Each year, the United States Army recruits and trains thousands of new soldiers to fill vacancies in Army organizations created by promotion, transfer, or termination of service. Installations responsible for training new recruits is conducted in two phases: Basic Combat Training followed by Advanced Individual Training. Proper management of the Army's initial entry training program is a very complex, practical military logistics problem that demands timely scheduling of a broad range of reusable training resources, such as, training companies. Currently, manual heuristic methods are used to schedule training companies throughout the planning horizon to support initial entry training, where training company scheduling also involves deciding how many recruits to assign to training companies each week. These methods have evolved over a number of years when there were few changes to the training base, and recruiting levels remained relatively stationary. Unfortunately, there are several severe shortcomings with these methods. For example, determining the number of recruits assigned per training company and the number of weeks a training company remains busy training recruits is a manual trial-and-error process. Second, it is possible for different analysts to generate different solutions for the same recruitment scenario. Third, no methods exist for conducting comparative analyses to appraise the quality of competing feasible training schedules. Finally, the temporal interdependence of decisions makes decision variables in the future periods depend on current decision variables. This complicates resource scheduling and makes the manual generation of week-by-week training schedules a tedious, time-consuming task. This dissertation: (1) formulates a mathematical dynamic model of the Basic Combat Training phase of initial entry training; (2) formulates a decision model for optimally scheduling training resources based on dynamic programming; (3) formulates an improved heuristic procedure for scheduling training resources; (4) incorporates a "training quality" performance measure into the formulation of the objective function making it possible to compare competing feasible training schedules obtained by various methods; and (5) designs, develops and implements a fully operational computer-based decision support system (DSS) for scheduling basic training resources. The computational experiments reveal that the heuristic procedures developed are indeed computationally efficient and provide "good" solutions in terms of training "quality," resources utilization, and training cost.
展开▼
