4 The Architecture of INSPECT

 

Figure 1 displays the architecture of INSPECT. The system takes the following inputs:

• An air campaign plan stored in ACPT. The plan does not need to be complete. However, since many of the problems in INSPECT's library refer to completeness, INSPECT may detect an inordinate number of errors in draft plans. Also, for several of its checks, INSPECT relies on the availability of structured descriptions of the objectives according to the grammar proposed in another related document and incorporated in an extended objectives editor in the IFD-4 version of ACPT [Valente et al., 1996]. If those descriptions are not available, INSPECT's evaluation capability may become severely limited.
• The friendly order of battle (FOB). The FOB contains information on the units available for the campaign. This information is used to check, for instance, if all the types of aircraft needed for executing the plan are effectively available.

  [IMAGE ]
Figure 1: The architecture of the INSPECT system.

Future versions of INSPECT will use other inputs, such as the Enemy Order of Battle (EOB, which contains information about enemy targets), measures of measure (MOM, also known as measures of effect or MOE), Rules of Engagement (RoE) and Commander's Guidance.

INSPECT produces as output an agenda of items which represent problems or inconsistencies detected in the plan. The INSPECT agenda is presented to the user, who need to revise and eventually solve the problems indicated.

INSPECT is seamlessly integrated with ACPT. As a result, INSPECT does not interact directly with the user (an air campaign planner); all interfacing is done through ACPT. The version of ACPT demonstrated in IFD-4 contains a new menu for controlling INSPECT that submits a plan to INSPECT and controls a window for displaying an agenda with the problems detected by INSPECT. Further, it has a specific window for specifying several preference parameters for INSPECT's operation, such as which types of inconsistencies INSPECT will report. Technical details about the integration of INSPECT and ACPT can be found in section 6.1.

4.1 The Role of EXPECT

As shown in figure 1, INSPECT is a knowledge-based system created by EXPECT. INSPECT includes a knowledge base (entered through EXPECT) of plan evaluation criteria that it uses to analyze air campaign plans. EXPECT provides support for acquiring knowledge about new criteria and changing existing knowledge. The main advantage of using EXPECT to build INSPECT is to achieve adaptability. To continue to be useful, INSPECT must be adaptable. Over time, it will be desirable to expand its library of problems and modify it as new techniques and military technologies become available. In addition, since INSPECT will be used in different theaters, it may be necessary to modify INSPECT's knowledge base to handle a particular campaign. In either case, we would like expert planners to be able to make these changes themselves, thus reducing their need to rely on programmers.

The EXPECT framework, which was used to construct INSPECT, was designed to allow non-programmers to modify a system, while freeing them from the need to understand implementation details. The key idea behind EXPECT is that it captures important information about how a knowledge-based system was put together and uses this info to guide a user in modifying the knowledge base. In the case of INSPECT, this empowers planning experts to change the system and thus helps assure that it will grow and remain relevant and useful in the years to come.

4.2 Scope of INSPECT

The following are assumptions that restrict the scope of the current version of INSPECT. Most were introduced to make feasible the effort of building a version of the system for IFD-4, and can be raised for future versions.

• INSPECT detects inconsistencies in Air Objectives and Air Tasks. It may be able to check some aspects of CINC's Objectives. Some inconsistencies can easily be extended for other levels of the plan, while others would require a specific knowledge acquisition effort.
• INSPECT does not try to find problems in the global strategy used by the planner in constructing the air campaign plan. Instead, it tries to find local errors and mistakes. In other words, INSPECT concentrates on the product (the air campaign plan) instead of the process (of producing this plan). As a consequence, many of the errors INSPECT tries to find have a ``grammatical' flavor. Therefore, the current version of INSPECT can be compared roughly to a spell and grammar-checker for air campaign plans.
• INSPECT does not guarantee that all the errors it finds are actual mistakes in the plan. This is due to the limitations in the current knowledge about air campaign plans: what looks like an error may be in fact a creative (and unusual) solution used by a planner. The last word on what is correct or good in ACPlans is always that of the experts in the Air Force.
• Using INSPECT does not create the best plan. INSPECT does not try to evaluate the quality of a plan with respect to the campaign it is related to and its goals, but rather concentrates on the quality of the formulation of the plan.
• INSPECT only finds the errors that are in its library of types of plan problems. This means that an error not in the library will most likely pass undetected.

4.3 Future Extensions

Within the same concept (consistency of an air campaign plan) several extensions of INSPECT can be proposed:

Evaluate doctrine

The system can be extended to evaluate which aspects of the doctrine are being followed and which are not. This evaluation can be used in a proactive way, i.e. would not only have available a report of the evaluation of the plan doctrinewise, but would allow the user to e.g. ask what changes can be made to improve the plan in terms of doctrine.

Commander's guidance

The system can be extended to evaluate whether or not the current plan is consistent with the policies and guidance defined by the commanders (CinC/JFACC).

These extensions have been discussed with Checkmate during the design of INSPECT. They explore the idea that some of the critiquing they would do can be seen in a way as a consistency verification.