Reactive Capabilities

This chapter will describe the different types of ``disruptions'' or disruptive events the reactive planner is capable of handling. By disruption we mean any action performed by the external user that will change the internal state of the system and will, eventually require some reactive or corrective behavior. We assume that all disruptions are introduced through the graphical user interface. For example, to introduce new patients in the system, the user has to load a file with the patient descriptions; these descriptions will be translated into internal objects. The creation of these objects tells the system that some disruption has occurred. In response to the disruption, the patient router will be triggered or executed to schedule the patients. In this sense, all the system behavior can be considered reactive, the generative behavior being only a special case.

The basic reactive mode of operation can be divided into three clearly defined steps:

• Introduction of a disruptive external event. The user, through the Graphical Interface, communicates to the system that changes have occurred in the real world. The types of disruption accepted are: new patients, mission delay, mission cancelled, patient goes sour, aircraft breakdown, and airport closed.
• Once the disruption is introduced, the system updates the internal state of the current solution. This involves the translation of the external event into changes in the schedule representation. For each type of disruption, some unconditional changes will be introduced. These unconditional changes represent the modeling assumptions on how the state of the system evolves in response to specific disruptions. The process of changing the state of the solution may cause conflicts or inconsistencies that are signaled by the system. Conflicts are signaled as a consequence of one or more constraint violations and are classified according to the type and severity of the violation. Even if no constraint has been violated, the system will still signal that an external disruption has occurred since some action can eventually be performed to improve the solution.
• In response to internal events being signaled, the problem-solving control architecture is invoked. The problem-solving mechanism cycles through a series of knowledge intensive actions on the set of pending events. These actions aims at identifying and performing the best possible correction and include event aggregation, event selection, event analysis, and conflict solution. The analysis process will acquire all the available information about the conflict and, based on its knowledge about available problem solving methods, suggest some corrective actions. Notice that no modification on the state of the solution is introduced in the analysis phase.

  The problem-solving methods are then selected based on a combination of user preferences and analysis results. These methods, also called knowledge-sources are the ones that will actually change the solution in order to resolve the conflicts.

• Disruptive Events