AFOSR/PK, 110 Duncan Avenue, Room B115, Bolling AFB, DC 20332-8050

A -- INNOVATIVE COMPUTATIONAL MATHEMATICS FOR PHYSICAL APPLICATIONS SOL AFOSR BAA 98-4 DUE 093098 POC Dr. Arje Nachman (AFOSR), (202) 767-4939; Dr. Anna Tsao (DARPA), (703) 696-2287; Dr. Dennis Healy (DARPA), (703) 696-0143 WEB: Doing Business with AFOSR, http://www.afosr.af.mil. E-MAIL: None, arje.nachman@afosr.af.mil. AFOSR, on behalf of DARPA's Applied and Computational Mathematics Program, is seeking projects demonstrating significant mathematical innovation and high DoD payoff potential in the following technical areas: 1) Physics-based modeling and signal processing applied to the end-to-end design and optimization of DoD sensor systems; 2) Modeling, data analysis, or scalable, high order numerical methods for electromagnetic, sensing, chemical, and biological applications; and 3) Optimized Portable Application Libraries. Proposers are strongly encouraged to discuss prospective proposals with the AFOSR and DARPA technical POCs in advance of proposal abstracts or full proposals to get an indication of potential interest. 1. Physics-based modeling and signal processing applied to the end-to-end design and optimization of DoD sensor systems. 1. The emphasis of this topic is to develop and apply innovative mathematical methods for achieving substantial and demonstrable improvements over currently available state-of-the-art capability in DoD sensor/processing systems. Candidate systems of interest include, but are not limited to, imaging radar and sonar including synthetic aperture radar (SAR) and synthetic aperture sonar (SAS), foliage and ground penetrating radar, laser detection and ranging (LADAR), infrared sensors, acoustic sensors, hyperspectral systems and adaptive array systems. Some examples of challenging applications include automatic target recognition/detection/identification (ATR/D/I), imaging in difficult clutter and noise environments such as that produced by landscape features and speckle, forward battlefield awareness, nondestructive evaluation, and unexploded ordinance/mine detection and recognition. Methods developed should have potential to significantly advance both theory and application in the areas proposed. 2. Modeling, data analysis, or scalable, high order numerical methods for electromagnetic, sensing, chemical, and biological applications. A major emphasis of this topic is on new mathematically innovative modeling and algorithmic approaches to electromagnetic, sensing, biological and chemical problems of interest to the DoD. In addition, innovative data representations are sought to allow effective approaches to managing the statistical "curse of dimensionality" associated with high dimensional modeling, data analysis and mining. Applications of interest include antenna design, tracking, materials processing, mine detection, and biological warfare defense. 3. Optimized Portable Application Libraries. This topic is aimed at the development of mathematical formulations to enable automatic compilation of scalable, high performance software libraries of key numerical kernels for sensor applications. In addition, algorithmic approaches should facilitate systematic and effective machine-specific tailoring of computational constructs such as data structures and communication protocols on platforms having deep memory hierarchies and multiprocessor architectures. One of the persistent problems in the current computational environment is insufficiently effective error control. The error in computational analyses is dependent not only on the quality of the algorithms for performing individual steps in a simulation, but on the ability to guarantee that the outputs at each step satisfy the initial requirements of subsequent steps. Approaches should include computational and mathematical models and theory for analyzing and predicting both performance and error properties for mathematically equivalent algorithmic formulations. Historically, it has been difficult to realize algorithms that simultaneously guarantee low computational complexity, low memory usage, accuracy, high efficiency, and portability across diverse computer architectures. Algorithmic design is further complicated by the disparity in mathematical, computational, and application considerations that often work at cross purposes when addressed in a stovepipe fashion. Finally, the need to re-design and implement algorithms to run efficiently on each new platform is extremely costly. The emphasis of OPAL is to surmount these difficulties with coordinated formulations and toolkits that would ultimately include automatic or semi-automatic code generation programs based on the research. Platforms of interest include high performance uniprocessors, tightly coupled multiprocessor systems, and adaptive computing platforms. The viability of the approach should be demonstrated through application to problems of real interest to the DoD. Note 26. (0071)

Loren Data Corp. http://www.ld.com (SYN# 0014 19980316\A-0014.SOL)