Over the last 10 to 20 years, heuristic search in the Operations Research and Artificial Intelligence communities has focused on developing high level general purpose algorithms, such as Tabu Search and Genetic Algorithms. However, understanding of when and why these algorithms perform well still lags. Our project extended the theory of certain combinatorial optimization problems to develop analytical characterizations of portions of search spaces and as the basis for creating ...
We investigated a new scheduling application under uncertainty: the Eglin AFB phased array radar tracks thousands of objects in space. Because of uncertainty associated with an object's location and radar cross-section, as well as the inherent power limits and weather, scheduling must trade-off maximizing the probability of detection against maximizing the total number of objects scheduled for observation; the system must also dynamically reschedule missed observations and incorporate new requests. ...
Because of the difficulties of obtaining data from real applications, researchers tend to develop their new algorithms on artificial problems and do not model what makes an algorithm successful. Developers then have little guidance on which algorithms are best for which applications. All of this makes it difficult for research results to transfer to deployment. Our project endeavored to ameliorate this situation by 1) modeling the topology of scheduling algorithms ...
Reports by Author
