Search

The U.S. Army's joint platform allocation tool (JPAT) is an integer linear program that was developed by the Army's Training and Doctrine Command Analysis Center and the Naval Postgraduate School to help inform acquisition decisions involving aerial reconnaissance and surveillance (R&S) resources. JPAT evaluates inputs such as mission requirements, locations of available equipment, and budgetary constraints to determine an effective assignment of unmanned aerial R&S assets to missions. As of September 2013, JPAT is solved using a rolling horizon approach, which produces a sub-optimal solution, and requires substantial computational resources to solve a problem of realistic size. Because JPAT is an integer linear program, it is a suitable candidate for using decomposition techniques to improve its computational efficiency. This thesis conducts an analysis of multiple approaches for increasing JPAT's computational efficiency. First, we reformulate JPAT using Benders decomposition. Then, we solve both the original and decomposed formulations using the simplex and barrier algorithms with multiple size datasets. In addition, we experiment with an initial heuristic solution and other techniques in our attempts to improve JPAT's runtime. We find that while Benders decomposition does not result in significant improvements in computation time for the instances considered in this thesis, initial solution heuristics and other modifications to the model improve JPAT's performance.

The U.S. Army's joint platform allocation tool (JPAT) is an integer linear program that was developed by the Army's Training and Doctrine Command Analysis Center and the Naval Postgraduate School to help inform acquisition decisions involving aerial reconnaissance and surveillance (R&S) resources. JPAT evaluates inputs such as mission requirements, locations of available equipment, and budgetary constraints to determine an effective assignment of unmanned aerial R&S assets to missions. As of September 2013, JPAT is solved using a rolling horizon approach, which produces a sub-optimal solution, and requires substantial computational resources to solve a problem of realistic size. Because JPAT is an integer linear program, it is a suitable candidate for using decomposition techniques to improve its computational efficiency. This thesis conducts an analysis of multiple approaches for increasing JPAT's computational efficiency. First, we reformulate JPAT using Benders decomposition. Then, we solve both the original and decomposed formulations using the simplex and barrier algorithms with multiple size datasets. In addition, we experiment with an initial heuristic solution and other techniques in our attempts to improve JPAT's runtime. We find that while Benders decomposition does not result in significant improvements in computation time for the instances considered in this thesis, initial solution heuristics and other modifications to the model improve JPAT's performance.

The U.S. Army's joint platform allocation tool (JPAT) is an integer linear program that was developed by the Army's Training and Doctrine Command Analysis Center and the Naval Postgraduate School to help inform acquisition decisions involving aerial reconnaissance and surveillance (R&S) resources. JPAT evaluates inputs such as mission requirements, locations of available equipment, and budgetary constraints to determine an effective assignment of unmanned aerial R&S assets to missions. As of September 2013, JPAT is solved using a rolling horizon approach, which produces a sub-optimal solution, and requires substantial computational resources to solve a problem of realistic size. Because JPAT is an integer linear program, it is a suitable candidate for using decomposition techniques to improve its computational efficiency. This thesis conducts an analysis of multiple approaches for increasing JPAT's computational efficiency. First, we reformulate JPAT using Benders decomposition. Then, we solve both the original and decomposed formulations using the simplex and barrier algorithms with multiple size datasets. In addition, we experiment with an initial heuristic solution and other techniques in our attempts to improve JPAT's runtime. We find that while Benders decomposition does not result in significant improvements in computation time for the instances considered in this thesis, initial solution heuristics and other modifications to the model improve JPAT's performance., http://archive.org/details/improvingarmysjo1094537635, Lieutenant, United States Navy, Approved for public release; distribution is unlimited.