A Decision Logic Approach to the Port-of-Entry Inspection Problem

Rutgers University New Brunswick, New Brunswick NJ

Investigators

Abstract

Finding ways to intercept illicit nuclear materials and weapons destined for the U.S. via the maritime transportation system is an exceedingly difficult task. Today, only a small percentage of ships entering U.S. ports have their cargoes inspected. The purpose of this study is to develop decision support algorithms that will help us to optimally intercept illicit materials and weapons and to test the algorithms on real data arising from port-of-entry inspection. The algorithms sought will find inspection schemes that minimize total cost, including "cost" of false positives and false negatives. Imagine a stream of entities arriving at a port and a decision maker having to decide how to inspect them, which to subject to further inspection and which to allow to pass through with only minimal levels of inspection. This is a complex sequential decision making problem. The ultimate goal of an inspection scheme is to classify an entity into important categories and states. In the simplest case, there are two categories ("ok" and "suspicious") and two states ("present" or "absent"). This classification can be described as involving a boolean decision function. Different binary tree representations for a boolean decision function have different inspection costs associated with them and an optimal decision tree representation is sought. Modeling sensors used to test for attributes makes the problem more realistic and brings in possible inspection errors and false positives and negatives. Extending the problem to more than two categories and more than two states also makes the problem more realistic. The project will be carried out in collaboration between a university team and a team from the Los Alamos National Laboratory and will initially follow an approach pioneered at Los Alamos. Four graduate students will be key participants in the project and many others will be involved through associated public events. The topic lends itself well to undergraduate research and participating faculty will coordinate topics with an existing undergraduate research program that caters to students from all over the U.S. To broaden the impact of the project, a seminar series will be publicized to neighboring institutions and a public workshop will be held. The workshop will include talks of an expository/tutorial nature, making it appealing to non-specialists and students. A project website and a project book will help to disseminate project results and ideas widely.