An Exploratory Prototype for Reactive Management of Aeromedical Evacuation Plans

This report summarizes the results of an initial, four month project demonstrating the applicability of the DITOPS scheduling system to USTRANSCOM’s Aeromedical Evacuation (MEDEVAC) re-planning problem. DITOPS is an advanced prototype system developed at Carnegie Mellon University for development, analysis and revision of large-scale schedules, applied originally to the logistics domain of strategic deployment. DITOPS implements a reactive, constraint-based approach to scheduling, providing techniques and a system architecture for efficient, localized revision of plans/schedules in response to changed constraints or decisions. Using DITOPS, a prototype medical evacuation re-planner has been designed and implemented for comparison with Carnegie Group’s TRAC2ES reactive planner module. From a system development perspective, DITOPS is a toolkit and a class library for configuring planning and scheduling applications; the approach to application design and construction relies on object-oriented programming techniques and software reuse, allowing applications to be constructed as a “differential” process, focusing primarily on the differences between existing software and the system being constructed. During this project, the medical evacuation domain was modeled using the core modeling primitives available in DITOPS, and rescheduling techniques were developed for responding to common medical evacuation replanning problems, again using existing components from DITOPS. The evacuation re-planning prototype is currently capable of handling a substantial set of disruptive events, while maintaining feasible patient itineraries on multi-leg missions. In this regard, the applicability of the DITOPS modeling and scheduling framework to this problem and the efficacy of the object-oriented approach to system configuration and customization has been clearly demonstrated. At the same time, it is important to recognize that the planner has been constructed in a very short period of time. Though the system’s current replanning capabilities are quite sophisticated, a systematic evaluation has not been performed and we would expect further work in this area to lead to algorithmic customizations and improvements. The underlying system architecture provides a flexible basis for system expansion and refinement.