Optimizing for Change : Mixed-Initiative Resource Allocation with the AMC Barrel Allocator

Most real world planning and scheduling domains are only partially amenable to optimized solutions due to size of the solution space and due to the impossibility of fully optimizing a problem that changes over time. In addition, even an “optimal” solution to a problem is of little use if it is not comprehensible to end users of the system. In this paper we describe the AMC Barrel Allocator, a resource allocation tool for day-to-day airlift and air refueling management for the USAF Air Mobility Command (AMC). We discuss the design elements that promote user management of the solution space in terms of volume, complexity, and change. This work has relevance even for domains where fully automated operational usage is envisaged, but where system validation and user acceptance