MSTAR Search
 

 

Customer DARPA ISO: Dr Robert Hummel

Agent Air Force Research Labs (AFRL): Joseph Diemunsch

Program Goals
A key to the computational feasibility of hypothesize-and-test processing in PEMS is the ability to search hypothesis space efficiently; indeed, having this ability is essential for scaling the model-based approach to realistic-sized problems.  The goal of our MSTAR Search program is to design and build in software a search engine that provides high performance target recognition with computational efficiency.

Key Technology
We are meeting our development goals for search through hierarchical application of numerical optimization techniques and intelligent search heuristics.  These are supported by rigorous Bayesian evidence accrual which summarizes evidence into a coherent statistical picture to aid decision making.  The architecture of our search engine is based on an "estimation-control" paradigm as illustrated below.

 

 
 
 The heart of search is an inference engine which selects algorithmic actions to perform based on the current state of hypothesis space, as conveyed by evidence accrual processes.  The actions which can be performed are defined by associating an algorithmic operator, such as a hypothesis refinement, to a data object, such as a particular hypothesis.  The operators which currently exist in the MSTAR system can broadly be organized into the "toolkit" shown in the companion figure.

Key Technology Products
Under this effort, ALPHATECH is developing the following re-usable technology products:

Points of Contact
Dr. John Wissinger, MSTAR Program Manager, Search PI: (781)-273-3388, x239
Mr. Robert Ristroph, Search Software Lead: (781)-273-3388, x339
Dr. Gil Ettinger, Technical Support: (781)-273-3388, x334
Dr. William Irving, Technical Support: (781)-273-3388, x285

Subcontractor and Consultant POCs:
Mr. Bill Severson, Lockheed Martin Astronautics: (303)-977-2577
Mr. Doug Morgan: (650)-964-7451

 Back to MSTAR Top Level Page

Last Modified 12/07/98