System Identification and Control Using Singular Valve Decomposition Systolic Arrays

Period of Performance: 08/24/1988 - 01/01/1991

$450K

Phase 2 SBIR

Recipient Firm

Computational Engineering, Inc.
14504 Greenbelt Rd - Ste 500
Laurel, MD 20706
Principal Investigator

Abstract

IN THE PREVIOUS RESEARCH PHASE, A NEW CLASS OF ALGORITHMS BASED UPON A GENERALIZED SINGULAR VALUE DECOMPOSITION (SVD) WAS DEVELOPED AND DEMONSTRATED FOR SYSTEM IDENTIFICATION, STATISTICAL MODEL ORDER DETERMINATION, MODEL ORDER REDUCTION, AND PREDICTIVE CONTROL. A CANNONICAL VARIATE ANALYSIS (CVA) METHOD WAS USED IN DETERMINING THE OPTICAL STATE OF A RESTRICTED ORDER IN SYSTEM IDENTIFICATION, REDUCED ORDER STOCHASTIC FILTERING, AND MODEL PREDICTIVE CONTROL (MPC). IN THE CURRENT RESEARCH EFFORT, A SYSTOLIC ARRAY PROCESSOR BASED UPON SVDs IS BEING DEVELOPED AND DEMONSTRATED IN SYSTEM IDENTIFICATION AND CONTROL OF A LARGE SCALE SYSTEM. THE RESEARCH INCLUDES: REFINEMENNT OF THE STOCHASTIC MPC AND CVA ALGORITHMS; CONFIGURATION OF A HARDWARE SYSTOLIC ARRAY FROM AVAILABLE EQUIPMENT; DEVELOPMENT OF SOFTWARE ON THE SYSTOLIC ARRAY FOR SVD-BASED SYSTEM IDENTIFICATION AND CONTROL; AND DEMONSTRATION OF THE SYSTOLIC ARRAY PROCESSOR ON A LARGE SCALE SYSTEM. WHEN SUCCESSFUL, THIS WOULD REPRESENT THE FIRST STEP TOWARD WIDESPREAD USE OF THESE METHODS FOR REAL TIME AND LARGE SCALE APPLICATIONS. BENEFITS INCLUDE DEFENSE SYSTEMS REQUIRING ONLINE ADAPTIVE CONTROL INCLUDING FLUTTER SUPPRESSION, FAILURE DETECTION, CONTROL OF LARGE SPACE STRUCTURES, AND TARGET DETECTION AND TRACKING. COMMERCIAL APPLICATIONS EXIST IN CHEMICAL PROCESS CONTROL, CONTROL AND IDENTIFICATION OF POWER PLANTS, AND ADAPTIVE CONTROL IN INDUSTRIAL MANUFACTURING AND ROBOTICS.