Szechtman, Roberto

We propose to develop models and algorithms to support surveillance and interdiction in naval and homeland security applications. The research will comprise three problems.

The first problem focuses on searching for a target (e.g., terrorist, adversary's ship), which moves among several given locations. The objective here is to study the pattern of this movement and thus infer the behavior and intentions of the target. In this setting the problem is how to dynamically employ a suite of sensors over time such that the movement pattern can be identified as fast as possible.

The second problem addresses a related situation where a searcher tries to detect targets that appear over a certain area according to a Poisson process. The search space is partitioned into area-cells, and targets not detected in a certain time period may disappear (renege). The problem for the searcher is to come up with a dynamic search policy that balances learning the intensity of arrivals into each area-cell, with exploiting the accumulated knowledge to-date in order to focus on the cells that appear most promising to have valid targets.