Planning for an adaptive evader with application to drug interdiction operations

• Main Author: Gift, Philip D.
• Other Authors: Royset, Johannes O.
• Published: Monterey, California. Naval Postgraduate School 2012
• Online Access: http://hdl.handle.net/10945/5161

Approved for public release; distribution is unlimited === In an effort to impede the flow of drugs from South America, a Coalition Force headed by Joint Interagency Task Force (JIATF) - South allocates its assets to detect and interdict drug smuggling vessels such as the self-propelled semi-submersible (SPSS) used by a Drug Trafficking Organization (DTO). In this thesis, we develop an interdiction model to place the Coalition Force assets optimally. We also develop a model - known as the Adaptive Evader Model - for a DTO that is able to learn the placement of the Coalition Force assets. This model is akin to the multiarmed bandit problem. We create two algorithms for the Adapting Evader Model. One algorithm uses an optimal learning policy and the other uses a heuristic learning policy. We also create an algorithm for the interdiction model using the Cross-Entropy method. Finally, we construct a case study that we use to draw some insights about how a DTO, that is capable of learning, reacts to different optimal plans. This information can be used by the Coalition Force to more effectively allocate their limited number of assets during drug interdiction operations.