationships between planning levels are complex, and each level must be able to convey information to other levels, without necessarily lining up each level exactly.
Abstraction must be driven by the nature of the planning at that level. This demands that tasks, objectives, and resources drive the abstraction, not the other way around. Time must be dealt with in a intelligent manner. Its granularity must not be fixed, it should be possible to plan with varying levels of granularity, especially as the plan progresses into the future.
All the aspects of the planning process must be quantifiable if the process is to be managed effectively. This includes judging the efficacy of plans at different levels, independently, and being able to use this information to drive both the planning and meta-planning process.