Toward a Portable Parallel Library for Space-Time Adaptive Methods
Lebak, James M.; Durie, Robert C.; Bojancyk, Adam W.
Space-time adaptive processing (STAP) refers to a class of methods for detecting targets using an array of sensors. The output of the array is weighted using data collected from the sensors over a given period of time. An optimal method of calculating weights exists; however, this method is usually computationally impractical. Therefore, various heuristic methods are used that approximate the optimal method. These heuristics use many of the same operations and are computationally demanding. We are in the process of constructing a portable, parallel library of subroutines useful for constructing STAP heuristics. As a first step in this process, we implemented one STAP heuristic, higher-order post-Doppler processing, using three different parallel methods on the IBM SP2 and the Intel Paragon: these methods characterize different parallel approaches to the STAP problem. From implementing these algorithms, we have been able to identify components for our parallel library. We propose models for some of the components and give preliminary timing results for the parallel methods.
Previously Published As