Advanced Computational Algorithms for Nonlinear Filtering for Real Time Environment

The techniques of linear filtering have contributed tremendously in missile guidance and modern radar technology. However, they can only serve as a first order approximation to the infinite-dimensional nonlinear filtering problem. Solutions to these problems are known in theory, and with the explosive growth of high performance computing, they have also increasingly become available in terms of practical algorithms. The objective of the proposed Phase II effort is to critically examine, categorize and evaluate existing nonlinear filtering algorithms, and develop software tools for using them. For target tracking as a focus area, we will develop suitable benchmark problems to compare filtering algorithms, avoiding many of the common pitfalls. The proposed team consists of Scientific Systems Company Inc. (SSCI), Professor Amarjit Budhiraja at University of North Carolina at Chapel Hill, and Dr. Ronald Mahler at Lockheed Martin Tactical Systems. SSCI has extensive expertise in control and signal processing and in algorithm/software development. Professor Budhiraja is one of the leaders in the study of stochastic filtering theory, and Dr. Mahler is a renowned expert in multi-sensor multi-target tracking. The Phase II tasks are: (1) investigate dynamic parameter estimation methods and develop corresponding algorithms, (2) code the algorithms developed in Phase I and Phase II, and construct a coherent and well-documented library, (3) develop a Particle Filters Wizard to help a general user in using Particle Filters, (4) develop a Multi-Environment Nonlinear Filtering Tracking Assessment Testbed (MENTAT), (5) investigate single-target (and multi-target) nonlinear filtering algorithms in the framework of Finite Set Statistics (FISST), and (6) work towards Phase III, commercialization, and transfer of technology.