COMPUTER ARCHITECTURE FOR KALMAN FILTERING

TO MEET THE HIGH THOUGHPUT REQUIREMENTS OF DOD SIGNAL PROCESSING APPLICATIONS, HIGHLY PARALLEL ARCHITECTURES NEED TO BE DEVELOPED. IN PARTICULAR, TO PERFORM MULTIDIMENSIONAL KALMAN FILTERING AT SAMPLE RATES APPROACHING 15 KHZ INNOVATIVE TECHNIQUES ARE NEEDED TO PARTITION THE PROBLEM FOR REAL-TIME IMPLEMENTATION. THEREFORE, A UNIQUE DECOUPLING OF THE PREDICTOR AND CORRECTOR EQUATIONS IN THE KALMAN FILTER IS PROPOSED TO SPEED-UP COMPUTATIONS BY TWO (2) ORDERS OF MAGNITUDE (100X). THE PROPOSED PARALLED ALGORITHMS AND ARCHITECTURES ARE WELL-SUITED FOR RECURSIVE FILTERING, IMAGE AND SIGNAL PROCESSING FOR NEXT GENERATION INSTRUMENT RADAR AND TELEMETRY EQUIPMENT.